Systems, devices and methods relating to drug dosage guidance - Patents.com

Abstract

Systems, devices, and methods are provided for determining a patient's or user's drug dosage. The dosage determination can take into account the patient's or user's recent and / or historical analyte levels. The dosage determination can also take into account other information about the patient or user, such as physiological information, dietary information, activity and / or behavior. Many different dosage determination embodiments are shown that relate to a wide variety of different aspects of the systems or environments in which the embodiments can be implemented.

Description

Related Applications

[0001] This application claims priority to and the benefit of U.S. Provisional Patent Application No. 62 / 882,249, filed August 2, 2019, U.S. Provisional Patent Application No. 62 / 979,578, filed February 21, 2020, U.S. Provisional Patent Application No. 62 / 979,594, filed February 21, 2020, U.S. Provisional Patent Application No. 62 / 979,618, filed February 21, 2020, and U.S. Provisional Patent Application No. 63 / 058,799, filed July 30, 2020, the entire contents of which are incorporated herein by reference for all purposes. [Technical Field]

[0002] The subject matter described herein relates generally to systems, devices and methods related to drug dosage guidance, such as, for example, determining insulin dosage to treat elevated glucose levels due to diabetes. [Background technology]

[0003] Detecting and / or monitoring analyte levels, such as glucose, ketones, lactate, oxygen, and hemoglobin A1C, can be crucial to the health of individuals with diabetes. Patients suffering from diabetes can suffer from complications, including loss of consciousness, cardiovascular disease, retinopathy, neuropathy, and nephropathy. Diabetic patients generally need to monitor their glucose levels to ensure that they are maintained within a clinically safe range, and can also use this information to determine whether and / or when insulin is needed to lower their glucose levels or whether additional glucose is needed to raise their glucose levels.

[0004] A growing body of clinical data demonstrates a strong correlation between frequency of glucose monitoring and glycemic control. However, despite this correlation, many individuals diagnosed with diabetes do not monitor their glucose levels as frequently as necessary due to a combination of factors, including the inconvenience, testing discretion, pain associated with glucose testing, and cost.

[0005] For patients who rely on the administration of medication (e.g., insulin) to treat or manage diabetes, it would be desirable to have a system, device, or method that can automatically utilize glucose information collected by an analyte monitoring system to provide medication dosage guidance as needed and in an easily accessible manner. It would further be desirable for such a system, device, or method to take into account the physiology, diet, activity, and / or behavior of the user or patient receiving treatment when providing such medication dosage guidance, which may itself improve accuracy and reliability. Furthermore, in some situations, it would also be desirable for such a system, device, or method to be able to automatically deliver a selected medication dose. Summary of the Invention [Problem to be solved by the invention]

[0006] For these and other reasons, there is a need for improved systems, methods and devices related to medication dosage guidance. [Means for solving the problem]

[0007] Exemplary embodiments of systems, devices, and methods related to medication dose guidance, and in some embodiments, medication delivery, are provided herein. According to one aspect, many of the embodiments described herein comprise a dose guidance system including a display device, a sensor control device, and a medication delivery device. The dose guidance system can include a dose guidance application (e.g., software) capable of determining and outputting dose guidance (e.g., recommendations regarding dosage, corrections, and titration) to the patient. Furthermore, according to some embodiments, the dose guidance system can learn the patient's dosing strategy during a learning period during which the dose guidance system can estimate key dosing parameters. According to some embodiments, the dose guidance system can also provide guidance for titration and corrections once the system is configured with the patient's current dosing strategy. The dose guidance system can also provide guidance for different meal dosing scenarios. For example, in some embodiments, the dose guidance system can provide dose guidance at or before the start of a meal or after the start of a meal. The dose guidance system can also provide dose guidance for additional meals (e.g., dessert) or for "touch-up" dosing to address high post-prandial glucose levels. Exemplary system and safety features of the dose guidance system are also described.

[0008] Many of the embodiments provided herein include improved software functionality or graphical user interfaces for use with analyte monitoring systems that are highly intuitive and user-friendly, providing rapid access to a user's physiological information. More specifically, these embodiments enable a user (or HCP) to quickly determine appropriate medication therapy based on information related to the user's physiological condition, past dosing patterns, and other factors, without having to go through the tedious task of reviewing large amounts of analyte data. Furthermore, some of the GUIs and GUI functionality enable users (and their caregivers) to better understand and improve their dosing patterns and subsequent hypoglycemic and hyperglycemic episodes. Similarly, many other embodiments provided herein include improved software functionality for dose guidance systems, improving the dose guidance provided to users by enabling safe titration strategies that minimize hypoglycemic episodes; methods for modifying dose guidance depending on the timing of dose administration relative to meal start times (e.g., before, at, or after meal start); consideration of real-world events that affect dosing strategies; and post-meal alarms based on predicted probability of occurrence rather than thresholds, to name just a few. Other improvements and advantages are also provided. Various configurations of these devices are described in detail in exemplary embodiments only.

[0009] Other systems, devices, methods, features, and advantages of the subject matter described herein will become apparent to one of ordinary skill in the art upon examination of the following figures and detailed description. All such additional systems, devices, methods, features, and advantages are intended to be included herein, be within the scope of the subject matter described herein, and be protected by the accompanying claims. Features of the exemplary embodiments should not be construed as limiting the scope of the appended claims unless those features are expressly recited in the claims. [Brief explanation of the drawings]

[0010] Details of the subject matter described herein, both as to its structure and operation, may become apparent from examination of the accompanying drawings, in which like reference numerals refer to like parts. The components in the drawings are not necessarily to scale, emphasis instead being placed upon illustrating the principles of the subject matter. Furthermore, all figures are intended to convey concepts, and relative sizes, shapes, and other detailed attributes may be shown diagrammatically, rather than literally or precisely. [Figure 1A] 1 shows a block diagram of an exemplary embodiment of a dose guidance system. [Figure 1B] 1 shows a block diagram of an exemplary embodiment of a dose guidance system. [Figure 2A] FIG. 1 is a schematic diagram illustrating an exemplary embodiment of a sensor control device. [Figure 2B] FIG. 1 is a block diagram illustrating an exemplary embodiment of a sensor control device. [Figure 3A] 1 is a schematic diagram illustrating an exemplary embodiment of a medication delivery device. [Figure 3B] 1 is a block diagram illustrating an exemplary embodiment of a medication delivery device. [Figure 4A] FIG. 1 is a schematic diagram illustrating an exemplary embodiment of a display device. [Figure 4B] FIG. 1 is a block diagram illustrating an exemplary embodiment of a display device. [Figure 5] FIG. 1 is a block diagram illustrating an exemplary embodiment of a user interface device. [Figure 6-1] FIG. 1 illustrates an exemplary glucose pattern report. [Figure 6-2] This is a continuation of Figure 6-1. [Figure 6-3] This is a continuation of Figure 6-2. [Figure 7] FIG. 1 is a flow diagram illustrating an example embodiment of a process flow of a portion of a dose guidance application directed to a learning method for estimating a patient's insulin dosing habits. [Figure 8A]FIG. 1 is a flow diagram illustrating an exemplary embodiment of a process flow for operation by a dose guidance application to evaluate meal bolus titration for multiple injection (MDI) dosing therapy. [Figure 8B] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a process flow for operation with a dose guidance application for glucose pattern analysis (GPA). [Figure 8C] FIG. 1 illustrates an exemplary embodiment of a graph showing information for determining hypoglycemic risk for GPA and other indicators. [Figure 8D] FIG. 1 is a flow diagram illustrating aspects of various exemplary embodiments of an algorithm for assessing meal bolus titration for MDI insulin dosing therapy. [Figure 8E] FIG. 1 is a flow diagram illustrating aspects of various exemplary embodiments of an algorithm for assessing meal bolus titration for MDI insulin dosing therapy. [Figure 8F] FIG. 1 is a flow diagram illustrating aspects of various exemplary embodiments of an algorithm for assessing meal bolus titration for MDI insulin dosing therapy. [Figure 8G] FIG. 1 is a flow diagram illustrating aspects of various exemplary embodiments of an algorithm for assessing meal bolus titration for MDI insulin dosing therapy. [Figure 8H] FIG. 1 is a flow diagram illustrating aspects of various exemplary embodiments of an algorithm for assessing meal bolus titration for MDI insulin dosing therapy. [Figure 9A] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for determining dose guidance using a physiological dose algorithm. [Figure 9B] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for determining dose guidance using a physiological dose algorithm. [Figure 9C] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for determining dose guidance using a physiological dose algorithm. [Figure 10A] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a process flow for operation by a dose guidance application for correction factor titration. [Figure 10B] FIG. 10 is a flow diagram illustrating an aspect of controlling a user interface device to output data for adjusting a correction factor in response to analyte data. [Figure 10C] FIG. 10 is a flow diagram illustrating an aspect of controlling a user interface device to output data for adjusting a correction factor in response to analyte data. [Figure 11] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for determining dose guidance for administration at or before the start of a meal. [Figure 12A] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for determining dose guidance for administration after the start of a meal. [Figure 12B] FIG. 10 is a flow diagram illustrating an exemplary embodiment of an alternative method for determining dose guidance for administration after the start of a meal. [Figure 12C] FIG. 10 is a flow diagram illustrating an exemplary embodiment of another alternative method for determining dose guidance for administration after the start of a meal. [Figure 13] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for determining dose guidance for administration to an additional meal. [Figure 14] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for determining dose guidance for administration of a correction dose. [Figure 15] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for alerting a subject. [Figure 16A] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for overriding dose guidance settings. [Figure 16B] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for sensor fault detection. [Figure 16C] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for detecting a change in medication strategy. [Figure 16D]FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for detecting discrepancies in medication strategies. [Figure 16E] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for managing multiple dosing. [Figure 16F] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for adjusting insulin dose guidance. [Figure 16G] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for managing a medication strategy. [Figure 17A] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for recommending insulin dose titration. [Figure 17B] FIG. 1 is a flow diagram illustrating another exemplary embodiment of a method for recommending insulin dose titration. [Figure 17C] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for recommending titration of insulin doses with nighttime periods. [Figure 17D] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for recommending titration of insulin doses with dinner periods. [Figure 17E] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for recommending insulin dosing. [Figure 17F] FIG. 1 is a flow diagram illustrating an exemplary embodiment of a method for determining whether insulin delivery is abnormal. [Figure 17G] FIG. 1 illustrates an exemplary embodiment of an exemplary tracking pair graph. [Figure 18A] FIG. 1 is a block diagram illustrating an exemplary embodiment of a site rotation system. [Figure 18B] FIG. 1 is a flow diagram of an exemplary embodiment of a method for monitoring a drug infusion site. DETAILED DESCRIPTION OF THE INVENTION

[0011] Before describing the present subject matter in detail, it is to be understood that this disclosure is not limited to particular embodiments described herein, as such may, of course, vary. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting, since the scope of the disclosure will be limited only by the appended claims.

[0012] Generally, embodiments of the present disclosure include systems, devices, and methods related to medication dosage guidance. Dose guidance can be based on a wide range of user-specific information and information categories, such as the user's current and past analyte levels, the user's current and past diet, the user's current and past physical activity, the user's current and past medication history, and other physiological information about the user. According to one aspect of the embodiments, dose guidance provided by the systems, devices, and methods of the present disclosure can be based not only on individual information categories but also on the predicted impact of such information categories on the user's future analyte levels.

[0013] The dose guidance functionality can be implemented as a dose guidance application (DGA) that includes software and / or firmware instructions stored in a computing device's memory and executed by at least one processor or its processing circuitry. The computing device can be owned by a user or a healthcare professional (HCP), and the user or HCP can interface with the computing device through a user interface. According to some embodiments, the computing device can be a server or trusted computer system accessible over a network, and the dose guidance software can be presented to the user in the form of an interactive web page via a browser running on a local display device (having the user interface) that is in communication with the server or trusted computer system over the network. In this and other embodiments, the dose guidance software runs across multiple devices, or runs partly on the processing circuitry of the local display device and partly on the processing circuitry of the server or trusted computer system. Those skilled in the art will understand that when a DGA is described as performing an action, such action is performed according to instructions stored in computer memory (including instructions hard-coded in read-only memory) that, when executed by at least one processor of at least one computing device, causes the DGA to perform the described action. In all cases, the operations may alternatively be performed by hardware (e.g., dedicated circuitry) hardwired to perform the operations, as opposed to being performed by instructions stored in a memory.

[0014] Additionally, as used herein, a system in which DGA is implemented may be referred to as a dose guidance system. A dose guidance system may be configured solely for the purpose of providing dose guidance, or may be a multi-function system in which dose guidance is only one aspect. For example, in some embodiments, a dose guidance system may also monitor a user's analyte levels. In some embodiments, a dose guidance system may also deliver medication to a user, such as with an injection or infusion device. In some embodiments, a dose guidance system is capable of both analyte monitoring and medication delivery.

[0015] These and other embodiments described herein represent improvements in the field of computer-based dose determination, analyte monitoring, and drug delivery systems. Particular features and potential advantages of the disclosed embodiments are further described below.

[0016] Before describing embodiments of dose guidance in detail, it is desirable to first describe an example of a dose guidance system in which the dose guidance application can be implemented.

[0017] Exemplary Embodiments of a Dose Guidance System 1A is a block diagram illustrating an exemplary embodiment of a dose guidance system 100. In this embodiment, dose guidance system 100 is capable of providing dose guidance, monitoring one or more analytes, and delivering one or more medications. This multi-functional example is used to illustrate the high degree of interconnectivity and performance provided by system 100. However, in the embodiments described herein, the analyte monitoring component, the medication delivery component, or both can be omitted as desired.

[0018] Here, system 100 includes a sensor control device (SCD) 102 configured to collect analyte level information from a user, a drug delivery device (MDD) 152 configured to deliver a drug to the user, and a display device 120 configured to present information to and receive input or information from the user, the structure and functionality of each device being described in detail herein.

[0019] The system 100 is configured for highly interconnected and highly flexible communication between devices. Each of the three devices 102, 120, and 152 can communicate with each other directly (without going through an intermediate electronic device) or indirectly with each other (e.g., via the cloud network 190 or via another device and then via the network 190). The bidirectional communication capabilities between devices and between a device and the network 190 are indicated by double-headed arrows in FIG. 1A. However, those skilled in the art will understand that any one or more devices (e.g., SCDs) may be capable of unidirectional communication, such as broadcast, multicast, or advertising communications. In all cases, whether bidirectional or unidirectional, communication can be wired or wireless. The protocols governing communication on each path may be the same or different, proprietary, or standardized. For example, wireless communication between the devices 102, 120, and 152 may be performed according to the Bluetooth (including Bluetooth Low Energy) standard, the NFC (near field communication) standard, the Wi-Fi (802.11x) standard, a mobile telephony standard, or the like. All communications over the various paths may be encrypted, and each device in Figure 1A may be configured to encrypt and decrypt those communications that are sent and received. In any case, the communication paths in Figure 1A may be direct (e.g., Bluetooth or NFC) or indirect (e.g., Wi-Fi, mobile telephony, or other Internet protocols). An embodiment of system 100 need not have the capability to communicate over all of the paths shown in Figure 1A.

[0020] 1A depicts a single display device 120, a single SCD 102, and a single MDD 152, those skilled in the art will appreciate that system 100 may include multiple of any of the aforementioned devices. By way of example only, system 100 may include a single SCD 102 in communication with multiple (e.g., two, three, four, etc.) display devices 120 and / or multiple MDDs 152. Alternatively, system 100 may include multiple SCDs 102 in communication with a single display device 120 and / or a single MDD 152. Furthermore, each of the multiple devices may be the same or different device types. For example, system 100 may include multiple display devices 120, including smartphones, handheld receivers, and / or smartwatches, each of which may communicate with the SCD 102 and / or MDD 152 as well as with each other.

[0021] Analyte data can be transferred between devices in system 100 autonomously (e.g., automatically according to a schedule) or in response to a request for analyte data (e.g., a request for analyte data from a first device to a second device, which then transmits the analyte data from the second device to the first device). To accommodate more complex systems, other technologies for data communication, such as cloud network 190, can also be employed.

[0022] 1B is a block diagram illustrating another exemplary embodiment of a dose guidance system 100. Here, system 100 includes an SCD 102, an MDD 152, a first display device 120-1, a second display device 120-2, a local computer system 170, and a trusted computer system 180 accessible by a cloud network 190. SCD 102 and MDD 152 can communicate with each other and with display device 120-1, which can function as a communications hub to aggregate information from SCD 102 and MDD 152, process and display that information at a desired location, and forward some or all of the information to cloud network 190 and / or computer system 170. Conversely, display device 120-1 can receive information from cloud network 190 and / or computer system 170 and communicate some or all of the received information to SCD 102, MDD 152, or both. Computer system 170 may be a personal computer, server terminal, laptop computer, tablet, or other suitable data processing device. Computer system 170 may include or present software for data management and analysis and communication with components within system 100. Computer system 170 may be used by a user or medical professional to display and / or analyze analyte data measured by SCD 102. Additionally, while FIG. 1B shows a single SCD 102, a single MDD 152, and two display devices 120-1 and 120-2, those skilled in the art will understand that system 100 may include multiple of any of the foregoing devices, and each multiple device may include devices of the same or different types.

[0023] 1B , according to some embodiments, trusted computer system 180, which may be physically owned or virtually owned via a secure connection by a manufacturer or distributor of components of system 100, may be used as a server to perform authentication of devices (e.g., devices 102, 120-n, 152) of system 100, to securely store user data, and / or to provide data analysis programs (e.g., accessible via a web browser) for performing analysis of users' measured analyte data and medication histories. Trusted computer system 180 may also function as a data hub for routing and exchanging data between all devices communicating with system 180 via cloud network 190. In other words, all devices of system 100 that can communicate with cloud network 190 (e.g., directly using an Internet connection or indirectly via other devices) can also communicate, directly or indirectly, with all other devices of system 100 that can communicate with cloud network 190.

[0024] Display device 120-2 is shown in communication with cloud network 190. In this example, device 120-2 may be owned by another user who is authorized to access the analyte and medication data of the person wearing SCD 102. For example, the person owning display device 120-2 may be, in one example, the parent of a child wearing SCD 102, or, in another example, the caregiver of an elderly patient wearing SCD 102. System 100 may be configured to communicate the analyte and medication data about the wearer via cloud network 190 (e.g., via trusted computer system 180) to another user who is authorized to access the data.

[0025] Exemplary Embodiments of an Analyte Monitoring Device The analyte monitoring functionality of dose guidance system 100 can be achieved by including one or more devices capable of collecting, processing, and displaying a user's analyte data. Exemplary embodiments of such devices and methods of using them are described in International Application No. 2018 / 152241 and U.S. Patent Application No. 2011 / 0213225, the contents of which are incorporated herein by reference in their entirety for all purposes.

[0026] Analyte monitoring can be performed in a number of different ways. A "continuous analyte monitoring" device (e.g., a "continuous glucose monitoring" device) can continuously or repeatedly transmit data from the sensor control device to the display device, for example, automatically according to a schedule, with or without prompting. A "flash analyte monitoring" device (e.g., a "flash glucose monitoring" device or simply "flash" device) can transfer data from the sensor control device in response to a user-initiated request for data (e.g., a scan) by the display device, using, for example, near-field communication (NFC) or radio frequency identification (RFID) protocols, as another example.

[0027] Analyte monitoring devices that utilize sensors configured to be partially or completely placed within a user's body may be referred to as in vivo analyte monitoring devices. For example, an in vivo sensor may be placed within a user's body such that at least a portion of the sensor is in contact with bodily fluid (e.g., interstitial (ISF) fluid, such as dermal fluid in the dermis layer or subcutaneous fluid below the dermis layer, blood, etc.) and may measure an analyte concentration in the bodily fluid. In vivo sensors may use various types of sensing technologies (e.g., chemical, electrochemical, or optical). Some systems utilizing in vivo analyte sensors may also operate without the need for fingerstick calibration.

[0028] An "in vitro" device is one that contacts a sensor with a biological sample outside the body (or "ex vivo"). These devices typically include a port for receiving an analyte test strip carrying a user's bodily fluid, which can be analyzed to determine the user's blood glucose level. Other ex vivo devices attempt to measure a user's internal analyte levels noninvasively, for example, by using proposed optical techniques that can measure internal analyte levels without mechanically intruding on the user's body or skin. In vivo and ex vivo devices often include in vitro functionality (e.g., an in vivo display device that also includes a test strip port).

[0029] Although the present subject matter is described with respect to sensors capable of measuring glucose concentrations, the detection and measurement of concentrations of other analytes is also within the scope of this disclosure. These other analytes may include, for example, ketones, lactate, oxygen, hemoglobin A1C, acetylcholine, amylase, bilirubin, cholesterol, chorionic gonadotropin, creatine kinase (e.g., CK-MB), creatine, DNA, fructosamine, glutamine, growth hormone, hormones, peroxides, prostate-specific antigen, prothrombin, RNA, thyroid-stimulating hormone, troponin, and the like. Drug concentrations may also be monitored, for example, antibiotics (e.g., gentamicin, vancomycin, and the like), digitoxin, gitoxin, drugs of abuse, theophylline, and warfarin. The sensor may be configured to measure two or more different analytes at the same or different times. In some embodiments, the sensor control device can be coupled to two or more sensors, where one sensor is configured to measure a first analyte (e.g., glucose) and the other sensor or sensors are configured to measure one or more different analytes (e.g., any of those described herein). In other embodiments, a user can wear two or more sensor control devices, each capable of measuring a different analyte.

[0030] The embodiments described herein can be used in any type of in vivo, in vitro, and ex vivo device capable of monitoring the aforementioned analytes and the like.

[0031] In many embodiments, the operation of the sensor can be controlled by the SCD 102. The sensor can be mechanically and communicatively coupled to the SCD 102, or can simply be communicatively coupled to the SCD 102 using wireless communication technology. The SCD 102 can include electronics and a power source that enable and control the analyte sensing performed by the sensor. In some embodiments, the sensor or SCD 102 can be self-powered so as not to require a battery. The SCD 102 can also include communication circuitry for communicating with another device (e.g., a display device), which may or may not be local to the user's body. The SCD 102 can be present on the user's body (e.g., attached to the user's skin, otherwise positioned, or carried in the user's clothing). The SCD 102 can also be implanted within the user's body along with the sensor. The functionality of the SCD 102 can be divided between a first component implanted within the body (e.g., a component that controls the sensor) and a second component present on the body (e.g., an intermediary component that communicates with the first component and further communicates with an external device such as a computer or smartphone). In other embodiments, the SCD 102 may be configured to be external to the body and to non-invasively measure a user's analyte levels. The sensor control device may also be referred to as a "sensor control unit," an "on-body electronics" device or unit, an "on-body" device or unit, an "in-body electronics" device or unit, an "in-body" device or unit, or a "sensor data communication" device or unit, to name a few, depending on the actual implementation or embodiment.

[0032] In some embodiments, the SCD 102 can include a user interface (e.g., a touchscreen) and can process analyte data and display the resulting calculated analyte levels to the user. In this case, the dose guidance embodiments described herein can be implemented, in whole or in part, directly by the SCD 102. In many embodiments, where the physical form factor of the SCD 102 is minimized (e.g., to minimize appearance on the user's body), the sensor control device may not be accessible to the user (e.g., if entirely implanted), or other factors, it may be desirable to have a display device available to the user to read analyte levels and interface with the sensor control device.

[0033] FIG. 2A is a side view of an exemplary embodiment of an SCD 102. The SCD 102 can include a case or mount 103 (FIG. 2B) for sensor electronics, which can be electrically coupled to an analyte sensor 101, here configured as an electrochemical sensor. According to some embodiments, the sensor 101 can be configured to reside partially within a user's body (e.g., through the outermost surface of the skin), where it is in fluid contact with the user's bodily fluids and, together with the sensor electronics, can be used to measure the user's analyte-related data. An attachment structure 105, such as an adhesive patch, can be used to secure the case 103 to the user's skin. The sensor 101 can extend through the attachment structure 105 and protrude away from the case 103. Those skilled in the art will understand that other forms of attachment to the body and / or case 103 can be used in addition to or instead of adhesive and are fully within the scope of the present disclosure.

[0034] The SCD 102 can be applied to the body in any desired manner. For example, an insertion device (not shown), sometimes referred to as an applicator, can be used to position all or a portion of the analyte sensor 101 through the exterior surface of the user's skin and in contact with the user's bodily fluids. In doing so, the insertion device can also position the SCD 102 on the skin. In other embodiments, the insertion device can first position the sensor 101 and then later couple (e.g., insert into a mount) the associated electronics (e.g., wireless transmission circuitry and / or data processing circuitry, etc.) to the sensor 101, either manually or with the aid of a mechanical device. Examples of insertion devices are described in U.S. Patent Publication Nos. 2008 / 0009692, 2011 / 0319729, 2015 / 0018639, 2015 / 0025345, and 2015 / 0173661, 2018 / 0235520, the entire contents of which are incorporated by reference herein for all purposes.

[0035] 2B is a block diagram illustrating an exemplary embodiment of an SCD 102 having an analyte sensor 101 and sensor electronics 104. The sensor electronics 104 can be implemented on one or more semiconductor chips (e.g., an application specific integrated circuit (ASIC), a processor or controller, memory, a programmable gate array, etc.). In the embodiment of FIG. 1B, the sensor electronics 104 interfaces to the sensor 101 in an analog manner and includes an analog front end (AFE) 110 configured to convert analog signals to and / or from digital format (e.g., with an A / D converter), a power supply 111 configured to provide power to the components of the SCD 102, processing circuitry 112, memory 114, timing circuitry 115 (e.g., oscillators and phase-locked loops for providing clocks or other timing to the components of the SCD 102, etc.), and communication circuitry 116 configured to communicate wired and / or wirelessly with one or more devices external to the SCD 102, such as the display device 120 and / or the MDD 152.

[0036] SCD 102 may be implemented in a highly interconnected manner, with power supply 111 coupled to each component shown in FIG. 2B , and those components that communicate or receive data, information, or commands (e.g., AFE 110, processing circuitry 112, memory 114, timing circuitry 115, and communication circuitry 116) may be communicatively coupled to all other such components, for example, via one or more communication connections or buses 118.

[0037] The processing circuitry 112 can include one or more processors, microprocessors, controllers, and / or microcontrollers, each of which can be a discrete chip or distributed among (and portions of) many different chips. The processing circuitry 112 can include on-board memory. The processing circuitry 112 interfaces with the communications circuitry 116 and can perform analog-to-digital conversion, encoding and decoding, digital signal processing, and other functions that facilitate converting data signals into a format suitable for wireless or wired transmission (e.g., in-phase and quadrature). The processing circuitry 112 also interfaces with the communications circuitry 116 and can perform the inverse functions necessary to receive wireless transmissions and convert them into digital data or information.

[0038] Processing circuit 112 can execute instructions stored in memory 114. These instructions can cause processing circuit 112 to process raw analyte data (or preprocessed analyte data) and arrive at a final calculated analyte level. In some embodiments, the instructions stored in memory 114, when executed, can cause processing circuit 112 to process the raw analyte data to determine one or more of a calculated analyte level, an average calculated analyte level within a predetermined time window, a calculated rate of change in analyte level within a predetermined time window, and / or whether a calculated analyte indicator exceeds a predetermined threshold condition. These instructions can also cause processing circuit 112 to read and act on received transmissions, adjust the timing of timing circuit 115, process data or information received from other devices (e.g., calibration information, encryption or authentication information received from display device 120, etc.), perform tasks to establish and maintain communications with display device 120, interpret voice commands from a user and transmit them to communication circuit 116, etc. In embodiments in which SCD 102 includes a user interface, the instructions may cause processing circuit 112 to control the user interface, read user input from the user interface, display information on the user interface, format data for display, etc. The functionality described herein that is encoded as instructions may instead be implemented by SCD 102 using hardware or firmware designs that do not rely on the execution of stored software instructions to achieve the functionality.

[0039] Memory 114 may be shared by one or more of the various functional units present in SCD 102, or may be distributed among two or more functional units (e.g., as separate memories present in different chips). Memory 114 may also be a separate chip itself. Memory 114 may be non-transitory and may be volatile memory (e.g., RAM, etc.) and / or non-volatile memory (e.g., ROM, flash memory, F-RAM, etc.).

[0040] The communications circuitry 116 may be implemented as one or more components (e.g., transmitters, receivers, transceivers, passive circuits, encoders, decoders, and / or other communications circuitry) that perform functions for communication over respective communications paths or links. The communications circuitry 116 may include or be coupled to one or more antennas for wireless communication.

[0041] The power source 111 can include one or more batteries, which can be rechargeable or single-use disposable batteries. Power management circuitry can also be included to regulate battery charging, monitor power source 111 usage, boost power, perform DC conversion, etc.

[0042] Additionally, an optional temperature sensor (not shown) can collect readings or measurements of the temperature on the skin or of the sensor. These readings or measurements can be communicated (individually or as aggregated measurements over time) from SCD 102 to another device (e.g., display device 120). However, the temperature readings or measurements can be used in conjunction with software routines executed by SCD 102 or display device 120 to correct or compensate the analyte measurements output to the user instead of, or in addition to, actually outputting temperature measurements to the user.

[0043] Exemplary Embodiments of a Medication Delivery Device The medication delivery functionality of the dose guidance system 100 can be achieved through the inclusion of one or more medication delivery devices (MDDs) 152. The MDD 152 can be any device configured to deliver a specific dose of medication. The MDD 152 can also include a device, such as a pen cap, that transmits data regarding the dose to the DGA, even if the device itself does not deliver the medication. The MDD 152 can be configured as a portable injection device (PID) capable of delivering a single dose per injection, such as a bolus. A PID is essentially a manually operated syringe, with medication either preloaded or drawn into the syringe from a reservoir prior to injection. However, in most embodiments, the PID includes electronics to interface with the user and effectuate medication delivery. PIDs are often referred to as medication pens, although they do not necessarily have a pen-like appearance. PIDs with user interface electronics are often referred to as smart pens. PIDs may be used to deliver a single dose and then discarded, or they may be durable and used repeatedly to administer multiple doses over a period of a day, week, or month. PIDs are commonly used by users of multiple-injection (MDI) therapy, which involves multiple injections per day.

[0044] The MDD can also include a pump and an infusion set. The infusion set includes a tubing cannula that resides at least partially within the recipient's body. The tubing cannula is in fluid communication with a pump, which can repeatedly deliver small amounts of medication into the recipient's body through the cannula. The infusion set can be applied to the recipient's body using an infusion set applicator, and the infusion set often remains implanted for 2-3 days or longer. The pump device includes electronics to interface with the user and control the gradual infusion of medication. Both the PID and the pump can store medication in a medication reservoir.

[0045] The MDD 152 can function as part of a closed-loop system (e.g., an artificial pancreas system that does not require user intervention for operation), a semi-closed-loop system (e.g., an insulin loop system that requires little user intervention for operation, such as confirming dose changes), or an open-loop system. For example, a diabetic patient's analyte levels can be repeatedly and automatically monitored by the SCD 102, and that information can be used by dose guidance embodiments described herein to automatically calculate or otherwise determine an appropriate drug dose to control the diabetic patient's analyte levels, and then deliver that dose to the diabetic patient's body. This calculation can be performed within the MDD 152 or any other device of the system 100, and the resulting determined dose can then be communicated to the MCD 152.

[0046] In many embodiments, the dose guidance provided by the embodiments described herein is for a type of insulin (e.g., rapid-acting (RA), short-acting insulin, intermediate-acting insulin (e.g., NPH insulin), long-acting (LA), very long-acting insulin, and mixed insulin) that is the same agent delivered by MDD152. Types of insulin include human insulin and synthetic insulin analogs. Insulin can also include premixed formulations. However, the dose guidance embodiments described herein and the drug delivery capabilities of MDD152 can also be applied to other non-insulin agents. Such agents can include, but are not limited to, exenatide, exenatide extended-release formulations, liraglutide, lixisenatide, semaglutide, pramlintide, metformin, SLGT1-i inhibitors, SLGT2-i inhibitors, and DPP4 inhibitors. Dose guidance embodiments can also include combination therapies. Combination therapies can include, but are not limited to, insulin and a glucagon-like peptide-1 receptor agonist (GLP-1RA), insulin and pramlintide.

[0047] To facilitate the description of the dose guidance embodiments herein, the MDD 152 is often described in the form of a PID, specifically a smart pen, although one skilled in the art will readily appreciate that the MDD 152 may alternatively be configured as a pen cap, a pump, or any other type of medication delivery device.

[0048] FIG. 3A is a schematic diagram illustrating an exemplary embodiment of a PID, specifically an MDD 152 configured as a smart pen. The MDD 152 can include a housing 154 for electronics, an injection motor, and a medication reservoir (see FIG. 3B) from which a medication can be delivered via a needle 156. The housing 154 can include a removable or detachable cap or cover 157 that can be attached to cover the needle 156 when not in use and then removed for injection. The MDD 152 can include a user interface 158 that can be implemented as a single component (e.g., a touchscreen for outputting information to and receiving input from the user) or as multiple components (e.g., a touchscreen or display in combination with one or more buttons, switches, etc.). The MDD 152 can also include an actuator 159 that can be moved, pressed, contacted, or otherwise actuated to initiate delivery of the medication from the internal reservoir through the needle 156 and into the recipient's body. According to some embodiments, cap 157 and actuator 159 may also include one or more safety mechanisms to prevent separation and / or activation to reduce the risk of injecting harmful agents. Details of these safety mechanisms and more are described in U.S. Patent Publication No. 2019 / 0343385 (the '385 publication), the entire contents of which are incorporated herein by reference for all purposes.

[0049] FIG. 3B is a block diagram illustrating an exemplary embodiment of an MDD 152 having electronics 160 coupled to a power source 161, and a powered injection motor 162 also coupled to the power source 161 and a drug reservoir. A needle 156 is shown in fluid communication with a reservoir 163; a valve (not shown) may be present between the reservoir 163 and the needle 156. The reservoir 163 may be permanent or may be removable and replaced with another reservoir containing the same or a different drug. The electronics 160 may be implemented on one or more semiconductor chips (e.g., an application specific integrated circuit (ASIC), a processor or controller, memory, a programmable gate array, etc.). In the embodiment of FIG. 3B, the electronics 160 may include higher-level functional units including processing circuitry 164, memory 165, communication circuitry 166 configured to communicate wired and / or wirelessly with one or more devices external to the MDD 152 (e.g., the display device 120), and user interface electronics 168.

[0050] MDD 152 may be implemented in a highly interconnected manner, with power source 161 coupled to each of the components shown in FIG. 3B , and those components (e.g., processing circuitry 164, memory 165, and communication circuitry 166) that communicate or receive data, information, or commands may be communicatively coupled to all other such components, for example, via one or more communication connections or buses 169.

[0051] The processing circuitry 164 can include one or more processors, microprocessors, controllers, and / or microcontrollers, each of which can be a discrete chip or distributed among (and portions of) many different chips. The processing circuitry 164 can include on-board memory. The processing circuitry 164 interfaces with the communications circuitry 166 and can perform analog-to-digital conversion, encoding and decoding, digital signal processing, and other functions that facilitate converting data signals into a format suitable for wireless or wired transmission (e.g., in-phase and quadrature). The processing circuitry 164 also interfaces with the communications circuitry 166 and can perform the inverse functions necessary to receive wireless transmissions and convert them into digital data or information.

[0052] Processing circuit 164 can execute software instructions stored in memory 165. These instructions can cause processing circuit 164 to receive a selection or provision of a specified dose from a user (e.g., entered via user interface 158 or received from another device), process a command to deliver the specified dose (e.g., a signal from actuator 159), and control motor 162 to deliver the specified dose. These instructions can also cause processing circuit 164 to read and act on received transmissions, process data or information received from other devices (e.g., calibration information, encryption or authentication information received from display device 120, etc.), perform tasks to establish and maintain communication with display device 120, interpret voice commands from the user and send them to communications circuit 166, etc. In embodiments in which MDD 152 includes user interface 158, the instructions can cause processing circuit 164 to control the user interface, read user input from the user interface (e.g., entering a drug dose for administration or entering a confirmation of a recommended drug dose), display information on the user interface, format data for display, etc. The functionality described herein that is coded as instructions may instead be implemented by MDD 152 using hardware or firmware designs that do not rely on the execution of stored software instructions to achieve the functionality.

[0053] Memory 165 may be shared by one or more of the various functional units present in MDD 152, or may be distributed among two or more functional units (e.g., as separate memories present in different chips). Memory 165 may also be a separate chip itself. Memory 165 may be non-transitory and may be volatile memory (e.g., RAM, etc.) and / or non-volatile memory (e.g., ROM, flash memory, F-RAM, etc.).

[0054] The communications circuitry 166 may be implemented as one or more components (e.g., transmitters, receivers, transceivers, passive circuits, encoders, decoders, and / or other communications circuitry) that perform functions for communication over respective communications paths or links. The communications circuitry 166 may include or be coupled to one or more antennas for wireless communication. Details of exemplary antennas can be found in the '385 publication, the entire contents of which are incorporated herein by reference for all purposes.

[0055] The power source 161 can include one or more batteries, which can be rechargeable or single-use disposable batteries. Power management circuitry can also be included to regulate battery charging, monitor power source 161 usage, boost power, perform DC conversion, etc.

[0056] The MDD 152 may also include an integrated or attachable in vitro glucose meter that includes an in vitro test strip port (not shown) for receiving in vitro glucose test strips for performing in vitro blood glucose measurements.

[0057] Exemplary Embodiments of a Display Device Display device 120 can be configured to display information related to system 100 to a user and to accept or receive input from a user also related to system 100. Display device 120 can display recently measured analyte levels to a user in any number of forms. The display device can display the user's past analyte levels as well as other indicators describing the user's analyte information (e.g., glucose control indicator (time in range), ambulatory glucose profile (AGP), hypoglycemic risk level, etc.). Display device 120 can display drug delivery information such as past dosage information and time and date of administration. Display device 120 can display alarms, alerts, or other notifications related to analyte levels and / or drug delivery.

[0058] Display device 120 can be dedicated for use with system 100 (e.g., an electronic device designed and manufactured primarily to interface with an analyte sensor and / or a drug delivery device), as well as a handheld or portable mobile communication device (e.g., a smartphone or tablet), or a multifunction, general-purpose computing device such as a laptop, personal computer, or other computing device. Display device 120 can be configured as a mobile smart wearable electronics assembly such as smart glasses or a smart watch or wristband. Display devices and variations thereof can be referred to as "reader devices," "readers," "handheld electronic devices" (or handhelds), "portable data processing" devices or units, "information receivers," "receiver" devices or units (or simply receivers), "relay" devices or units, or "remote" devices or units, to name a few.

[0059] 4A is a schematic diagram illustrating an exemplary embodiment of display device 120. Here, display device 120 includes user interface 121 and case 124 in which display device electronics 130 (FIG. 4B) are held. User interface 121 can be implemented as a single component (e.g., a touchscreen capable of input and output) or multiple components (e.g., a display and one or more devices configured to receive user input). In this embodiment, user interface 121 includes touchscreen display 122 (configured to display information and graphics and accept user input via touch) and input buttons 123, both of which are coupled to case 124.

[0060] Display device 120 may have stored software (e.g., software downloaded by the manufacturer or by the user in the form of one or more "apps" or other software packages) that forms an interface to SCD 102, MDD 152, and / or a user. Additionally or alternatively, the user interface may be controlled by web pages displayed in a browser or other internet interface software executable on display device 120.

[0061] 4B is a block diagram of an exemplary embodiment of display device 120 with display device electronics 130. Here, display device 120 includes user interface 121 including display 122 and input components 123 (e.g., buttons, actuators, touch-sensitive switches, capacitive switches, pressure-sensitive switches, jog dials, microphones, speakers, etc.), processing circuitry 131, memory 125, communication circuitry 126 configured to communicate with and / or from one or more other devices external to display device 120, power supply 127, and timing circuitry 128 (e.g., oscillators and phase-locked loops for providing clocks or other timing to components of SCD 102, etc.). Each of the aforementioned components may be implemented as one or more distinct devices or may be integrated into a multi-function device (e.g., integration of processing circuitry 131, memory 125, and communication circuitry 126 on a single semiconductor chip). Display device 120 may be implemented in a highly interconnected manner, with power supply 127 coupled to each of the components shown in Figure 4B, and those components that communicate or receive data, information, or commands (e.g., user interface 121, processing circuitry 131, memory 125, communication circuitry 126, and timing circuitry 128) may be communicatively coupled to all other such components, for example, via one or more communication connections or bus 129. Figure 4B is a simplified representation of typical hardware and functionality present in a display device, and those skilled in the art will readily recognize that other hardware and functionality (e.g., codecs, drivers, glue logic) may also be included.

[0062] Processing circuitry 131 can include one or more processors, microprocessors, controllers, and / or microcontrollers, each of which can be a discrete chip or distributed among (and portions of) many different chips. Processing circuitry 131 can include on-board memory. Processing circuitry 131 interfaces with communications circuitry 126 and can perform analog-to-digital conversion, encoding and decoding, digital signal processing, and other functions that facilitate converting data signals into formats suitable for wireless or wired transmission (e.g., in-phase and quadrature). Processing circuitry 131 interfaces with communications circuitry 126 and can also perform the inverse functions necessary to receive wireless transmissions and convert them into digital data or information.

[0063] Processing circuit 131 can execute software instructions stored in memory 125. These instructions can cause processing circuit 131 to process raw analyte data (or preprocessed analyte data) and arrive at corresponding analyte levels suitable for display to the user. These instructions can cause processing circuit 131 to read, process, and / or store dosing instructions from the user, which in turn can communicate the dosing instructions to MDD 152. These instructions can cause processing circuit 131 to execute user interface software adapted to present an interactive group of graphical user interface screens to the user for the purposes of configuring system parameters (e.g., alarm thresholds, notification settings, display preferences, etc.), presenting current and historical analyte level information to the user, presenting current and historical medication delivery information to the user, collecting other non-analyte information from the user (e.g., information regarding meals consumed, activities performed, medications administered, etc.), and presenting notifications and alarms to the user. These instructions may also cause the processing circuit 131 to transmit to the communications circuit 126, read and act on received transmissions, read input from the user interface 121 (e.g., entering a drug dose to be administered or confirming a recommended drug dose), display data or information on the user interface 121, adjust the timing of the timing circuit 128, process data or information received from other devices (e.g., analyte data, calibration information, encryption or authentication information received from the SCD 102, etc.), perform tasks to establish and maintain communications with the SCD 102, interpret voice commands from a user, etc. The functionality described herein that is encoded as instructions may instead be implemented by the SCD 102 using hardware or firmware designs that do not rely on the execution of stored software instructions to achieve their functions.

[0064] Memory 125 may be shared by one or more of the various functional units present in display device 120, or may be distributed among two or more functional units (e.g., as separate memories present in different chips). Memory 125 may also be a separate chip itself. Memory 125 may be non-transitory and may be volatile memory (e.g., RAM, etc.) and / or non-volatile memory (e.g., ROM, flash memory, F-RAM, etc.).

[0065] The communications circuitry 126 may be implemented as one or more components (e.g., transmitters, receivers, transceivers, passive circuits, encoders, decoders, and / or other communications circuitry) that perform functions for communication over respective communications paths or links. The communications circuitry 126 may include or be coupled to one or more antennas for wireless communication.

[0066] The power source 127 may include one or more batteries, which may be rechargeable or single-use disposable batteries. Power management circuitry may also be included to regulate battery charging, monitor power source 127 usage, boost power, perform DC conversion, etc.

[0067] Display device 120 may also include one or more data communication ports (not shown) for wired data communication with external devices such as computer system 170, SCD 102, or MDD 152. Display device 120 may also include an integrated or attachable in vitro glucose meter that includes an in vitro test strip port (not shown) for receiving in vitro glucose test strips for performing in vitro blood glucose measurements.

[0068] The display device 120 can display the measured analyte data received from the SCD 102 and can also be configured to output alarms, alert notifications, glucose values, etc., which may be visual, audible, tactile, or any combination thereof. In some embodiments, the SCD 102 and / or MDD 152 can also be configured to output alarm or alert notifications in visual, audible, tactile form, or any combination thereof. Further details and other display embodiments can be found, for example, in U.S. Patent Publication No. 2011 / 0193704, the entire contents of which are incorporated herein by reference for all purposes.

[0069] Exemplary Embodiments Related to Dose Guidance The following exemplary embodiments relate to dose guidance functionality provided by dose guidance system 100. In many embodiments, the dose guidance functionality is implemented as a set of software instructions stored and / or executed on one or more electronic devices. This dose guidance functionality will be referred to herein as a dose guidance application (DGA). In some embodiments, the DGA is stored, executed, and presented to a user on the same single electronic device. In other embodiments, the DGA can be stored and executed on one device and presented to a user on a different electronic device. For example, the DGA can be stored and executed on trusted computer system 180 and presented to a user via a web page displayed via an internet browser executing on display device 120.

[0070] Thus, there are many different embodiments related to the number and types of electronic devices used to store, execute, and present the DGA to a user. With respect to presentation to a user, a device configured to implement this capability will be referred to herein as a user interface device (UID) 200. FIG. 5 is a block diagram illustrating an example of one embodiment of a UID 200. In this embodiment, the UID 200 includes a case 201 coupled to a user interface 202. The user interface 202 can output information to the user and receive input or information from the user. In some embodiments, the user interface 202 is a touchscreen. As shown here, the user interface 202 includes a display 204, which may be a touchscreen, and input components 206 (e.g., buttons, actuators, touch-sensitive switches, capacitive switches, pressure-sensitive switches, a jog dial, a microphone, a touchpad, softkeys, a keyboard, etc.).

[0071] Many of the devices described herein can be implemented as UID 200. For example, display device 120 is used as UID 200 in many embodiments. In some embodiments, MDD 152 can be implemented as UID 200. In embodiments in which SCD 102 includes a user interface, SCD 102 can be implemented as UID 200. Computer system 170 can also be implemented as UID 200.

[0072] Detecting MDI dosing strategies Turning now to the DGA aspect, more particularly, DGA can use knowledge of a patient's dosing strategy and analyte levels to provide accurate dose guidance. Described herein are exemplary embodiments for automated detection of a patient's dosing strategy, which allows for easy and rapid setup of a DGA. Detection of a dosing strategy can be based on numerous characteristics of the monitored drug (e.g., insulin) dose. For example, in this embodiment, a dose can be identified as basal or bolus based on the MDD 152 used to administer the dose. Some patients can have multiple MDDs 152. For example, a patient can have one MDD for administering long-acting insulin (e.g., a basal dose) and another MDD for administering fast-acting insulin (e.g., a prandial dose). Additionally, the count (e.g., number of doses) and timing of basal doses per administration can be used to classify the basal strategy as a "single" or "split" basal dosing strategy. For example, in a "split" basal dosing strategy, a daily basal dose of 20 U can be split into two 10 U doses, one dose can be administered before bedtime and the other upon awakening.

[0073] When consecutive bolus doses are administered in quick succession, the system can attempt to distinguish between a primary meal dose, a boost to a primary meal dose, or a correction dose for high glucose between meals. When the DGA detects a small dose followed by a large dose (both occurring near the start of a meal), it can group the doses as a single meal dose, even if the first dose was a priming dose not injected into the patient. If a dose subsequently occurs long after a group of doses tagged as known meals and / or meal doses, the DGA can tag the later dose as a correction dose for high glucose after a meal or as a dose that boosts the previous meal dose to account for the extra food consumed. Once a meal event is recognized based on the meal detector algorithm or a user-entered meal event, the DGA can use the volume of the previous dosing event and its timing relative to the currently detected meal to help disambiguate whether the previous dose is the first of multiple meal doses or a correction for high glucose between meals. Correction doses are expected to be smaller in size than mealtime doses. Furthermore, if the time elapsed between the previous dose and the current meal event is sufficiently long, it is reasonable to assume that these two events are not associated with the treatment of the same glucose excursion event, eliminating the possibility that the previous dose is the first of multiple doses for a given meal. Thus, if the previous dose is sufficiently smaller than the meal doses recorded within this window in recent days and sufficiently distant from the current meal, the previous dose can be classified as a correction dose event.

[0074] The DGA can be configured to use real-time meal detection algorithms and dosing times to identify doses added to the basal as breakfast, lunch, and / or dinner bolus doses and / or correction doses. The DGA can also be configured to use the number of daily bolus doses to identify dosing strategies as basal only, basal +1, basal +2, etc.

[0075] These different scenarios and aspects of DGA are described in more detail elsewhere in this specification.

[0076] Onboarding To enhance the safety profile of the DGA, the HCP can approve the learned insulin dosing parameters and subsequent titration calculated by the DGA. Embodiments of the DGA include multiple methods of interaction between the HCP and the DGA, thereby providing the HCP with relevant evidence for approving the proposed dose learning and titration in a concise and informative manner that improves workflow.

[0077] For diabetic patients already on an insulin dosing regimen, HCPs can leverage existing reports that provide insight into the patient's glucose patterns to identify users who may benefit from dose guidance. DGA embodiments provide a learning period (e.g., while using the DGS100) during which the patient's dosing strategy and trends can be classified. If the combined insulin and glucose data further confirms that the user is a suitable candidate for a DGA, e.g., a candidate for which the DGA can learn a particular dosing strategy, the insulin dosing parameters learned during the learning period can serve as the initial conditions for dose guidance that can be titrated as needed by the DGA. HCP notification methods for initializing and titrating DGA dose parameters can also be presented. This process streamlines DGA onboarding and titration, helping both HCPs and users while ensuring that the DGA is used only by those indicated. If the DGA is unable to learn the patient's dosing parameters, the DGA can indicate a patient dosing discrepancy, which the HCP can use to address the patient's dosing discrepancy.

[0078] The first step in identifying potential users of DGA can involve an initial analysis of a patient's glycemic control through a glucose concentration profile. To make DGA accessible to as many users as possible, this process can be independent of the user's current glucose monitoring method.

[0079] For diabetic patients currently using the SCD 102, a glucose pattern report is available that includes key indicators, a glucose concentration profile (e.g., an ambulatory glucose profile (AGP)), patterns identified over different time periods, and titration and lifestyle suggestions to improve glucose levels consistently outside the target range. The patterns can be identified using a GPA algorithm, as described in detail elsewhere. As seen in FIG. 6, an exemplary glucose pattern report 250 is provided. Those skilled in the art will appreciate that the glucose pattern report 250 can be a graphical user interface output on a computing device display. The glucose pattern report 250 can include a glucose control index (TIR) ​​display 252 that indicates the percentage of time the patient's glucose levels were below the target range (e.g., below 70 mg / dL), within the target range (e.g., 70-180 mg / dL), and above the target range (e.g., above 180 mg / dL). The TIR display 252 can also report the amount of time the patient's glucose level was below a low threshold below the lower limit of the target range (e.g., below 54 mg / dL) or above a high threshold above the upper limit of the target range (e.g., above 250 mg / dL). The TIR display 252 can include a histogram in which different ranges are displayed in different colors. For example, time below the target range can be displayed in red, time within the target range in green, and time above the target range in yellow or orange. The glucose pattern report 250 can also display an average glucose level 254 for the time period 264 of the report, e.g., approximately 14 days. The glucose pattern report 250 can also display a glucose concentration profile 256, such as an ambulatory glucose profile (AGP). The glucose concentration profile 256 is a graph of glucose data for the time period of the report, and various data points on the graph can be color-coded to correspond to whether the glucose analyte level is below the target range, within the target range, or above the target range. This color coding can correspond to the color coding of the TIR display 252.Boxes 258 surrounding different portions of the glucose concentration profile 256 highlight patterns (e.g., hyperglycemia, hypoglycemia, moderate or moderate hyperglycemia, moderate or moderate hypoglycemia, and combinations thereof) detected according to the GPA algorithm described elsewhere herein.

[0080] Dosage guidance 260 can also be provided in the glucose pattern report 250 if the patient's current therapy (e.g., basal + RA insulin, basal only, basal + SU, etc.) is known. Dosage guidance can be provided in the form of text recommendations. General advice regarding insulin dose titration can be provided based on identified patterns of high and low glucose, highlighted in box 258 in the glucose concentration profile 256. Note that this general advice could be determined without access to data regarding the actual insulin doses administered. Recommendations can generally follow the rule of first mitigating any low patterns before mitigating any high patterns. If the glucose pattern report includes suggestions regarding insulin dose titration, the glucose pattern report 250 can also include a suggestion that the patient is a good candidate for DGS 100, facilitating a conversation between the HCP and the patient prior to the learning period.

[0081] If the GPA algorithm identifies a highly variable pattern, self-care guidance 262 may be displayed in the glucose pattern report 250. Alternatively, the glucose concentration profile 256 may have such high variability values ​​that the logic behind the report cannot make specific recommendations, instead defaulting to the user consulting an HCP about lifestyle or treatment changes.

[0082] As described above, for those who do not currently use a device or system (e.g., SCD102) associated with an application capable of generating glucose pattern report 250, the HCP can suggest that the patient be monitored by another device or system so that report 250 or similar can be generated. For example, a patient may wear an SCD102 configured in a masking or blind mode to collect glucose data over a period of days or weeks, preventing the user from accessing measured glucose levels and therefore from modifying their behavior during this time. From these data, a glucose pattern report can be generated. If a suggested insulin titration is included in the glucose pattern report, the glucose pattern report 250 can also include a suggestion that the patient is a good candidate for DGS100 and can suggest a learning period for a medication dosing strategy.

[0083] During the learning period, the MDD 152 can be integrated into the glucose sensing system used for initial screening to provide a more complete portrait of insulin-intensive diabetes management. The learning period can utilize algorithms, such as those described elsewhere herein, to detect the user's insulin dosing strategy. During the learning period, the DGA can be configured to determine how the user determines mealtime doses. For example, the DGA can determine whether the user determines mealtime doses based on carbohydrate counting, whether the user determines mealtime doses based on an empirical approach such as learning appropriate doses based on past or similar meal experiences, whether the user administers a fixed amount of insulin at mealtimes, whether the user varies the insulin meal dose (determined from fixed dosing, carbohydrate counting, or empirical dosing) based on pre-meal glucose levels, whether the user considers residual insulin from previous injections (IOB) or other techniques when determining the dose (determined from fixed dosing, carbohydrate counting, or empirical dosing). The DGA can also determine whether the user's mealtime doses are constant or vary depending on the type of meal (e.g., breakfast, lunch, and dinner). A determination that the mealtime dose has changed may indicate that the user is basing their mealtime dose on carbohydrate counting technology. The DGA can also determine whether the user is adjusting their mealtime dose to account for pre-meal high glucose. In some embodiments, the DGA can also determine a target glucose level, and the user adjusts or corrects their mealtime dose if their level is above or predicted to be above the target glucose level. The DGA can also determine which meals are associated with insulin dosing. The DGA can also determine patterns of missed meal doses. For example, the DGA can detect whether the user has missed at least two, or alternatively at least three, insulin doses associated with a meal or time period over a period (e.g., one or two weeks).

[0084] The learning period can last any period of time sufficient to obtain the necessary information. In many embodiments, this period is at least two days, more preferably one week or longer (e.g., 14 days), and can vary depending on how well the DGA is able to learn trends. Results can be compiled into a summary report for both the user and the physician.

[0085] Learning methods Manual configuration of the DGS100 requires time from the HCP, which may not always be available. Even if the HCP has the time, the configuration can be complex and error-prone. To mitigate these issues, the DGA can include a patient parameter initialization (PI) module, which requires no or minimal configuration. The PI module learns the patient's dosing strategy, including, for example, basal only, basal +1, or basal +2, and parameterizes the patient's dosing habits to configure the DGA's dose guidance settings.

[0086] According to one aspect of an embodiment, the PI module's learning process can include automatically configuring the patient's dose guidance settings from observed data. Once the settings are successfully learned, the DGS 100 can enter a guidance mode in which the patient can request dose guidance and receive dosing notifications. During the learning process that precedes guidance mode, the DGA can process glucose and insulin data collected by the patient's SCD 102, UID 202, and / or other devices and determine dosing information based on the processed data.

[0087] The administration information can include, for example, a dosing regimen, a meal-dose type, a dose parameter, and a dose range. The dosing regimen can include, for example, basal dose + BF, basal dose + LU, basal dose + DI, basal dose + BF / LU, basal dose + BF / DI, basal dose + LU / DI, and basal dose + 3, where BF indicates "breakfast," LU indicates "lunch," and DI indicates "dinner." Additional regimens, such as an afternoon snack dose, can also be included. The meal dose type can include, for example, a fixed meal dose or a variable meal dose. The dose parameters can include, for example, a nominal fixed dose or carbohydrate ratio for each meal, a pre-meal correction factor (CF), and a post-meal CF. The dose range can include an estimate of the minimum meal dose.

[0088] For each of the above dosing information types, the DGA can determine whether the cumulative data is sufficient or insufficient to determine the dosing information. In some embodiments, the patient's SCD 102 can be configured to operate for a predetermined period of time, e.g., 14 days. In these embodiments, the DGA can determine after the predetermined period of time (or earlier if the sensor stops operating before the end of the period) whether the available analyte and dosing data is sufficient to determine each of the above dosing information. If so, the DGA can execute the parameterization method 300 and initiate the dose guidance mode. In an alternative embodiment, periodically during the learning period (e.g., once daily), the DGA can determine whether the data is sufficient to determine each of the above dosing information. In either case, if the collected data is sufficient, the DGA can terminate the learning period, perform parameterization, and initiate the guidance period. Otherwise, the DGA can continue the learning process.

[0089] Referring to FIG. 7 , the DGA can be configured to execute method 300 on a suitable computing device, such as the UID 200, SCD 102, or MDD 152, alone or in any combination. Program instructions for executing method 300 can be grouped in a PI module or any other suitable code configuration. Generally, method 300 can include classifying, by the DGA, each of the doses of medication received by the patient during the analysis period, at step 302, based on data characterizing the patient's analytes and the doses of medication received by the patient during the analysis period. Method 300 can further include grouping each of the doses into one of a set of mealtime groups, at step 304. The method can further include creating dose parameters for the patient at least in part by applying the data for each of the mealtime groups to a model, at step 306. The method can include storing the dose parameters in computer memory and configuring dose guidance settings, at step 308. In the embodiments described herein, the analyte can be glucose or can include an indicator of the patient's glucose level, and the medication can be insulin or can include insulin. The dose guidance settings can be used by the DGA to formulate dose guidance or provided for output to an interface device, such as the UID 200 or a healthcare professional's terminal. More detailed aspects of each operation in method 300 are described below. As used herein, a "PI module" refers to a portion or part of a DGA that performs the operations of method 300 and any ancillary operations. The PI module is not limited to a particular configuration and can encompass various arrangements of computer code.

[0090] In one embodiment, the classification operation 302 can include classifying each dose of a medication (e.g., insulin) as either a meal dose, a correction dose, and / or an ambiguous dose. If the DGA cannot classify the medication dose as a meal dose or a correction dose with a defined confidence, the DGA can classify the medication dose as ambiguous and omit the medication dose from use in generating dose parameters for dose guidance.

[0091] DGA can classify medication doses through a sequence of two operations, referred to herein as feature extraction and classification. With reference to FIG. 7 , the classification operation 302 can include creating a feature matrix correlating a set of classification features to each of the doses. In some embodiments, the DGA can take a vector of insulin injection timestamps, a data file containing analyte measurements from the patient's SCD 102, and results from the meal detection algorithm module discussed elsewhere herein as inputs to a function that outputs a feature matrix for insulin dose classification. The number of rows in the feature matrix can indicate the amount of injections, or equivalent medication administration events, during the associated analysis period. Each row in the feature matrix can be or can include a feature vector for a single medication event. In embodiments for classifying insulin injections, each vector can include elements, described below, referred to herein as classification features. The DGA can determine each of the elements of the feature vector based on a corresponding segment of glucose monitoring data from a time range relative to the insulin injection time, e.g., from -2.5 hours to 1.5 hours.

[0092] In an embodiment, the classification features may include the time of administration of each administration, for example, the time period when the insulin injection was recorded by the MDD 152 or by the patient using the UID 200.

[0093] The classification features may further include time-filtered analyte values, e.g., glucose values ​​filtered using a Savitsky-Golay filter, a low-pass filter, a band-pass filter, a non-parametric smoothing filter such as locally estimated scatterplot smoothing, or other filters. In one embodiment, the Savitsky-Golay filter may be of order 2, frame length 7 with a sampling interval of 15 minutes.

[0094] Classification features can further include the rate of change of the analyte value closest to the time of dosing, for example, the rate of change of the analyte (e.g., glucose) value calculated by linear regression of five analyte data points (e.g., using a 15 minute sampling interval) centered on the data point closest to the time of dosing (e.g., injection).

[0095] The classification feature may further include a left-sided area under the curve (AUC) index that indicates the difference between the analyte value and the analyte value closest to the time of dosing, integrated over the interval before the time of dosing. For example, to obtain the left-sided AUC index, the DGA may collect all data points from the filtered analyte data within a time window (e.g., 2.5 hours), count back from the time of injection, and then calculate the difference between the average analyte value of the collected data points and the data point closest to the time of injection (i.e., the reference data point), and multiply this difference by the duration of the time window to calculate the left-sided AUC increment.

[0096] The classification feature may further include a right-hand AUC index that indicates the difference between the analyte value and the analyte value closest to the time of dosing, integrated over an interval after the time of dosing. For example, the DGA may calculate the right-hand AUC index by collecting all data points from the filtered analyte data within a time window (e.g., 1.5 hours), counting back from the time of injection, then calculating the difference between the average analyte value of the collected data points and the data point closest to the time of injection (the reference data point), and multiplying this difference by the duration of the time window to calculate the increment in the right-hand AUC.

[0097] The classification feature can further include the elapsed time between dosing times. For example, for each injection time, the DGA can calculate the elapsed time between the previous injection time and the current injection time by subtracting the previous injection time from the current injection time. For the first injection time in the insulin log, there is no previous injection time available, so the DGA can calculate the elapsed time from the first SCG time data point to the current injection time. Additionally, as a further example, the DGA can calculate the elapsed time between the current injection time and the next injection time by subtracting the current injection time from the next injection time. For the last injection time in the insulin log, there is no next injection time available, so the DGA can calculate the elapsed time from the current injection time to the last SCG time data point. In both backward and forward calculations, if the elapsed time is greater than a predetermined maximum value (e.g., 12 hours), the DGA can set the elapsed time value equal to the maximum time.

[0098] The classification features may further include the probability that a meal begins within a defined interval before the time of dosing, e.g., the maximum probability of meal initiation within a time window (e.g., 1.5 hours) before the injection, which may be calculated by a meal detection module described elsewhere herein.

[0099] Classification features may further include the most probable interval of time elapsed since the most recent meal, for example, the time elapsed since the point of maximum probability of meal onset relative to the time of injection (eg, as determined by a meal detection module).

[0100] The classification features may further include the probability that a meal will be initiated within a defined interval after the time of administration, for example, the maximum probability of a meal being initiated within two hours after injection (as determined by the meal detection module).

[0101] The classification features may further include the most probable interval until the next meal, for example, the predicted elapsed time from the injection time to the point of maximum probability of meal onset after meal injection (e.g., as determined by a meal detection module).

[0102] As mentioned above, calculating a portion of the classification features includes estimating the time of each meal eaten by the patient during the analysis period; methods for estimating meal times are described in further detail below. Briefly, estimating the time of each meal can further include generating a feature matrix based on the time-correlated analyte data using DGA, where the feature matrix correlates a set of analyte (e.g., glucose) data features with each of the distinct regions classified as rise, pre-fall, and fall. The set of analyte data features can be or include maximum analyte rate of change, maximum analyte acceleration, analyte value at the point of maximum analyte acceleration, duration of the region, height of the region, maximum deceleration, average rate of change within the region, and time of maximum analyte acceleration. The estimating step can further include generating estimated meal times based on the feature matrix using an algorithm described below.

[0103] More detailed aspects of retrospective meal time detection algorithms for use in method 300 or elsewhere are described in the following paragraphs, followed by a description of other aspects of method 300. The DGA can perform retrospective meal time detection based on time-correlated analyte data by executing one or more code modules, e.g., a feature extraction module and a meal detection module. The feature extraction module, when executed by the DGA, causes the DGA to receive a glucose time series as input and output a feature matrix that is passed to the retrospective meal detection module to detect glucose excursions in response to meal events.

[0104] DGA can perform feature extraction using the following operation, described below, which can be divided into a sequence of three sub-operations: smoothing, segmentation and extraction.

[0105] In the smoothing sub-operation, the DGA can smooth the analyte (e.g., glucose) time series using a Savitzky-Golay filter (order 2) and calculate the rate of change and acceleration at each analyte data point. The filter's frame length parameter can be the number of data points collected in a first time interval (e.g., 60 minutes), and thus sampling is interval-dependent. The DGA can calculate the rate of change by averaging the difference between the smoothed analyte value before and after the point of interest and a point within a second interval (e.g., 15 minutes) before and after the point of interest, where the second interval is smaller than the first interval, e.g., equal to one-quarter of the first interval. Similarly, the DGA can calculate the acceleration rate by averaging the difference between the rate of change of the analyte before and after the point of interest and a point within a second interval (e.g., 15 minutes) before and after the point of interest.

[0106] In the segmentation sub-operation, the DGA can segment the smoothed analyte trace into monotonically increasing (i.e., rise) and decreasing (i.e., fall) regions, with each rising region considered as a candidate glucose excursion in response to a meal event.

[0107] In the extraction sub-operation, the DGA can extract features from the data. For example, 16 features can be extracted, which may be or include features from each ascending region (e.g., eight features), the preceding descending region (e.g., four features), and the following descending region (e.g., four features). Features that the DGA can extract from the ascending features can include, for example, the following: 1) maximum analyte rate of change, 2) maximum analyte acceleration, 3) analyte value at the point of maximum analyte acceleration (reference point), 4) duration of the ascending region (time elapsed from the reference point to the last point of the region), 5) height of the region (difference in smoothed analyte value between the last point and the reference point), 6) analyte value at the point of maximum analyte acceleration (reference point), 7) average rate of change within the region (height / duration), and 8) time of the data at the reference point. Further by way of example, four features extracted from the preceding and following descending regions can include: 1) height of the falling region, 2) duration of the falling region, 3) average rate of change of the region (height / duration), and 4) maximum absolute value of the rate of change of glucose. The number of rows in the feature matrix output by the feature extraction module may be equal to the number of rising regions in the smoothed glucose time series.

[0108] According to another aspect of the embodiment, the retrospective meal detection module can receive a feature matrix as input and output a binary detection result for each elevated region. Such output can include a binary classification result and a probability value that each elevated region is an analyte (e.g., glucose) excursion in response to a meal event. The DGA can assign the probability value of each elevated region to its reference point. In some embodiments, for example, a pre-trained machine learning model for meal detection can be implemented using a random forest classifier by scikit learn (https: / / scikitlearn.org / stable / modules / generated / sklearn.ensemble.RandomForestClassifier.html). The meal detection module can detect meal-induced postprandial glucose excursions based on multiple decision trees constructed and optimized during the training process. In alternative embodiments, the DGA can construct pre-trained models based on alternative classification algorithms, including, for example, gradient boosting, ADA boosting, artificial neural networks, linear discriminant analysis, and extra trees.

[0109] Referring again to method 300 of FIG. 7 , classification operation 302 can receive a patient's feature matrix as input and output a binary classification result for each associated medication event (e.g., for each insulin injection). For example, DGA can output binary data "1" indicating a meal dose and "0" indicating a non-meal dose. According to some embodiments, classification operation 302 can use meal detection results, in which case meal detection can occur before insulin dose classification. As mentioned for retrospective mealtime detection, classification operation 302 can include a pre-trained machine learning model, such as a model implemented using a random forest classifier by Scikit Learn (see above). The machine learning model implemented by DGA can perform classification based on tree-building rules and thresholds for various features in each decision tree, optimized during the training process. Alternatively, the model can be trained using other machine learning algorithms, including gradient boosting, ADA boosting, artificial neural networks, linear discriminant analysis, and extra-trees. Once DGA successfully classifies each dose, the drug can proceed to determine the dosing regimen and administration parameters.

[0110] In step 304, method 300 can include a step in which the DGA groups each of the doses into one of a set of mealtime groups or clusters. For example, the DGA can determine a dosing strategy by clustering an analysis of the administration (e.g., injection) times of the prandial doses. The DGA can execute a clustering module implemented using a K-means algorithm with the elbow method, which receives the injection times as input and outputs an optimal number of clusters K (up to 3) and a cluster index for each injection time. The optimal number of clusters K can be the number of prandial doses the patient takes per day. Using the cluster index for each injection, the DGA can classify the prandial doses into K groups according to their cluster indexes.

[0111] The DGA can identify these groups as breakfast, lunch, or dinner (B, L, D) as follows: For each group, the DGA can determine a typical time of day (TOD) by calculating the median TOD for the group. Alternatively, the DGA can use other centroid metrics. If K=3, the DGA can associate breakfast with the group after the longest period between the group's typical TODs. Then, the next group is lunch, and the last group is dinner. If K=2, the DGA can use assumption rules regarding the time between each meal to infer which group is associated with breakfast, lunch, or dinner. For example, if two groups are more than six hours apart, the DGA can identify the groups as breakfast and dinner. Otherwise, if the first group occurs before 10:00 a.m., the DGA can identify the groups as breakfast and lunch; otherwise, it can identify the groups as lunch and dinner. In an alternative embodiment, after the DGA identifies typical times for meal events, the user can be prompted to identify meals associated with each typical time. As a further example, in an alternative embodiment, the DGA can combine the two methods described herein by estimating meal associations and then prompting the user for confirmation. A further alternative method could include analyzing glucose data to identify meals and clustering meal times to detect typical meal times. This could be useful for distinguishing meals when K=2, i.e., identifying meals for which no doses were taken.

[0112] Once the doses have been grouped into mealtime clusters, in step 306, the DGA can perform a step of generating patient dose parameters at least in part by applying the data for each mealtime group to a model. For example, for each meal group (B, L, D), the DGA can pair each set of corresponding preprandial glucose levels with a corresponding meal dosage. The DGA can fit each group with an appropriate model, such as a linear function with zero slope, a linear function with non-zero slope, a piecewise linear function connected at a point, or a nonlinear function that approximates the connected piecewise model but has smooth curvature around the connection point. Other models are also suitable.

[0113] DGA can perform model fitting and parameter estimation by minimizing the sum of squared residuals (SSR) of the model parameters. DGA can then use a grinding algorithm to find the optimal parameters that minimize the SSR value. For linear models, DGA can perform fitting using the Nelder-Mead simplex method. For nonlinear models, DGA can use the Levenberg-Marquardt algorithm. That is, DGA can use the Nelder-Mead simplex numerical optimization method for linear models and the Levenberg-Marquardt optimization method for nonlinear models. Alternative methods for fitting data to these models are also possible.

[0114] If the number of iterations during optimization exceeds a convergence criterion, the model will not fit, and the DGA can eliminate the non-fitting model as a candidate model. Additionally, the DGA can apply certain rules to minimize the uncertainty in the parameter estimates, such as by requiring at least three pre-meal glucose data points greater than an estimated threshold glucose to validate the estimated correction factor, or by requiring at least three pre-meal glucose data points less than an estimated threshold glucose to validate the estimated fixed volume, or by requiring that the 95% confidence interval for the parameter intercept exclude zero, or by requiring that the 95% confidence interval for the model slope exclude zero.

[0115] Insufficient data can lead to model fitting failure, resulting in certain models being excluded as candidate models.DGA can evaluate each model using the Akaike Information Criterion (AIC) and select the model with the smallest AIC value as the desired model for each dietary group.

[0116] After selecting a model for each mealtime cluster, the DGA can then determine dose parameters, including, for example, fixed dose insulin amounts, target glucose levels, and correction factors, based on the selected model for each mealtime cluster. The DGA can determine the target glucose levels and correction factors as single values ​​for all groups, as described in more detail in the next paragraph. In an alternative embodiment, the DGA can determine the target glucose levels and correction factors separately for each group and use the separately determined parameters for downstream dose guidance operations.

[0117] According to another aspect of the embodiment, the DGA can form a combined data group to obtain a more accurate correction factor for the patient. For example, after fitting the dose data to various models for each meal group to select the best-fit model and estimate the fixed insulin dose, the DGA can subtract the fixed dose insulin amount from the associated meal dose for each meal group. The remaining non-zero values ​​correspond to doses with correction doses. These non-zero values ​​can then be combined from all three meal groups (B, L, and D) to form a combined group. If the fixed dose insulin amount cannot be determined for a group, the DGA can exclude that group's data from the combined group. The system can then repeat the process to find the best model for the combined group, or use the same model identified when analyzing the groups individually. Using this combined group approach, the system assumes that the patient has the same (or constant) correction factor and target glucose for all meals, which can provide a more accurate fit with a larger sample size for the combined group. After the DGA determines the target glucose level and correction factor based on the best-fit model, the dose parameter estimation is complete. The DGA can then store the dose parameters in computer memory and configure the dose guidance settings in step 308.

[0118] In additional embodiments, the DGA can determine whether the patient is potentially carb counting (e.g., varying meal dosage to account for carbohydrate consumption) by comparing the AIC value of the preferred model to a threshold value such as 50, or 75, or 100. If the AIC value is greater than the threshold value, the DGA can determine that the patient is carb counting and ask the patient for confirmation via UID 200.

[0119] In alternative embodiments, one or more of the above-described operations may be omitted and replaced by requiring the patient or HCP to manually provide the information, or by extracting the information from another source, such as an EMR or other software program. Nevertheless, method 300 should be useful for a variety of applications that do not have more information than the SCD and MDD can provide.

[0120] User feedback during the learning period Next, an exemplary embodiment of a method for obtaining user feedback during or after the DGA is learned is described. The user can be prompted for feedback during the initial learning phase of the DGA. The user feedback can indicate to the user that the system is making progress. The DGA can prompt the user for feedback (e.g., input or confirmation) regarding any aspect of the dose guidance, such as administered doses, analyte history, patient behavior or activity, general dosing strategy, specific dose types, or confirmation that the dose type or strategy determined by the DGA (e.g., learned by the system) is correct.

[0121] During (or after) the learning period, the DGA can output prompts or other indications to the UID 200 requesting user feedback. This feedback can be regarding the dosing strategy, for example, the strategy regarding insulin action type (e.g., long-acting and / or short-acting or rapid-acting). If the feedback (or other determination) indicates the use of a long-acting strategy, the DGA can monitor the patient's basal dosing pattern during a first time period, e.g., the first three days, and classify each dose or dosing pattern as a single or split dose type and / or characterize the dose by time period (e.g., a single dose in the morning, a single dose in the evening, or split doses (e.g., both morning and evening)). The DGA can also determine trends (e.g., median, mean) and associated dose variability values ​​for the dosage. From this information, the DGA can develop an expected basal dose. After the first time period, if the actual administered dose (e.g., automatically registered by the MDD 152 or entered by the user) differs from expectations, the user can be prompted for feedback.

[0122] According to one aspect of the embodiment, the user can be prompted in many different situations. For example, the DGA can be configured to detect a missed dose, such as when the user fails to administer a basal or bolus dose during the time period in which the previous basal or bolus dose is administered. If a missed dose is detected, the DGA can be configured to prompt the user for input regarding whether a basal dose was administered during that time period. According to some embodiments, the DGA can be configured to detect differences in administration timing. For example, the DGA can be configured to detect when the user administers a basal dose at a different time than the time in which the previous basal dose was administered (e.g., a basal dose normally administered in the morning is administered in the evening). If such a difference in administration timing is detected, the DGA can be configured to prompt the user for input regarding whether the basal dose was administered during a different time period. In another aspect of this embodiment, the DGA can also be configured to detect when an extra dose has been administered. For example, the DGA can be configured to detect a change in the number of basal doses administered per day. In yet another aspect of the embodiment, the DGA can be configured to detect whether the dosing strategy on day 1 (e.g., one basal dose was administered) differs from the dosing strategy on day 2 (e.g., two basal doses were administered). If a different dosing strategy is detected, the DGA can be configured to prompt the user for input regarding whether the user has adopted the dosing strategy used on day 2 as a new dosing strategy. In yet another aspect of the embodiment, the DGA can also be configured to detect whether a different dosage has been administered. For example, the DGA can be configured to detect whether a first amount of a dose administered in a time period is different (smaller or larger) than a preceding dose administered in a time period on the previous day. If a different dosage is detected, the DGA can be configured to prompt the user for input regarding whether the user has changed the dosage.

[0123] The user's response to these prompts can either confirm that the DGA has identified the correct pattern (e.g., the user confirms that they forgot to take their morning basal dose but normally take it) or provide the user with an opportunity to correct the pattern (e.g., the user informs the DGA that they will adjust the basal dose based on glucose before taking it).

[0124] For fast-acting insulin dosing strategies, in addition to the prompts above, the DGA can include prompts for dose classification, which can include, but are not limited to, bolus, correction, split dose, bolus + correction, and bolus + carbohydrate count + correction classifications.

[0125] The user can be prompted in various situations regarding the administration of fast-acting insulin. The DGA can be configured to detect whether a dose not associated with a meal was administered. For example, the DGA can be configured to determine whether a dose was taken during a time period in which no meal was identified or detected. If the DGA detects the administration of a dose and no meal was detected within the time period of administration (e.g., within about one hour of administration), the DGA can prompt the user for input regarding the reason the dose was administered (e.g., because they ate, lowered their glucose, or completed the administration of a previous meal). The DGA can also be configured to detect when a meal dose does not match a previous meal dose associated with the same meal type. For example, the DGA can be configured to determine whether a bolus dose associated with a first meal type and administered during a time period is not the same as a previous bolus dose associated with the first meal type and administered during that time period on the previous day. If such a difference in bolus dose is detected, the DGA can be configured to determine the reason for the different dose. For example, the DGA can be configured to determine the difference in pre-meal glucose values ​​associated with the bolus dose and the previous bolus dose and determine whether the detected difference is a correction. The DGA can also prompt the user for input regarding the reason for the difference in bolus dose (e.g., under / overeating and / or correcting hyperglycemia and / or correcting other factors).

[0126] In addition to allowing the DGA to determine what types of immediate-acting doses are being taken throughout the day, it can facilitate the timing of expected doses. After a learning period in which no prompts are provided, the DGA can provide these prompts to the user when the dose differs from the expected dose in order to improve the DGA's model of the user's medication strategy.

[0127] For both long-acting and fast-acting treatments, DGAs can aim to minimize the number of prompts over time and as the user responds, focusing on frequent prompts early on and tapering off if repetitive patterns are observed.

[0128] Glucose pattern analysis and meal bolus titration for MDI insulin therapy An exemplary embodiment of a method for determining meal bolus titration will now be described. Once the system learns (or is configured for) a patient's current dosing strategy, it can provide titration guidance for multiple injection (MDI) dosing therapy. For patients using fixed meal dosing, fixed doses (e.g., breakfast, lunch, dinner, snacks, etc.) can be titrated. For patients using carbohydrate counting, carbohydrate ratios can be adjusted for the same meal or different time periods. For patients using empirical dosing, dose titration can be performed for each meal. Titration guidance via DGA can provide recommendations for changing the dose or carbohydrate ratio in a specific direction. The change can be an appropriate percentage, e.g., 5%, 10%, 15%, etc. Dose guidance can also include the initiation of meal dosing. For example, if a patient is using a basal+1 (e.g., lunch dosing regimen) and breakfast exhibits a high pattern, DGA can provide a recommendation for administering RA insulin at breakfast.

[0129] The DGA may require defining administration categories such as time of day (TOD) period, meal type (e.g., breakfast), and meal composition (e.g., cereal with milk). For example, an administration category may be a time of day period defined by the duration associated with prandial insulin dosing during the time of day period. In a further example, a post-breakfast time period may be defined as beginning when a prandial insulin dose is taken during a defined time period, e.g., between 5:00 AM and 10:00 AM, and ending after a defined post-prandial period (e.g., 6 hours later) or when the next prandial insulin dose is taken, whichever occurs first. One or more metrics may be required to define whether a post-prandial glycemic response is nominal or requires correction, or to rank one post-prandial glycemic pattern as more favorable or unfavorable than another. The low glucose likelihood (LLG) index and median glucose level may be used to quantify the degree of hypoglycemic and hyperglycemic risk, respectively.

[0130] U.S. Patent Publication No. 2018 / 0188400 (the '400 publication), the entire contents of which are incorporated herein by reference for all purposes, describes an embodiment for deriving and determining a risk index that can be utilized in glucose pattern analysis (GPA) of DGA embodiments. This embodiment utilizes, among other things, central tendency (e.g., mean, median, etc.) and variability value data from a multi-day period to determine a risk index that corresponds to the degree of hypoglycemic risk ("hypo risk"). This embodiment is summarized herein, and a more comprehensive description of the embodiment and its variations can be obtained by reference to the '400 publication.

[0131] An alternative to the embodiment described in the '400 publication is described in U.S. Patent Publication No. 2014 / 0350369, which is also incorporated herein by reference in its entirety for all purposes. For example, instead of using a median and a variability value, the method can employ any two statistical measures that define the distribution of the data. As described in the '369 publication, the statistical measures can be used to determine the glucose target range (e.g., G LOW = 70 mg / dL and G HIGH = 140 mg / dL). Common measurements related to the target range are time in target range (TIR), time above target (t AT ), and the time to fall below the target (t BT ) If the glucose data is modeled as a distribution (e.g., a gamma distribution), then a predefined threshold G LOW and G HIGH About t AT and t BT For the threshold, the algorithm calculates t BT_HYPO We can also define t BT If the blood glucose level exceeds G, the patient can be determined to be at high risk of hypoglycemia. LOW = 70 mg / dL, t BT can be defined as when t exceeds 5%. AT_HYPER We can define an index called t AT If the blood glucose level exceeds 100mg / dL, the patient can be judged to be at high risk of hyperglycemia. The degree of risk of hypoglycemia and hyperglycemia is determined by the following equations: LOW or t BT_HYPO , G HIGH or t AT_HYPER It can be adjusted by adjusting either TIR or t BT and t AT Any two of the three measures can be used to define the control grid. These options (and others) can be used to determine the risk index for the DGA embodiments described herein.

[0132] The DGA embodiments described herein can operate based on a quantitative evaluation of a user's analyte data during a TOD period. This quantitative evaluation can be performed in a variety of ways. For example, the embodiments described herein can evaluate analyte data across multiple days to determine one or more indicators that describe the associated risk indicated by the analyte data for the corresponding TOD. These indicators can be used to classify the analyte data during the TOD period into one of a number of patterns. For example, these patterns can indicate common or generalized glucose behaviors or trends for that TOD. DGA embodiments can utilize any number of two or more patterns. For ease of reference, these patterns are referred to herein as glucose pattern types, and although the embodiments described herein refer to implementations utilizing three glucose pattern types (e.g., low pattern, high / low pattern, and high pattern), other embodiments can utilize only two types or more types, and these types can differ from those described herein.

[0133] Taking a fixed meal dose as an example, once the DGA has learned the dosing strategy and dose or carbohydrate ratio amount, it can begin titration assessment, which can be classified into four titration categories: "nighttime," "after breakfast," "after lunch," and "after dinner." For each of these categories, the DGA can map the two indicators (LLG and median glucose) mentioned above to four logical "pattern" variables using the GPA method described below. FIG. 8A illustrates an exemplary method 400 operation by DGA for evaluating meal-bolus titration for multiple-injection (MDI) dosing therapy. Method 400 can include, at 402, determining at least one TOD analyte pattern type by executing a glucose pattern analysis (GPA) algorithm that receives as input time-correlated analyte data originating from a sensor-control device worn by the patient during the analysis period. Method 400 can also include, at 404, selecting an MDI dosing recommendation based on the analyte pattern type and the patient's defined dosing strategy for the analysis period by executing a recommendation algorithm. Method 400 may further include, at 406, storing a recommended course of action in computer memory by the DGA for output to at least one of the UID 200 or MDD 152, which administers the medication to the patient. The UID 200 may use the recommended course of action to control a user interface, for example, by causing a human-readable representation of the guideline to appear on a display or by generating an audio output of the guideline in a human language. The MDD 152 may use the guideline to adjust or maintain the next associated dose administration. Further details of method 400 are provided below.

[0134] 8B is a flowchart illustrating an example embodiment of a GPA method 410 that may be implemented as the GPA algorithm referenced at 402. Method 410 may be performed for a specific TOD period, which may be an entire day (e.g., a 24-hour period) or a portion of a day defined by time blocks (e.g., three 8-hour periods) or user activities (e.g., eating, exercise, sleep, etc.). In many embodiments, the TOD periods may correspond to meals (e.g., after breakfast, lunch, and dinner) and sleep (e.g., overnight). These TOD periods may correspond to fixed times of day when activities would typically occur (e.g., after breakfast from 5:00 AM to 10:00 AM), and such time blocks may be user-configurable or may be contingent on a meal or activity actually occurring, as determined by automatic meal or activity detection or user instruction indicating the same (e.g., using UID 200).

[0135] The DGA can perform method 410 independently for each TOD period to obtain an individual pattern assessment for that period. At 412, the DGA can determine central tendency and variability values ​​from the user's analyte data for a particular TOD period. The user's analyte data can be available from the user's own records or the user's medical professional's records, or the user's analyte data can be collected, for example, by the DGS 100. The analyte data preferably spans a multi-day period (e.g., 2 days, 2 weeks, 1 month, etc.) so that there is enough data within the TOD period to make a reliable determination. In other embodiments, the method can be performed in real time on limited data. The DGA can use any type of central tendency measure that correlates to the central tendency of the data, including, but not limited to, the median or mean. Any variability measure can also be used, including, but not limited to, a variability range that spans the entire data set (e.g., minimum to maximum), a variability range that spans the majority of the data but less than the entire data set to reduce the significance of outliers (e.g., from the 90th percentile to the 10th percentile, from the 75th percentile to the 25th percentile), or a variability range that covers a particular asymmetric range (e.g., low range variability, which can span, for example, from the central tendency value to a lower value of the data, e.g., the 25th percentile, the 10th percentile, or the minimum). The choice of measures representing central tendency and variability can vary depending on the implementation.

[0136] At 414, the DGA can assess a hypoglycemia risk (“hypo risk”) indicator based on the central tendency and variability values. One such methodology for determining hypoglycemia risk is described with respect to FIG. 8C , which shows an exemplary embodiment of a framework for determining hypoglycemia risk and other indicators. FIG. 8C is intended to convey the framework to the reader; however, this framework can be implemented electronically in a number of different ways, such as a software algorithm (e.g., a mathematical formula, a set of if-else statements, etc.), a lookup table, firmware, or a combination thereof.

[0137] 8C is a graph of central tendency versus variability (e.g., low-range variability) that can be used to evaluate or identify a region or zone that holds or corresponds to a data pair of central tendency and variability determined for a particular TOD. Any number of two or more zones can be used. In this embodiment, the data pair can correspond to a target zone 425 or one of three hypoglycemic risk regions: low 426, medium 428, or high 430. A first hypoglycemic risk function (e.g., a curve or linear boundary), referred to as medium risk function 422, distinguishes between low 426 and medium 428. A second hypoglycemic risk function, referred to as high risk function 424, distinguishes between medium 428 and high 430. The central tendency versus variability data pair can be evaluated or compared to the zones to determine a hypoglycemic risk index for the corresponding TOD period.

[0138] The hypoglycemia risk functions 422 and 424 can be explicitly implemented in the DGA as mathematical functions (e.g., polynomials), or can be implicitly implemented by defining each zone with an included pair, using a lookup table, a set of if-else statements, threshold comparisons, etc. The hypoglycemia risk functions 422 and 424 can be preloaded into the DGA, downloaded from a trusted computer system 480, or set by another party such as an HCP. Once implemented in the DGA, the hypoglycemia risk functions 422 and 424 can be treated as fixed or can be adjusted by the user or HCP. An exemplary methodology for determining hypoglycemia risk functions is described in the '400 publication.

[0139] At 416, the DGA can assess a hyperglycemic risk index (“hyper risk”) based on the central tendency value. In this embodiment, hyperglycemic risk can be assessed by comparing the central tendency value for a particular TOD period to a central tendency target or threshold 432. The magnitude and / or sign of the difference of the central tendency value from the target 432 can identify the amount of hyperglycemic risk. For example, a low hyperglycemic risk may exist if the central tendency value is less than the target 432 (e.g., a negative value). A moderate hyperglycemic risk may exist if the central tendency value is less than a threshold (e.g., 5 percent, 10 percent, etc.) but exceeds the target 432 (e.g., a positive value). A high hyperglycemic risk may exist if the central tendency value exceeds the target 432 by a value greater than the threshold. The use of three distinct groupings for hyperglycemic risk (e.g., low, medium, high) is exemplary, and any number of two or more groupings can be used.

[0140] In other embodiments, the DGA can assess the hyperglycemic risk indicator at 416 before assessing the hypoglycemic risk at 414. Alternatively, in another embodiment, the assessment of hypoglycemic risk at 414 and the assessment of hyperglycemic risk at 416 can be performed simultaneously in parallel.

[0141] Other indicators, such as volatility risk, can also be evaluated. For example, a variability value less than a first variability threshold 434 can indicate a low volatility risk, a variability value greater than the first variability threshold 434 and less than a second variability threshold 436 can indicate a medium volatility risk, and a variability value greater than the second variability threshold 436 can indicate a high volatility risk. Again, the use of three distinct groupings for volatility risk is exemplary. The DGA can use any number of groupings greater than two.

[0142] In step 418, the DGA can determine a pattern type for the TOD period based on the evaluated one or more risk indicators. In an exemplary embodiment, the pattern determination can be evaluated using a hypoglycemia risk indicator and a hyperglycemia risk indicator. If the hypoglycemia risk indicator is high, the pattern can be set as a low pattern. Otherwise, if the hypoglycemia risk is moderate and the hyperglycemia risk is high or moderate, the pattern can be set as a high / low (or moderate) pattern. Otherwise, if the hyperglycemia risk is high or moderate and the hypoglycemia risk is low, the pattern can be set as a high pattern. If both the hyperglycemia risk and the hypoglycemia risk are low, the identified pattern can be No Problem (e.g., an "OK" message is displayed and output).

[0143] Thus, method 410 is one example of how the DGA outputs one of multiple pattern types for each TOD period. The number of pattern types themselves may be different than those described in this embodiment (e.g., low, high / low, high). Once the pattern type for a TOD period is determined, the DGA can store an indication of the pattern type in a memory location for use in determining a titration recommendation. Referring again to FIG. 8A, once the DGA has completed GPA for each associated TOD period at 404, it can proceed to determine a titration recommendation.

[0144] The recommendation method can branch depending on other factors such as pattern type (e.g., low, high / low, high), TOD period, dosing strategy, adherence to the strategy (e.g., whether the dose is underdose), and whether sufficient data is available to conduct the evaluation. The DGA will not make a titration recommendation until sufficient data is available for the corresponding TOD period; for example, the DGA can omit the evaluation and generate an error message if the amount of available data is below a threshold, e.g., if the number of distinct days with more than a minimum portion (e.g., 90%) of available data is below a threshold (e.g., 5).

[0145] 8D-8H show example branches of a recommendation algorithm or method for determining dose titration recommendations based on the input information described above. Other branches may also be useful. FIG. 8D shows a recommendation method branch 440 for TOD where sufficient data is available and a possible cause of a low pattern type includes one or more of a basal dose, prandial dose, pre-prandial correction dose, or post-prandial dose that is higher than optimal. At 442, the DGA evaluates whether the pattern type for the nighttime TOD period is low. If the pattern is low, at 444, the DGA generates a recommendation to reduce all associated doses, including at least the basal dose and, optionally, one or more of a prandial dose, pre-prandial correction dose, or post-prandial dose, by the same amount, e.g., 10%. Titration recommendation rules for a low pattern may include generating a recommendation to reduce the long-acting insulin dose or basal rate for the nighttime TOD period at 444. If other TOD periods are low patterns at 446, the DGA may generate a recommendation to reduce the fixed prandial dose only for the associated TOD period at 448.

[0146] In this embodiment, if there is at least one low pattern, no titration guidance is provided for the TOD period with a high pattern. The idea here is to emphasize hypoglycemia prevention and only increase the dose if the risk of hypoglycemia is low for all TOD periods. Also, in some cases, if a TOD period has a high pattern, this may be because the previous TOD period had a low pattern and the patient is overeating to compensate; therefore, addressing the low pattern may itself help address any subsequent high patterns. If the pattern is not high at 449, process 440 waits or ends without generating a recommendation, or passes high pattern evaluation 450.

[0147] Therefore, in the case of a high / low pattern, the DGA will not generate titration guidance. If there are no TOD periods for which titration guidance can be given and data for all time periods is sufficient, the DGA can provide a message to the patient indicating that glucose variability needs to be addressed before further titration guidance can be given. The DGA can also provide a report to the patient's HCP to consider alternative medications or therapies that can address glucose variability.

[0148] FIG. 8E illustrates operations by the DGA for generating titration recommendations for a high pattern when there are no TOD periods with a low pattern. If an overnight period has a high pattern at 452 and there are no other periods with a moderate risk of hypoglycemia, the DGA can increase the long-acting insulin dose or basal rate recommendation at 454. If an overnight period has a high pattern and there is at least one other non-dinner period with a moderate risk of hypoglycemia at 456, the DGA can decrease the prandial insulin dose associated with any period with a moderate risk of hypoglycemia at 458. If an overnight TOD period does not have a moderate risk of hypoglycemia and does not have a high pattern at 460, the DGA can generate a recommendation to increase the prandial insulin amount associated with the first TOD period with a high pattern at 462. If an overnight period has a moderate risk of hypoglycemia and the only postprandial period with a high pattern at 464 is dinner, the DGA can generate a recommendation to increase the long-acting insulin dose or basal rate at 466. If the nighttime period has a moderate risk of hypoglycemia and the post-dinner period does not, then at 462 the DGA can generate a recommendation to increase the meal insulin dose associated with the first TOD period with a high pattern.

[0149] In alternative embodiments, pre-meal glucose may be higher or lower than a target glucose (e.g., 120 mg / dL). The glucose data for each meal that contributes to the calculation of hypoglycemia and hyperglycemia risk indicators can be corrected to compensate for the effects of previous meals or conditions that affect glucose not attributable to the current meal. The DGA can correct these data by subtracting an offset so that the resulting starting glucose is at the target level. Alternatively, the DGA can correct these data with a "trigonometric" function, in which the difference between the meal start glucose and the target glucose is subtracted for the start time of the meal, but this correction is reduced linearly over a defined period (e.g., 3 hours) or by another decay function.

[0150] Alternatively, the function may itself be a function of the glucose level or glucose trend at the start of the meal and / or when the previous meal dose was taken.

[0151] According to another aspect of the embodiment, the algorithm for generating meal bolus titration recommendations may become more complex when considering additional aspects, such as, for example, missed meal doses, missed basal doses, post-meal corrections, and pre-meal corrections, etc. An algorithm for providing appropriate recommendations in the presence of these factors may need to exclude some data while still meeting the data sufficiency threshold after excluding the data to provide guidance.

[0152] For example, referring to FIG. 8F, if a high pattern is detected at 461 and there are several days with missed meal doses, the days with missed meal doses are excluded at 463 and the GPA analysis 410 is repeated. Then, if a high pattern is detected at 465, the system can either increase the dose at 467 based on patterns identified in other TODs or wait for further input or return at 469. Alternatively, the system can evaluate only the high pattern using data that excludes the days with missed meal doses. An algorithm 470 with this branching pattern is shown in FIG. 8F. If the system detects a low pattern at 473, it can execute the low pattern algorithm 472, described in the next paragraph. If the system does not detect a high or low pattern, it may return to block 469 for further input or return.

[0153] At 472, if a meal administration is missed, the DGA can detect a low pattern during the TOD period and generate recommendations to reduce the dose if meal administrations are missed on any days during the TOD. Recommendations can include, for example, reducing the fixed dose or compensation dose portion.

[0154] Regarding missed basal doses, if the DGA detects a low pattern 473 on a nighttime TOD, the missed basal dose should not impact the dose titration logic. Similarly, if a low pattern is detected on a non-nighttime TOD, the missed basal dose should not impact the dose titration logic.

[0155] If the DGA detects a high pattern 461 in TODs using data that includes at least one day (or TOD) with a missed basal dose, the data for any day (or TOD) with a missed basal dose can be excluded 463, and the pattern analysis 410 can be repeated. Subsequent actions can depend on the particular TOD for which the high pattern was detected. For example, if the DGA detects a high pattern in nighttime TODs for data that includes at least one day with a missed basal dose, the data for any day with a missed basal dose can be excluded, and the pattern analysis can be repeated. If a high pattern is detected in nighttime TODs when days with missed basal doses are excluded, the results of the nighttime TODs can be used to adjust the basal dose, and the basal dose can be increased. If a high pattern is detected in TODs other than nighttime TODs, the days with missed basal doses can be excluded, and the pattern analysis can be repeated. If a high pattern is detected when days with missed basal doses are excluded, the meal doses associated with the TODs with the high pattern can be analyzed and titrated, as described herein. In either case, the logic flow 470 is as illustrated in Figure 8F.

[0156] FIG. 8G shows an example logic flow 474 for formulating recommendations involving post-meal corrections. After GPA 410, if the DGA detects a low pattern 479 of TODs over several days that include post-meal corrections, the following analysis can be used to titrate the correction or meal dose. If the DGA first detects a low pattern 479 at 475, it can first test that sufficient data is available at 487 to exclude data from days without post-meal corrections. If sufficient data is not available, the DGA may perform an error recovery routine 489, such as displaying an error message. If sufficient data is available, the DGA may repeat pattern analysis 410. Subsequently, if the DGA detects a low pattern, it can decrease the post-meal correction dose (i.e., increase the correction factor) at 476, contingent on the pattern analysis results at other TODs. If the DGA does not subsequently detect a low pattern, it can decrease the meal dose at 477.

[0157] In all embodiments described herein, a correction dose modification (e.g., titration) in one direction can be achieved by a correction factor modification in the opposite direction. The two parameters have an inverse relationship, such that an increase in correction dose can be achieved by a decrease in correction factor, and a decrease in correction dose can be achieved by an increase in correction factor. Thus, in all embodiments described herein, the DGA can recommend or implement a correction by either a correction factor modification or a correction dose modification. Thus, to the extent that a correction factor modification or titration is described herein, the embodiment can be configured to achieve the same effect by an inverse correction dose modification, and conversely, to the extent that a correction dose modification or titration is described herein, the embodiment can be configured to achieve the same effect by an inverse correction factor modification. Given this compatibility, both options are available in all embodiments described herein, although not all embodiments are described solely for ease of explanation.

[0158] Additionally or alternatively, starting with the original data set 491, the DGA can exclude days on which meal administrations were missed at 478. After finding sufficient data at 487, if pattern analysis 410 of these data excluding days on which post-meal corrections were taken at 490 does not indicate a low pattern, the DGA can recommend reducing the post-meal correction dose at 476. Alternatively, the DGA can implement logic 510 of FIG. 10B, thereby recommending reducing either the prandial insulin or pre-meal correction portion of the dose guidance. If the DGA does not detect a low value at 479 or a high glucose pattern at 492, it can wait for further input or return at 469. If the DGA detects a high pattern at 492, it may implement process 480 at block 471 (FIG. 8H).

[0159] Referring to FIG. 8H, if the DGA detects a high pattern 493 for a TOD where several days include postprandial corrections, the following procedure 480 can be performed to develop recommendations for corrections and titration of meal doses. At 481, the DGA can repeat pattern analysis 410, including data from days when meals were skipped and days when postprandial corrections were made. If the DGA then detects a high pattern at 494, it can increase the postprandial correction dose (i.e., decrease the correction factor) at 482, contingent on the pattern analysis results for other TODs. If the DGA does not detect a high pattern at 494, it can check for a low pattern at 495 and, if a low pattern is detected, return to 474 of FIG. 8G or otherwise wait for further input or return at 469. Although not shown in FIG. 8H, after filtering out any data for GPA 410 and before performing GPA, the DGA may test for data sufficiency and perform an error recovery routine if insufficient data is available.

[0160] Alternatively, or additionally, starting with the original data set at 493, if pattern analysis of data excluding days with missed meal administrations at 483 indicates a high pattern along either one of branches 2.1 or 2.2, the DGA can continue with procedure 480 as follows: For branch 2.1, if pattern analysis 410 of data excluding days with missed meal administrations at 484 (i.e., bolus-only data) does not indicate a high pattern at 497, the DGA can generate a recommendation to increase the post-meal correction dose at 482, contingent on pattern analysis of the other TODs. Otherwise, the DGA can generate a recommendation to increase either the prandial insulin or the pre-meal correction portion according to procedure 550 of FIG. 10C .

[0161] For branch 2.2, if the pattern analysis of data for only days that include post-meal corrections at 485 does not indicate a high pattern at 496, the DGA may increase either the prandial insulin or the pre-meal correction portion according to procedure 550 of Figure 10C. If no high pattern is detected at 496, the DGA may return to block 484.

[0162] If the correction factor titration recommendations from different TODs are conflicting and the patient is currently using the same correction factor across all TODs, the DGA can increase the correction factor. Procedure 480 can increase the meal dose first if all three components - meal dose, pre-meal correction, and post-meal correction - are not optimal. The pre-meal correction can be titrated and increased after the meal dose has been titrated. The post-meal correction can be titrated and increased after the meal dose and pre-meal correction have been completed.

[0163] During subsequent analysis, if a TOD for which the DGA generated a recommendation to "increase the correction factor" using the method described above now receives a recommendation to "keep the correction factor unchanged," and conversely, if a TOD for which the DGA generated a recommendation to "decrease the correction factor" using the method described above now receives a recommendation to "decrease the correction factor," it is likely that different TODs were optimized using different correction factors.

[0164] 8D-8H illustrate various embodiments of recommendation algorithms 404 for use in method 400, but it should be understood that these are exemplary. Various other algorithms may also be suitable.

[0165] Food Bolus Titration Hysteresis Next, an exemplary embodiment of a method for reducing oscillations around the optimal meal bolus dose will be described.When DGA reaches or approaches the optimal titration, a situation may occur in which titration is still required but is not actually necessary.That is, the indicators used to determine whether titration is necessary may have a level of error or variability, and this variability may cause the titration algorithm to oscillate around the optimal dose.To reduce this oscillation problem, a PPC (Patient Parameter Convergence Tracking) module can be installed in DGA.

[0166] In one embodiment, the PPC module can collect historical information from the user, including sufficient data to determine the impact of various events at different times on the patient's glucose levels. The PPC module can be configured to calculate and track how multiple outcome indicators change over time. The outcome indicators may include, but are not limited to, hypoglycemia risk, hyperglycemia risk, and glucose control indicators. The PPC module can also be configured to calculate and track changes in dose guidance parameter estimates and / or dose guidance suggestions. Dose guidance parameter estimates include, but are not limited to, estimated insulin sensitivity factor / insulin correction factor, estimated insulin-to-carbohydrate ratio, and estimated average carbohydrates for each mealtime. Dose guidance suggestions include, but are not limited to, recommended basal insulin dose, recommended meal dose, recommended touch-up dose, and recommended correction dose.

[0167] In one embodiment, the PPC module of the DGA can be configured to create a correlation for each outcome indicator. The correlation can be multidimensional. The PPC module can be configured to map changes in each dose guidance parameter over time and / or changes in guidance suggestions over time to changes in the outcome indicator over time. The PPC module can also be configured to track the slope of a particular indicator. For example, if changes in multiple dose guidance parameters and / or changes in guidance suggestions cause a predicted small change in the outcome indicator, the PPC module can be configured to determine to delay adjustment suggestions for a period of time to prevent unnecessary adjustment requests to the user.

[0168] In another embodiment, the PPC module can be configured to use multiple gradients and compare each gradient to a set of predetermined gradient thresholds. The PPC module can be configured to determine to delay the adjustment recommendation if the number of gradients exceeding the thresholds is not greater than a predetermined value. The predetermined value can be a universal value or a user-specific value.

[0169] Physiological Dose Guidance Algorithm Next, we describe an exemplary embodiment of a method for determining dose guidance based on physiologically relevant processes. Many model-based control systems, particularly those utilizing model predictive control (MPC) algorithms, are based on black-box diagrams and lack a rigorous physiological basis. As a result, these models may not account for the specific pharmacokinetic and pharmacodynamic differences between different insulin analogs used in intensive MDI therapy. Furthermore, while MDI therapy requires a given insulin dose to manage a user's blood glucose levels over several hours, many dose guidance algorithms currently in use are intended for use with insulin pumps that continuously deliver only one type of insulin analog (usually fast-acting) to the user. By communicating with a CGM, the MPC algorithm can receive frequent glucose feedback (e.g., every 5 minutes) regarding the current insulin delivery rate and then modify the pump's flow rate in real time to maintain euglycemia. As a result, current MPC algorithms that can frequently modify pump-driven insulin delivery based on glucose feedback do not work with MDI therapy, which requires modest insulin delivery, and must forecast over a longer time horizon. As these horizons become longer (e.g., hours), the pharmacokinetics and pharmacodynamics of drugs can play a larger role in generating accurate predictions. This requirement necessitates the design of more physiologically informed insulin dose guidance and glucose control strategies.

[0170] Described herein is an insulin dose guidance algorithm that takes into account physiologically relevant processes such as insulin diffusion, subcutaneous pharmacokinetics (PK), and glucose-insulin dynamics. Parameters can be determined by numerically solving a modified minimal model of insulin absorption and glucose control (see Bergman, et al. 1978 and Dalla Man, et al., 2007, the entire contents of which are expressly incorporated herein by reference) during a "learning phase" to develop user-specific model parameters. Once these parameters are determined, the optimal current insulin dose can be determined by solving the user-specific minimal model for both prandial and correction doses.

[0171] As similarly described in other embodiments, the DGA can receive or otherwise access glucose values ​​and trends from a glucose monitoring system. The DGA can be configured to query the user for upcoming doses and for which meals dose guidance is needed. The DGA can be configured to output dose recommendations based on an algorithm, such as a physiological dose guidance algorithm or another algorithm described herein. The DGA can then be configured to monitor the user's adherence to dose guidance and can track the resulting glucose trace for consideration for future dose guidance.

[0172] The DGA can be configured to include a physiological dose guidance algorithm that determines the optimal dose, which is output in the dose guidance. In one embodiment, dose guidance can be provided for a user with a frequent insulin injection (MDI) regimen with a once-daily long-acting analog and a three-times-daily rapid-acting prandial analog correction. This methodology can also be applied to other MDI strategies, such as basal only or basal with a single rapid-acting injection.

[0173] In an exemplary embodiment, as depicted in the flow diagram of FIG. 9A, in exemplary method 900, beginning at step 904, DGA determines possible bolus insulin doses (u1, u2, ..., u1) based on past administration history. n In step 906, the physiological dose guidance algorithm processes the possible insulin doses and generates a glucose time course (g) for each bolus insulin dose input. 1, g2,...,g n ) (e.g., a data array) can be generated and output. Then, in step 908, the DGA generates cost function values ​​(C1, C2, ..., C n ) can be calculated. A cost function can be defined to minimize the time outside a range of time courses for each glucose time course (e.g., outside a range of about 70 mg / dL to about 180 mg / dL). In step 910, the DGA can determine the optimal insulin bolus dose that minimizes the cost function value.

[0174] The optimal insulin bolus amount (u) to be delivered at time t can be determined from the estimated glucose trace after simulating insulin dosing and carbohydrate meal input (if required). This dose is calculated using the following simple bolus formula:

[0175]

number

[0176] can be used to decompose it into its component parts, such as a glucose correction portion, a meal compensation portion, and an insulin onboard portion. However, the physiological dose guidance algorithm does not have to use equation (1) to determine the optimal dose. The physiological dose guidance algorithm can instead use the individual components of the equation to allow the user and HCP to understand what dose is intended for the current glucose level in relation to meal compensation. In the equation, u is the optimal insulin bolus dose. The first term of the equation in parentheses is TIFF0007662611000002.tif13114

[0177] is the glucose correction part, which is the user's current glucose (BG(t)) and target value (BG target ) The correction factor (CF) represents the user's specific sensitivity to insulin, i.e., how effective one unit of insulin is in lowering the user's blood glucose level. The second term (CHO * The IC represents the fraction of the insulin dose required to cover a meal if one is eaten. The CHO value indicates the amount of carbohydrates contained in the upcoming meal, and the IC value indicates the subject's specific insulin:carbohydrate ratio, i.e., how many grams of carbohydrates can be covered with a given insulin dose. Finally, the IOB value aims to account for active "insulin onboard" that can still derive a therapeutic effect to avoid insulin stacking. This value can be based on insulin pharmacokinetic models and endogenous insulin production. BG target The values ​​for IC and total daily insulin dose (TDD) can be defined by the user's HCP. In one embodiment, the initial condition for CF can be defined according to the "1800 rule," which is defined as 1800 / TDD. Additionally, in one embodiment, during the learning phase, CF can be improved based on correction dose data.

[0178] Physiological dose guidance algorithms can use different methods to determine dose guidance for correction doses, prandial doses, and basal doses. However, for all three types of dose guidance, DGAs can assume that rapid-acting insulin analogs (e.g., insulin lispro, aspart, or glulisine) are used, and because all three have similar subcutaneous PK profiles, their duration of action can be considered equivalent.

[0179] correction dose When dose guidance is requested from the DGA for a correction dose, the correction dose is not meal-related, so there are no additional carbohydrates to consider, resulting in CHO = 0. Therefore, the second term in equation (1) is equal to 0. The component parts of the correction dose are therefore the glucose correction part and IOB.

[0180]

number

[0181] In one exemplary embodiment, as described in the flow diagram of Figure 9B, once a correction dose is prompted in exemplary method 911, DGA may generate multiple candidate insulin doses, beginning at step 912. In one embodiment, the multiple candidate insulin doses may be generated based on the subject's past history of insulin dosing.

[0182] In step 914, the DGA can determine multiple glucose time courses corresponding to multiple candidate insulin doses. In one embodiment, the DGA can calculate subject-specific glucose time courses using a physiological dose guidance algorithm with minimal model correction for the range of possible dosages. Each candidate dose can be considered the sum of the injected insulin dose and the amount of long-acting insulin currently on-board. The amount of IOB can be calculated from the pharmacokinetics specific to the subject's long-acting analog and can be the same across all candidate input doses. Numerical differences between all candidate insulin doses can be due to differences in the fast-acting component of the overall dose.

[0183] In step 916, the DGA can calculate multiple cost function values ​​corresponding to multiple glucose time courses. In one embodiment, for each glucose time course, a different glucose control index is calculated and can be used to determine a cost function value that describes the risk of time outside a target range (e.g., about 70 mg / dL to about 180 mg / dL).

[0184] An exemplary cost function is shown below in Equation 5, although many other forms can be used. The area under the curve (AUC) is used in these calculations to incorporate both the magnitude and duration of hyperglycemic / hypoglycemic events.

[0185]

number

[0186]

number

[0187]

number

[0188] AUC HYPO and AUC HYPER The threshold values ​​of C can be defined as the minimum time and duration spent in either regime. Each term in C can have an associated weighting factor w, the whole set {w1, w2, ..., w n} must sum to 1. This weighting allows prioritizing protection from hypoglycemic events. The insulin dose associated with the smallest cost function value can be the suggested dose in the dose guidance output by the DGA.

[0189] In step 918, the DGA can determine an optimal insulin dose, the optimal insulin dose having the lowest cost function value among the multiple cost function values. In one embodiment, the DGA can be configured to determine the dose candidate that minimizes the out-of-range time-related cost function C as the optimal insulin dose.

[0190] In step 920, the DGA may output dose guidance including the determined optimal insulin dose.

[0191] In another embodiment, instead of simulating the resulting glucose trace from a guess set of possible insulin doses and selecting the best option, in an alternative embodiment, DGA can provide an initial dose guess, from which the glucose horizon and cost function value can be tracked. Constrained minimization can then be performed on the cost function using either a gradient-based or non-gradient-based approach (e.g., a genetic algorithm). Thus, each insulin dose guess can be selected following the initial dose to minimize the cost function.

[0192] Once the optimal insulin dose is determined, it can be further broken down to help the user understand what aspects of the correction it covers and how much insulin is currently onboard.

[0193] Dietary Dosage If dose guidance is requested from the DGA for a meal dose, similar to the calculation of the correction dose described above, the DGA can calculate a user-specific glucose time course over the range of possible dosages using minimal model modifications. Because dose guidance is meal-related, unlike the process of determining a correction dose, the DGA can also consider additional meal carbohydrate information when determining dose guidance. Because dose guidance is meal-related, CHO is greater than zero and the meal dose component is calculated based on the additional meal carbohydrate information (CHO). * Thus, the components of a meal dose include a glucose correction portion, a meal compensation portion, and an IOB, according to formula (1).

[0194] In one exemplary embodiment, as described in the flow diagram of FIG. 9C , in exemplary method 921, beginning at step 922, in response to a user query regarding dosage guidance for a meal, a DGA can determine a distribution of carbohydrate values ​​for the meal. The distribution of carbohydrate values ​​can include a carbohydrate value of central tendency, low carbohydrate values ​​less than the carbohydrate value of central tendency, and high carbohydrate values ​​greater than the carbohydrate value of central tendency. In one embodiment, a meal can be represented as a meal-specific distribution of carbohydrate values ​​with known descriptive statistics describing the central tendency (e.g., mean or median) and variation (e.g., standard deviation, coefficient of variation, 25 / 75% values). These data can be transformed to provide these summary statistics. As a result, each meal can have a unique carbohydrate distribution.

[0195] In one embodiment, the carbohydrate value of central tendency may be the mean or median. In one embodiment, the low carbohydrate value may be, for example, the 25th percentile of the glucose data, alternatively the 30th percentile, alternatively the 35th percentile, alternatively a value between the 20th and 40th percentile, or alternatively μ-σ if the data are normally distributed. In one embodiment, the high carbohydrate value may be, for example, the 75th percentile, alternatively the 80th percentile, alternatively the 65th percentile, alternatively the 60th percentile, alternatively a value between the 60th and 80th percentile of the glucose data, or alternatively μ+σ if the data are normally distributed.

[0196] In step 924, the DGA may determine a plurality of candidate insulin doses for each of the central tendency carbohydrate value, the low carbohydrate value, and the high carbohydrate value. In one embodiment, the plurality of candidate insulin doses may be generated based on the subject's past history of insulin dosing.

[0197] In step 926, the DGA can determine multiple glucose time courses corresponding to multiple candidate insulin doses for each of the central tendency carbohydrate value, low carbohydrate value, and high carbohydrate value. In one embodiment, the DGA can calculate subject-specific glucose time courses using a physiological dose guidance algorithm with minimal model correction for the range of possible dosages, as described above for the correction dose. Each candidate dose can be considered the sum of the injected insulin dose and the amount of long-acting insulin currently on-board. The amount of IOB can be calculated from the pharmacokinetics specific to the subject's long-acting analog and can be the same across all candidate input doses. Numerical differences between all candidate insulin doses can be due to differences in the fast-acting component of the overall dose.

[0198] In step 928, the DGA can calculate multiple cost function values ​​corresponding to multiple glucose time courses for each of the central tendency carbohydrate value, low carbohydrate value, and high carbohydrate value. In one embodiment, for each glucose time course, a different glucose control index is calculated and used to determine a cost function value describing the risk of time outside a target range (e.g., about 70 mg / dL to about 180 mg / dL). Although many other configurations can be used, in one embodiment, the DGA can be configured to calculate the multiple cost function values ​​using the exemplary AUC analysis described with respect to the correction dose calculation. In other embodiments, for each glucose time course, a different glucose value at a predetermined percentile of glucose values ​​within a predetermined (rolling) time window can be determined. For example, instead of the percent time of hypoglycemia or instead of the AUC below the low glucose threshold, the 5th percentile value (or other percentile below 30%) within the time window can be calculated and compared to the low glucose threshold. The time window can be a fixed function of the start and / or end of the event, e.g., about 30 minutes after a meal, alternatively up to about 300 minutes after a meal, or other definitions, a time of day, e.g., 9:00 AM to 11:00 AM. Also, different low percentiles with corresponding low glucose thresholds can be used. Instead of percent time in a hyperglycemic state or instead of the AUC above the high glucose threshold, the 90th percentile (or other percentile greater than 70%) can be calculated within the time window and compared to the high glucose threshold.

[0199] In step 930, the DGA may determine an optimal insulin dose for each of the central tendency carbohydrate value, the low carbohydrate value, and the high carbohydrate value. The optimal insulin dose may be the dose having the lowest cost function value among the multiple cost function values ​​for each of the central tendency carbohydrate value, the low carbohydrate value, and the high carbohydrate value.

[0200] In step 932, the DGA can output multiple dose guidances, including optimal insulin doses determined for each of the carbohydrate value of central tendency, the low carbohydrate value, and the high carbohydrate value. The dose guidance associated with the carbohydrate value of central tendency can correspond to median meal dose guidance. The dose guidance associated with the low carbohydrate value can correspond to "smaller than usual" meal dose guidance. The dose guidance associated with the high carbohydrate value can correspond to "larger than usual" meal dose guidance. In one embodiment, the DGA can output a distribution of meal carbohydrate values ​​as well as a range of meal dose guidance that takes current glucose values ​​into account. Extreme cases can be modified or skewed as needed. For example, the optimal insulin dose associated with a larger-than-usual meal may need to be based on the 60th percentile rather than the 75th percentile to avoid postprandial hypoglycemia as a result of an excessive bolus dose. Using equation (1), the optimal insulin dose can be further divided to display to the user and HCP the amount allocated to the meal or the amount allocated to preprandial glucose levels. In one embodiment, each component of the dose guidance (eg, glucose correction portion, meal compensation portion, and IOB) may be output and displayed to the user and / or HCP.

[0201] Basal dose Basal doses of long-acting insulin analogs are given once or twice daily, depending on the drug selected. For example, insulin glargine has a 24-hour duration of action and is administered once daily, while insulin detemir has a 12-hour duration of action and is administered twice daily. Because these time scales are much longer than the 4-5 hour duration of action associated with rapid-acting analogs, long-acting analogs can have different subcutaneous insulin PK profiles. The time course of action is a difference for basal doses compared to the other two. Like correction doses, basal doses are not associated with meals but rather are present during all meals due to their extended pharmacokinetic profile.

[0202] 9B, beginning at step 912, when the DGA is prompted for basal dose guidance, the DGA can generate multiple candidate basal insulin doses. In one embodiment, the multiple candidate basal insulin doses can be generated based on the subject's past basal dose history.

[0203] In step 914, the DGA can determine multiple glucose time courses corresponding to multiple candidate basal insulin doses. In one embodiment, the DGA can calculate subject-specific glucose time courses using a physiological dose guidance algorithm with minimal model correction for the range of possible dosages. The amount of IOB can be calculated from the pharmacokinetics specific to the subject's insulin analog dose and can be the same across all candidate input doses.

[0204] In one embodiment, the DGA can be configured to scale the multiple glucose time courses in time to reflect the longer duration of action (e.g., 12 or 24 hours) associated with long-acting analogs. Because basal insulin is typically a once-daily medication, long-acting basal guidance can be developed to prompt the user at the same time each day. Similar to determining correction and meal doses, the DGA can be configured to generate a glycemic profile for a potential basal dose. Each of the multiple glucose time courses includes three meal events, each of which may include an associated fast-acting dose input to mimic fluctuations within a 24-hour window of basal action. Each meal event may include a carbohydrate input as well as a fast-acting insulin input. In one embodiment, the value of the insulin input can be based on subject-specific median fast-acting analog requirements from MDD152 data. In one embodiment, the meal input can be represented by a carbohydrate amount of central tendency (e.g., a median or average carbohydrate amount), as described in Meal Dose Guidance Determination. In one embodiment, the meal event can be the same for all basal simulation events within a single dose guidance. In one embodiment, the basal dose is the only input value that varies.

[0205] In step 916, the DGA can calculate multiple cost function values ​​corresponding to the multiple glucose time courses. In one embodiment, for each glucose time course, a different glucose control index is calculated and used to determine a cost function value describing the risk of time outside a target range (e.g., about 70 mg / dL to about 180 mg / dL). In one embodiment, the cost function analysis for the multiple glucose time courses generated for the multiple basal dose candidates can be the same as that described for the correction dose analysis. In another embodiment, the cost function analysis for the multiple glucose time courses generated for the multiple basal dose candidates can be different from the cost function analysis used for the correction dose analysis.

[0206] In step 918, the DGA may determine the optimal basal insulin dose having the lowest cost function value among the multiple cost function values. In one embodiment, the DGA may be configured to determine the basal dose candidate that minimizes the out-of-range time-related cost function C as the optimal insulin dose.

[0207] In step 920, the DGA may output dose guidance including the determined optimal insulin dose.

[0208] Correction Factor Titration An exemplary embodiment of a method for determining correction factor titration is now described. While initiated by a physician, insulin dosing for diabetic patients has largely been at the patient's mercy for management throughout the course of the disease. These management methods are often empirical, with patients learning through trial and error to improve diabetes management. More quantitative determination methods when using MDIs often rely on simplistic approaches like insulin bolus calculators, which rely on frequent blood glucose monitoring and values ​​like insulin:carbohydrate ratios or insulin correction factors that are difficult to determine and understand from the patient's perspective. The algorithmic subroutine described herein leverages granular glucose data from a CGM device and insulin dosing information from a Bluetooth-enabled insulin pen to titrate patient-specific dosing parameters previously learned during passive observation periods to optimal levels, thereby providing personalized dose guidance that evolves with the user with minimal input from the user.

[0209] As described elsewhere herein, to provide initial dose guidance, the DGA can undergo a "learning phase" in which user-specific dosing parameters, such as a fixed prandial dose, target glucose, and insulin correction factor, can be ascertained. A method for the learning phase is described in connection with FIG. 7 herein. These parameter values ​​serve as initial estimates and can be further titrated to more optimal, patient-specific values ​​as the system is used. This initial learning followed by a continuous parameter titration approach can accommodate not only disease progression but also external changes, both lifestyle and pharmacotherapy, that the user may undergo. With the increasingly popular class of insulin-based diabetes treatments, which can enhance both endogenous insulin production and insulin sensitivity, a user's correction factor cannot remain static for the lifetime of the system. A method for titrating a user's insulin correction factor (also referred to as insulin sensitivity) is described herein.

[0210] As used in many embodiments of DGA, a user's correction factor need not be a single value but may be a distinct set of values ​​specific to a given meal type (e.g., breakfast, lunch, dinner). The correction factor may have units of mg / dL glucose per insulin unit. A high correction factor indicates a user's high sensitivity to insulin, as a small dose will result in a larger glucose drop. Conversely, a low correction factor suggests a user's low sensitivity to insulin. An effective prandial insulin dose should return postprandial blood glucose levels to a safe range within its pharmacodynamic window. For example, given the dosing regimen of a bolus calculator such as that described in connection with FIGS. 8A-8H, an accurate correction factor should account for both meal carbohydrates and pre-meal glucose rises, contributing to an insulin dose that returns glucose levels to a safe and stable target value. Thus, if GPA plots all insulin doses for a particular meal as an ordinate against a particular metric that quantifies glucose readings for a time period associated with the meal (e.g., a metric that quantifies glucose data or excursions associated with the meal, such as a set amount of time post-dose (e.g., 4 hours), or percentiles of glucose values ​​during the mealtime period (e.g., 5th percentile, 10th percentile)), then theoretically, the best fit for these data is a straight line with a slope of zero and a y-intercept equal to the user's target glucose. In practice, all prandial insulin doses will return the user's blood glucose to the target value or range.

[0211] A zero slope for the relationship between insulin and postprandial glucose represents an ideal scenario for dosing effectiveness. The DGA can assume that it has previously learned the user's fixed dose, for example, via user input or transmission from the MDD 152. For doses greater than the fixed dose, the system can observe the relationship between insulin dose and postprandial glucose from the SCD 102 or other sources. The DGA can assess the effectiveness of the dose using one or more of several methods, including, but not limited to:

[0212] Steps for determining the centroid (median / mean) of postprandial glucose values This method uses the difference between the centroid of postprandial glucose values ​​for an insulin dose without any correction and the centroid of postprandial glucose values ​​for an insulin dose with correction. The ideal scenario is zero or no significant difference between the two centroids. A positive difference above a predetermined threshold is considered to indicate a suboptimal correction factor. Similarly, a negative difference below a predetermined threshold is considered to indicate a correction factor that is higher than optimal.

[0213] Applying a linear fit to the data If the fit meets a certain goodness of fit, DGA can perform statistical tests to determine whether there is a statistically significant difference between the slope of the best fit and the ideal case where the slope is zero. A slope less than zero indicates that too much insulin was administered. This corresponds to the correction factor being too low and needing to be increased. A slope greater than zero indicates that postprandial glucose is elevated and the insulin dose can be increased. This corresponds to the correction factor being too high and needing to be decreased. The amount by which these values ​​are titrated can vary depending on the slope of the fit. A steeper slope requires more aggressive titration than one closer to zero. This titration method is independent of and can be performed in parallel with fixed-dose titration.

[0214] Steps for determining the area under the postprandial glucose / insulin curve An ideal dosing scenario would have a constant area, representing a return to a constant glucose level after every dose. Therefore, DGA can evaluate any change in area as an indication of an error in the correction factor. The integral method can be used to determine whether the prandial component corresponding to the fixed meal dose is effective or whether further titration is required. An increase in the integral level compared to the ideal scenario indicates that the fixed dose can be increased to bring the postprandial value closer to the target level.

[0215] During the learning phase, which can last up to 14 days depending on the SCD characteristics or other factors, the DGA can learn the user's dosing strategy or other user-specific parameters, including, but not limited to, dosage, correction factor, and insulin action time (IAT) / insulin on board (IOB). It may be desirable to titrate the correction factor for one or more of the following reasons: (1) if the correction factor is insufficient; or (2) if it changes due to a therapeutic intervention or environmental change. For bolus administration (no carbohydrate counting) involving correction only, the DGA can analyze postprandial glucose values ​​and incrementally titrate the correction factor. For example, if the postprandial glucose peak is greater than 180 mg / dL or the difference between the postprandial peak and the preprandial value is greater than a threshold and there is no risk of postprandial hypoglycemia, the correction factor can be incrementally titrated with each measurement cycle. For example, the correction factor can be titrated in preset increments, such as 1 unit, or alternatively, in 2 unit increments. In the case of a separate correction dose administered to treat hyperglycemia when glucose levels are high (e.g., without a meal), the difference between the value at the time of administration and the glucose value at a predetermined time (e.g., 4 hours after administration) can be used to titrate the correction factor.

[0216] Mealtime insulin dosing consists of two different amounts: a) a portion intended to cover the meal consumed, and b) a portion intended to address pre-meal glucose levels above the target range. The meal-related portion is usually fixed (e.g., the portion consumed for breakfast is always a specific amount, which can be titrated over time) or varies according to the amount of carbohydrates the patient expects to consume. In the following examples, a fixed meal dose is assumed. However, the following explanation can also apply to a variable portion, where the actual portion is determined by a fixed carbohydrate-to-insulin ratio, which itself can be titrated. As used herein, the portion intended to address pre-meal glucose refers to a correction dose or correction portion determined by a pre-meal correction factor.

[0217] The pre-meal correction factor can be used to control the MDD 152. Reducing or down-titrating the correction factor is comparable to increasing the correction dose, as described herein. For example, FIG. 10A illustrates a DGA method 500 for providing a pre-meal correction factor in response to analyte data for use in controlling the MDD 152. At 502, the method 500 can include determining, by at least one processor, an analyte pattern type for at least one TOD period by executing a GPA algorithm that receives as input time-correlated analyte data originating from a patient-worn sensor control device over an analysis period. At 504, the method 500 can include determining, by the at least one processor executing the algorithm, a pre-meal correction factor based on the analyte pattern type and the patient's defined dosing strategy for the analysis period. At 506, the method 500 can include storing, by the at least one processor, an indicator of the pre-meal correction factor in computer memory for output to the user or at least one of the dosing devices. In contrast to the linear fit embodiment described above, the embodiment described here is a unified approach in which fixed dose and correction factor titration can be performed simultaneously via one technique or software function (e.g., a single logic tree).

[0218] In a related embodiment, FIG. 8B and the related description above disclose a GPA method 410 for classifying a patient's glucose pattern over various TOD periods as high, low, or high / low. FIG. 8A and the related description disclose a method for a DGA to use output from the GPA 410 to provide MDI dose guidance recommendations. FIGS. 8D-8H and the related description show an example algorithm for a DGA to provide specific MDI dose guidance recommendations based on the glucose pattern type output by the GPA 410 and the patient's defined dosing strategy over the analysis period. Recommendations may be provided for automatic or semi-automatic control of the MDD 152, or for operation of a user interface to guide manual dosing control.

[0219] FIG. 10B illustrates a method 510 for evaluating recommendations for titrating meal doses and pre-meal corrections to increase high analyte levels, with or without pre-meal corrections, for example, when GPA indicates a low pattern. In a related embodiment, FIG. 8G illustrates an example algorithm 474 for formulating recommendations for post-meal corrections. When implemented by DGA, the algorithm 474 prompts consideration of whether to perform a pre-meal correction if GPA indicates a low pattern when considering analyte data excluding data from TOD periods in which a bolus was missed. Other low glucose pattern conditions 512 during TOD periods may also be appropriate to trigger execution of the method 510. If both the meal dose and pre-meal correction are higher than optimal, the method 510 can output a pre-meal correction factor that can only reduce the meal dose. If a low pattern is present, this may be due to a high fixed dose, a high pre-meal correction, or both. In one embodiment, the fixed dose and pre-meal correction can be titrated sequentially, with the fixed dose being titrated first because it is the basis for medication. Once the fixed dose has been appropriately titrated, the correction dose can be titrated if the low pattern is still observed. In another embodiment, the fixed dose and the pre-prandial correction dose can be titrated in parallel.

[0220] If a low pattern is detected during the TOD period, the DGA can determine at 514 whether the meal dosing of the defined dosing strategy includes a pre-meal correction. If so, at 534, the DGA can exclude the missed bolus days from the original data set or pre-meal corrections, for example, by including a portion of the analyte data only for days when a meal administration occurred and no bolus was missed. At 536, the DGA can test whether sufficient data remains after excluding data from days when no bolus or pre-meal corrections were missed to achieve a minimum confidence level. If sufficient data exists, at 410, the DGA can repeat GPA 410 on the input analyte data set excluding data from days when no bolus or pre-meal corrections were missed.

[0221] At 538, the DGA can determine whether the subsequent glucose pattern is still low, and if so, reduce only the meal portion of the insulin dose at 540. If the pattern is not low, at 542 the DGA can reduce the pre-meal correction factor, which, if implemented by the user or MDD, will reduce the corresponding pre-meal correction dose.

[0222] If the DGA determines at 536 that there is insufficient data to determine a pattern, then at 516, the DGA can include data from days when meal administrations and pre-meal corrections were made. The DGA can then retest the data set for sufficiency at 518. If the data is sufficient, then at 410, the DGA can repeat the GPA on the expanded data set. If the resulting glucose pattern is not low, then at 540, the DGA can reduce the recommended meal portion of each dose for the associated TOD period without reducing the pre-meal correction factor. If the DGA determines at 520 that the glucose pattern is low, then at 526, the DGA can include data on only the meal doses. Then, at 530, the DGA can retest the data set for sufficiency. If the data set is not sufficient, then at 528, the DGA can reduce both the meal dose and the correction dose. If the data set is sufficient, then the DGA can perform the GPA at 410 and determine if the glucose pattern is low at 532. If the glucose pattern is low, the DGA can reduce the recommended meal portion of each dose for the associated TOD period without reducing the pre-meal correction factor at 540. If the glucose pattern is not low, the DGA can reduce the pre-meal correction factor at 542, which, when implemented by the user or the MDD, can reduce the corresponding pre-meal correction dose.

[0223] FIG. 10C illustrates a method 550 for evaluating recommendations for titrating meal doses and pre-meal corrections to reduce high analyte levels, with or without pre-meal corrections, for example, when GPA 410 indicates a high pattern. In a related embodiment, FIG. 8H illustrates an example algorithm 480 for formulating recommendations for post-meal corrections. When implemented by DGA, algorithm 480 prompts consideration of whether to perform a pre-meal correction if GPA indicates a high pattern. Other high glucose pattern conditions 552 during the TOD period may also be appropriate to trigger execution of method 550. If both the meal dose and pre-meal correction are lower than optimal, method 550 can first increase the meal dose. High glucose patterns can be addressed sequentially (patterns that are not high in the uncorrected data) so that the correction dose can be up-titrated after the meal dose is titrated to avoid hypoglycemia from both increasing the fixed dose and decreasing the correction factor (similar to increasing the correction dose at a given blood glucose level). High patterns can be addressed first by increasing the fixed dose. If the high pattern still persists and only occurs when the correction dose is included, the correction factor can be titrated.

[0224] If a high pattern is detected for a particular TOD, the DGA can determine at 554 whether the analyte data includes a meal dose with a pre-meal correction. If a pre-meal correction is not included, the DGA can exclude data for days on which a pre-meal correction was taken and days on which a bolus was missed from the original data set at 564 and repeat the glucose pattern analysis at 410. At 566, the DGA can determine whether the subsequent glucose pattern is still high. If not, at 572, the DGA can increase the pre-meal correction factor, which will cause a decrease in the pre-meal correction dose by outputting it to the UID or MDD. Otherwise, if the glucose pattern is high, at 562, the DGA can increase the recommended meal portion of each dose for the associated TOD period. If a pre-meal correction is determined to be included at 554, the DGA can include data for days on which a meal dose and a pre-meal correction was taken and exclude data for days on which a bolus was missed from the original data set at 556 and repeat the glucose pattern analysis at 410. At 558, the DGA can determine whether the subsequent glucose pattern is still high. If the pattern is not high, the DGA can increase the meal portion recommendation for each dose for the associated TOD period at 562. If the pattern is high at 558, the DGA can exclude data for days on which pre-meal corrections were made and exclude data for days on which a bolus was missed at 564, and repeat the pattern analysis at 410 and the subsequent steps described above.

[0225] 10B-10C illustrate aspects of various algorithms for determining pre-meal correction factors, it should be understood that these are exemplary and that various other algorithms may also be suitable.

[0226] Dosage guidance method for insulin doses according to the timing of administration (at the start of a meal or before and after the start of a meal) Next, an exemplary embodiment of a method for determining dose guidance based on administration timing will be described. To maximize glycemic control, it is desirable to synchronize insulin administration with meals so that the peak of circulating insulin coincides with the postprandial glucose rise. However, patients sometimes forget to take prandial insulin before eating. This can result in a mismatch between the glucose rise associated with a meal and insulin action, leading to confusion about the appropriate dose. Taking a meal dose after the start of a meal rather than just before the start of the meal can lead to hypoglycemia because active insulin may still be circulating even after the postprandial glucose rise has subsided. Furthermore, patients correcting glucose trends may be inclined to increase their dose due to the upward trend in glucose at the start of a meal. This behavior can also lead to hypoglycemia. To minimize timing-related hypoglycemic episodes, the DGA can be configured to determine a dose delay and take the dose delay into account when providing dose guidance. A positive delay indicates that the insulin dose is administered after the start of the meal. A zero delay indicates that the insulin dose is administered approximately simultaneously with the start of the meal. A negative time delay indicates that the insulin dose is administered before the start of the meal.

[0227] In many embodiments, the DGA can be configured to include a method for detecting missed meal doses in real time for the purpose of determining whether the meal dose guidance calculation should be an "on-time" calculation (i.e., zero or negative time delay) or a "late dose" calculation (i.e., positive time delay). In some embodiments, the DGA can also be configured to determine and output an estimate of the start time of the meal that can be used in delayed dose calculations. In some embodiments, if a prandial insulin dose is not detected +X or -Y minutes after the estimated start of the meal, the DGA can be configured to notify the patient that a dose may have been missed.

[0228] Algorithm for detecting missed meal administration in real time In many embodiments, the DGA can be configured to detect missed meals using a real-time meal detection algorithm. Systems and processes for detecting missed meals in real time and subsequently alerting the patient are described herein. The process for detecting missed meals can be performed periodically (e.g., whenever new glucose data becomes available to the system). Alternatively, the process can be performed whenever it is appropriate to provide a "missed dose" alert to the patient or whenever an alert is enabled.

[0229] In an exemplary embodiment, real-time meal detection can be performed by a feature extraction module and a meal detection module. The feature extraction module can receive CGM data points one at a time as the data points become available. When the feature extraction module detects that glucose levels are increasing, the feature extraction module can extract multiple features and pass the multiple features to the meal detection module for meal detection.

[0230] In one embodiment, the feature extraction module is configured to perform data smoothing each time a new glucose data point is received by fitting the data within a time window using a quadratic function and working backward from the current data point. The time window may be approximately 60 minutes. The feature extraction module may be configured to store the fitted value at the center of the time window as the current smoothed data. The feature extraction module may also be configured to store the coefficients of the linear and quadratic terms of the fitted value at the center of the time window as the current glucose rate of change and acceleration values, respectively. In addition to being configured to store the fitted value at the center point, the feature extraction module may also be configured to store the fitted value at the current point for feature extraction. The feature extraction module may be configured to compare the current smoothed glucose value with a previous smoothed glucose value (e.g., the smoothed glucose value immediately preceding the current smoothed glucose value) to determine whether the smoothed glucose data is rising or falling. The feature extraction module may be configured to extract a plurality of features and then pass the plurality of features to the meal detection module after the feature extraction module determines that the comparison between the current smoothed glucose value and the previous smoothed glucose value is rising.

[0231] The feature extraction module can be configured to extract multiple features from two segments of the smoothed data, which can be a current ascending segment and a previous descending segment. The multiple features extracted from the current ascending segment can include, but are not limited to, 1) maximum acceleration, 2) time of the maximum acceleration point, 3) glucose value at the maximum acceleration point, 4) height calculated by the difference in glucose values ​​between the current time point (fitted value) and the maximum acceleration point (base point), 5) duration of the current segment calculated by the elapsed time from the base point to the current point, 6) average rate of ascent of the current segment calculated by dividing height by duration, 7) maximum acceleration increase (the increase in acceleration at a given time point is obtained by subtracting the acceleration at that time point from the acceleration at the previous time point), and 8) increment of area under the curve (average glucose value minus the glucose value at the base point multiplied by the duration of the segment). The features extracted from the previous descent segment may include, but are not limited to, 1) duration, 2) height, 3) average descent rate (height / duration), 4) maximum descent rate (maximum absolute value of rate of change), and 5) maximum deceleration (maximum absolute value of acceleration). The feature extraction module may be configured to pass the extracted features to the meal dosage module.

[0232] The meal detection module can be configured to receive a feature vector as input and output a binary detection result indicating whether the current upsegment is a meal-responsive glucose excursion. The meal detection module can also be configured to output a probability value along with the binary detection result. In one embodiment, the pre-trained machine learning model in the meal detection module can be implemented using a random forest classifier by scikit learn (https: / / scikit-learn.org / stable / modules / generated / sklearn.ensemble.RandomForestClassifier.html). The meal detection module can be configured to detect the onset of a meal based on tree-building rules and feature thresholds for each feature in each tree, which can be optimized during the training process. In one embodiment, the pre-trained model can also be built based on alternative classification algorithms, including gradient boosting, ADA boosting, artificial neural networks, linear discriminant analysis, and extra-trees.

[0233] The meal detection model can also be configured to estimate the start time of a meal if a meal is detected. In one embodiment, the start time of the meal can be estimated as the point in time of the maximum increase in glucose value acceleration, working backward from the point of detection within a time window size of approximately 1.25 hours. For example, if a missed meal is detected by the algorithm at 1:15 PM, the model can work backward to approximately 12 PM to determine the start time of the meal. The acceleration of glucose values ​​at each point can be calculated using a quadratic function, i.e., y = ax 2 +bx+c can be calculated by fitting five data points centered on the data point of interest. The fitted parameter "a" is the acceleration at the point of interest. The acceleration increase of the glucose value at time point k can be defined as a(k+1)-a(k).

[0234] The meal detection model may also be configured to output a missed dose notification to the user on UID 200 if no mealtime insulin administration is detected within a period around the estimated start of the meal. In one embodiment, if the estimated start of the meal is less than two hours, the notification may also indicate that the patient can still obtain dose guidance for the meal and that the dose will be delayed.

[0235] Details of other types of meal detection methods and algorithms are described in U.S. Patent Publication No. 2017 / 0185748 and PCT Application No. PCT / US2020 / 12134, the entire contents of which are incorporated herein by reference.

[0236] Dosage guidance to be administered at the start of a meal If the dosing time delay is negative or zero (i.e., the dose guidance is for administration at or before the start of a meal), the DGA can be configured to consider other factors that may modify hypoglycemic and hyperglycemic risk, such as the time span during which the bolus insulin can take effect before the presence of a meal. For example, the DGA can be configured to include risk factors related to circadian rhythm.

[0237] In one exemplary embodiment, as described in the flow diagram of Figure 11, in exemplary method 600, the DGA can provide dose guidance for administration to a subject at the start of a meal, for example, when the time delay is negative or zero. Starting at step 602, the DGA can determine a first dose guidance for a meal to be administered during a time period at the start of the meal. The first dose guidance can be a fixed meal dose (with or without compensation) or can be based on the carbohydrate content of the meal.

[0238] In step 604, the DGA can then determine whether a risk of hypoglycemia exists based on at least the first dose guidance and the time of day when the dose is administered. The DGA can be configured to determine the risk of hypoglycemia by referencing a risk map. In one embodiment, historical user data (e.g., glucose level and insulin dosing data), population data, or a combination of both historical user data and population data can be used to develop a risk map associated with time of day, day of week, and / or other available patterns. In one embodiment, the DGA can be configured to use the risk map to identify the edge of the distribution that is closest to the highest risk, rather than identifying typical behavior.

[0239] In step 606, the DGA can output second dose guidance that differs from the first dose guidance. The second dose guidance is output to UID 200 and can include a lower drug amount than the first dose guidance. The second dose guidance can also be associated with a lower risk of hypoglycemia than the first dose guidance. For example, if a first time period has a higher near-term hypoglycemia risk than another (second) time period and the user requests dose guidance during the first time period at or before the start of a planned meal, the DGA can be configured to output dose guidance that is lower than the nominal proposed meal dose guidance (e.g., a fixed meal dose (with or without correction) or a meal dose based on carbohydrate content).

[0240] If the dose is administered after the start of a meal (i.e., if the dose time delay is positive), the DGA can be configured to determine the dose time delay and output dose guidance that takes into account the dose time delay and any associated risks.

[0241] In one exemplary embodiment, as described in the flow diagram of FIG. 12A , in exemplary method 607, beginning at step 608, the DGA can determine a first dose guidance for a meal in response to a query from the subject, the first dose guidance being determined for administration to the subject at the start of the meal. In one embodiment, the DGA can be configured to assume that the initial optimal prandial dosing occurs simultaneously with the start of the meal. Thus, in one embodiment, when a user opens the DGA and requests dose guidance, the algorithm can first calculate the optimal insulin dose if administered at the start of the meal. The optimal insulin dose for a meal can be determined in a number of ways, including, but not limited to, a fixed meal dose (with or without correction) or a meal dose based on carbohydrate content.

[0242] In step 610, the DGA can determine whether there is a time delay between the start of a meal and the subject's query (i.e., the start of dose guidance). In one embodiment, the DGA can be configured to identify a dosing time delay in multiple ways. In one embodiment, the DGA can be configured to determine the estimated start of a meal. In one embodiment, the DGA can be configured to identify a dosing time delay using the meal detection algorithm described above, which can detect the estimated start of a meal, and the DGA can be configured to record the time between the start and the user-prompted dose guidance. In another embodiment, the DGA can be configured to identify a dosing time delay between the start of a meal and the start of a bolus insulin dose, which can be recorded by the user. In this embodiment, the DGA can be configured to prompt the user for input regarding the dosing time delay. For example, the DGA can be configured to ask the user whether the requested dose guidance is for a meal or a correction dose to correct high glucose. If the requested dose guidance is for a meal, the DGA can be configured to ask whether the meal has already started, and if so, for how long. This input can then be used as the meal-to-medication time delay for dose guidance.

[0243] In step 612, the DGA can determine whether a risk of hypoglycemia exists based on at least the first dose guidance and the time delay in response to determining the time delay. In one embodiment, the DGA can determine whether the patient is in a high hypoglycemia risk zone or at risk for hypoglycemia compared to the rest of the user population with the first dose guidance and the time delay.

[0244] In one embodiment, the DGA can determine the risk of hypoglycemia by reference to a multidimensional surface map. In one embodiment, the map can be population-based. For example, the map can include observations across all DGA users until sufficient data is collected to personalize recommendations based on the specific user requesting dose guidance. In one embodiment, the DGA can be configured to receive and store multiple types of data related to a DGA user's prandial dose. The multiple types of data related to prandial doses include, but are not limited to, a dose time delay, a proposed or recommended dose guidance, an administered dose, and a glucose time series for a predetermined time period after the dose is administered. The glucose time series can have a fixed time interval between glucose value samples (e.g., from a CGM), a varying time interval, or a combination thereof. Postprandial glucose data in the glucose time series can be used to calculate a hypoglycemic index specific to the meal / dosage episode. This hypoglycemic index can be, but is not limited to, time below approximately 70 mg / dL, time below approximately 54 mg / dL, or a calculated risk factor such as a hypoglycemic index (LBG). The hypoglycemic index can also be normalized to the glucose concentration value at the start of the meal. Dose guidance and administered doses can be treated as variables individually, as well as their difference. A multidimensional surface map can then be derived, in which the resulting hypoglycemic index is a function of various variables. Taking into account different meal sizes and the influence of circadian rhythm, DGA can create multiple multidimensional surface maps, including multidimensional surface maps for breakfast, lunch, and dinner. Once the maps are created, DGA can also determine a cutoff value for acceptable hypoglycemic risk. One of the key variables used in this system may be the meal-to-medication time delay.

[0245] The DGA can be configured to output dose guidance to the UID 200 based on whether a risk of hypoglycemia is determined. If the DGA determines that the first dose guidance and time delay do not expose the subject to a risk of hypoglycemia, then in step 614, the DGA can output the first dose guidance. In step 616, the DGA can output second dose guidance in response to determining that a risk of hypoglycemia exists, the second dose guidance being associated with a lower risk of hypoglycemia than the first dose guidance. The second dose guidance can be output to the UID 200. In one embodiment, if the DGA determines that the first dose guidance at the time delay poses a risk of hypoglycemia, then the DGA can search for a dose change along the delta dose / hypoglycemia risk isopotential of the multidimensional surface map that reduces the risk below a predetermined cutoff value. This dose change can be applied to the first dose guidance to provide updated dose guidance to minimize hypoglycemia. The system can continue to collect information to refine and / or update the multidimensional surface map as a function of multiple variables (e.g., dosing time delay, proposed or recommended dose guidance, administered dose, and glucose time series for a predetermined time period after administration of the dose).

[0246] In another embodiment, the DGA can be configured to determine delayed dose guidance that does not take into account endogenous insulin production. The problem with using traditional meal dose calculations to determine dose guidance after a meal has started is that the user's glucose typically increases, even if the user has administered insulin. If the user administers a meal dose after the start of a meal and uses current glucose values ​​to determine the meal dose, the user may inject too much insulin, ultimately resulting in hypoglycemia. The DGA can be configured to determine dose guidance for administration after the start of a meal that mitigates the problem of elevated glucose levels in the user by considering the user's glucose value at the estimated start time of the meal to determine the correction portion of the dose guidance.

[0247] In one exemplary embodiment, as shown in the flow diagram of FIG. 12B, in an exemplary method 617 beginning at step 618, the DGA may receive a dose guidance query for a meal having a start time. In step 620, the DGA may determine whether the dose guidance query was received after the start of the meal. In one embodiment, in response to the dose guidance query from the user, the DGA may determine whether a dose is late using the delayed dose detection algorithm described above. In one embodiment, the delayed dose detection algorithm may determine and output an estimated start time of the meal, which the DGA may compare to the time the user query was received.

[0248] In step 622, the DGA may determine the user's glucose level associated with the start time of the meal. In one embodiment, the DGA may determine the glucose value associated with the start time of the meal (e.g., the glucose value closest in time to the start time of the meal).

[0249] In step 624, the DGA can output delayed dose guidance, where the delayed dose guidance includes meal dose guidance and correction dose guidance. In one embodiment, the correction dose guidance can be based on the determined glucose level at the start time of the meal. The correction dose guidance can be determined using a bolus calculator to correct for high glucose and can include a glucose correction portion. As discussed above with respect to equation (1), the glucose correction portion of the correction dose guidance can be determined based on the following equation, where (BG(t)) is the current glucose value and BG target is the target glucose.

[0250]

number

[0251] In one embodiment, for purposes of determining the glucose correction portion for the delayed dose guidance, the current glucose value is the glucose value associated with the estimated start time of the meal. In one embodiment, the correction factor may be the user's insulin sensitivity factor. In one embodiment, the delayed dose guidance is the meal dose guidance plus the correction dose guidance. The meal dose guidance can be determined in a number of ways. In one embodiment, the meal dose guidance can be determined based on the estimated carbohydrate content of the meal. In another embodiment, the meal dose guidance can be based on a "fixed" or "fixed + correction" dosing strategy, where the user takes a predetermined amount of insulin for the meal dose regardless of the carbohydrate content of the meal.

[0252] In another embodiment, the DGA can be configured to determine delayed dosing that takes into account endogenous insulin production. The problem with using traditional meal dose calculations when dosing after a meal has begun is that type 2 diabetes patients often still produce their own (endogenous) insulin. Therefore, if a user takes a dose late after a meal has begun, the user may have additional "insulin" produced by the user's own pancreas, which could ultimately lead to hypoglycemia if the dose calculation does not take this into account. The DGA can be configured to determine dose guidance that takes into account the production of additional endogenous insulin.

[0253] In one exemplary embodiment, as set forth in the flow diagram of Figure 12C, in exemplary method 630 beginning at step 632, the DGA can determine a first dose guidance for a meal in response to a query from the subject. In one embodiment, the first dose guidance is determined for administration to the subject at the start of the meal. In other embodiments, the first dose guidance can be a fixed meal dose (with or without an adjustment) or a meal dose determined based at least on the carbohydrate content of the meal.

[0254] In step 634, the DGA can calculate a time delay between the start of the meal and the query from the subject. In one embodiment, the time delay can be calculated using the delayed dose detection algorithm described above. In one embodiment, the delayed dose detection algorithm can estimate the start time of the detected meal and calculate the time delay by comparing the estimated start time with the time the query for dose guidance from the user is received by the DGA. The estimated start time can be determined by the DGA as described with respect to other embodiments.

[0255] In step 636, the DGA can determine a factor corresponding to the estimated amount of endogenous insulin and the amount of time delay in response to determining that the time delay is greater than zero (>0). In one embodiment, the factor can be a number of minutes. In one embodiment, the factor can be determined by the amount of time between the query and the estimated start time of the meal, for example, from a universal lookup table where the fractional value decreases as the time delay from the start of the meal increases.

[0256] In step 638, the DGA can output a second dose guidance associated with the first dose guidance and the factor. The second dose guidance can be output to UID 200. The second dose guidance can be calculated by multiplying the first dose guidance by the factor determined in step 636, which can be a fraction. According to some embodiments, for example, 1 minus the fraction can account for endogenous insulin production up to the time of injection. In one embodiment, this fraction can be based on a simulated type 2 metabolic response to a meal. This value can be summed with a fraction of the delayed dose to deliver a dose that accounts for endogenous production, which may not be known for each subject, and deliver a corrective dose that keeps the user's post-prandial blood glucose levels in a safe range.

[0257] Dosage guidance for included meals (e.g., dessert) In some situations, patients may administer a mealtime insulin dose sufficient to cover a planned meal, but then later eat an additional meal in the form of a "refill" or dessert. A common strategy is to re-administer a new dose regardless of glucose levels and trends, and this is common practice. However, there may be situations in which this is not the correct course of action. For example, the patient may not have originally consumed enough insulin to cover the first dose, resulting in a hypoglycemic episode if the new dose is too large. Conversely, if the patient takes a conservative approach and does not administer extra insulin to cover the extra meal, a hyperglycemic episode may occur. In one embodiment, the DGA can provide dosage guidance for the additional meal by (1) ensuring the user continues to extend the original meal with more food, (2) notifying the user of the risk of hypoglycemia before the extra dose is administered, and (3) monitoring such risk after the meal.

[0258] The DGA can be configured to output an initial query to the user at UID 200 to confirm that the meal has been extended with extra food. If the user wishes to receive dose guidance, it is assumed that the user first opens the dose guidance app. In one exemplary embodiment, as described in the flow diagram set forth in FIG. 13 , in exemplary method 660 beginning at step 662, the DGA can be configured to determine whether the user's query for dose guidance occurred within a first episode of time, the first episode including a meal having a start time. The start of the time period can be the start time of the meal determined from a real-time meal detection algorithm, and the length of the time period can also be determined by the real-time meal detection algorithm. The start time of the meal can be determined in several ways, including, but not limited to, the last time a meal was detected (e.g., by the real-time meal detection algorithm) or the last time dose guidance and / or insulin dosing was provided in conjunction with meal detection.

[0259] If the user queries the DGA for dose guidance within the time period, as seen in step 664, the DGA may request input from the user to confirm whether the meal was extended with additional food. The DGA may output a prompt or other indication to UID 200 requesting user feedback. In one embodiment, the DGA may also request input from the user to confirm whether the purpose of the dose guidance was to correct high glucose. In another embodiment, the DGA may prompt the user for an explanation regarding the most recent dose guidance request and, optionally, provide selectable options as a response. Options may include, for example, an option that the additional request was made in response to an extended meal and an option that the additional request was made to correct high glucose not associated with additional food intake.

[0260] As seen in step 666, the DGA can determine the risk of hypoglycemia from the start time of the meal. In one embodiment, to avoid hypoglycemic episodes due to insulin dose stacking, the DGA can be configured to determine the risk of hypoglycemia by determining the point at which the user is in the current glycemic excursion before extending the meal with additional food. The DGA can be further configured to create a forward prediction of the glucose level from the determined point in the current excursion to determine whether the user's glucose level is still rising or falling. As a safety measure to avoid insulin stacking, the DGA can be configured to not provide dose guidance until the glucose level reaches a post-prandial maximum. For example, if dose guidance for the added meal is requested while the user's glucose is still rising, the DGA can output a notification that guidance cannot be provided at this time for safety reasons. In one embodiment, if a risk of hypoglycemia is determined, in step 668, the DGA can be configured to not administer any doses or to notify the user to pay close attention to their medication. In another embodiment, if there is no risk of hypoglycemia, the DGA can be configured to notify the user to dose according to the HCP recommendations for the extra food in step 670. Additionally, in one embodiment, if the DGA detects that an extra dose has been delivered, the DGA can be configured to advise the user to check their glucose levels after at least about two hours to ensure there is no hypoglycemia.

[0261] Dosage guidance method for corrective doses (taking the dose) In certain circumstances, external factors may affect the effectiveness of insulin dosing, causing the dose to have a greater glucose-lowering effect than expected. In these cases, the user may conservatively estimate the insulin dose to avoid hypoglycemia. Also, if the user is unsure of how much insulin will be required for a particular meal type or amount, the user may administer the dose conservatively. For example, a certain meal type may be higher in fat and / or protein than the meals the user typically consumes. To enable conservative dosing, the DGA can be configured to provide the user with dose guidance after a meal, accompanied by an initial meal insulin dose. In one embodiment, the DGA can be configured to perform at least four functions following initial meal dose guidance and administration. The DGA can be configured to confirm that the user has administered a lower dose than indicated by the initial meal dose guidance. The DGA can then be configured to determine the risk of hypoglycemia and notify the user of the risk before an additional dose is suggested. The DGA can be further configured to provide additional dose guidance and monitor subsequent hypoglycemia risk for a period following the introduction of the additional dose guidance.

[0262] In one exemplary embodiment, at or near the beginning of a meal, a user queries the DGA for a recommended dosage. As depicted in FIG. 14 , in exemplary method 700, beginning at step 702, the DGA can output a first dose guidance in response to the first user query. The first dose guidance can be output to UID 200. In one embodiment, the first dose guidance can be calculated to be administered at the beginning of the meal. The first dose guidance can also be a fixed dose (with or without offset) meal guidance. The first dose guidance can also be determined based on the carbohydrate content of the meal.

[0263] However, the user may choose whether to follow the recommended dose guidance when administering insulin. If the administered dose differs from the recommended dose, the system records the difference. In step 704, the DGA may determine whether the administered first dose differs from the first dose guidance. The administered first dose may be lower or higher than the first dose guidance. In one embodiment, the administered first dose is lower than the first dose guidance. The DGA may be configured to record if the administered dose is lower than the recommended dose guidance.

[0264] In step 708, the DGA can determine whether a second user query for second dose guidance is received within a period of time from administration of the first dose, which can be determined by a real-time meal detection algorithm or can be a predetermined time from administration of the first dose.

[0265] In step 710, the DGA may request input from the user to determine whether the second user query is to adjust high glucose levels after meal completion if it is determined that a second user query was received within the time period. The DGA may output a prompt or other notification to the UID 200 requesting user feedback. For example, if the DGA subsequently queries for a further dose recommendation following conservative dosing (e.g., a dose less than the recommended dose was administered) within a certain amount of time from the first meal dose, the DGA may use both the query time and the dosing discrepancy record to prompt the user as to the reason for the second dose guidance. In one embodiment, the DGA may be configured to request input to determine whether the later-requested dose is to cover additional mealtime foods or to account for post-prandial glucose hypertension. If the user requests second dose guidance to cover additional mealtime foods, the DGA may follow the flow described in the "Dose Guidance for Added Meals" section elsewhere herein.

[0266] If the user requests second dose guidance to correct a postprandial high glucose, as seen in step 712, the DGA can determine the risk of hypoglycemia by at least determining whether the user's blood glucose is elevated. In one embodiment, to avoid hypoglycemic episodes due to insulin dose stacking, the DGA can be configured to determine the risk of hypoglycemia by determining the point at which the user is in their current blood glucose excursion. The DGA can be further configured to create a forward prediction of the glucose level from the determined point in the current excursion to determine whether the user's glucose level is still rising or falling. In one embodiment, as a safety measure to avoid insulin stacking, the DGA can be configured to not provide dose guidance until the glucose level reaches a postprandial maximum. If a dose recommendation for a correction is requested while the user's glucose is still elevated, the DGA can be configured to provide a notification that guidance cannot be provided at this time for safety reasons. In one embodiment, the DGA can be configured to provide additional reasons for not providing dose guidance. For example, the DGA can be configured to provide an updated estimate of insulin onboard, when available, determined from a combination of population-based aggregate parameters and user-specific parameters.

[0267] In one embodiment, if the DGA determines that the user's blood glucose is declining, the DGA can be configured to calculate the risk of future hypoglycemia in the absence of a corrective "touch-up" dose. In one embodiment, the DGA can be configured to calculate the risk of hypoglycemia by calculating a forward prediction of current glucose levels to determine the likelihood of a hypoglycemic episode. In another embodiment, the DGA can be configured to analyze past events to observe how frequently a given dose induces postprandial hypoglycemia for a given preprandial glucose range. In one embodiment, if the DGA determines that the current hypoglycemia risk exceeds a predefined threshold, the DGA can be configured to output a recommendation to the UID 200 in step 716 not to take any more insulin at this time.

[0268] If the DGA determines that the user is not currently at risk of hypoglycemia, then in step 714, the DGA can be configured to calculate and output dose guidance as if the subsequent dose were a post-prandial correction. Thus, in one embodiment, the subsequent dose can be determined without considering new carbohydrate intake because the meal has already ended. As discussed with respect to equation (2), the DGA can be configured to calculate the second dose (correction dose guidance) as a function of the glucose correction portion minus the residual insulin on-board (IOB) of the initial prandial dose, i.e.,

[0269]

number

[0270] wherein: TIFF0007662611000009.tif18114

[0271] is the glucose correction portion. In one embodiment, the DGA calculates the glucose correction portion of the correction dose guidance as the ratio of current glucose (BG(t)) to target glucose (BG target) divided by a correction factor. The correction factor can be the user's insulin sensitivity factor (ISF), which is a metric used to measure how much one unit of insulin lowers fasting blood glucose levels. The ISF can be personalized for each user during the algorithm training period. In one embodiment, for the initial estimation, population-based statistics related to insulin pharmacokinetics (PK) can be used to estimate IOB. Rapid-acting insulin analogs typically achieve peak plasma concentrations in approximately 45 minutes followed by an exponential decay profile, although this time may be shorter if the user is taking an ultrarapid-acting insulin analog. This time window may correspond to a rest period during which rising glucose levels prevent DGA from providing guidance. IOB can be estimated from this profile by measuring the exponential decay directly or by estimating a linear decay from the peak insulin concentration to the pre-meal value. The IOB value can account for the current decline in glucose levels from the initial insulin dose and can be subtracted from the glucose correction dose to minimize insulin stacking. If an extra dose is delivered, the application's user interface advises the user to scan their glucose after two hours to ensure there is no hypoglycemia.

[0272] Postprandial hypoglycemia and hyperglycemia alarm method in postprandial insulin therapy The DGA can be configured to generate and / or output alarms or otherwise proactively notify the user of predicted or potential future hypoglycemic and / or hyperglycemic episodes. These alarms allow the user to take action to maintain blood glucose levels within the normoglycemic range, a primary goal of diabetes management.

[0273] In contrast to threshold-based alarm methods, predictive alarm methods for DGA can be based on the predicted probability, timing, and severity of hypoglycemic / hyperglycemic episodes. There are several advantages associated with predictive alarm methods. Predictive alarm methods overcome one of the most common problems with threshold-based alarms: determining the optimal threshold. Setting the threshold too low can result in many false alarms, while setting it too high can result in insufficient patient notification. Furthermore, predictive alarm methods can provide patients with more specific information about upcoming glycemic episodes, including the probability, timing, and severity. The additional specificity provided by the alarm allows patients to take more appropriate actions. Furthermore, predictive alarm methods can provide patients with personalized options for appropriate actions.

[0274] In one exemplary embodiment, as depicted in the flow diagram of FIG. 15 , in exemplary method 720, beginning at step 722, a DGA can be configured to receive a plurality of data sets including analyte (e.g., glucose) time series data and event data. The DGA can be configured to learn specific patterns from an individual patient's historical data. The plurality of data sets received by the DGA can include timestamps associated with glucose time series data and other event markers. In one embodiment, higher-order derivatives and integrals of the glucose time series data are also inputs associated with the alarm system. Furthermore, the plurality of data sets can include, but are not limited to, patient location data, calendar date data, TOD data, and stress level data. Event data can include, but are not limited to, meal data, snack data, exercise data, and medication data, as well as associated timestamps for each event. Medication data can include bolus insulin dosage and amounts and / or basal insulin dosage and amounts.

[0275] In step 724, the DGA can be configured to process at least a portion of the plurality of data to determine the probability, predicted timing, and predicted severity of a future hypoglycemic or hyperglycemic episode. In one embodiment, historical records of each event type can be profiled to generate a prediction of which is most likely to occur in the near future. Predictions of future glycemic episodes can be made with reference to TOD, time of week, or association with other events. For example, the DGA can be configured to predict association with exercise events occurring before or after a particular meal on a particular day of the week.

[0276] The DGA can be configured to predict future hypoglycemic or hyperglycemic episodes using a glucose value prediction algorithm. In one embodiment, the algorithm can be implemented using a naive Bayes classifier. In another embodiment, the algorithm can use a recurrent neural network (RNN) long short-term memory (LSTM) architecture, a random forest, or a combination of various methods. In one embodiment, a machine learning model can be initially trained using glucose time series data collected from clinical studies and real-world databases to develop a population-based model. In one embodiment, the population-based model can be an initial model for each patient at a starting point, and the model can be trained to develop subject-specific patterns by continuously training the algorithm using data from the patient. Thus, the performance of the DGA can be improved for each subject as they use the DGA.

[0277] In step 726, the DGA can be configured to alert the patient of a predicted future hypoglycemic or hyperglycemic episode. In one embodiment, the DGA can be configured to output an alarm to the UID 200 to notify the patient so that the patient can act on the alert, for example, by consuming carbohydrates to treat a future hypoglycemic condition or by consuming insulin to treat a future hyperglycemic condition. In another embodiment, the DGA can be configured such that the alarm system is coupled with a dosing algorithm to suggest an appropriate amount of treatment. In one embodiment, if a future hyperglycemic episode is predicted, the DGA can be configured to output an alarm to the UID 200 that includes recommended dose guidance. In another embodiment, the DGA can be configured to provide a user-accessible display to the UID 200 when determining whether postprandial insulin dosing is recommended. For example, if a future hypoglycemia is calculated to be highly likely, the DGA can be configured to output dose guidance indicating that additional insulin is not recommended. The DGA can also be configured to output recommendations to the UID 200 that are unrelated to dose guidance. For example, the DGA can be configured to output a message recommending carbohydrate consumption, to recheck glucose levels after a short period of time (e.g., 15 minutes), to set a reminder to check glucose after a pre-set or user-set time, and / or to activate a hypoglycemic threshold alarm. In addition to outputting alarms, the DGA can also be configured to include detailed status displays of future hypoglycemic / hyperglycemic episodes.

[0278] In one embodiment, the prediction of a particular episode can be qualified by the probability of occurrence. Different levels of probability can result in different system outputs. For example, initiation of an alarm can require a higher probability than displaying patient-requested post-prandial dose guidance.

[0279] In another embodiment, the DGA can be configured to allow the sensitivity or specificity of the prediction method to be adjustable. In one embodiment, different levels of sensitivity and / or specificity allow the user or the DGA to select a level of sensitivity and / or specificity appropriate to the user's level of engagement at the time. For example, when the user needs to focus on other aspects of their life, the user can select a higher specificity so that the user is only alerted to situations with high urgency. In another embodiment, a higher sensitivity can be selected. For example, when the user decides to allocate more time to improving glucose management, a prediction system with higher sensitivity can be proactively selected to prevent urgent situations. In one embodiment, the sensitivity and specificity settings, as well as the learning module itself, may initially be based on population data. As more user preferences are recorded, pattern recognition can attempt to correlate days of the week (e.g., individual days of a seven-day period, or weekdays vs. weekends), time of day, and potentially the density of the user's calendar activity log to evaluate the most likely sensitivity and specificity setting preferences at any given time.

[0280] System Features The DGS100 can include system considerations that address common events that occur during the management of insulin-rich diabetes. Current insulin bolus dose calculators do not account for these real-world events, leaving users to modify recommended doses according to their best judgment. While it would be desirable for diabetics to be able to modify ideal dose recommendations to reflect real-world events, these considerations can place a significant cognitive burden on diabetics. To reduce this burden and create an even more user-friendly dosing system, the DGS100 leverages insulin and glucose data to make appropriate dose adjustments based on real-world conditions. These system features are further enhanced by the ability for physicians to modify dosing parameters.

[0281] HCP override of dose guidance settings The DGA can provide dose guidance based on individualized parameters such as fixed dosage, target glucose, correction factor, and insulin action duration. The user can view these various parameters in a variety of ways. For example, the user can view the parameters on the UID200 via the settings tab within the DGA or as a viewable breakdown of the dose recommendation. While the user can view these parameters that are part of the recommended dose, these values ​​are strictly informational and may not be editable by the user.

[0282] In contrast, the DGA can be configured to allow the user's HCP to view these values, for example, as part of a clinician-facing system web application, and can also be configured to allow the HCP to edit one or more of these parameters. This web application can display patient performance indicators through a set of the patient's glucose values, such as glucose concentration profiles as well as insulin dosing statistics. If the HCP wishes to modify the dose recommendations to accommodate other treatment regimen changes, the clinician-facing web application can allow the HCP to edit user-specific dosing parameters, including but not limited to those described above.

[0283] In one exemplary embodiment, as depicted in the flow diagram of FIG. 16A , in exemplary method 800, at step 801, an HCP can adjust at least one parameter used to provide insulin dose guidance for a subject in a dose guidance application to create new insulin dose guidance. For example, the HCP can adjust at least one of the fixed dosage, target glucose, correction factor, and insulin action duration. Before adjusting the at least one parameter to create new insulin dose guidance, the HCP can display glucose concentration profiles and insulin statistics to determine whether the subject is experiencing a pattern of low or high glucose for any time period of the day and notify the HCP if the insulin dose guidance needs to be adjusted. A high or low glucose pattern can be determined by GPA, as described elsewhere herein.

[0284] After the HCP changes at least one parameter, the subject may be notified of the new insulin dose guidance in step 802. Additionally, an explanation of why the dose guidance has changed may also be provided to the subject. In one embodiment, either the user or the HCP or both may approve any changes to the dose guidance before the changes take effect. The parameters and dose guidance may be fixed for a period of time (e.g., 14 days to coincide with the life of the sensor), although any parameters that have not changed may be further varied as part of the ongoing learning of the algorithm.

[0285] In step 804, the DGA can determine whether the subject will experience a hypoglycemic episode during a time period, e.g., 14 days, after at least one parameter has been adjusted. If any instances of hypoglycemia are observed during this time period, the HCP can be immediately notified and can determine whether the parameters should be reverted to pre-override or pre-adjustment values. At the end of this time period, the HCP can be asked to review the subject's performance during the time period, e.g., the subject's glucose concentration profile and insulin statistics, to determine whether the adjustment of the dosing parameters should be maintained. If the HCP decides to maintain the adjustment of the dosing parameters, these values ​​can be used as the initial conditions for adjusting future doses. Any historical data related to these values ​​is either negated or significantly discounted in future dose recommendations.

[0286] If the HCP only verbally notifies the user of a dosage change without updating the DGA, for example, via a web application, discrepancies may occur between the printed dosage guidance and the administered dose. If such discrepancies are consistently observed over an extended period of time (e.g., 3 days), the DGA can provide a notification to both the user and the HCP and inquire about any changes to the dosage. If both the user and the HCP confirm, the algorithm can employ the strategy described above. Additionally, the system can prompt the user to enter a reason for such persistent changes.

[0287] The DGA can also feature a "10%" decrease button. When selected by the HCP in association with a specific dose recommendation, the dose recommendation automatically decreases by 10% (e.g., 1 U). For example, an HCP may want to decrease a patient's insulin dose as a result of a change in the dose of a non-insulin medication. This dose change may be known to make the patient more sensitive to insulin. Therefore, the patient's current insulin therapy may be too effective in lowering glucose and may cause hypoglycemia. As a preventative measure, the HCP can decrease the insulin dose. In this example, the DGA can be titrated based on the adjusted value in the future.

[0288] On-board glucose sensor fault detection from insulin and glucose data While the glucose sensor 101 can have built-in fault detection to notify the user to remove and replace the current sensor, the DGS 100 can also be configured to provide fault detection via insulin totals.

[0289] In one exemplary embodiment, an exemplary method 807 beginning at step 808 may receive subject insulin dosage data from the MDD 152 as described in the flow diagram of FIG. 16B.

[0290] In step 810, the DGA can determine whether the multiple recommended insulin dosages differ compared to multiple previous insulin doses administered over a period of time. Each of the multiple insulin dose guidances and each of the multiple previous insulin doses administered are associated with a TOD period, and each of the multiple insulin dose guidances can be compared to one of the multiple previous insulin doses administered associated with the same TOD period to determine a difference. A difference is detected when the dose guidance differs from the corresponding previous dose administered during the same time period. If the sensor 102 is reading a higher glucose level compared to previous sensors, the recommended dose can be similarly increased to compensate for the increased glucose elevation. Conversely, if the sensor 102 has a much lower reading compared to previous sensors, the recommended dose can be immediately lowered to prevent hypoglycemia.

[0291] In step 811, the DGA may determine whether a new sensor was connected to the SCD 102 proximate to the start of the period during which the difference was detected. If it is determined that a new sensor was recently connected to the SCD, in step 812, the DGA may recommend replacing the sensor with a new one. While there are multiple reasons why a sensor may have high or low glucose readings, if these abnormal glucose levels and the onset of dose guidance coincide with the placement of a new sensor and persist while wearing the sensor 102, a faulty sensor may be the underlying cause of such deviations. In this case, the DGA may notify the user to replace the glucose sensor 102.

[0292] Changes in dosing strategy in response to combination therapy or new insulin types Intensive diabetes treatment for patients with type 2 diabetes is often highly complex. In addition to insulin, patients often take a variety of complementary or synergistic medications to improve glucose homeostasis amid a gradual decline in insulin sensitivity. Changes in supportive therapy, such as secretagogue or incretin-based therapy, can affect both endogenous insulin production and tissue insulin sensitivity. Consequently, changes in supportive therapy can affect the efficacy of subsequent exogenous insulin and should be considered as part of insulin dose guidance. A similar situation can arise when a user switches to a different insulin (e.g., from a rapid-acting to an ultrarapid-acting insulin, or from a once-daily basal to a twice-daily basal).

[0293] A change in dosing strategy may occur if the HCP changes the patient's adjunctive therapy or insulin type. In this case, the HCP can provide adjustments to insulin dosing parameters to minimize hypoglycemic episodes, as described above. However, in some cases, the DGA may not be notified of changes in adjunctive therapy or dosing recommendations. In this case, the system can monitor trends in the associated deltas of insulin doses over a period of time.

[0294] In one exemplary embodiment, the subject's insulin dosage data may be received from the MDD 152 in an exemplary method 813 beginning at step 814, as described in the flow diagram of FIG. 16C.

[0295] In step 816, the DGA can detect a differential trend associated with the multiple insulin doses administered in the first time slot over a period of time. The differential trend can include, but is not limited to, the difference between the proposed dose and the administered dose, as well as the difference in the effectiveness of that dose compared to the previous administration at a given dose value (in either magnitude or longevity of response). If a differential trend over a period of time is observed, in step 818, the DGA can provide a notification regarding the differential trend to the user, the HCP, or both. This period can be about two days, or about three days, or about four days. In one embodiment, the DGA can also output a prompt to the HCP and / or user on the UID 200 to confirm a change in therapy, explaining the differential trend.

[0296] As a trend for differences, if the administered dose is consistently different from the DGA recommended dose, it is possible that the HCP has overridden the user's dosing parameters, as described above in the section "HCP Override of Dose Guidance Settings." As a trend for differences, if the recommended dose has a significant and consistent difference in postprandial glucose control compared to past administration, it is possible that insulin effectiveness has changed. This change in insulin effectiveness can be either in magnitude (indicating a change in adjunctive therapy) or duration (indicating a change in insulin analog).

[0297] If ancillary medications are changed, the DGA can transition to a conservative mode and set dose suggestions to a fraction of the previous dose guidance once the user and HCP confirm the change. The DGA can then titrate dosing parameters and dosages to optimize for these new conditions. If insulin type is changed (e.g., from rapid-acting to rapid-acting), the magnitude of the response should remain the same because the market difference between the two is the drug's rapid onset / offset. Rather, the duration of response may change. To address this, population-based values ​​for the new insulin type can be used to estimate insulin duration of action until the system can determine a new, individualized value for this.

[0298] If the user takes a dose that differs from the recommended dosage guidance Although the DGA can recommend doses to users, the DGS100 does not guarantee that users will strictly adhere to dose guidance. The DGA can record the difference between the administered dose and the recommended dose and detect trends in that difference. Observed persistent trends are discussed above in the sections "HCP override of dose guidance settings" and "Changes in dosing strategy in response to adjunctive therapy or new insulin type."

[0299] In an exemplary embodiment, as described in the flow diagram of FIG. 16D , in exemplary method 820, beginning at step 822, the DGA can detect a difference associated with an insulin dose administered to a subject during a first time period compared to dose guidance provided for the first time period. The DGA can compare the dose guidance to insulin dose data received from MDD 152 to determine whether a dose other than that indicated in the dose guidance was administered to the user. The administered dose may be a different amount of insulin taken or a different type of insulin taken compared to the dose guidance provided.

[0300] A specific example of taking a dose that differs from the dose guidance is when the user takes the wrong insulin but in the correct amount (e.g., injecting a long-acting insulin instead of a fast-acting one, or vice versa). The DGA can calculate the optimal dose of a given insulin analog type for any dosing scenario, i.e., using fast-acting insulin for mealtime and correction doses and long-acting insulin for basal doses. To avoid any errors, the DGA can output both the type and amount of insulin during the dose guidance on the UID 200. The MDD 152 can record and transmit both the amount and type of insulin administered. The DGA collects information about which insulin type was used for the injection and can detect any discrepancies. If a discrepancy is detected, a notification can be sent to the user on the UID 200. This notification can indicate that the DGA detected that a different insulin type was used and prompt the user for confirmation.

[0301] If a user administers the wrong type of insulin, an insulin discrepancy notification can be output to the UID 200 to alert the user as soon as possible to a potential severe hypoglycemic episode, even before the hypoglycemia alarm is triggered. For example, if a rapid-acting dose is taken instead of a long-acting dose, severe hypoglycemia may be imminent. This is due to three reasons: (1) rapid-acting insulin, as the name suggests, exerts a more pronounced effect sooner after administration than long-acting insulin; (2) a single daily long-acting dose can be much larger than a single mealtime fast-acting dose; and (3) depending on the timing of the last mealtime dose, an incorrect rapid-acting bolus can cause insulin stacking. Taking a long-acting dose instead of a fast-acting mealtime dose can lead to more unpredictable results. Because long-acting insulins have slower pharmacokinetic and pharmacodynamic profiles than fast-acting insulins, postprandial blood glucose levels after an incorrect long-acting insulin dose can cause hyperglycemia in patients shortly after injection. Depending on the timing of the previous long-acting dose, the potential for insulin stacking and subsequent hypoglycemia exists, especially since the current dose reaches its maximum plasma concentration approximately six hours after injection. The user can be made aware of the possibility of severe hypoglycemia within the UID200 notification. In either situation, the DGA may not make any insulin dose recommendations until the insulin onboard value for injection is near zero (e.g., approximately 5–6 hours for rapid-acting doses, 12–24 hours for long-acting doses). Glucose data during these periods may also be flagged by the system to prevent further parameter refinement and dose titration.

[0302] In step 824, the DGA may detect a hypoglycemic episode associated with an insulin dose administered during a first time period. The DGA may determine a hypoglycemic episode, for example, when a glucose level falls below 70 mg / dL.

[0303] In step 826, the DGA can notify the subject and / or HCP that any hypoglycemic episodes associated with insulin doses administered during the first time period have been detected. This notification can be output to UID 200. If the DGA observes that the user has experienced hypoglycemia due to persistently taking higher than recommended insulin doses, a predictive hypoglycemic alert can be output to UID 200 to notify the user in advance of the actual hypoglycemic event itself and mitigate any associated side effects immediately. However, if hypoglycemia occurs due to persistently dosing above recommended values ​​(defined as, for example, three separate episodes, or four separate episodes), notifications can be sent to both the user via the system phone application and the HCP via the clinician web application alerting both of this trend.

[0304] Any discrepancies that do not appear to be part of a broader trend are noted by the DGA, and the given dose can be incorporated into a continuous learning system of individual dosing patterns. Post-meal indicators can be noted, as can any administered dose. Whenever a different dose is associated with post-prandial hypoglycemia (e.g., less than 70 mg / dL) or other negative outcome, that dose can be flagged, allowing both the user and HCP to improve their respective dosing approaches. This flag can be displayed to the HCP as a note in the glucose pattern report and to the user as a note in the DGA insulin logbook.

[0305] How a system for administering multiple medications consecutively works To ensure accurate dose recording, the DGA must be able to correctly interpret situations where multiple insulin injections are given in close succession, such as priming (potentially multiple times) before the actual dose and multiple injections for a given dosing recommendation.

[0306] In one exemplary embodiment, an exemplary method 827 beginning at step 828 may receive subject insulin dosage data from the MDD 152 as described in the flow diagram of FIG. 16E.

[0307] In step 830, the DGA can detect the administration of multiple insulin doses. The multiple insulin doses include at least a first dose and a last dose, with the last dose administered within the time period of the first dose. If multiple doses are administered within a short time period, at least one of the administered doses can be a priming dose. Procedural best practice in insulin delivery is to use a new needle for each injection and prime each new needle with insulin before administering a dose. This is done by dispensing a small amount of insulin (typically 2 U) into the air until insulin is visible at the tip of the needle, indicating that the needle is full of the injectate. To distinguish between priming doses, the DGA can assume that (1) the priming dose is typically much smaller than the actual dose, (2) the priming dose is the same each time, or (3) little time elapses between priming and administration. Alternatively, if multiple doses are administered within a short time period, the amount that must be injected may actually be a split dose. For example, the user's MDD 152 may have less insulin remaining than the requested dosage, necessitating a cartridge change and subsequent extra injection.

[0308] In step 832, the DGA can record the last dose as the administered dose if a first criterion is met. The first criterion can be met if the last dose is administered within about one minute, two minutes, or three minutes of the first dose. In such a situation, the DGA can consider the first dose and any subsequent intermediate doses (excluding the last dose) to be priming doses and therefore not considered actual administered doses. Therefore, the DGA can be configured to record only the last dose as the administered dose. Because priming doses typically contain a small amount of insulin, in another embodiment, the first condition can be met if the first dose is substantially smaller than the last dose. For example, the first condition can be met if the first dose is about one-tenth, one-fifth, one-quarter, or one-third of the last dose.

[0309] Alternatively, in step 833, the DGA can record the combined amount of the first and last doses as the administered dose if a second criterion is met. The second criterion can be met if a single dose is delivered via multiple smaller injections over a longer time period (e.g., about 4 to about 35 minutes, or about 5 to about 30 minutes) than the priming time period (e.g., 1 to 3 minutes). For example, the user's MDD 152 may have less insulin remaining than the requested dosage, necessitating a cartridge change followed by an extra injection. In this case, the DGA can interpret these injections as two separate actions within the same dosing episode rather than as separate episodes. To do this, the DGA can employ a delay in recording dosing, recording the insulin dose value only approximately 30 minutes after the first dose administration. For example, if the DGA recommends a 10U mealtime dose but only 4U remains in the insulin cartridge, two separate injections would be administered: the first 4U and the remaining 6U. An intermediate priming dose may also occur if a new needle is used for the second injection. Once the first 4U is given, the DGS100 searches for the remaining 6U. If the 6U dose arrives within 30 minutes of the first dose, the DGA can consider this as one 10U dose. If one or more prime doses are detected between the actual doses, the DGA can consider all doses between the first and last dose as primes and therefore not include them in the final dosage. Because the DGA communicates with the MDD152, the DGA can be configured to import the remaining amount of insulin in the insulin cartridge along with the insulin dose information. By knowing the amount of insulin remaining in the cartridge, the DGA can be configured to predict split doses due to cartridge switches and even notify the user about the remaining insulin amount.

[0310] In another embodiment, the DGA can be configured to consider the priming dose and the subsequent dose as split doses and add the two doses together to obtain a single mealtime value that includes the prime dose. If the prime dose is much smaller than the meal dose, the prime dose may have little effect on the titration of the algorithm's dose guidance. This is similar to the split dose logic described below. While not exact, the assumption that the prime dose is much smaller than the injection volume may be reasonable for type 2 diabetes patients suffering from increasing insulin resistance.

[0311] Titration using glucose data when medication is missed To provide optimal dosing advice for progressive disease, DGA can continuously refine its estimates of a user's specific dosing parameters. As a result, it is necessary to identify the appropriate data streams on which to base this algorithm training. To avoid confounding results, DGA can be configured to train only on insulin and glucose data that aligns with the user's own clinically recommended dosing strategy. Such strategies include, but are not limited to, basal-only, basal and one prandial rapid-acting insulin, basal and two prandial rapid-acting insulin, and a fully multiple-injection strategy of basal and three prandial rapid-acting insulin.

[0312] In one exemplary embodiment, the subject's insulin dosage data may be received from the MDD 152 in an exemplary method 833 beginning at step 834, as described in the flow diagram of FIG. 16F.

[0313] In step 836, the DGA can detect a missed insulin dose, where the insulin dose has a time period associated with a duration of action. For example, the DGA can initially identify the user's dosing strategy during an initial learning period before dose guidance can be provided. Using automated meal detection methods and data from a Bluetooth-connected MDD 152, the system can also identify meal events and their associated doses. Thus, the DGA can determine whether a dose was missed for a given meal. A missed basal dose can be detected in the absence of data reported from the user's long-acting MDD 152.

[0314] In step 838, the DGA can ignore glucose analyte data associated with that time period when determining adjustments to insulin dose guidance. A single missed mealtime dose can cause elevated blood glucose levels and insulin boluses compared to those occurring in the past. These changes in mealtime doses, as well as glucose levels after the missed meal, can distort the current dose titration determined by the DGA for a given dosing strategy. Similarly, a missed basal dose can cause a sustained rise in glucose for the duration of insulin action, typically assumed to be one day. As a result, the DGA algorithm will only include glucose and insulin data obtained from meals that involve insulin doses. For fast-acting insulins, the duration of action can be approximately four hours, or approximately five hours, or approximately six hours, or between approximately four and six hours. For example, if a breakfast dose is missed, four hours of glucose data obtained after the start of breakfast may not be included in the dose titration. For long-acting insulins, the duration of action can be about 18 hours, or about 20 hours, or about 24 hours, or about 20 hours to about 24 hours. Because long-acting insulins are useful for both maintaining euglycemia between meals and preventing diabetic ketoacidosis, the system algorithm may not include data from the duration of action if a basal dose is missed. For example, insulin glargine has a reported duration of insulin action of 24 hours. If a user taking glargine misses a basal dose for the day, all data from the following 24 hours will not be used by the system for dose titration.

[0315] Prompts users to avoid missing doses Adherence to a properly titrated insulin dosing regimen improves diabetes management by reducing not only the hyperglycemia associated with missed doses but also the hypoglycemia caused by overcompensatory correction doses. DGAs can be configured to provide users with actionable, easily interpretable data that highlights the positive impact of medication adherence. One such method would be to provide periodic updates comparing a glucose control index or another relevant statistic between time periods with missed doses and time periods without missed doses.

[0316] In one exemplary embodiment, an exemplary method 840 beginning at step 841 may receive subject insulin dosage data from the MDD 152 as described in the flow diagram of FIG. 16G.

[0317] In step 842, the DGA can detect a missed insulin dose. In step 846, the DGA can determine the amount of time the subject's glucose level remained in target range (TIR) ​​during first and second time periods. The target range can be set by the DGA, a user, or an HCP. The target range can be about 70 mg / dL to about 180 mg / dL, or about 70 mg / dL to about 190 mg / dL, or about 70 mg / dL to about 200 mg / dL. The first and second time periods can be the same amount of time, with the first time period not including a missed insulin dose and the second time period including a missed insulin dose. In this manner, a comparison of the TIR for periods including missed doses and periods not including missed doses can be provided.

[0318] In step 848, the DGA may notify the user of the determined TIRs for the first and second time periods. If the difference between the two determined TIRs is greater than a threshold (i.e., the TIR for the first time period minus the TIR for the second time period), a positive message may be displayed to the user on the UID 200 to encourage good medication behavior as well as educate the user about the benefits of medication adherence. For example, the user may be notified that a glucose control index will increase by 10% on days when all doses are taken compared to days when meal doses are missed.

[0319] safety features Exemplary embodiments of the DGS 100 safety features described herein prioritize user safety during the guidance and titration process. Current insulin bolus dose calculators utilize static dosing parameters to calculate dose suggestions. Updating these values ​​by the user or HCP can be a trial-and-error process that can potentially lead to hypoglycemia. As an automated process, the DGA can utilize the user's insulin and glucose data to titrate these parameters and provide customizable dose guidance for the user. Safeguards can be built in to ensure that the recommended titration does not result in hypoglycemia after dosing.

[0320] Safe titration methods The DGA's titration logic can be configured to not increase dose titration until the hypoglycemic window is alleviated. When titration is performed manually by an HCP, the HCP may want to increase some doses and decrease others to achieve a complete titration as quickly as possible to minimize time wastage. However, an automated system can titrate less aggressively and require less HCP time, which takes longer (and is more conservative and safer). Thus, a hypoglycemic patient may initially experience higher average glucose values ​​when titration begins, but once the hypoglycemia is alleviated, insulin doses may be safely increased to achieve glycemic targets.

[0321] One problem with detecting high glucose patterns after a meal is that the previous meal may have postprandial glucose levels, resulting in the next meal starting with high glucose levels. If this occurs, the next meal may falsely indicate a high pattern due to the high starting glucose. To address this issue, the DGA titration strategy can optionally include titrating the overnight dose first before titrating the meal dose. Furthermore, meal doses can be titrated in a sequential order, titrating the earliest meal with a high glucose pattern first, followed by subsequent meals of the day with high glucose patterns. For example, recommendations for titrating insulin doses associated with the overnight period may be made first. Next, for any high glucose patterns detected during any postprandial time period, recommendations for titrating insulin doses associated with breakfast may be provided before recommendations for doses associated with lunch, which in turn may be provided before recommendations for doses associated with dinner. Titrating the previous meal first minimizes preprandial glucose levels for the next meal, reducing the likelihood that the recommended titration will affect or interfere with the titration of the next meal.

[0322] For the embodiments described herein, the DGA can detect high / low patterns as described with respect to the GPA described elsewhere herein.

[0323] For the embodiments described herein, the DGA can be configured to recommend a change (e.g., an increase or decrease) in insulin dosage. The recommended change can be any desired amount of insulin, such as a fraction of a unit (0.1 or 0.5 units), or a unit (1.0 unit), or two or more units of insulin (2.0 or more), or any combination thereof. For ease of explanation, the embodiments described herein refer to adjustments in one-unit intervals.

[0324] The DGA can perform the various steps described in the safe titration embodiment in a variety of different ways, for example, before each meal, or at the beginning or end of the day, or daily, or every other day, or every third day, or when the user queries the DGA for a dose recommendation, or any combination thereof.

[0325] In an exemplary embodiment, the DGA can access measured glucose data (e.g., from SCD 102). The DGA can determine whether there is a high glucose pattern during a nighttime period. If a high pattern is detected during a nighttime period, the DGA can modify the dose guidance. For example, the DGA can increase the amount of medication in the dose guidance for a basal dose if it is safe to do so, e.g., if it can be done without causing any low glucose patterns during any period of the day. The DGA can then determine whether there is a high glucose pattern during at least one postprandial period of the day. If a high pattern is detected, the DGA can increase the medication amount in the dose guidance associated with the earliest of the at least one postprandial period of the day. The DGA can then be configured to increase the medication amount in the dose guidance associated with the next earliest of the at least one postprandial period of the day determined to have a high glucose pattern.

[0326] In another exemplary embodiment, as described in the flow diagram of FIG. 17A , in step 852 of exemplary method 850, the DGA can output first dose guidance that is less than the preceding dose for at least a first time period in response to detecting a low glucose pattern in the subject's analyte data for at least a first time period. The first dose guidance can be output to UID 200. If a low glucose pattern is detected, the DGA can output dose guidance related to the time period during which the low glucose pattern was detected. The DGA can be configured to not recommend any reduction in insulin dose to address the high glucose pattern until the low glucose pattern is no longer detected.

[0327] In step 856, the DGA can output a second dose guidance that is less than the preceding dose for the overnight period in response to detecting a high glucose pattern during the overnight period in the subject's analyte data. For example, the DGA can recommend an increase in the basal dose if it is safe to do so, e.g., if such an increase would not cause a low glucose pattern during another time period of the day.

[0328] In step 860, the DGA can output a third dose guidance that is less than the preceding dose for at least one postprandial period in response to detecting a high glucose pattern for at least one postprandial period in the subject's analyte data. In one embodiment, the DGA can be configured to detect all high glucose patterns found in any time period of the day. In another embodiment, the DGA can be configured to detect whether a post-breakfast period has a high glucose pattern, then the DGA can be configured to detect whether a post-lunch period has a high glucose pattern, then the DGA can be configured to detect whether a post-dinner period has a high glucose pattern.

[0329] If a high glucose pattern is determined to exist in multiple post-prandial periods, the third dose guidance output by the DGA can be associated with the post-prandial period with the highest occurring high glucose pattern of the day. For example, if a high glucose pattern is detected in both the post-breakfast and post-lunch periods, the DGA can increase the recommended insulin dose associated with breakfast before recommending an increase in the recommended insulin dose associated with lunch. Additionally, the DGA can reassess whether the high glucose pattern in the post-breakfast period has abated before recommending an increase in the recommended insulin dose associated with lunch.

[0330] In another exemplary embodiment, as described in the flow diagram of FIG. 17B , in exemplary method 862, beginning at step 864, the DGA can detect a low glucose pattern during any period of a day. The low pattern can be detected based on GPA, as described elsewhere. If a low glucose pattern is detected, in step 866, the DGA can output a first dose guidance for the period of the day during which the low glucose pattern was detected. The first dose guidance can include a lower drug amount compared to the previous dose administered during the period of the day during which the low glucose pattern was detected. The DGA can be configured to not recommend any increase in insulin dose to address the high glucose pattern until a low glucose pattern is no longer detected for any time period.

[0331] In step 868, the DGA can detect whether a high glucose pattern is present during the nighttime period. If a high glucose pattern is detected, in step 870, the DGA can output second dose guidance for the basal dose. For example, the second dose guidance can include a higher drug amount than the prior basal dose if the DGA determines that such a modification is safe, e.g., if such an increase would not cause a low glucose pattern during another time period of the day. In one embodiment, the second dose guidance can include a higher drug amount than the basal dose administered the previous day.

[0332] In step 872, the DGA can detect whether a high glucose pattern is present in the post-breakfast period. If a high glucose pattern is detected, in step 874, the DGA can output third dose guidance. The third dose guidance can increase the recommended insulin dose associated with breakfast, i.e., the third dose guidance can include a higher drug amount than the previous post-breakfast dose. In one embodiment, the third dose guidance can include a higher drug amount than the post-breakfast dose administered the previous day.

[0333] In step 876, the DGA can detect whether a high glucose pattern is present in the post-lunch period. If a high glucose pattern is detected, in step 878, the DGA can output fourth dose guidance. The fourth dose guidance can increase the recommended insulin dose associated with lunch, i.e., the fourth dose guidance can include a higher drug amount than the previous post-lunch dose. In one embodiment, the fourth dose guidance can include a higher drug amount than the post-lunch dose administered the previous day.

[0334] In step 880, the DGA can detect whether a high glucose pattern is present during the post-dinner period. If a high glucose pattern is detected, in step 882, the DGA can output a fifth dose guidance, i.e., the fifth dose guidance can increase the recommended insulin dose associated with the evening meal, i.e., the fifth dose guidance can include a higher drug amount than the previous post-dinner dose. In one embodiment, the fifth dose guidance can include a higher drug amount than the post-dinner dose administered the previous day. In one embodiment, the DGA can increase the recommended insulin amount associated with the evening meal only if it is safe to do so, e.g., can be increased without causing a low glucose pattern during the overnight period.

[0335] In another exemplary embodiment, as described in the flow diagram of FIG. 17C, in step 885 of exemplary method 883, DGA can detect whether a high glucose pattern is present during the nighttime period, as described with reference to GPA.

[0336] If a high glucose pattern is detected during the overnight period, in step 886, the DGA can determine whether increasing the basal dose will cause a low glucose pattern during any time period of the day. If increasing the basal dose can cause a low glucose pattern during any time period of the day, in step 887, the DGA can output first dose guidance for the period of the day determined to have a low glucose pattern. The first dose guidance can reduce the recommended insulin dose associated with the period determined to have a low glucose pattern, i.e., the first dose guidance can include a lower drug amount than the previous dose administered during the same period of the day. In one embodiment, the first dose guidance can include a lower drug amount than the dose administered the previous day during the same period of the day.

[0337] If an increase in the basal dose is not determined to cause a hypoglycemic pattern in any time period, the DGA can output second dose guidance in step 889. The second dose guidance can increase the recommended basal dose to address the high glucose pattern during the nighttime period, i.e., the second dose guidance can include a higher drug amount than the previous basal dose. In one embodiment, the second dose guidance can include a higher drug amount than the basal dose administered the previous day. The DGA can then be configured to output first dose guidance that decreases the recommended insulin amount associated with the daily period associated with the time period determined to have a low glucose pattern in step 887, and then perform a next titration iteration in step 888, and then determine whether a high glucose pattern is present during the nighttime period in step 885.

[0338] In another exemplary embodiment, as set forth in the flow diagram of Figure 17D, in exemplary method 890, beginning at step 891, DGA can detect whether a high glucose pattern is present in a post-dinner period. If a high glucose pattern is detected in the post-dinner period, then in step 892, DGA can determine whether an insulin dose associated with dinner can be safely increased. For example, if an insulin dose associated with dinner could cause a low glucose pattern in the overnight period, then the dinner dose cannot be safely increased.

[0339] If the insulin dose associated with the dinner can be safely increased, the DGA can output first dose guidance in step 893. The first dose guidance can increase the recommended insulin dose associated with the dinner to address the high glucose pattern in the post-dinner period, i.e., the first dose guidance can include a higher drug amount than the previous post-dinner dose. In one embodiment, the first dose guidance can include a higher drug amount than the post-prandial dose administered the previous day.

[0340] If the dinner dose cannot be safely increased, the DGA can output second dose guidance in step 894a. The second dose guidance can decrease or lower the recommended basal insulin dose, i.e., the second dose guidance can include a lower drug amount than the previous basal dose. In one embodiment, the second dose guidance can include a lower drug amount than the basal dose administered the previous day. The DGA can also be configured such that after the DGA decreases the recommended basal dose in step 894a, it can then perform the next titration iteration in step 894b and then determine whether a high glucose pattern is present in the postprandial period in step 891.

[0341] MDD connectivity issues Accurate dose guidance requires that DGAs have access to up-to-date glucose analyte and insulin dose data. Gaps in either data can lead to inaccurate recommendations and potentially serious hypoglycemic episodes.

[0342] 17E, in exemplary method 895 beginning at step 896, the DGA may receive or otherwise access the subject's insulin data (e.g., from the MDD 152). For example, the DGA may check for the latest insulin delivery information by requesting delivery information from various sources, including but not limited to the MDD 152, an MDD-related application, or an interface that stores the latest insulin delivery information (e.g., an MDD application web server), or by checking the memory of various applications for the latest insulin delivery information.

[0343] In step 897, the DGA may determine whether it has received data relating to the last dose administered to the subject. This step may be particularly applicable to embodiments in which the device or software responsible for recording dose administration is different from the DGA or the device that executes the DGA. In embodiments in which the DGA is automatically provided with dose administration data (e.g., the DGA is executed by the MDD 152), this step may not apply.

[0344] The DGA can determine whether it has the most recent data available based on various factors. In one embodiment, this determination can be based on a time gap in received insulin dose data. For example, if a user is on a full multiple injection regimen (basal + 3 mealtime boluses), the DGA can communicate with the MDD 152 or its associated application at least approximately every 6 hours. If no communication has occurred during that time, the DGA can be configured to determine that the MDD 152 needs to connect to the DGA before providing dose guidance. In one embodiment, the DGA can assume that it has not received data related to the last administered dose if the time gap since receiving the last dose is longer than the expected time between meals. For example, the expected time between meals can be approximately 5 hours, or approximately 6 hours, or approximately 6.5 hours, or approximately 7 hours, or approximately 7.5 hours, or approximately 8 hours. In another embodiment, the DGA can detect whether Bluetooth communication is enabled between the user's display device 120, e.g., a smartphone, and the MDD 152. For example, Bluetooth communication may not be enabled on either the device 120, the MDD 152, or both. In another embodiment, the DGA can detect whether a power supply associated with a medication dispensing device needs to be replaced.

[0345] If it is determined that data related to the last dose administered to the subject has not been received, then in step 898, the DGA may notify the user that it cannot provide dose guidance. In one embodiment, the UID 200 may display a message to the user indicating that the DGA cannot output dose guidance until the DGA receives the latest insulin delivery information. In one embodiment, the UID 200 may display a message to the user indicating that the DGA cannot provide dose guidance until the DGA receives the latest insulin delivery information. In another embodiment, the DGA may also generate and display a prompt on the UID 200 informing the user to turn on Bluetooth on the display device 120, the MDD 152, or both. In another embodiment, the DGA may also indicate that the battery of the MDD 152 needs to be replaced. Additionally, the DGA may determine the remaining battery life in the MDD 152 and output a warning to the patient that may be displayed on the UID 200 if the battery life falls below a certain threshold, for example, if the battery life falls below 10%. In another embodiment, the DGA may also output a notification that is displayed on UID 200 to inform the user of the last recorded insulin dose and timestamp, and to warn the user that dose guidance is not based on any doses that may follow after that last dose and timestamp.

[0346] If it is determined that data relating to the last dose administered to the subject has been received, then in step 899 the DGA may output dose guidance to the UID 200 based on the received glucose analyte data and insulin dose data.

[0347] Recommendations for additional testing in cases of insulin delivery abnormalities The DGA can be configured to perform statistics corresponding to various measurements of administered insulin and measured glucose levels. Correlations between various insulin and glucose measurements can be used to identify DGA anomalies, which may include, but are not limited to, incorrect recording of insulin doses, glucose readings that may be biased to low or high, and decreased or increased insulin resistance.

[0348] The insulin index used can be a rolling insulin index. In one embodiment, the rolling insulin index can be the total dose of insulin within a time period. The time period can be about 24 hours, about 48 hours, or about 72 hours. Furthermore, the total dose of insulin can be the total dose of long-acting and fast-acting insulin within a time period. In another embodiment, the rolling insulin index can be the insulin onboard time elapsed since the start of a meal. Such an index can have different predetermined insulin-related parameters, such as DIA or duration of insulin action, for different prandial insulin types.

[0349] The glucose index used can be a rolling glucose index. In one embodiment, the rolling glucose index can be a rolling mean glucose, a rolling median glucose, or a rolling mode glucose. In another embodiment, the rolling glucose index can be a meal start normalized glucose AUC or a change in meal glucose (meal delta).

[0350] In another exemplary embodiment, as described in the flow diagram of Figure 17F, in step 851 of exemplary method 849, the DGA can determine a first rolling insulin index associated with a first time from the insulin dose data. In step 853, the DGA can determine a first rolling glucose index associated with the first time from the glucose dose data. The first rolling insulin index and the first rolling glucose index can be associated together to form a first complementary pair.

[0351] Many different complementary pairs of rolling insulin and glucose indicators can be formed. For example, in one embodiment, a rolling insulin indicator of total insulin dose within a rolling time period can be paired with one of rolling average glucose, rolling median glucose, or rolling mode glucose for the same or similar time window. In another embodiment, complementary pairs such as a rolling past 48-hour glucose mode and a rolling past 48-hour total insulin dose delivered can follow the same general procedure outlined above. In another embodiment, an IOB at a specific time elapsed since the start of a meal can be paired with either a meal start normalized glucose AUC or a meal delta (e.g., the change in glucose due to the meal). Time horizons for pairing IOB and AUC can include, but are not limited to, about 60 minutes, about 120 minutes, about 150 minutes, or about 300 minutes after a meal.

[0352] In step 855, the DGA can reference a data space including the first, second, and third zones to determine which of the first, second, and third zones contains the first complementary pair. The data space can be defined by multiple complementary pairs, each containing a rolling insulin indicator and a rolling glucose indicator associated with the same time. For multiple complementary pairs, paired values ​​can be collected at regular time intervals, such as about every 2 hours, about every 6 hours, about every 12 hours, or about every 24 hours, or other intervals that balance appropriate data density with minimal data storage requirements. In some embodiments, the amount of stored complementary pairs can be maintained in a first-in-first-out (FIFO) buffer implemented in software or hardware. Correlations can be performed among the complementary pairs in the FIFO buffer, similar to fitting a curve to a scatter plot created with insulin and glucose indicator pairs collected over time. This correlation can be a curve of predetermined structure (e.g., a third-order polynomial) potentially with one or more parameters determined based on the paired data in the FIFO buffer and one or more parameters predetermined from the prior population data. This curve can represent the nominally expected relationship between paired observations. In addition to the nominally expected relationship, two safety boundary curves with predetermined structures can be constructed above and below the nominally expected relationship curve. Also, some parameters can be determined based on the paired data in the FIFO buffer, while others can be predetermined from the prior population data. To improve numerical stability, the parameter fit of the nominal curve, upper curve, and lower curve for different time horizons can be related by prior rules. To cover this insulin-glucose balance check, other instances can be interpolated from existing time horizons. The nominally expected relationship curve and the two safety boundary curves can form three zones in the data space. The first zone is defined as the zone between the upper and lower safety boundary curves and contains the curve of the nominally expected relationship.The second zone is defined as the zone above the upper safety boundary curve, and the third zone is defined as the zone below the lower safety boundary curve.

[0353] In step 857, the DGA may output a notification regarding checking at least one of the SCD102 or the MDD152 in response to determining that the complementary pair including the first rolling insulin indicator and the first rolling glucose indicator is included in the second zone or the third zone.

[0354] FIG. 17G illustrates an example of tracking complementary pairs where the rolling insulin index is the estimated IOB 90 minutes after each meal and the rolling glucose index is the meal start normalized glucose AUC 90 minutes after each meal. The complementary pair 867 being analyzed (e.g., the most recent complementary pair, shown by the solid circle in FIG. 17G) is excluded from the curve fit of the nominally expected relationship 861 and represents an event of interest, such as the user's glucose-insulin balance state 90 minutes after a meal. The most recent complementary pair 867 is analyzed to determine which zone it corresponds to. Its position relative to the two safety boundaries 863, 865 can be examined to determine which zone contains this complementary pair 867. For example, the solid circle 867 shown in FIG. 17G appears "below" the lower safety boundary of the third zone 873. There are several possibilities for this to occur: (1) the recorded amount of insulin may be erroneously higher than the amount of insulin actually delivered, (2) the glucose readings may be biased toward low, or (3) other confounding factors may have occurred, such as a decrease in insulin resistance as a result of exercise or a significantly different composition of the meal consumed. Depending on the actual combination of events, there may be a risk of false notification of impending hypoglycemia or a risk of undetected postprandial hyperglycemia if the most recent pair is significantly below the lower safety boundary.

[0355] In contrast, if the location of the complementary pair of interest is above the upper safety boundary of the second zone 871, one or more of the opposite scenarios may have occurred: (1) the recorded amount of insulin may be erroneously lower than the actual amount of insulin delivered, including a missed meal dose; (2) the glucose reading may be biased high; or (3) other confounding factors may have occurred, such as increased insulin resistance as a result of illness or a significantly different composition of the meal that may have been consumed. If the combination of actual events caused the most recent pair to be significantly above the upper safety boundary, there may be a risk of a false alarm of impending hyperglycemia or a risk of postprandial hyperglycemia going undetected.

[0356] A preliminary determination was made with sufficient research data to conclude that if a complementary pair is mapped above the upper safety boundary in the second zone or below the lower safety boundary in the third zone, a consistency check of the DGA 100 may be necessary. Therefore, the DGA can notify the user to perform a self-monitoring of blood glucose (SMBG). If a blood glucose monitoring (SMBG) reading can be linked to the DGS 100 or otherwise input into the DGA, this can be used to determine whether the sensor 102 needs to be replaced using one or more threshold comparisons not covered in this paper. Otherwise, the DGA can notify the user to check the cap of the MDD 152 or smart insulin pen for possible sources of error.

[0357] Incorporating trends into bolus calculations Conventional bolus calculators determine a dose based on the difference between a user's current glucose and their target glucose. These conventional bolus calculators consider the current glucose value as a discrete snapshot in time and do not take into account trends in glucose values ​​at that time. However, during periods of high glucose change, current glucose values ​​may rise or fall rapidly from one point in time to the next, which can result in large fluctuations in the recommended dose.

[0358] In December 2018, the Endocrine Society published consensus guidelines for using glucose trend arrows in diabetes management (see YC Kudva, et al., “Approach to Using Trend Arrows in the Freestyle Libre Flash Glucose Monitoring Systems in Adults,” Journal of the Endocrine Society, vol. 2, pp. 1320-1337, 2018, the entire contents of which are incorporated herein by reference). Glucose trend arrows can be classified into five categories based on the rate of glucose change: (1) rapidly rising (rising by >2 mg / dL / min), (2) rising (1-2 mg / dL / min), (3) slowly changing (no change greater than 1 mg / dL / min), (4) falling (decreasing by 1-2 mg / dL / min), and (5) rapidly falling (decreasing by >2 mg / dL / min).

[0359] To attempt to account for such rapid changes, heuristics can be developed for each trend arrow category that incorporate the rate of glucose change in addition to the current glucose value to calculate the recommended dose. Depending on the reported rate of change, the recommended dose may be increased or decreased, utilizing an additional rate of change term to accommodate dynamic glucose responses. It is envisioned that these heuristics will be developed to minimize the incidence of post-dosing hypoglycemia, thereby improving glucose control indices.

[0360] Insulin site rotation The method described herein facilitates proper site rotation for insulin injections by automatically detecting insulin pen needles and infusion set injection sites.

[0361] The most common method of insulin delivery is via individual injections or continuous infusion into subcutaneous tissue. Proper insulin injection and infusion set placement require "rotation" of injection sites, i.e., cycling through various body sites to avoid any local skin reactions due to repeated and persistent needle sticks. Two common consequences associated with poor injection site rotation are scar tissue formation and lipohypertrophy. Scar tissue is a fibrous tissue formed as a result of sustained trauma or improper healing of an initial wound, characterized by collagen-rich avascular tissue. Lipohypertrophy is clinically defined as the localized accumulation of fat deposits at insulin injection sites. While scar tissue is a ubiquitous problem across injection therapies, lipohypertrophy is an almost exclusive manifestation associated with subcutaneous insulin administration. Both manifest as firm nodules beneath the skin and are primarily associated with localized avascular areas, adversely affecting systemic insulin uptake and action. Therefore, their effects on insulin absorption are often considered together. These effects include reduced insulin absorption (as much as 40%), increased total daily doses (TDD) of insulin, and poorer glycemic control, and are highly prevalent, estimated to affect approximately 50% of insulin-treated individuals.

[0362] Currently, there is no technological solution to the problem of poor site rotation, which relies on patient education by physicians and diabetes educators and subsequent patient adherence.

[0363] As described herein, the DGS 100 can include a site rotation application that can be configured to detect common injection sites and provide guidance to the user to rotate to a new injection site if the same location is repeatedly detected. By presenting site rotation information to the user as part of a dose guidance regimen, the DGS 100 can promote better injection habits, greater insulin efficacy, and greater time-in-range.

[0364] 18A and 18B, in an exemplary method 275 beginning at step 276, a site rotation application can be configured to determine a first distance 270 between an electronic device, such as a display device 120 (e.g., a smartphone), and the SCD 102 and a second distance 272 between the electronic device and the MDD 152. In one embodiment, the site rotation application can employ Bluetooth communication (BLE) between three separate devices: (1) the display device 120, such as a smartphone; (2) the SCD 102; and (3) the connected MDD 152, such as a pen needle. As seen in FIG. 18A, the display device 120 can be configured to serve as a central connection point for communication. When dose guidance is requested from the DGA, the display device 120 can determine the first distance 270 between the display device 120 and the SCD 102 and the second distance 272 between the display device 120 and the MDD 152. In one embodiment, the first distance 270 can be determined by a first signal strength between the display device 120 and the SCD 102, and the second distance 272 can be determined by a second signal strength between the display device 120 and the MDD 152.

[0365] In step 278, the site rotation application can be configured to calculate a third distance 274 between the SCD 102 and the MDD 152, where the SCD 102 is at a fixed location on the subject. Because the display device 120 is not fixed in one location, the first and second signals can be triangulated to determine the third distance 274 between the SCD 102 and the MDD 152, where the SCD 102 is at a fixed location on the subject's body. The calculated third distance 274 between the MDD 152 and the SCD 102 can then be recorded within the site rotation application of the display device 120.

[0366] In step 282, the site rotation application can be configured to output a message recommending that the user move the MDD to a new injection site for the subject in response to determining that the calculated third distance is substantially similar to the pre-calculated third distance. In one embodiment, if the calculated third distance 274 between the SCD 102 and the MDD 152 repeatedly exceeds a predetermined threshold, the site rotation application can be configured to provide a guidance message suggesting that the user inject insulin into the new site location. The threshold can be one, two, or three times. In one embodiment, the site rotation application can further be configured to provide the user with a list of acceptable location sites to suggest new injection site ideas. These alternative regions can include the arms, thighs, abdomen, and buttocks.

[0367] Additional Exemplary Embodiments Additional exemplary embodiments related to medication delivery are also described herein. These embodiments are described in the context of an insulin pen, but are not limited to such. In many embodiments, the MDD 152 (e.g., an insulin pen) can communicate directly with the SCD 102, thus eliminating the need for an additional reader device. In some embodiments, the reader device can also communicate with a sensor control device that is in communication with the MDD 152. In these embodiments, the reader device can also communicate with the MDD 152. In other embodiments, the reader device can communicate with the MDD 152, which itself is in communication with the SCD 102, which is not in communication with the reader device. In other embodiments, the SCD 102 is in communication with a reader device, which itself is in communication with the MDD 152, but which is not in communication with the SCD 102. These various communication schemes allow any type of information (e.g., analyte measurements, alarms, user information or settings, dose guidance, etc.) to be communicated from one device to another in the system via an intermediate device. Communication can occur via Bluetooth or Bluetooth Low Energy, or another wireless protocol (e.g., NFC, RFID, Wi-Fi, etc.) Communication can be maintained (e.g., active Bluetooth pairing) or intermittent (e.g., NFC proximity scanning).

[0368] In some embodiments, the SCD 102 or display device 120, e.g., a reader device (RD), can check whether the MDD 152 maintains a wireless connection with the SCD 102 or reader device for safety purposes to notify the user if the MDD 152 is potentially out of the immediate area. If the SCD 102 or RD detects a loss of connection, an alarm is generated at the SCD 102 (if configured to generate an audible or visual alarm) or the RD (or the SCD 102 can notify the RD to generate an alarm). Similar alarms can be generated when detecting a priming problem or failure or a depleted drug supply (e.g., a low cartridge). If the SCD 102, RD, or MDD 152 detects a high analyte condition (e.g., hyperglycemia), the SCD 102 or RD (through communication with the MDD 152), or the MDD 152 itself, can determine whether a dose was recently administered and refrain from generating an alarm (at the SCD 102, RD, or MDD 152).

[0369] Other notifications or alarms that can be generated for the user may be related to the remaining amount of medication in the pen, including estimating and / or reminding the user that a sufficient amount of medication remains to be administered within a certain time period (e.g., one day). If the next scheduled dose is determined to be greater than the remaining amount of medication, a message can be output recommending the use of a new pen or the loading of a new cartridge. The MDD 152 can be configured to monitor the amount of medication and generate an alarm or notification if the expected amount does not match the actual (sensed) amount. The MDD 152 can also be configured to monitor the duration the delivery button is pressed to ensure complete delivery and notify the user if incomplete delivery is suspected. If the MDD 152 or DGS 100 detects that the button is pressed for too short a time, the MDD 152 or DGS 100 can assume that the user has administered less than recommended. The DGA or MDD 152 can output a query to the user to verify whether the administered dose is an incomplete dose or an intended underdose. If the MDD 152 or DGS 100 detects that the delivery button has been pressed too long (e.g., the amount of time required to administer the recommended dose or the dosage indicated in the dose guidance is too long), the MDD 152 or other system device (e.g., a DGA installed on the display device 120) can notify or alert the user that too large a dose may have been administered. In one embodiment, the MDD 152 can be configured to output an audible notification (e.g., a beep), a tactile notification (e.g., a vibration or click), and / or a visual notification (e.g., an LED, a light) when the dose is complete so that the user can stop pressing the actuator or button.After a dose is administered, the SCD 102 or RD can be configured to send a blackout time to the MDD 152 so that if the MDD 152 is disconnected or otherwise unable to communicate with the SCD 102 or RD 120 configured to provide dose guidance, the MDD 152 knows that further dose administration should be prevented or limited, at least during the blackout period. Suggested feed data can be sent to the integrated pen based on the latest best estimate.

[0370] If the connection between the MDD 152 and the SCD 102 and / or RD 120 is lost and then re-established, the user can be queried to enter the amount of any doses administered during the lack of connection. In embodiments in which the MDD 152 monitors the duration of time the injection or delivery button is pressed, an accelerometer or other press-sensing mechanism can be included. Based on the duration of the press (or the pattern and / or orientation of the pen movement when the button is pressed) measured by the press-sensing mechanism, the MDD 152 can determine whether the button press was for priming or dose administration. The MDD 152 can be configured to distinguish between different types of presses. For example, the MDD 152 can assume that a short set of presses is for priming purposes and thus does not count as dose administration.

[0371] In one embodiment, the DGS 100 (e.g., the SCD 102, the display device 120, or the MDD 152) can be configured to detect the occurrence of meal consumption and then prompt the user to administer if they have not already done so. The DGS 100 can update dose guidance as analyte data is collected during or immediately after a meal and before dose administration.

[0372] The MDD 152 may also include a temperature sensor, allowing the DGS 100 to adjust insulin dose guidance based on temperature fluctuations (if insulin effectiveness or insulin on-board (IOB) levels decrease). The system may also be configured to generate a notification if a temperature range is exceeded.

[0373] The SCD 102 or RD 120 can identify the MDD 152 via communication and determine whether the MDD 152 is of the correct type for administering a dose. For example, if the MDD 152 includes a fast-acting insulin and the dose guidance is for a long-acting insulin, the SCD 102 or RD 120 can generate an alarm or notification, or have the same generated by the MDD 102 itself. The SCD 102 or RD 120 can support simultaneous connection to multiple different MDDs 152, including multiple pens of the same type and / or different types.

[0374] Any device in the DGS 100 (e.g., a "system device," such as the SCD 102, the RD 120, or the MDD 152) can estimate the time until the next medication is needed and notify the user. The system device can track the location of the MDD 152 (e.g., geotracking, geofencing), for example, using a GPS monitor in the MDD 152, and notify the user of the location of the MDD 152.

[0375] The MDD 152 can detect the application of a new needle and prompt the user to apply a new needle if a needle change is not detected. The MDD 152 can be configured with a sensor that detects the presence of air or gas in the needle. The MDD 152 can be configured to detect whether the needle has been tapped (e.g., using a sensor or accelerometer) and prompt or request the user to tap the needle to remove the air or gas. A durable pen can detect the presence of a reservoir or cartridge and prompt the user to enter the medication type (e.g., fast-acting insulin, long-acting insulin). The MDD 152 can be configured to detect whether the medication has been properly mixed (e.g., using a sensor or accelerometer) and prompt the user to mix or shake the MDD 152 to properly mix the medication.

[0376] The MDD 152 may be configured to display the time remaining before the battery of the MDD 152 is depleted or lost and / or the time remaining before the medication within the MDD 152 is depleted or lost.

[0377] The MDD 152 can be configured to lock out a user in certain circumstances, such as if the user is not authenticated (failed login / password entry on the device), the user's analyte levels are too low or falling rapidly, a large dose was administered recently, the insulin in the MDD 152 is expired or too old or has been exposed to excessive temperatures, the system status time has exceeded since the last dose (possibly combined with glucose levels), any combination of these, or other circumstances.

[0378] A system device, e.g., a DGA on display device 120, can calculate and send dose guidance to MDD 152 and / or cloud 190 (e.g., a trusted server) if connectivity is lost or communication fails. MDD 152 can send a confirmation message that the dose guidance was received, and if the confirmation message is not received, the system device can generate an alarm or notification to the user.

[0379] Before administering a dose, for example, when the user picks up or otherwise activates the MDD 152, the MDD 152 (or other system device in communication with the MDD 152) can prompt the user to verify that the last dose known by the system is correct (e.g., a query asking the user to confirm that the last dose, e.g., 5 ml @ 5 pm, was correct and that no other doses were taken). This situation can arise if a different MDD 152 (e.g., a different pen or pump) is used. If the user indicates that no other doses have been taken, the DGS 100 can continue. If the user indicates that other doses have been taken (e.g., ones unknown to the DGS 100), the system device can prompt the user to enter the dosage, time, and medication type and determine new dose guidance. The system device can also be configured to periodically prompt the user to determine whether other pens or medication delivery devices are being used and request that the user integrate those devices into the system, for example, by establishing Bluetooth pairing.

[0380] In many embodiments, a method of parameterizing a patient's medication habits for configuring dose guidance settings is provided, the method including: classifying, by at least one processor, based on time-correlated data characterizing the patient's analytes and medication doses received by the patient during an analysis period, wherein each of the medication doses is classified into a medication class; grouping, by the at least one processor, each of the doses into one of a set of mealtime groups; creating, by the at least one processor, dose parameters for the patient, at least in part, by applying the data for each of the mealtime groups to a model; and storing, by the at least one processor, the dose parameters in computer memory and configuring the dose guidance settings.

[0381] In some embodiments, the classifying step of the method further comprises creating a feature matrix correlating a set of classification features to each of the doses. The method can further include classification features selected from the group including: dosing time for each dose; a time-filtered analyte value; a rate of change of the analyte value closest to the time of dosing; a left-hand AUC (area under the curve) that indicates the difference between the analyte value and the analyte value closest to the time of dosing integrated over the interval before the dosing time; a right-hand AUC that indicates the difference between the analyte value and the analyte value closest to the time of dosing integrated over the interval after the dosing time; time elapsed between dosing times; a probability that a meal will start within a defined interval before the dosing time; a most likely time interval since the most recent meal; a probability that a meal will start within a defined interval after the dosing time; and a most likely time interval until the next meal. The method can further include estimating the time of each meal eaten by the patient during the analysis period. The method may provide that estimating the time of each meal further includes creating a feature matrix based on the time-correlated analyte data, the feature matrix correlating a set of analyte data features to each of the distinct regions classified as rise, pre-fall, and fall. The method may provide that estimating the time of each meal further includes using an algorithm to create an estimated meal time based on the feature matrix. The method may further provide that the set of analyte data features is selected from the group including maximum analyte rate of change, maximum analyte acceleration, analyte value at point of maximum analyte acceleration, duration of region, height of region, maximum deceleration, average rate of change within the region, and time point of maximum analyte acceleration.

[0382] In some embodiments, the mealtime groups include breakfast, lunch and dinner. In some embodiments, the method further comprises grouping by cluster analysis.

[0383] In some embodiments, the model for fitting the data pairs is selected from a linear model with zero slope, a linear model with non-zero slope, a piecewise model with single-point connections, or a nonlinear model that approximates a piecewise model.

[0384] In some embodiments, fitting the data pairs further comprises minimizing the sum of squared residuals.

[0385] In some embodiments, fitting the data pairs further includes evaluating each model with the Akaike Information Criterion (AIC) and selecting the model with the smallest AIC value. In some embodiments, the dose parameters include a fixed dose dosage, an analyte level, and a correction factor from the model selected for each group. In some embodiments, the method further includes combining, by at least one processor, data from multiple mealtime groups to form a combined group, and selecting the best-fitting model and correction factor from the combined group. In some embodiments, the method further includes comparing, by at least one processor, the AIC value of the selected model to a threshold, and requesting user input if the AIC value exceeds the threshold.

[0386] In some embodiments, the analyte comprises an indicator of a glucose level and the agent com...

Claims

1. 1. A method for parameterizing a patient's dosing habits to configure dose guidance settings, the method comprising: classifying, by at least one processor, each of the doses of the plurality of medications into a medication class based on the time correlation data characterizing the analytes of the patient; grouping, by the at least one processor, each of the doses into one of a set of mealtime groups; generating, by the at least one processor, dosage parameters for the patient, at least in part, by applying the data for each of the mealtime groups to a model; storing, by the at least one processor, the dose parameters in a computer memory and configuring dose guidance settings; A method comprising:

2. 2. The method of claim 1, wherein the classifying step further comprises creating a feature matrix correlating a set of classification features to each of the doses.

3. 3. The method of claim 2, wherein the classification features are selected from the group including: time of administration of each dose, a time filtered analyte value, the rate of change of the analyte value closest to the time of administration, a left AUC (area under the curve) which is the integral of the difference between the analyte value and the analyte value closest to the time of administration over the interval before the dose time, a right AUC which is the integral of the difference between the analyte value and the analyte value closest to the time of administration over the interval after the dose time, the time elapsed between dose times, the probability that a meal will start within a defined interval before the dose time, the most likely time interval since the most recent meal, the probability that a meal will start within a defined interval after the dose time, and the most likely time interval until the next meal.

4. 3. The method of claim 2, wherein the classifying step further comprises estimating the time of day of each meal consumed by the patient during an analysis period.

5. 5. The method of claim 4, wherein estimating the time of each meal further comprises creating a feature matrix based on the time-correlated analyte data, the feature matrix correlating a set of analyte data features to each of the distinct regions classified as up, pre-down, and down.

6. The method of claim 5 , wherein estimating a time of each meal further comprises using an algorithm to generate estimated meal times based on the feature matrix.

7. 6. The method of claim 5, wherein the set of analyte data features is selected from the group including maximum analyte rate of change, maximum analyte acceleration, analyte value at point of maximum analyte acceleration, duration of region, height of region, maximum deceleration, average rate of change within the region, and time of maximum analyte acceleration.

8. The method of claim 1 , wherein the meal time groups include breakfast, lunch and dinner.

9. The method of claim 8 , wherein the grouping step further comprises grouping by cluster analysis.

10. The method of claim 1, wherein applying the data for each of the meal time groups to the model includes fitting the data pairs to the model, the model being selected from a linear model with zero slope, a linear model with non-zero slope, a piecewise model with single point connections, or a nonlinear model that approximates a piecewise model.

11. The method of claim 10 , wherein fitting the data pairs further comprises minimizing a sum of squared residuals.

12. 11. The method of claim 10, wherein fitting the data pairs further comprises evaluating each model with the Akaike Information Criterion (AIC) and selecting the model with the smallest AIC value.

13. 13. The method of claim 12, wherein the dose parameters include fixed dose dosages, analyte levels, and correction factors from a model selected for each group.

14. 14. The method of claim 13, further comprising the steps of: combining, by the at least one processor, data from multiple meal time groups to form a combined group; and selecting, from the combined group, a best fitting model and correction coefficients.

15. 13. The method of claim 12, further comprising the steps of: comparing, by the at least one processor, the AIC value of the selected model to a threshold; and requesting user input if the AIC value exceeds the threshold.

16. The method of claim 1 , wherein the analyte comprises an indicator of a glucose level and the agent comprises insulin.

17. The method of claim 1 , further comprising providing the dose guidance settings to a user interface device for output to a user.

18. The method of claim 1 , further comprising receiving, by the at least one processor, time correlation data characterizing analytes of the patient and doses of medication received by the patient during an analysis period.

19. 10. The method of claim 1, wherein the at least one processor selects the medication class for each of the doses from among a prandial dose, a non-prandial dose, and an ambiguous dose based on the time correlation data.

20. The method of claim 1 , wherein applying the data for each of the mealtime groups to a model comprises fitting data pairs to the model.

21. 1. An apparatus for parameterizing a patient's dosing habits to configure dose guidance settings, the apparatus comprising: an input configured to receive measured analyte data, dietary data, and medication data; a display configured to visually present information; one or more processors coupled to the input, the display, and a memory for storing instructions and time-correlated data characterizing the patient's analytes and the dose of medication received by the patient during an analysis period; the instructions, when executed by the one or more processors, cause the apparatus to: classifying each of said doses into a drug class based on said time correlation data; each of said doses is grouped into one of a set of mealtime groups; generating, at least in part, dosage parameters for said patient by applying said data for each of said mealtime groups to a model; Storing said dose parameters to configure dose guidance settings An apparatus for:

22. 22. The device of claim 21, wherein the memory further holds instructions for classifying each of the doses, at least in part, by creating a feature matrix that correlates a set of classification features to each of the doses.

23. 23. The device of claim 22, wherein the memory further retains instructions for correlating the classification features selected from the group including: time of administration of each dose, a time filtered analyte value, a rate of change of the analyte value closest to the time of administration, a left AUC (area under the curve) indicating the difference between the analyte value and the analyte value closest to the time of administration integrated over the interval before the dose time, a right AUC indicating the difference between the analyte value and the analyte value closest to the time of administration integrated over the interval after the dose time, time elapsed between dose times, a probability that a meal will start within a defined interval before the dose time, a most likely time interval since the most recent meal, a probability that a meal will start within a defined interval after the dose time, and a most likely time interval until the next meal.

24. 23. The apparatus of claim 22, wherein the memory further holds instructions for classifying each of the doses at least in part by estimating a time of day for each meal consumed by the patient during the analysis period.

25. 22. The device of claim 21, wherein the memory further retains instructions for grouping each of the doses into the mealtime groups comprising breakfast, lunch and dinner.

26. 26. The device of claim 25, wherein the memory holds instructions for grouping each of the doses at least in part by cluster analysis.

27. 22. The apparatus of claim 21, wherein the memory further holds instructions for fitting the data pairs using a model, the model being selected from a linear model with zero slope, a linear model with non-zero slope, a piecewise model with single point connections, or a nonlinear model that approximates a piecewise model.

28. 30. The apparatus of claim 27, wherein the memory further retains instructions for fitting the data pairs at least in part by minimizing a sum of squared residuals.

29. 30. The apparatus of claim 27, wherein the memory further holds instructions for fitting the data pairs at least in part by evaluating each model with the Akaike Information Criterion (AIC) and selecting the model having the smallest AIC value.

30. 30. The apparatus of claim 29, wherein the memory holds dose parameters including fixed dose dosages, analyte levels, and correction factors from a model selected for each group.

31. 22. The device of claim 21, wherein the memory holds time-correlated data characterizing analytes and medications including insulin of the patient, including indicators of glucose level.

32. 22. The apparatus of claim 21, wherein the memory further retains instructions for providing the dose guidance settings to a user interface device for output to a user.

33. 22. The apparatus of claim 21, wherein the memory further holds instructions for receiving time-correlated data characterizing analytes of the patient and doses of medication received by the patient during an analysis period.

34. 22. The apparatus of claim 21, wherein the memory further retains instructions for selecting the medication class for each of the doses from among a prandial dose, a non-prandial dose, and an ambiguous dose based on the time correlation data.

35. 22. The apparatus of claim 21, wherein the memory further retains instructions for applying data for each of the meal time groups to a model, at least in part, by fitting data pairs to the model.

36. The method of claim 1, wherein the classifying step is performed by a machine learning model.

37. The method of claim 1, wherein the data for each of the mealtime groups includes pre-meal analyte levels and corresponding meal dosages.

Images