Patient-specific glucose prediction system and method

Abstract

Infusion devices and related medical devices, patient data management systems and methods for monitoring a physiological condition of a patient are provided. An exemplary method of monitoring a physiological condition of a patient involves obtaining current measurement data of the physiological condition of the patient provided by a sensing device; obtaining a user input indicative of one or more future events associated with the patient; and determining, in response to the user input, a prediction of a future physiological condition of the patient based at least in part on the current measurement data and the one or more future events using one or more prediction models associated with the patient; and displaying the predicted graphical representation on a display device.

Description

[0001] This application is a divisional application of the PCT international application (PCT application number: PCT / US2018 / 024099) entering the national phase with application number 201880032844.0, application date March 23, 2018, and invention name “Patient-specific glucose prediction system and method”.

[0002] priority

[0003] This PCT application claims the benefit of and priority to the following patent applications: U.S. Patent Application Serial No. 15 / 933,264, filed March 22, 2018; U.S. Patent Application Serial No. 15 / 933,258, filed March 22, 2018; U.S. Patent Application Serial No. 15 / 933,266, filed March 22, 2018; U.S. Patent Application Serial No. 15 / 933,268, filed March 22, 2018; U.S. Patent Application Serial No. 15 / 933,272, filed March 22, 2018; U.S. Patent Application Serial No. 15 / 933,275, filed March 22, 2018; U.S. Patent Application Serial No. 15 / 933,277, filed March 22, 2018; U.S. Patent Application Serial No. 15 / 933,278, filed July 18, 2017 No. 62 / 476,468, filed on March 24, 2017; U.S. Provisional Patent Application Serial No. 62 / 476,493, filed on March 24, 2017; U.S. Provisional Patent Application Serial No. 62 / 476,506, filed on March 24, 2017; and U.S. Provisional Patent Application Serial No. 62 / 476,517, filed on March 24, 2017. Technical Field

[0004] Embodiments of the subject matter described herein relate generally to medical devices and related patient monitoring systems, and more particularly, embodiments of the subject matter relate to database systems that facilitate improved patient-specific queries, predictions, and recommendations. Background Art

[0005] Infusion pump devices and systems are well known in the medical field for delivering or dispensing agents such as insulin or another prescription drug to a patient. The use of infusion pump therapy has been increasing, particularly for insulin delivery to diabetic patients. Continuous insulin infusion provides better control of diabetic symptoms, and therefore control schemes have been developed that allow insulin infusion pumps to monitor and regulate a patient's blood glucose level in a substantially continuous and autonomous manner, for example, throughout the night while the patient sleeps.

[0006] Regulating blood glucose levels is complicated by the variation in response time of the type of insulin used and the individual insulin response of each patient. In addition, the patient's daily activities and experiences may cause the patient's insulin response to vary during the day or from one day to the next. Therefore, there is a need to help improve glucose control, which takes into account many different variables in a personalized manner. In addition, the effects and efficacy of different treatment regimens may vary from patient to patient. Therefore, it is also desirable to better understand how the condition of an individual patient can be affected by various behaviors, or how different therapies or behaviors can improve the regulation of the patient's condition. In conjunction with the accompanying drawings and the previous background technology, according to the subsequent detailed description and the appended claims, other desired features and characteristics of the methods, apparatus and systems described herein will become apparent. Summary of the Invention

[0007] The present invention provides an infusion device and related medical devices, a patient data management system, and a method for monitoring a patient's physiological condition. One embodiment of the method for monitoring a patient's physiological condition includes: obtaining, at a computing device, current measurement data of the patient's physiological condition provided by a sensing device; obtaining, at the computing device, user input indicating one or more future events associated with the patient; and, in response to the user input, using one or more predictive models associated with the patient, determining a prediction of the patient's future physiological condition based, at least in part, on the current measurement data and the one or more future events; and displaying, at the computing device, a graphical representation of the prediction on a display device.

[0008] In another embodiment, an apparatus of an electronic device is provided. The electronic device includes: a communication interface for receiving current measurement data of a patient's physiological condition from a sensing device; a display device having a graphical user interface display presented thereon; a user interface for obtaining user input indicating one or more future events; and a control system coupled to the communication interface, the display device, and the user interface to determine a prediction of the patient's future physiological condition based at least in part on the current measurement data and the one or more future events using one or more predictive models associated with the patient, and to display a graphical representation of the prediction within the graphical user interface display of the display device.

[0009] In another embodiment, a method for monitoring a patient's glucose level is provided. The method includes: obtaining, at a computing device, current sensor glucose measurement data of the patient from a glucose sensing device; obtaining, at the computing device, user input indicating one or more future events for the patient via a user interface; determining, at the computing device, a simulated future glucose level for the patient based at least in part on the current sensor glucose measurement data and the one or more future events using a plurality of different prediction models associated with the patient, wherein the plurality of different prediction models includes an hourly forecast model associated with the patient; and displaying, on a display device associated with the computing device, a graphical representation of the simulated glucose level relative to future time.

[0010] In another embodiment, a method for monitoring a patient's physiological condition includes: obtaining current measurement data of the patient's physiological condition from a sensing device; predicting one or more events that may affect the patient's physiological condition at one or more different times in the future based at least in part on historical event data associated with the patient; using a predictive model associated with the patient, determining a plurality of predicted values of the patient's physiological condition associated with a plurality of different time periods in the future based at least in part on the current measurement data and the one or more events; and displaying the plurality of predicted values relative to the plurality of different time periods in the future on a display device.

[0011] In another embodiment, a system is provided, comprising: a display device; a sensing device for obtaining current measurement data of a patient's physiological condition; and a control system coupled to the display device and the sensing device to determine a plurality of forecast hourly averages of the patient's future physiological condition based at least in part on the current measurement data using an hourly forecast model associated with the patient, and to display the plurality of forecast hourly averages on the display device.

[0012] In another embodiment, a method for monitoring a patient's glucose level is provided. The method includes: determining an hourly forecast model for the patient based at least in part on a relationship between historical glucose measurement data of the patient and historical event data associated with the patient; obtaining current glucose measurement data of the patient from a glucose sensing device; predicting one or more events that may affect the patient's glucose level at one or more different times in the future based at least in part on the historical event data associated with the patient; determining, using the hourly forecast model associated with the patient, a plurality of future hourly forecast average glucose values for the patient based at least in part on the current glucose measurement data and the one or more events; and displaying a graphical representation of the plurality of future hourly forecast average glucose values on a display device.

[0013] In another embodiment, a method for monitoring a patient's physiological condition includes: obtaining current measurement data for the patient's physiological condition from a sensing device; determining a first plurality of prediction values for the patient's physiological condition in the future based at least in part on the current measurement data using a first prediction model; determining a second plurality of prediction values for the patient's physiological condition in the future based at least in part on the current measurement data using a second prediction model different from the first prediction model; determining an aggregate prediction for the patient's physiological condition relative to a future time based at least in part on the first plurality of prediction values, the second plurality of prediction values, and weighting factors associated with the respective first and second prediction models, wherein the weighting factors vary relative to the future time based on a relationship between a first reliability metric associated with the first prediction model and a second reliability metric associated with the second prediction model; and displaying a graphical indication of the aggregate prediction for the patient's physiological condition relative to the future time on a display device.

[0014] Another method for monitoring a patient's physiological condition includes: obtaining current measurement data for the patient's physiological condition from a sensing device; determining a plurality of prediction values indicative of the physiological condition for a future time based at least in part on the current measurement data using a plurality of different prediction models associated with the patient, wherein each of the plurality of prediction values is associated with a corresponding prediction model in the plurality of different prediction models, determining a reliability metric associated with the corresponding prediction model for each corresponding prediction model in the plurality of different prediction models based at least in part on a relationship between the future time of day and the current time, determining a weighting factor associated with the corresponding prediction model for each corresponding prediction model in the plurality of different prediction models based at least in part on the reliability metric associated with the corresponding prediction model, determining a collective prediction value for the patient's physiological condition as a weighted average of the corresponding prediction values of the plurality of prediction values and the weighting factors associated with the corresponding prediction models, and displaying a graphical indication of the collective prediction value of the patient's physiological condition associated with the future time.

[0015] In another embodiment, an apparatus of an electronic device is provided. The electronic device includes: a communication interface for receiving current measurement data of a patient's physiological condition from a sensing device; a display device having a graphical user interface display including a graphical representation of the current measurement data; a user interface for obtaining user input for adjusting the graphical user interface display to view the future; and a control system coupled to the communication interface, the display device, and the user interface to determine a first plurality of predicted values of the patient's future physiological condition based at least in part on the current measurement data using a first prediction model; determine a second plurality of predicted values of the patient's future physiological condition based at least in part on the current measurement data using a second prediction model different from the first prediction model; determine an aggregate prediction of the patient's physiological condition relative to a future time based at least in part on the first plurality of predicted values and the second plurality of predicted values; and display a graphical representation of the aggregate prediction on the graphical user interface display in response to the user input.

[0016] In another embodiment, a database system is provided. The database system includes: a database for maintaining data related to a plurality of entities; and a computing device coupled to the database to identify relationships between different pairs of the plurality of entities, generate metadata defining a graph structure maintaining relationships between different entities in the plurality of entities at different logical levels, and store the metadata in the database.

[0017] In another embodiment, a method for managing a database maintaining data related to a plurality of patients is provided. The method includes: analyzing, by a computing device, a graph data structure of a logical layer in the database, the graph data structure being defined by metadata in the database and comprising a plurality of entities, wherein each entity in the plurality of entities maintains a logical relationship with one or more fields of observation data associated with a corresponding patient in a plurality of patients; identifying, by the computing device, a relationship between a pair of entities in the plurality of entities in the logical layer; and updating, by the computing device, the metadata in the database to create a link between the pair of entities.

[0018] In another embodiment, a system includes: a plurality of medical devices for obtaining observation data related to a plurality of patients; a database for maintaining data related to a plurality of entities, wherein each of the plurality of entities maintains a logical relationship between one or more fields of the observation data stored in the database; and a computing device coupled to the database to identify relationships between different entities of the plurality of entities, generate metadata defining a graphical structure that maintains relationships between different entities of the plurality of entities at a plurality of different logical layers, and store the metadata in the database.

[0019] In another embodiment, a method for querying a database is provided. The method includes: receiving, by a computing device coupled to the database, an input query from a client device; identifying, by the computing device, a logical layer among a plurality of different logical layers of the database to search based at least in part on the input query; generating, by the computing device, a query statement based at least in part on the input query, the query statement for searching the identified logical layer among the plurality of different logical layers of the database; querying the identified logical layer of the database using the query statement to obtain result data; and providing, by the computing device, search results influenced by the result data to the client device.

[0020] In another embodiment, a database system is provided. The database system includes a database for maintaining data related to a plurality of entities and metadata defining a graph structure, the graph structure maintaining relationships between different entities in the plurality of entities in each of a plurality of different logical layers; a client device coupled to a network for transmitting session input from a user of the client device; and a computing device coupled to the database and the network for receiving the session input from the client device, determining a logical layer in the plurality of different logical layers of the database to search based at least in part on the session input, generating a query statement based at least in part on the session input, the query statement for searching the identified logical layer in the plurality of different logical layers of the database to obtain result data from the logical layer of the database, and providing search results influenced by the result data to the client device via the network.

[0021] In another embodiment, a method for querying a database includes: providing a graphical user interface display at a client electronic device that prompts a user for conversational interaction; receiving conversational input from the user at the client electronic device; transmitting the conversational input from the client electronic device to a remote device over a network, wherein the remote device analyzes the conversational input based at least in part on the conversational input to identify a logical layer among a plurality of different logical layers of a database to search, and queries the identified database logical layer to obtain result data; and providing conversational search results within the graphical user interface display at the client electronic device in response to the conversational input, wherein the conversational search results are affected by the result data.

[0022] In another embodiment, an apparatus for an infusion device is provided. The infusion device includes: an actuator operable to deliver a fluid to a user, the fluid affecting a physiological condition of the user; a communication interface for receiving measurement data indicative of the physiological condition of the user; a sensing device for obtaining background measurement data; and a control system coupled to the actuator, the communication interface, and the sensing device to determine commands for autonomously operating the actuator in a manner affected by the measurement data and the background measurement data, and autonomously operating the actuator in accordance with the commands to deliver the fluid to the user.

[0023] In another embodiment, a method of operating an infusion device to regulate a physiological condition of a patient is provided. The method includes: obtaining measurement data indicative of a physiological condition from a first sensing device at the infusion device; determining, at the infusion device, a delivery command for autonomously operating an actuator of the infusion device to deliver a fluid that affects the physiological condition to the patient; obtaining background measurement data from a second sensing device of the infusion device at the infusion device; adjusting the delivery command in a manner influenced by the background measurement data to obtain an adjusted delivery command; and automatically operating the actuator to deliver the fluid according to the adjusted delivery command.

[0024] In another embodiment, a method for monitoring a patient's physiological condition includes: obtaining, by a computing device, measurement data related to the patient's physiological condition from a sensing device; obtaining, by the computing device, medical record data associated with the patient from a database; determining, by the computing device, a risk score for a medical condition associated with the patient based at least in part on the measurement data, the medical record data, and one or more relationships between the group measurement data and the group medical record data; and initiating one or more actions at the computing device based at least in part on the risk score.

[0025] In another embodiment, a system is provided that includes: a sensing device for obtaining measurement data related to a patient's physiological condition; a database for maintaining medical record data associated with the patient, group measurement data associated with multiple patients, and group medical record data associated with multiple patients; and a computing device communicatively coupled to the sensing device and the database, for determining a risk score for the medical condition associated with the patient based at least in part on the measurement data, the medical record data, and one or more relationships between the group measurement data and the group medical record data, and performing one or more actions when the risk score is greater than a threshold.

[0026] In another embodiment, a method for monitoring a patient includes: obtaining population glucose measurement data associated with multiple patients from a database; obtaining population medical record data associated with the multiple patients from a database; determining a risk model for the medical condition for a subset of multiple patients having the medical condition based on a relationship between the population glucose measurement data and the population medical record data; obtaining sensor glucose measurement data of the patient from an interstitial glucose sensing device; obtaining medical record data associated with the patient from a database; using the risk model, determining a risk score for the medical condition associated with the patient based at least in part on the sensor glucose measurement data and the medical record data; and generating a treatment recommendation for the patient when the risk score is greater than a threshold.

[0027] In another embodiment, a method for managing a patient's physiological condition includes: obtaining, by a computing device, measurement data related to the patient's physiological condition from a sensing device; obtaining, by the computing device, medical record data associated with the patient from a database; classifying, by the computing device, the patient into patient groups based at least in part on the measurement data and the medical record data; obtaining, by the computing device, a plurality of different boost models associated with the patient groups, wherein each of the plurality of different boost models corresponds to a corresponding therapeutic intervention among a plurality of different therapeutic interventions, determining, by the computing device, a plurality of boost metric values associated with the patient for the plurality of different therapeutic interventions based on the measurement data and the medical record data using the plurality of different boost models, and providing, by the computing device, an indication of a recommended therapeutic intervention to the patient based at least in part on the corresponding boost metric values associated with the recommended therapeutic intervention.

[0028] In another embodiment, a method for managing a patient's physiological condition includes: obtaining, by a computing device coupled to a database, population measurement data associated with multiple patients from the database; obtaining, by the computing device, population medical record data associated with the multiple patients from the database; determining, by the computing device, a patient group including a subset of the multiple patients for modeling based on at least one of a subset of the population measurement data associated with the subset of the multiple patients and a subset of the population medical record data associated with the subset of the multiple patients; determining, by the computing device, a plurality of different boost models associated with the patient group for different therapeutic interventions based at least in part on one or more relationships between the subset of the population measurement data associated with the subset of the patients and the subset of the population medical record data; obtaining, from a sensing device, measurement data related to the patient's physiological condition; obtaining, from a database, medical record data associated with the patient; determining, using a plurality of different boost models, a plurality of boost metric values associated with the patient based on the measurement data and the medical record data; selecting a recommended therapeutic intervention from the different therapeutic interventions based at least in part on the plurality of boost metric values; and providing an indication of the recommended therapeutic intervention to the patient.

[0029] In another embodiment, a system is provided, comprising: a sensing device for obtaining measurement data related to a patient's physiological condition; a database for maintaining medical record data associated with the patient and a plurality of different boost models associated with patient groups based on a relationship between group measurement data and group medical record data associated with a plurality of patients, wherein each of the plurality of different boost models corresponds to a respective therapeutic intervention among a plurality of different therapeutic interventions; and a computing device communicatively coupled to the sensing device and the database for classifying the patient into patient groups based at least in part on the measurement data and the medical record data; determining a plurality of boost metric values associated with the patient for a plurality of different therapeutic interventions based on the measurement data and the medical record data using the plurality of different boost models; and generating a user notification of a recommended therapeutic intervention for the patient based at least in part on the respective boost metric values associated with the recommended therapeutic interventions.

[0030] In another embodiment, a method for managing a patient's physiological condition includes: obtaining, by a computing device, measurement data related to the patient's physiological condition from a sensing device; obtaining, by the computing device, medical record data associated with the patient from a database; obtaining, by the computing device, a plurality of different adherence models associated with a plurality of different treatment regimens, wherein each of the plurality of different adherence models corresponds to a respective treatment regimen among the plurality of different treatment regimens; determining, by the computing device, a plurality of adherence metrics associated with the patient for the plurality of different treatment regimens based on the measurement data and the medical record data using the plurality of different adherence models; and providing, by the computing device, an indication of a recommended treatment regimen to the patient based at least in part on the respective adherence metrics associated with the recommended treatment regimens.

[0031] In another embodiment, a method for managing a patient's physiological condition includes: obtaining, by a computing device, population medical record data for a plurality of patients to whom a treatment regimen is prescribed; obtaining, by the computing device, population medical claims data for the plurality of patients; obtaining, by the computing device, population measurement data for the plurality of patients; determining, by the computing device, an adherence model based at least in part on a relationship between the population medical claims data, the population medical record data, and the population measurement data; obtaining, by the computing device, measurement data related to the patient's physiological condition from a sensing device; obtaining, by the computing device, medical record data associated with the patient from a database; determining, by the computing device, an adherence metric value for the treatment regimen for the patient based at least in part on the measurement data and the medical record data using the adherence model; and recommending, by the computing device, a treatment regimen for the patient based on the adherence metric value.

[0032] In another embodiment, a system is provided, comprising: a sensing device for obtaining measurement data related to a physiological condition of a patient; a database for maintaining medical record data associated with the patient and a plurality of different adherence models associated with a plurality of different treatment regimens, wherein each of the plurality of different adherence models corresponds to a respective treatment regimen in the plurality of different treatment regimens; and a computing device communicatively coupled to the sensing device and the database for determining, using the plurality of different adherence models, a plurality of adherence metrics associated with the patient for the plurality of different treatment regimens based on the measurement data and the medical record data, and generating a user notification of a recommended treatment regimen for the patient based at least in part on the respective adherence metrics associated with the recommended treatment regimens.

[0033] This summary is provided to introduce some concepts in a simplified form that are further described below in the detailed description. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] A more complete understanding of the present subject matter may be obtained by referring to the detailed description and claims when considered in conjunction with the following drawings, in which like reference numerals indicate similar elements throughout the drawings, which are illustrated for simplicity and clarity and are not necessarily drawn to scale.

[0035] Figure 1 Depicts an exemplary embodiment of a patient data management system;

[0036] Figure 2 is a combination of one or more exemplary embodiments Figure 1 A flowchart of a suitable implementation of an exemplary data management process for a patient data management system;

[0037] Figure 3 is a combination of one or more exemplary embodiments Figure 1 A flowchart of a suitable implementation of an exemplary query process for a patient data management system;

[0038] Figure 4 Describes the basis Figure 3 An exemplary embodiment of the query process is suitable for Figure 1 an exemplary graphical user interface (GUI) presented on a client electronic device in a patient data management system;

[0039] Figure 5 Describes the Figure 1 an exemplary graph data structure implemented at a logical database layer in a patient data management system;

[0040] Figure 6 is a block diagram of an exemplary infusion system suitable for use with a fluid infusion device in one or more embodiments;

[0041] Figure 7 In one or more embodiments, Figure 6 A block diagram of an exemplary pump control system for use in an infusion device in an infusion system;

[0042] Figure 8 is a flow chart of a suitable implementation of an exemplary prognostic process in conjunction with a patient data management system in one or more exemplary embodiments;

[0043] Figure 9 Describes the basis Figure 8 an exemplary GUI display suitable for presentation on a client electronic device of an exemplary embodiment of a forecast process;

[0044] Figure 10 Describes a suitable combination Figure 8 A block diagram of an hourly forecast model used in an exemplary embodiment of a forecast process;

[0045] Figure 11 is a flow chart of a suitable implementation of an exemplary ensemble prediction process in conjunction with a patient data management system in one or more exemplary embodiments;

[0046] Figure 12 Describes the basis Figure 11 an exemplary GUI display suitable for presentation on a client electronic device of an exemplary embodiment of an ensemble prediction process;

[0047] Figure 13 is a flow chart of a suitable implementation of an exemplary patient simulation process in conjunction with a patient data management system in one or more exemplary embodiments;

[0048] Figures 14 to 16 Describes the basis Figure 13 exemplary GUI displays suitable for presentation on a client electronic device of various exemplary embodiments of a patient simulation process;

[0049] Figure 17 is a flow chart of a suitable implementation of an exemplary risk management process in conjunction with a patient data management system in one or more exemplary embodiments;

[0050] Figure 18 is a flow chart of a suitable implementation of an exemplary enhancement suggestion process in conjunction with a patient data management system in one or more exemplary embodiments;

[0051] Figure 19is a flow chart of a suitable implementation of an exemplary adherence recommendation process in conjunction with a patient data management system in one or more exemplary embodiments;

[0052] Figure 20 An exemplary embodiment of an infusion system is shown;

[0053] Figure 21 Shows the application Figure 20 a plan view of an exemplary embodiment of a fluid infusion device of an infusion system;

[0054] Figure 22 yes Figure 21 An exploded perspective view of a fluid infusion device;

[0055] Figure 23 It is along Figure 22 The line 23-23 in FIG. 23 shows the assembly of the reservoir inserted into the infusion set. Figures 21 to 22 A cross-sectional view of a fluid infusion device;

[0056] Figure 24 is a block diagram of an exemplary patient monitoring system; and

[0057] Figure 25 Depicts an embodiment of a computing device of a diabetes data management system adapted for use with a Figure 1 、 Figure 6 、 Figure 20 and Figure 24 Any one or more systems and Figures 2 to 3 、 Figure 8 、 Figure 11 、 Figure 13 and Figures 17 to 19 Any one or more processes can be used in combination. DETAILED DESCRIPTION

[0058] The following detailed description is merely illustrative in nature and is not intended to limit the embodiments of the subject matter or the application and uses of these embodiments. As used herein, the word "exemplary" means "serving as an example, instance, or illustration." Any specific implementation described herein as exemplary is not necessarily to be construed as preferred or advantageous over other specific implementations. Furthermore, no one wishes to be bound by any expressed or implied theory presented in the preceding technical field, background technology, summary of the invention, or the following detailed description.

[0059] For illustrative purposes, the subject matter herein may be described primarily in the context of infusion systems and devices configured to support monitoring and / or regulating glucose levels in a user in a personalized and / or context-sensitive manner. That is, the subject matter described herein is not necessarily limited to glucose regulation or insulin infusion and, in practice, may be implemented in an equivalent manner with respect to any number of other medications, physiological conditions, etc.

[0060] Although the subject matter described herein can be implemented in the context of any electronic device, the exemplary embodiments described below are implemented in conjunction with a medical device (such as a portable electronic medical device). Although many different applications are possible, the following description may primarily focus on a fluid infusion device (or infusion pump) as part of an infusion system deployment. For the sake of brevity, conventional techniques related to the operation of the infusion system, the operation of an insulin pump and / or an infuser, and other functional aspects of the system (and the individual operating components of the system) may not be described in detail herein. Examples of infusion pumps may be of the type described in, but not limited to, U.S. Patents 4,562,751, 4,685,903, 5,080,653, 5,505,709, 5,097,122, 6,485,465, 6,554,798, 6,558,320, 6,558,351, 6,641,533, 6,659,980, 6,752,787, 6,817,990, 6,932,584, and 7,621,893, each of which is incorporated herein by reference. A fluid infusion device typically includes a motor or other actuating device for linearly displacing a plunger (or stopper) of a reservoir disposed within the fluid infusion device to deliver a dose of fluid, such as insulin, to a user's body. In one or more exemplary embodiments, the delivery command (or dosage command) that controls the operation of the motor is determined in a substantially autonomous manner and on a substantially continuous basis based on a difference between a measured value of a physiological condition of the user's body and a target value using closed-loop control to adjust the measured value to the target value.

[0061] As follows Figures 1 to 5As described in more detail in the context of , in one or more embodiments, historical observation patient data (e.g., measurement data, insulin delivery data, event log data, background data, etc.), electronic medical record data, and medical insurance claims data associated with multiple different patients are stored or otherwise maintained in a database and organized into multiple different logical layers. Each logical layer has its own associated directed graph data structure that maintains associations or relationships between different entities within the logical layer. In this regard, an entity typically represents a container or logical grouping of fields, attributes, or other information that characterizes the entity. Therefore, an entity can maintain a logical association between one or more fields of a patient's historical observation data, a patient's electronic medical record data, and / or a patient's medical insurance claims data. For example, a patient identifier and one or more additional data fields associated with a single patient can be mapped to different entities within a specific logical database layer, which in turn serve as nodes within a directed graph data structure associated with the logical database layer, which are linked to other nodes (or entities) within the logical database layer. Thus, similarities or commonalities between different patients or entities within a logical database layer can be exploited to establish links between different patients, lifestyle events, treatment regimens, patient outcomes, etc., which in turn can be used to provide improved recommendations related to the management of a given patient's condition, or otherwise improve the control, regulation, or understanding of a given patient's condition. Similarly, in some embodiments, similarities, commonalities, or causal relationships can be used to establish links between an entity within one logical database layer and another entity in a different logical layer, thereby establishing links or edges that span logical database layers.

[0062] When the corresponding data for an entity is loaded, created, or otherwise instantiated in a database, links or edges between different nodes (or entities) may be initially created. For example, when a new patient is introduced into a database system, a corresponding entity for the patient may be created within the logical database layer for the patient. Thereafter, the logical database layer may be searched to identify other entities that are related or associated with some aspect of the new entity. For example, if the entity for the new patient includes an identifier for the patient's healthcare provider, a bidirectional link may be created to a node corresponding to an existing entity associated with the patient's healthcare provider (which may be in the same or different logical layer of the database).

[0063] In one or more exemplary embodiments, for each logical database layer, entities or nodes in a graph data structure associated with that layer are periodically analyzed to identify and create new causal or logical relationships between different nodes of the graph data structure. In one or more embodiments, a generated recursive neural network or other machine learning or artificial intelligence technology can periodically scan the nodes of the graph data structure to identify causal pairs and establish causal links (or edges) between such nodes of the graph data structure. For example, in response to a causal relationship engine employing machine learning to identify causal relationships based on a common sequence of events occurring relative to one or more patients, directed links can be created between entities corresponding to different types of meals and corresponding to different types of glucose excursion events (e.g., hyperglycemic events, hypoglycemic events, acute diabetic ketoacidosis, etc.). In one or more embodiments, the generated recursive neural network technology is applied by randomly starting from different outcome nodes of interest and tracing back to the links or edges of that node in a "random" manner to determine whether a particular pattern or sequence leads to that particular outcome node. Additionally, query logs associated with queries executed on or at a particular logical database layer can be analyzed to detect repeated associations or query paths involving at least a threshold number of nodes, thereby establishing new edges between the end nodes of the query paths to improve query performance. In some embodiments, new edges or relationships between entities or nodes in the graph data structure can also be manually established (e.g., based on new research, clinical evidence, data scraping and manual verification, or other external knowledge).

[0064] As follows Figures 3 to 5 As described in more detail in the context of , different logical database layers allow observed patient data, electronic medical record data, and health insurance claims data to be efficiently converted into different forms with different interrelationships between different data subsets, thereby adapting to different types of queries. In addition, the query results may be more personalized or otherwise produce better patient outcomes, recommendations, or understanding of the patient's physiological condition. For example, natural language processing or other artificial intelligence techniques can be applied to an input query or search string to determine the intent or target associated with the input query, and then based on this, one or more logical database layers are identified to perform a search based on the query intent. Then, a query statement is constructed and executed on the identified logical database layers to obtain results for the input query. In one or more embodiments, the initial query results are filtered or otherwise parsed based on information related to the current operating context (e.g., time of day, day of the week, geographic location, environmental conditions, etc.) to obtain context-relevant query results, which are then output or otherwise provided in response to the input query.

[0065] The following are mainly Figures 3 to 7As described in more detail in the context of [ 15 ], in one or more embodiments, a medical device (such as an infusion device, a sensing device, a monitoring device, etc.) includes or otherwise supports a user interface capable of receiving a conversational input query, which is then parsed or otherwise analyzed at the medical device to obtain an input query to be analyzed, thereby identifying a logical database layer for searching and generating a corresponding query statement. For example, in one or more embodiments, the medical device includes a microphone or similar audio input device suitable for receiving audio input from a user, which is then processed, parsed, or otherwise analyzed to identify the conversational input query in the audio input. The query results can be presented to the user or otherwise provided to the user in a conversational manner or otherwise within the context of a conversation or dialogue with the user within the user interface. Thus, a patient or user can interact conversationally with and query a database system, which can then be transformed to allow queries to be executed quickly at different logical layers and provide personalized and contextually relevant results, while also leveraging the interrelationships between different types of data and data subsets (e.g., different patients with similar overall demographic characteristics, different patients with similar medical histories, different patients with similar treatment regimens, etc.).

[0066] Diabetes Intelligence Network

[0067] Figure 1 An exemplary embodiment of a patient management system 100 is shown, which includes, but is not limited to, a computing device 102 coupled to a database 104, which is also communicatively coupled to one or more electronic devices 106 via a communication network 108 (e.g., the Internet, a cellular network, a wide area network (WAN), etc.). It should be understood that for purposes of illustration, Figure 1 A simplified representation of patient data management system 100 is shown and is not intended to limit the subject matter described herein in any way.

[0068] In an exemplary embodiment, the electronic device 106 includes one or more medical devices, such as an infusion device, a sensing device, a monitoring device, etc. In addition, the electronic device 106 may include any number of non-medical client electronic devices, such as a mobile phone, a smart phone, a tablet computer, a smart watch, or other similar mobile electronic devices, or any type of electronic device capable of communicating with the computing device 102 via the network 108, such as a laptop or notebook computer, a desktop computer, etc. One or more electronic devices 106 may include or be coupled to a display device (such as a monitor, a screen, or another conventional electronic display) that is capable of graphically presenting data and / or information related to the patient's physiological condition. In addition, one or more electronic devices 106 also include or are otherwise associated with a user input device (such as a keyboard, a mouse, a touch screen, a microphone, etc.) that is capable of receiving input data and / or other information from a user of the electronic device 106.

[0069] In an exemplary embodiment, one or more electronic devices 106 transmit, upload, or otherwise provide data or information to the computing device 102 for processing at the computing device 102 and / or storage in the database 104. For example, when the electronic device 106 is implemented as a sensing device, a monitoring device, or other device including a sensing element that is inserted into a patient or otherwise worn by the patient to obtain measurement data indicative of a physiological condition within the patient, the electronic device 106 may periodically upload or otherwise transmit the measurement data to the computing device 102. In other embodiments, when the electronic device 106 is implemented as an infusion device or similar device capable of delivering fluids or medications to a patient, the electronic device 106 may periodically upload or otherwise transmit delivery data indicating the timing and amount of the fluids or medications delivered to the patient. In still other embodiments, the client electronic device 106 may be utilized by the patient to manually define, enter, or otherwise record meals, activities, or other events experienced by the patient, and then transmit, upload, or otherwise provide these event log data to the computing device 102.

[0070] The computing device 102 generally represents a server or other remote device configured to receive data or other information from the electronic device 106, store or otherwise manage the data in the database 104, and analyze or otherwise monitor the data received from the electronic device 106 and / or stored in the database 104, as described in more detail below. In practice, the computing device 102 may reside in a physically distinct and / or separate location from the electronic device 106, for example, at a facility owned and / or operated by or affiliated with the manufacturer of one or more medical devices utilized in conjunction with the patient data management system 100. For purposes of illustration, but not limitation, the computing device 102 may alternatively be referred to herein as a server, a remote server, or variations thereof. The server 102 generally includes a processing system and a data storage element (or memory) capable of storing programming instructions for execution by the processing system, which programming instructions, when read and executed, cause the processing system to create, generate, or otherwise facilitate an application or software module configured to perform or otherwise support the processes, tasks, operations, and / or functions described herein. Depending on the embodiment, the processing system can be implemented using any suitable processing system and / or device, such as one or more processors, central processing units (CPUs), controllers, microprocessors, microcontrollers, processing cores and / or other hardware computing resources configured to support the operation of the processing system described herein. Similarly, the data storage element or memory can be implemented as random access memory (RAM), read-only memory (ROM), flash memory, magnetic or optical mass storage, or any other suitable non-transitory short-term or long-term data storage or other computer-readable media and / or any suitable combination thereof.

[0071] In an exemplary embodiment, database 104 is used to store or otherwise maintain historical observed patient data 120, electronic medical record data 122, and health insurance claims data 124 for a plurality of different patients. In this regard, a subset of patients having associated data in one of datasets 120, 122, 124 may also have associated data in another of datasets 120, 122, 124. That is, some, but not necessarily all, patients associated with one of datasets 120, 122, 124 may be common to another of datasets 120, 122, 124. In an exemplary embodiment, database 104 also stores or maintains metadata 126 for characterizing or otherwise defining a directed graph data structure corresponding to different logical layers within database 104. In this regard, graph metadata 126 may define the nodes (or entities) that make up the graph data structure associated with a particular logical database layer, wherein each of these nodes (or entities) is mapped to one or more fields of datasets 120, 122, 124. Additionally, the graph metadata 126 characterizes or defines edges or links between nodes within a graph data structure associated with a particular logical database layer that establish logical or causal relationships between nodes within that logical database layer. In various embodiments, a node (or entity) may exist in multiple different logical database layers, or a node (or entity) in one logical database layer may link to another node (or entity) in a different logical database layer.

[0072] In the illustrated embodiment, the server 102 implements or otherwise executes a data management application 110 that receives or otherwise obtains data from the electronic device 106, stores the received data in the database 104, and generates or otherwise creates entities that logically associate the different fields of the stored data 120, 122, 124. The data management application 110 also generates or otherwise creates graph metadata 126 to maintain relationships between different entities in the database 104, as described below. Figures 2 to 5 In the illustrated embodiment, the server 102 further implements or otherwise executes a query management application 112 that receives or otherwise obtains input queries from one or more electronic devices 106 and, in response to the corresponding input queries, generates, executes, or otherwise performs corresponding query statements on one or more different logical layers of the database 104 to obtain results provided to the corresponding electronic devices 106, as described below in Figures 2 to 5 described in more detail in the context of .

[0073] Still refer to Figure 1In an exemplary embodiment, the historical observation data 120 maintained in the database 104 includes historical measurement data associated with a particular patient (or patient identifier) indicating the patient's physiological condition with respect to time (e.g., historical blood glucose values, historical interstitial fluid glucose values, etc.), historical delivery data indicating fluid or medication doses delivered to the patient with respect to time (e.g., historical meal or correction boluses, basal doses, or other automatically delivered amounts, etc.), historical meal data and / or other event log data associated with the patient, historical contextual data related to the measurement data, delivery data, and event log data, etc. For example, the server 102 may receive measurement data values associated with a particular patient (e.g., sensor glucose measurements, acceleration measurements, etc.) obtained using a sensing element from a medical device via the network 108, and the server 102 may store or otherwise maintain the historical measurement data in the database 104 associated with the patient as patient data 120 (e.g., using one or more unique patient identifiers). In addition, the server 102 may also receive meal data or other event log data from or via the client device 106 that may be input by the patient or otherwise provided (e.g., via a client application at the client device 106), and store or otherwise maintain historical meal data and other historical event or activity data associated with the patient in the database 104. In this regard, meal data includes, for example, a time or timestamp associated with a particular meal event, a meal type or other information indicating the composition or nutritional characteristics of the meal, and an indication of the size associated with the meal. In an exemplary embodiment, the server 102 also receives historical fluid delivery data (e.g., insulin delivery dose amounts and corresponding timestamps) corresponding to basal or bolus doses of fluid delivered to the patient by the infusion device 106. The server 102 may also receive geographic location data and possibly other contextual data associated with the electronic device 106 that provided the patient data 120, and store or otherwise maintain historical operational contextual data associated with the particular patient. In this regard, one or more of the devices 106 may include a Global Positioning System (GPS) receiver, or similar module, component, or circuitry capable of outputting or otherwise providing data representative of the geographic location of the respective device 106 in real time.

[0074] Electronic medical record (EMR) data 122 typically includes information associated with one or more identifiers for a given patient within an EMR dataset that indicates the following: medical diagnoses or conditions the patient has been diagnosed with, medications or drugs administered or taken by the patient, prescription information, changes in therapy for the patient, laboratory results or measurements of the patient's physiological condition, the patient's immunization records, microbiology results or other observations related to the patient, healthcare utilization information (e.g., hospitalizations, emergency room visits, outpatient visits, etc.), overall demographic information associated with the patient (e.g., age, income, education, location, gender), past medical procedures, clinical observations or other habitual behavioral information (e.g., smoking, alcohol consumption, etc.), family medical history, physician notes and care plans, etc. The EMR data 122 may also include data about healthcare providers associated with various aspects of the patient's medical record, the patient's insurance information, etc. In various embodiments, the server 102 may receive or obtain the EMR data 122 from another server computing device, another database different from the database 104 (e.g., by copying from another database), individual computing devices associated with healthcare providers, patients, etc. The claims data 124 generally includes information related to health insurance claims submitted by or on behalf of a patient, associated with one or more identifiers of a given patient within a claims dataset, including charge information, prescriptions filled or refilled by the patient, and the like. Similar to the EMR data 122, the server 102 may receive or obtain the claims data 124 from another server computing device, another database different from the database 104 (e.g., by copying from another database), various computing devices associated with healthcare providers, patients, pharmacies, and the like. In an exemplary embodiment, the claims data 124 includes medical, pharmaceutical, and obstetric-related claims data, including corresponding diagnoses, procedures, prescription codes, charges (e.g., net amount plus allowable amount, etc.), and the like.

[0075] Figure 2 Describes a method suitable for use by a computing device such as Figure 1 The exemplary data management process 200 implemented by the server 102 in the patient data management system 100 of FIG. The various tasks performed in conjunction with the data management process 200 can be performed by hardware, firmware, software executed by processing circuitry, or any combination thereof. For illustrative purposes, the following description relates to the above-mentioned Figure 1elements mentioned. In practice, portions of the data management process 200 may be performed by different elements of the patient data management system 100; however, for purposes of explanation, the data management process 200 may be described primarily in the context of implementation at or by the server 102 and / or data management application 110. It should be understood that the data management process 200 may include any number of additional or alternative tasks, which need not be performed in the order shown and / or which may be performed simultaneously, and / or the data management process 200 may be incorporated into a more comprehensive program or process having additional functionality not described in detail herein. Furthermore, portions of the data management process 200 may be omitted from an actual implementation of the data management process 200 so long as the intended overall functionality remains unchanged. Figure 2 One or more tasks shown and described in the context of.

[0076] In an exemplary embodiment, the data management process 200 is executed, facilitated, or otherwise supported by the data management application 110 at the server 102 to generate graph metadata 126 for the different logical layers supported by the database 104. For example, in one embodiment where the database system 104 maintains data 120, 122, 124 related to diabetic patients, the database 104 supports five different logical layers: a patient layer, a lifestyle layer, a treatment layer, a diabetes management layer, and a diabetes knowledge layer. The patient layer contains a subset of the patient data 120 and EMR data 122 associated with each patient, including, but not limited to, historical patient blood glucose measurements, information characterizing historical glucose excursion events, and information characterizing complications or improvements to the respective patient's physiological condition. In this regard, the graph metadata 126 can indicate which fields of the patient data 120 and EMR data 122 associated with an individual patient should be mapped or otherwise used for nodes of the patient layer graph data structure, as well as corresponding edges or links between those nodes. The patient layer can be queried to obtain information related to the patient's health history, such as glucose measurements, excursion events, year-over-year improvements, comorbidities, complications, and the like. The lifestyle layer includes event log data and other possible subsets of patient data 120 and EMR data 122 related to the respective individual patients. The treatment layer includes subsets of data 120, 122, 124 that indicate which medications, drugs, or other treatments are associated with the respective patients and may include, for example, indications of the type of medical device the patient may be using to manage or monitor their treatment (e.g., infusion device, continuous glucose monitoring device, etc.) and the costs associated with the patient's treatment. The diabetes management layer includes subsets of data 120, 122, 124 that support maintaining relationships between the different individuals or entities represented within the datasets 120, 122, 124, including patients, healthcare providers, physicians, payers, hospitals, etc. The diabetes knowledge layer contains subsets of data 120, 122, 124 that support queries for patient-independent general knowledge, such as queries related to specific physiological conditions or diagnoses (e.g., type 1 diabetes, type 2 diabetes, etc.), the pharmacodynamics of insulin or other fluids, medications, or drugs, excursion events, meal types, etc.

[0077] For each logical database layer, the data management process 200 shown periodically scans or otherwise analyzes the nodes or entities within the graph data structure associated with the corresponding logical database layer to identify causal relationships between the entities within the logical database layer (tasks 202, 204). In one or more embodiments, the data management application 110 implements or otherwise performs a machine learning-based causal relationship analysis to discover repeated causal pairings of nodes within the graph data structure. For example, a timestamp or other temporal relationship between a meal event entity and a glucose excursion event entity associated with a particular patient or across multiple different patients can be used to identify a causal relationship between the meal event and the glucose excursion event, and establish a causal link between the meal event and glucose excursion event entities in the lifestyle layer. In response to discovering a relationship between previously unconnected nodes or entities within the graph data structure associated with the corresponding logical database layer, the data management process 200 creates or otherwise generates updated graph metadata representing the relationship between the identified nodes, and stores or otherwise maintains the updated graph metadata in a database associated with the logical database layer (tasks 206, 208). In this regard, the data management application 110 updates the graph metadata 126 associated with a particular logical database layer to create new directed edges or links between previously unconnected nodes or entities identified as having a causal relationship within that logical database layer.

[0078] As an example, the data management application 110 may detect a pattern in which meals with a fat content exceeding a threshold value (e.g., exceeding 50 grams) result in a hyperglycemic excursion event lasting longer than a threshold duration (e.g., exceeding 45 minutes), and thereby establish directed links or edges between one or more meal event nodes with a fat content exceeding the threshold value and corresponding hyperglycemic excursion outcome nodes. The newly created edges may be assigned weights or other quantitative values that correspond to or otherwise reflect the strength of the relationship between the nodes (e.g., based on a probabilistic analysis of the incidence of the outcome). As another example, it may be determined that for a particular group of patients sharing certain characteristics, exercising more than a threshold number of times per week (e.g., three or more times per week) results in increased insulin sensitivity, and thereby establish directed links or edges between certain exercise event nodes and an increased insulin sensitivity outcome node, where the weights correspond to the relative probabilities that the respective exercise events result in increased insulin sensitivity.

[0079] Still refer to Figure 2In an exemplary embodiment, the data management process 200 also analyzes a query log associated with a corresponding logical database layer to identify relationships between previously unconnected nodes or entities within the logical database layer based on results of previously executed queries. In this regard, the database 104 may store or otherwise maintain a query log, wherein, in response to executing a query statement, a corresponding log entry is created that maintains an association between the logical database layer being queried and the query path resulting from executing the query statement (e.g., the order of nodes and edges traversed in the logical database layer during execution of the query statement).

[0080] In an exemplary embodiment, for each logical database tier, the illustrated data management process 200 periodically analyzes the query log associated with the logical database tier to identify logical relationships between nodes or entities within the logical database tier based on repeated queries that traverse those nodes or entities (tasks 210, 212). For example, in one or more embodiments, the data management application 110 analyzes the query log associated with a particular logical database tier to identify or retrieve query paths that traverse more than a threshold number of nodes or entities within the database tier (e.g., more than three nodes). Of the subset of query paths that traverse more than the threshold number of nodes, the data management application 110 identifies repeated query paths that have common end nodes and establishes logical relationships between those end nodes within the logical database tier. In this regard, the data management process 200 creates or otherwise generates updated graph metadata that establishes logical relationships between previously unconnected end nodes and stores or otherwise maintains the updated graph metadata in a database associated with the logical database tier (tasks 214, 216). In this manner, the data management application 110 creates new edges or links between end nodes of repeated query paths that have a common end node and traverse more than a threshold number of nodes.

[0081] By creating or otherwise establishing relationships between previously unconnected nodes within the graph data structure of a particular logical database tier via the data management process 200, subsequent queries of that logical database tier may be conducted or executed more efficiently, or otherwise provide improved results reflecting possible causal and / or logical relationships between the nodes of the graph data structure. In an exemplary embodiment, the data management process 200 is performed for each different logical database tier, and the data management process 200 may be repeated periodically (e.g., daily, weekly, monthly, etc.) to continuously analyze and update the relationships between nodes or entities within the corresponding logical database tier.

[0082] Figure 3 Describes a database suitable for querying multiple different logical layers (such as Figure 1104 in the patient data management system 100). For illustrative purposes, the query process 300 may be described herein primarily in the context of input queries received in a conversational format from a human user or patient. However, it should be understood that the query process 300 is not limited to conversational input queries received from a user, and that the query process 300 may be implemented in an equivalent manner for queries that are neither submitted nor initiated by a user and that are not provided in a conversational format. The various tasks performed in conjunction with the query process 300 may be performed by hardware, firmware, software executed by processing circuitry, or any combination thereof. For illustrative purposes, the following description relates to the query process 300 described above in conjunction with Figure 1 elements mentioned. In practice, portions of the query process 300 may be performed by different elements of the patient data management system 100; however, for purposes of explanation, the query process 300 may be described primarily in the context of implementation at or by the server 102 and / or the query management application 112. It should be understood that the query process 300 may include any number of additional or alternative tasks, which need not be performed in the order illustrated and / or which may be performed simultaneously, and / or the query process 300 may be incorporated into a more comprehensive program or process having additional functionality not described in detail herein. Furthermore, the query process 300 may be omitted from an actual implementation of the query process 300 so long as the intended overall functionality remains unchanged. Figure 3 One or more tasks shown and described in the context of.

[0083] In the embodiment shown, the query process 300 receives or otherwise obtains an input query from a patient or other user (task 302). For example, the patient can interact with or otherwise manipulate a user interface associated with a client application on an electronic device 106 to create an input query, which is then transmitted or otherwise provided to the server 102 via the network 108. In one or more embodiments, the input query is implemented as a conversational string of words or text provided to the server 102. In this regard, the user can use natural language rather than predefined grammar to create or provide the input query in a free form or unstructured manner. For example, in one or more embodiments, the electronic device 106 includes an audio input device and a speech recognition engine or vocabulary that supports parsing or otherwise decomposing the conversational speech or audio input by the user of the device 106 into a corresponding text representation to be provided to the server 102. In various embodiments, the conversational input query can be received silently, or alternatively, the user can manipulate the device 106 to select or otherwise activate a graphical user interface (GUI) element that enables or initiates the query process 300. For example, in one or more embodiments, the query process 300 can be initiated in response to a user selecting a GUI element for a search feature, digital assistant, or similar feature supported by a client application at device 106. In response, the client application at device 106 can generate or otherwise provide a GUI display or other GUI element, prompting the user to indicate what he or she wants to know or ask. Thereafter, the user can enter a conversational string (e.g., via voice, typing, swiping, touch, or any other suitable input method), where a textual representation of the conversational input query is provided by device 106 to server 102 via network 108.

[0084] In an exemplary embodiment, the query process 300 also receives or otherwise obtains context information associated with the input query from the client electronic device that provided the input query (task 304). The operational context information provided with the input query characterizes the current operational state or environment at the time the query was entered. For example, in association with the submitted input query, the client device 106 may also transmit or otherwise provide context information related to the operation of the client device 106, such as the current location of the client device 106, the current local time and current day of the week at the location of the client device 106, the current environmental conditions at the location of the client device 106, and the like. In addition, in some embodiments, the client device 106 may also provide information indicating the current physiological condition of the user and / or the current operational state of an infusion device or other medical device associated with the user. For example, in conjunction with the input query, the client device 106 may transmit or otherwise provide one or more of the following: a current or recent glucose measurement associated with the user, an indication of whether to pause insulin delivery by an infusion device associated with the user, a current or recent insulin delivery to the user, a current or recent heart rate measurement associated with the user, an acceleration measurement, or other measurement of activity level associated with the user, and the like.

[0085] The illustrated query process 300 continues by identifying or otherwise determining which logical database tier or tiers should be queried based at least in part on the input query, and then generating or otherwise constructing one or more corresponding query statements to be executed on the identified logical database tiers based at least in part on the input query (tasks 306, 308). In this regard, the query management application 112 at the server 102 may analyze the input query to identify or otherwise determine the likely intent or purpose of the query, and then determine the logical database tier to be queried based on the intent or purpose of the query. In some embodiments, when determining which logical database tier should be queried, the query management application 112 at the server 102 may also analyze contextual information associated with the input query as well as the content of the input query. Once the logical tier to be queried is identified, the query management application 112 at the server 102 analyzes the content of the input query to obtain parameters or criteria to be used for the query, and then uses these parameters or criteria to generate or otherwise construct query statements for execution on the identified logical database tier. Additionally, in some embodiments, the query management application 112 at the server 102 may also utilize operational context information associated with the input query for one or more parameters or criteria when constructing the query statement.

[0086] After constructing the query statement, the query process 300 executes or otherwise initiates execution of the constructed query statement on the identified logical database tier to obtain results for the input query from the identified logical database tier (task 310). In this regard, when the constructed query statements are linked or dependent on each other, the query management application 112 at the server 102 can initiate a first query statement on a first logical tier of the database 104 to obtain results for the intermediate query statement, which in turn is utilized by a second query statement executed on a different logical tier of the database 104. For example, the results obtained from querying one logical database tier can be used as parameters or criteria in subsequent query statements on a different logical database tier. Additionally, in some embodiments, the results obtained from querying one logical database tier can be filtered, processed, analyzed, or otherwise optimized to determine parameters or criteria for use in subsequent query statements on a different logical database tier, as described below. Figure 4 described in more detail in the context of .

[0087] To execute a query statement, the database 104 utilizes the graph metadata 126 to traverse the nodes or entities within the logical database layer of the query to obtain the results of the query statement based on the established edges or links between the nodes or entities within the logical database layer. It should be noted that, by virtue of the weighted directed graph data structure used to maintain the data in the database 104, by supporting point-based index references that do not require complex table scan sequences, the response time for executing a query statement at the database 104 is generally less than that of a conventional database that relies on table scans based on primary keys and / or foreign keys. In an exemplary embodiment, as described above, the query path that details the nodes or entities traversed during the execution of the query statement and the corresponding edges is also stored or otherwise maintained in a query log at the server 102 or one of the databases 104 associated with the logical database layer of the query.

[0088] In one or more exemplary embodiments, the query process 300 filters the initial query results based on the operational context information associated with the input query, and then generates output query results based on the filtered query results in response to the received input query (task 312). In this regard, the initial query results can be analyzed relative to the current operational context associated with the input query to select or otherwise identify a subset of the initial query results that are most relevant to one or more aspects of the current operational context. The query management application 112 at the server 102 can select from the initial query results obtained by executing the query statement: a subset of information that is most likely relevant to the current location of the client device 106, the current local time at the location of the client device 106, the day of the week, the current environmental conditions at the location of the client device 106, the current physiological condition of the patient, the current operational state of an infusion device or other medical device, etc. For example, the query management application 112 can select one of the initial query results that is closest to the current location of the client device 106 or within a threshold distance of the current location. As another example, the query management application 112 may select one of the initial query results that is most likely to produce the best patient outcome based on the patient's current glucose level, the current operating state of the patient's infusion device (e.g., delivery paused, reservoir depleted, etc.).

[0089] The query process 300 generates or otherwise constructs a response to the input query based on the filtered query results, and then presents or otherwise provides the query response in response to the input query (tasks 314, 316). For example, in one or more embodiments, the query management application 112 uses the filtered query results to generate a session query response, and then transmits the session query response to the query client device 106 for presentation or reproduction in the context of the session that includes the session input query.

[0090] Figure 4 Describes the combinable Figure 3 The query process 300 in Figure 11 . An exemplary embodiment of a graphical user interface (GUI) display 400 presented at a query client device 406 (e.g., one of the devices 106) in the patient data management system 100 of FIG. The GUI display 400 includes a conversational dialog box or one or more similar GUI elements that prompt a user to conversationally interact with the client device 106, 406 to query the database 104. In the illustrated embodiment, a patient using the query client device 106, 406 utilizes an input device or user interface at the client device 106, 406 to enter or otherwise provide a conversational input query. In response, an application at the client device 106, 406 updates the GUI display 400 to graphically depict a textual representation of the conversational input query 402 received by the client device 106, 406. The application at the client device 106, 406 transmits, submits, or otherwise provides the conversational input query text to the query management application 112 at the server 102 for execution.

[0091] As above Figure 3 As described in the context of , the query management application 112 analyzes the conversational input query text "What should I eat now?" to determine the intent or purpose of the input query (e.g., intent = finding food), the subject of the input query (e.g., subject = patient identifier), and any other time or context parameters contained in or associated with the input query (e.g., time = now). Based on identifying the subject of the input query as the patient, the query management application 112 can determine that the lifestyle logic layer of the database 104 should be queried to obtain lifestyle information related to the patient. In this regard, the query management application 112 can construct an initial query statement for querying the lifestyle logic layer of the database 104 using the patient's unique identifier in order to retrieve lifestyle information related to the patient. For example, executing a query on the lifestyle logic layer of the database 104 can return information indicating the current or recent type of diet being consumed by the patient (e.g., low carbohydrate), information related to the patient's exercise habits or other recent activities performed by the patient, and / or other contextual information that may describe the patient's lifestyle. Additionally, based on identifying the subject of the input query as a patient, the query management application 112 may also query the patient logical layer of the database 104 to identify other patients that are similar to the patient that is the subject of the input query (e.g., based on edges or links in the graph metadata 126 for the patient logical layer that link those other patients to the current patient via more than a threshold number of common nodes or entities).

[0092] Using the lifestyle information obtained by querying the lifestyle logic layer and identifiers of other patients similar to the subject patient, the query management application 112 generates or otherwise constructs a query statement for querying the patient logic layer to obtain meal logs or other meal information associated with a subset of similar patients who have similar lifestyle information associated with them (e.g., patients with similar exercise or activity behaviors, geographic locations, etc.) and whose meal outcomes are favorable (e.g., no hypoglycemic or hyperglycemic events or other excursions after meals for patients with similar lifestyle contexts, post-meal glucose levels within a threshold range of the patient's target glucose value, etc.). After obtaining information from the patient logic layer about meals consumed by similar patients with favorable outcomes, the query management application 112 may generate or otherwise construct a query statement for querying the lifestyle logic layer to identify a subset of those meals that best match or are most closely associated with the patient's lifestyle information (e.g., meals associated with a low-carbohydrate diet or other meals related to a low-carbohydrate diet, etc.). In one or more exemplary embodiments, the query statement may also take into account the patient's current operating context (e.g., meals within a threshold distance of the current location of the client device 106, 406, etc.). In still other embodiments, the current operating context is used to filter or otherwise exclude query results and identify meal results that best match the querying patient's current operating context and lifestyle.

[0093] After obtaining the meal results that best match the patient's lifestyle and current operational context from the query lifestyle logic layer and achieving positive outcomes for one or more similar patients, the query management application 112 generates or otherwise constructs a query response that includes or incorporates the meal results in a conversational format. In this regard, in one or more embodiments, based on the identified meal type and the current location associated with the input query, the query management application 112 queries a database of restaurant information, including geographic location information and menu data associated with a plurality of restaurants, to identify restaurants that are closest to or near the current location of the querying device 106, 406 and offer items that match or correspond to the identified meal. The query management application 112 can then generate query response text that indicates the identified restaurant and menu item that best matches the meal results obtained from the query database 104. The query management application 112 transmits or otherwise provides the conversational query response text to the querying client device 106, 406 for presentation at the client device 106, 406. The application at the client device 106 , 406 updates the GUI display 400 to graphically depict a textual representation of a session query response 404 received by the client device 106 , 406 from the server 102 in the context of a session including the session input query 402 .

[0094] Figure 5 An exemplary graphical representation of a portion of a graph data structure 500 corresponding to a subset of the patient logical layer in the database 104 is depicted, the graphical representation depicting relationships between different patients that may be related to the query subject patient based on the graph metadata 126. In this regard, Figure 5 A node 502 within the patient logic layer associated with an inquiry patient is depicted. Based on the graph metadata 126, the inquiry patient node 502 is associated with a plurality of different entity nodes 504 within the patient logic layer, wherein those entity nodes 504 correspond to different fields or subsets of the data 120, 122, 124 in the database 104 that are associated with the inquiry patient and that are mapped to the respective nodes based on the graph metadata 126 of the patient logic layer. Additionally, the graph metadata 126 of the patient logic layer may further define edges or links between the entity node 504 associated with the inquiry patient 502 and one or more other patient nodes 506 associated with different patients that have similar values for their related fields or subsets of the data 120, 122, 124 in the database 104 that are mapped to those entity nodes 504. In some embodiments, weights may be assigned to edges between an entity node 504 and similar patient nodes 506 based on differences or similarities between the values of associated fields or subsets of the query patient's data 120, 122, 124 that map to that node 504 and the values of those fields or subsets of the data 120, 122, 124 associated with the respective patients whose patient nodes 506 are linked to the respective entity node 504. Thus, the number of edges between respective pairs of related patient nodes 502, 506 and the respective weights assigned to those edges may be used to calculate or otherwise determine a metric indicative of the relative similarity between a query subject associated with a patient node 502 and a different patient associated with one of the patient nodes 506.

[0095] As above Figure 4 As described in the context of , when a query statement is executed on the patient logical layer, the graph metadata 126 associated with the patient logical layer in the database 104 can be utilized to obtain patient identifiers associated with the patient nodes 506 that are most similar to the query subject (e.g., based on a similarity metric characterizing the relationship between the corresponding pair of patient nodes 502, 506). The patient identifiers associated with the similar patient nodes 506 can then be used to query other logical layers of the database 104 to obtain information indicating the meals, activities, medications, therapies, etc. of these patients and how such variables affect the physiological conditions of these patients (e.g., glucose measurements, excursion events, etc.), so as to generate recommendations or otherwise provide query results that are most likely to achieve the best outcome with respect to the physiological conditions of the query subject patients.

[0096] In some embodiments, the nodes 504 may reside in a different logical layer of the database 104 than the patient nodes 502, 506, where corresponding pairs of patient nodes 502, 506 have at least a threshold number of nodes 504 that are common or shared within another logical layer, which are used to establish a relationship between the corresponding pairs of patient nodes 502, 506, or otherwise classify the pairs of patient nodes 502, 506 into a common group or cohort, as described in more detail below.

[0097] cognitive pump

[0098] Now refer to Figure 6 According to one or more exemplary embodiments, the infusion device 602 in the infusion system 600 is used as a Figure 1 In this regard, the infusion device 602 is capable of receiving session user input and capturing simultaneous, concurrent, or temporally related operational context information associated with the session user input, and responding accordingly. Figure 3 The query process 300 provides the corresponding session input query text and associated context information to the query management application 112 at the server 102 .

[0099] In exemplary embodiments, the infusion system 600 is also capable of automatically or autonomously controlling or otherwise regulating a physiological condition within the user's body 601 to a desired (or target) value, or otherwise maintaining the condition within a range of acceptable values. In one or more exemplary embodiments, the regulated condition is sensed, detected, measured, or otherwise quantified by a sensing device 604 (e.g., sensing device 604) that is communicatively coupled to the infusion device 602. However, it should be noted that in alternative embodiments, the condition regulated by the infusion system 600 may be related to a measurement obtained by the sensing device 604. That is, for clarity and illustration purposes, the subject matter may be described herein in the context of a glucose sensing device in which the sensing device 604 is implemented as a glucose sensing device that senses, detects, measures, or otherwise quantifies a user's glucose level, which is regulated in the user's body 601 by the infusion system 600.

[0100] In an exemplary embodiment, the sensing device 604 includes one or more interstitial glucose sensing elements that generate or otherwise output an electrical signal (alternatively referred to herein as a measurement signal) having a signal characteristic that is related to, affected by, or otherwise indicative of the relative interstitial fluid glucose level in the user's body 601. The output electrical signal is filtered or otherwise processed to obtain a measurement value indicative of the user's interstitial fluid glucose level. In an exemplary embodiment, a blood glucose meter 630, such as a finger-prick device, is used to directly sense, detect, measure, or otherwise quantify blood glucose in the user's body 601. In this regard, the blood glucose meter 630 outputs or otherwise provides a measured blood glucose value that can be used as a reference measurement result for calibrating the sensing device 604 and converting the measured value indicative of the user's interstitial fluid glucose level into a corresponding calibrated blood glucose value. For illustrative purposes, the calibrated blood glucose value calculated based on the electrical signal output by one or more sensing elements of the sensing device 604 may alternatively be referred to herein as a sensor glucose value, a sensed glucose value, or variations thereof.

[0101] In an exemplary embodiment, the infusion system 600 further includes one or more additional sensing devices 606, 608 configured to sense, detect, measure, or otherwise quantify characteristics of the user's body 601 that are indicative of a condition of the user's body 601. In this regard, in addition to the glucose sensing device 604, one or more additional sensing devices 606 may be worn, carried, or otherwise associated with the user's body 601 to measure characteristics or conditions of the user (or the user's activity) that may affect the user's blood glucose level or insulin sensitivity. For example, a heart rate sensing device 606 may be worn on or otherwise associated with the user's body 601 to sense, detect, measure, or otherwise quantify the user's heart rate, which in turn may indicate exercise (and its intensity) that may affect glucose levels or insulin response within the user's body 601. In yet another embodiment, another invasive, interstitial, or subcutaneous sensing device 606 may be inserted into the user's body 601 to obtain measurements of another physiological condition that may be indicative of exercise (and its intensity), such as, for example, a lactate sensor, a ketone sensor, or the like. Depending on the embodiment, the auxiliary sensing device 606 may be implemented as a separate component worn by the user, or alternatively, the auxiliary sensing device 606 may be integrated with the infusion device 602 or the glucose sensing device 604.

[0102] The illustrated infusion system 600 also includes an acceleration sensing device 608 (or accelerometer) that can be worn on or otherwise associated with the user's body 601 to sense, detect, measure, or otherwise quantify the acceleration of the user's body 601, which in turn can indicate movement or some other condition of the body 601 that may affect the user's insulin response. Figure 6 608 is shown as being integrated into the infusion device 602, but in alternative embodiments, the acceleration sensing device 608 may be integrated with other sensing devices 604, 606 on the user's body 601, or the acceleration sensing device 608 may be implemented as a separate independent component worn by the user.

[0103] In an exemplary embodiment, the infusion device 602 further includes one or more environmental sensing devices 650 to sense, detect, measure, or otherwise quantify the current operating environment surrounding the infusion device 602. In this regard, the environmental sensing devices 650 may include one or more of a temperature sensing device (or thermometer), a humidity sensing device, a pressure sensing device (or barometer), etc. In an exemplary embodiment, the infusion device 602 further includes a location sensing device 660, such as a global positioning system (GPS) receiver, to sense, detect, measure, or otherwise quantify the current geographic location of the infusion device 602.

[0104] In the illustrated embodiment, the pump control system 620 generally represents the electronics and other components of the infusion device 602 that control the operation of the fluid infusion device 602 according to a desired infusion delivery program in a manner influenced by a sensed glucose value indicative of the current glucose level in the user's body 601. For example, to support a closed-loop operating mode, the pump control system 620 maintains, receives, or otherwise obtains a target or commanded glucose value and automatically generates or otherwise determines a dose command for operating an actuator, such as a motor 632, to displace the plunger 617 and deliver insulin to the user's body 601 based on the difference between the sensed glucose value and the target glucose value. In other operating modes, the pump control system 620 may generate or otherwise determine a dose command configured to maintain the sensed glucose value below an upper glucose limit, above a lower glucose limit, or other value within a desired range of glucose values. In practice, the infusion device 602 may store or otherwise maintain a target value, one or more upper and / or lower glucose limits, one or more insulin delivery limits, and / or one or more other glucose thresholds in a data storage element accessible to the pump control system 620.

[0105] Still refer to Figure 6, target glucose values and other threshold glucose values utilized by the pump control system 620 may be received from an external component or may be input by a user via a user interface element 640 associated with the infusion device 602. In practice, the one or more user interface elements 640 associated with the infusion device 602 typically include at least one input user interface element, such as, for example, a button, a keypad, a keyboard, a knob, a joystick, a mouse, a touch panel, a touch screen, a microphone or other audio input device, or the like. In addition, the one or more user interface elements 640 include at least one output user interface element for providing notifications or other information to the user, such as, for example, a display element (e.g., a light emitting diode, etc.), a display device (e.g., a liquid crystal display, etc.), a speaker or other audio output device, a tactile feedback device, or the like. It should be noted that although Figure 6 The one or more user interface elements 640 are shown as being separate from the infusion device 602, but in practice, the one or more user interface elements 640 can be integrated with the infusion device 602. Furthermore, in some embodiments, in addition to and / or in lieu of the one or more user interface elements 640 being integrated with the infusion device 602, the one or more user interface elements 640 are also integrated with the sensing device 604. A user can manipulate the one or more user interface elements 640 as needed to operate the infusion device 602 to deliver a correction bolus, adjust a target value and / or threshold, modify a delivery control scheme or operating mode, and the like.

[0106] Still refer to Figure 6 In the illustrated embodiment, the infusion device 602 includes a motor control module 612 coupled to a motor 632 that is operable to displace a plunger 617 in a reservoir and provide a desired amount of fluid to a user's body 601. In this regard, displacement of the plunger 617 causes a fluid capable of affecting a physiological condition of the user (such as insulin) to be delivered to the user's body 601 via a fluid delivery path (e.g., via tubing of an infusion set). A motor driver module 614 is coupled between an energy source 618 and the motor 632. The motor control module 612 is coupled to the motor driver module 614 and generates or otherwise provides command signals that operate the motor driver module 614 to provide current (or power) from the energy source 618 to the motor 632 to displace the plunger 617 in response to receiving a dose instruction from the pump control system 620 indicating a desired amount of fluid to be delivered.

[0107] In an exemplary embodiment, the energy source 618 is implemented as a battery housed within the infusion set 602 that provides direct current (DC) power. In this regard, the motor driver module 614 generally represents a combination of circuitry, hardware, and / or other electrical components configured to convert or otherwise transform the DC power provided by the energy source 618 into an alternating current (AC) signal that is applied to the phases of the stator windings of the motor 632, causing current to flow through the stator windings, thereby generating a stator magnetic field and causing the rotor of the motor 632 to rotate. The motor control module 612 is configured to receive or otherwise obtain a commanded dose from the pump control system 620, convert the commanded dose into a commanded translational displacement of the plunger 617, and command, signal, or otherwise operate the motor driver module 614 to rotate the rotor of the motor 632 by an amount that produces the commanded translational displacement of the plunger 617. For example, the motor control module 612 may determine the amount of rotor rotation required to produce a translational displacement of the plunger 617 that achieves the commanded dose received from the pump control system 620. Based on the current rotational position (or orientation) of the stator relative to the rotor as indicated by the output of the rotor sensing device 616, the motor control module 612 determines the appropriate sequence of AC electrical signals to be applied to the phases of the stator windings that should cause the rotor to rotate a determined amount relative to its current position (or orientation). In embodiments where the motor 632 is implemented as a BLDC motor, the AC electrical signals commutate the phases of the stator windings at the appropriate orientation of the rotor poles relative to the stator and in the appropriate sequence to provide a rotating stator magnetic field that causes the rotor to rotate in the desired direction. The motor control module 612 then operates the motor driver module 614 to apply the determined AC electrical signals (e.g., command signals) to the stator windings of the motor 632 to achieve the desired fluid delivery to the user.

[0108] When the motor control module 612 is operating the motor driver module 614, current flows from the energy source 618 through the stator windings of the motor 632 to generate a stator magnetic field that interacts with the rotor magnetic field. In some embodiments, after the motor control module 612 operates the motor driver module 614 and / or the motor 632 to achieve the commanded dose, the motor control module 612 stops operating the motor driver module 614 and / or the motor 632 until a subsequent dose command is received. In this regard, the motor driver module 614 and the motor 632 enter an idle state, during which the motor driver module 614 effectively disconnects or separates the stator windings of the motor 632 from the energy source 618. In other words, when the motor 632 is idle, current does not flow from the energy source 618 through the stator windings of the motor 632, and therefore the motor 632 does not consume power from the energy source 618 in the idle state, thereby improving efficiency.

[0109] Depending on the embodiment, the motor control module 612 may be implemented or realized using a general purpose processor, microprocessor, controller, microcontroller, state machine, content addressable memory, application specific integrated circuit, field programmable gate array, any suitable programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. In an exemplary embodiment, the motor control module 612 includes or otherwise has access to a data storage element or memory, including any type of random access memory (RAM), read-only memory (ROM), flash memory, registers, a hard disk, a removable disk, magnetic or optical mass storage, or any other short-term or long-term storage medium or other non-transitory computer-readable medium capable of storing programming instructions for execution by the motor control module 612. When read and executed by the motor control module 612, the computer-executable programming instructions cause the motor control module 612 to perform or otherwise support the tasks, operations, functions, and processes described herein.

[0110] It should be understood that the subject matter described herein is for illustrative purposes only and is not intended to limit the subject matter described in any way. Figure 6 is a simplified representation of the infusion device 602. In this regard, depending on the embodiment, some features and / or functions of the sensing device 604 may be implemented by or otherwise integrated into the pump control system 620, or vice versa. Similarly, in practice, features and / or functions of the motor control module 612 may be implemented by or otherwise integrated into the pump control system 620, or vice versa. Furthermore, features and / or functions of the pump control system 620 may be implemented by control electronics located in the fluid infusion device 602, while in alternative embodiments, the pump control system 620 may be implemented by a remote computing device that is physically distinct and / or separate from the infusion device 602 (e.g., a mobile computing device coupled to the infusion device 602 via a personal area network, etc.).

[0111] Figure 7 According to one or more embodiments, a Figure 6 6. The pump control system 700 of FIG. 6 shows an exemplary embodiment of the pump control system 700 of FIG. 6. The illustrated pump control system 700 includes, but is not limited to, a pump control module 702, a communication interface 704, and a data storage element (or memory) 706. The pump control module 702 is coupled to the communication interface 704 and the memory 706, and the pump control module 702 is suitably configured to support the operations, tasks, and / or processes described herein. In various embodiments, the pump control module 702 is also coupled to one or more user interface elements (e.g., user interface 640) for receiving user input (e.g., a target glucose value or other glucose threshold) and providing notifications, alerts, or other treatment information to the user.

[0112] The communication interface 704 generally represents the hardware, circuitry, logic, firmware, and / or other components of the pump control system 700 that are coupled to the pump control module 702 and configured to support communication between the pump control system 700 and one or more of the various sensing devices 604, 606, 608, 650, 660. In this regard, the communication interface 704 may include or otherwise be coupled to one or more transceiver modules capable of supporting wireless communication between the pump control system 620, 700 and the external sensing devices 604, 606. For example, the communication interface 704 may be used to wirelessly receive sensor measurements or other measurement data from each external sensing device 604, 606 in the infusion system 600. In other embodiments, the communication interface 704 may be configured to support wired communication to / from the external sensing devices 604, 606. In various embodiments, the communication interface 704 may also support communication with a remote server (e.g., server 102) or another electronic device in the infusion system (e.g., to upload sensor measurements, receive control information, etc.).

[0113] The pump control module 702 generally represents the hardware, circuitry, logic, firmware, and / or other components of the pump control system 700 that are coupled to the communication interface 704 and the sensing devices 604, 606, 608, 650, 660 and configured to determine dosage commands for operating the motor 632 to deliver fluid to the body 601 based on measurement data received from the sensing devices 604, 606, 608, 650, 660 and to perform various additional tasks, operations, functions, and / or operations described herein. For example, in an exemplary embodiment, the pump control module 702 implements or otherwise executes a command generation application 710 that supports one or more autonomous operating modes and calculates or otherwise determines dosage commands for operating the motor 632 of the infusion device 602 in the autonomous operating mode based at least in part on current measurements of the condition of the user's body 601. For example, in a closed-loop operating mode, the command generation application 710 may determine a dose command for operating the motor 632 to deliver insulin to the user's body 601 to adjust the user's blood glucose level to a target reference glucose value based at least in part on the current glucose measurement value most recently received from the sensing device 604. In various embodiments, the dose command may also be adjusted or otherwise influenced by background measurement data, i.e., measurement data that characterizes, quantifies, or otherwise indicates the context of the simultaneous or concurrent operation of the dose command, such as environmental measurement data obtained from the environmental sensing device 650, current location information obtained from the GPS receiver 660, and / or other contextual information characterizing the current operating environment of the infusion device 602. Additionally, the command generation application 710 may generate a dose command for a bolus that is manually initiated or otherwise directed by the user via a user interface element.

[0114] In one or more exemplary embodiments, the pump control module 702 also implements or otherwise executes a prediction application 708 (or prediction engine) that is configured to estimate or otherwise predict future physiological conditions and potentially other future activities, events, operating environments, etc. in a personalized, patient-specific (or patient-specific) manner. In this regard, in some embodiments, the prediction engine 708 is cooperatively configured to interact with the command generation application 710 to support adjusting dosage commands or controlling information indicating how dosage commands are generated in a predictive or proactive manner. In this regard, in some embodiments, based on correlations between current or recent measurement data and the current operating context relative to historical data associated with the patient, the prediction engine 708 can forecast or otherwise predict the patient's future glucose levels at different times in the future, and correspondingly adjust or otherwise modify the values of one or more parameters utilized by the command generation application 710 in determining dosage commands in a manner that takes into account the predicted glucose levels, for example, by modifying the parameter values at registers or locations in the memory 706 referenced by the command generation application 710. In various embodiments, the prediction engine 708 can predict meals or other events or activities that the patient is likely to engage in and output or otherwise provide an indication of how the patient's predicted glucose level is likely to be affected by the predicted event, which indication can then be reviewed or considered by the patient to proactively adjust his or her behavior, and / or used to adjust in a personalized manner how dosing instructions are generated to regulate glucose in a manner that takes into account the patient's behavior.

[0115] In one or more exemplary embodiments, the pump control module 702 further implements or otherwise executes a conversational interaction application 712 configured to support conversational interaction with a patient or other user. For example, the conversational interaction application 712 can generate or otherwise provide a GUI display on a display device 640 associated with the infusion device 602, the GUI display including a conversational box prompting a user to conversationally interact with the infusion device 602. In this regard, in one or more embodiments, the conversational interaction application 712 can generate a GUI display, such as the GUI display 400, that prompts a user to conversationally query or search the database system 104. The conversational interaction application 712 can also support conversational monitoring or management of a patient's physiological condition, as described above. Figures 3 and 4 described in the context of and below in Figures 13 to 16 In one or more exemplary embodiments, the pump control module 702 also implements or otherwise executes a recommendation application 714 (or recommendation engine) that is configured to support providing treatment recommendations to the patient, as described below in Figures 17 to 19 described in more detail in the context of .

[0116] Still refer to Figure 7 Depending on the embodiment, the pump control module 702 may be implemented or realized as a general-purpose processor, microprocessor, controller, microcontroller, state machine, content addressable memory, application specific integrated circuit, field programmable gate array, any suitable programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof, designed to perform the functions described herein. In this regard, the steps of the methods or algorithms described in connection with the embodiments disclosed herein may be embodied directly in hardware, in firmware, in a software module executed by the pump control module 702, or in any practical combination thereof. In an exemplary embodiment, the pump control module 702 includes or otherwise accesses a data storage element or memory 706, which may be implemented using any type of non-transitory computer-readable medium capable of storing programming instructions for execution by the pump control module 702. The computer-executable programming instructions, when read and executed by the pump control module 702, cause the pump control module 702 to implement or otherwise generate the application programs 708, 710, 712, 714 and perform the tasks, operations, functions, and processes described herein.

[0117] It should be understood that the subject matter described herein is for illustrative purposes only and is not intended to limit the subject matter described in any way. Figure 7 is a simplified representation of the pump control system 700. For example, in some embodiments, the features and / or functionality of the motor control module 612 may be implemented by or otherwise integrated into the pump control system 700 and / or pump control module 702, such as by converting dose commands into corresponding motor commands via the command generation application 710, in which case a separate motor control module 612 may not be present in embodiments of the infusion device 602.

[0118] Glucose Prediction and Forecast

[0119] In one or more exemplary embodiments, the patient-specific forecast model for physiological condition is determined based on the historical data relevant to the patient, and is used to predict future values or the level of physiological condition based at least in part on the current measured value of current operating context and physiological condition. In addition, historical event data and associated background information can be used to predict one or more future events of different future times within the forecast range based at least in part on current measured data and / or current operating context (e.g., the current time of the day, the current date of the week, the current geographical location, etc.), which can then be input into the patient-specific forecast model so that the forecast value or the level of physiological condition are adjusted to reflect the forecast event at the appropriate time in the future. Although this paper describes this theme in the context of glucose forecast and prediction, this theme is not necessarily limited to glucose levels and may be implemented in an equivalent manner to forecast or predict other physiological conditions of individuality.

[0120] In an exemplary embodiment, a patient-specific glucose prediction model is determined that allows the patient's glucose level to be predicted for discrete time intervals in the future. For illustrative purposes, this paper describes the subject matter in the context of hourly forecasts that allow the patient's glucose level to be predicted on an hourly basis; however, it should be noted that the subject matter described herein is not limited to hourly forecasts and can be used for different forecast time intervals (e.g., every 15 minutes, every 30 minutes, every 4 hours, etc.).

[0121] Figure 8 Describes a method suitable for use by a computing device such as Figure 1 The exemplary forecasting process 800 implemented by the server 102 or the client electronic device 106 in the patient data management system 100 is shown in FIG. The various tasks performed in conjunction with the forecasting process 800 may be performed by hardware, firmware, software executed by a processing circuit, or any combination thereof. For illustrative purposes, the following description relates to the aforementioned forecasting process 800 in conjunction with FIG. Figure 1 and Figures 6 and 7 elements mentioned. In practice, portions of the forecasting process 800 may be performed by different elements of the patient data management system 100 or the infusion system 600 (e.g., the server 102, the one or more electronic devices 106, the infusion device 602, and / or the pump control system 620, 700). It will be understood that the forecasting process 800 may include any number of additional or alternative tasks, which need not be performed in the order illustrated and / or which may be performed simultaneously, and / or the forecasting process 800 may be incorporated into a more comprehensive program or process having additional functionality not described in detail herein. Furthermore, the components described in the forecasting process 800 may be omitted from an actual implementation of the forecasting process 800 as long as the intended overall functionality remains unchanged. Figure 8 One or more tasks shown and described in the context of.

[0122] The illustrated embodiment of the prediction process 800 initializes or otherwise begins by retrieving or otherwise obtaining historical data associated with a patient of interest to be modeled, and using the historical data associated with the patient to develop, train, or otherwise determine a prediction model for the patient (tasks 802, 804). In one or more exemplary embodiments, for an individual patient within the patient data management system 100, the server 102 periodically retrieves or otherwise obtains historical patient data 120 associated with the patient from the database 104, and analyzes relationships between different subsets of the historical patient data 120 to create a patient-specific prediction model associated with the patient. Depending on the embodiment, the patient-specific prediction model can be stored on the database 104 associated with the patient and utilized by the server 102 to determine a glucose forecast for the patient (e.g., in response to a request from a client device 106) and provide the resulting glucose forecast to the client device 106 for presentation to the user. In other embodiments, the server 102 pushes, provides, or otherwise transmits the patient-specific prediction model to one or more electronic devices 106 associated with the patient (e.g., infusion device 602) for implementing and supporting glucose prediction at the end-user device (e.g., via prediction engine 708).

[0123] In one or more exemplary embodiments, a recurrent neural network is used to create hourly neural network units that are trained based on a subset of historical patient data corresponding to the hourly interval across multiple different days prior to model development to predict the average glucose level of the patient associated with the corresponding hourly interval. For example, in one embodiment, for each hourly interval in a day, a corresponding long short-term memory (LSTM) unit (or cell) is created, wherein the LSTM unit outputs the average glucose value for the hourly interval based on the subset of historical patient data corresponding to the hourly interval and the variables of one or more LSTM units preceding the current LSTM unit. For example, the LSTM unit associated with the 1-2 PM time interval is configured to calculate the average glucose value for the patient in the 1-2 PM time frame based on a subset of historical patient data timestamped in or associated with the 1-2 PM time frame and the inputs and / or outputs of one or more previous LSTM units (e.g., the average glucose value for the patient in the 12-1 PM time frame output by the 12-1 PM LSTM unit, a relevant portion of the subset of historical patient data timestamped in or associated with the 12-1 PM time frame, etc.).

[0124] For each LSTM unit, machine learning can be utilized to determine that corresponding equation, function or model are used to calculate the average glucose value of the patient in this time interval based on the patient's history insulin delivery data, history meal data and history exercise data during this time interval at least in part. In this regard, the model of specific hourly interval can characterize or map the average glucose value of the sensor of hourly interval being modeled by the insulin delivery data during hourly interval, the meal data during hourly interval, the exercise data during hourly interval and the average glucose value of previous hourly interval. In addition, hourly model can consider historical auxiliary measurement data (for example, historical acceleration measurement data, historical heart rate measurement data etc.), historical drug data or other historical event log data, historical geographical location data, historical environmental data, and / or may be relevant to the patient's average glucose level during this time interval or predict other history or background data of this average glucose level. Therefore, because different variables produce more or less influence on patient's glucose level in one day, each function or equation associated with corresponding LSTM unit can suitably increase or decrease the weighting or emphasis of specific input variable to the average glucose value calculated by corresponding LSTM unit. It should be noted that any number of different machine learning techniques may be utilized to determine which input variables predict the current patient of interest and the current hourly interval of the day, such as artificial neural networks, genetic programming, support vector machines, Bayesian networks, probabilistic machine learning models or other combinations of Bayesian techniques, fuzzy logic, heuristic derivations, and the like.

[0125] The forecasting process 800 continues by receiving, retrieving, or otherwise obtaining recent patient data, identifying or otherwise obtaining a current operational context associated with the patient, and predicting the patient's future behavior based on the recent patient data and the current operational context (tasks 806, 808, 810). In this regard, a prediction model for future insulin delivery, future meals, future exercise events, and / or future medication doses can be determined that characterizes or maps a particular combination of one or more of current (or recent) sensor glucose measurement data, auxiliary measurement data, delivery data, geographic location, meal data, exercise data, patient behavior or activity, etc., to a value representing a current probability or likelihood of a particular event and / or a current value associated with the particular event or activity (e.g., predicted meal size, predicted exercise duration and / or intensity, predicted bolus amount, etc.). Thus, the forecasting process 800 can obtain current or recent sensor glucose measurements associated with the patient, as well as data or information quantifying or characterizing recent insulin deliveries, meals, exercise, and other events, activities, or behaviors that the user may have engaged in within a previous time interval (e.g., within the previous 2 hours) from one or more of the sensing devices 604, 606, 608, the infusion device 602, and / or the database 104. The forecasting process 800 can also obtain data or information quantifying or characterizing the current or recent operating context associated with the infusion device 602 from one or more of the sensing devices 650, 660, the infusion device 602, and / or the database 104.

[0126] Based on current and recent patient measurement data, insulin delivery data, meal data and exercise data, as well as the current time of day, the current date of the week and / or other current or recent environmental data, the forecast process 800 determines the event probability and / or characteristics of the future hourly time interval. For example, for each future hourly time interval, the forecast process 800 can determine the meal probability and / or predicted meal size during the future hourly time interval, which can be used as the input of the LSTM unit within the hourly interval. Similarly, the forecast process 800 can determine the predicted insulin delivery amount, predicted exercise probability and / or predicted exercise intensity or duration, predicted drug dosage, etc. within each corresponding future hourly time interval, based on the relationship between the historical patient data and background data before the previous instance of these events occurred. Some examples of predicting patient behavior or activity are described in U.S. patent application serial number 15 / 847,750, which is incorporated herein by reference.

[0127] Still refer to Figure 8After predicting future patient behavior that may affect the patient's future glucose level, the forecast process 800 continues to calculate or otherwise determine the forecast glucose level for the next hourly interval based at least in part on the current or most recent glucose measurement data and the predicted future behavior, and to generate or provide a graphical representation of the forecast glucose level associated with different future hourly intervals (tasks 812, 814). Based on the current time of day, a forecast model for the next hourly interval of the day can be selected and utilized to calculate the forecast glucose level for the hourly interval based at least in part on the most recent sensor glucose measurement value and the predicted meal, exercise, insulin delivery, and / or medication dosage for the next hourly interval of the day. For example, the current sensor glucose measurement value obtained within the current hourly interval and the previous sensor glucose measurement value can be averaged or otherwise combined to obtain an average sensor glucose measurement value for the current hourly interval, which can be input into the forecast model for the next hourly interval of the day. The forecast model is then utilized to calculate the forecast average glucose value for the next hourly interval of the day based on the average sensor glucose measurement value for the current hourly interval and the predicted patient behavior for the next hourly interval. The predicted average glucose value for the next hourly interval may then be input into a prediction model for subsequent hourly intervals for use in calculating predicted glucose values for that subsequent hourly interval based on its associated predicted patient behavior, among other things.

[0128] Figure 9 Described exemplary GUI display 900 including glucose forecast area 902, this glucose forecast area includes the graphical representation of the forecast glucose level of the patient associated with the subsequent hourly interval in one day.In shown GUI display 900, the graphical representation 904 of the patient's most recent sensor glucose measurement data is presented near glucose forecast area 902.In exemplary embodiments, sensor glucose measurement value display area 904 includes a line graph or a line graph of the patient's historical sensor glucose measurement data, which has a visually distinguishable coverage area of the target range of the sensor glucose measurement value of the patient's indication.Depending on the embodiment, GUI display 900 can be presented on the display device 640 associated with infusion device 602 or on another electronic device 102,106 in patient data management system 100.In one or more embodiments, execution forecast process 800 is so that in response to patient selection, be configured to cause the GUI element presenting forecast area 902 or otherwise request presenting glucose forecast (e.g., by requesting glucose forecast via conversational interaction application 712 conversationally) and generate forecast area 902 on GUI display 900.

[0129] refer to Figures 8 and 9In one or more exemplary embodiments, based on the current time of day (e.g., 9:45 AM), the current sensor glucose measurement (e.g., 110 mg / dL), and possibly other recent patient data (e.g., a recent meal, exercise, or bolus) and / or the current operating context, the forecasting process 800 calculates or otherwise determines predicted patient behavior for the 10 AM hourly interval and subsequent hourly intervals for which the patient's glucose level is to be forecast. Figure 10 Depicts a graphical representation of a portion of a recurrent neural network comprising hourly LSTM cells configured to compute using predicted patient behavior for future hourly intervals and the patient's current sensor glucose measurement data Figure 9 . In this regard, the average sensor glucose value for the current interval 1001 is input into the 10 AM hourly interval LSTM cell 1002 along with the predicted patient behavior 1003 for the 10 AM hourly interval (e.g., the predicted amount of carbohydrates consumed, insulin delivered, exercise, medication, etc., by the patient between 10 AM and 11 AM). Depending on the embodiment, the current interval sensor glucose value 1001 can be implemented as the current or most recent sensor glucose measurement value (e.g., 110 mg / dL), the average of the current and previous sensor glucose measurements obtained during the current interval (e.g., the average of sensor glucose measurements timestamped between 9:00 AM and 9:45 AM), or another sensor glucose value calculated at least in part based on the current sensor glucose measurement value. For example, the current sensor glucose measurement and other recent behavior can be used to predict the patient's glucose level for the remainder of the current time interval (e.g., from 9:45 AM to 10 AM), which can then be averaged, weighted, or otherwise combined with the average of previous sensor glucose measurements obtained during the current time interval (e.g., from 9 AM to 9:45 AM) to obtain an estimated average sensor glucose measurement for the current time interval.

[0130] Based on the average sensor glucose value for the current interval 1001 and the predicted patient behavior 1003 for the 10 AM interval, the LSTM cell 1002 calculates or otherwise determines an average glucose value 1005 associated with the 10 AM interval using a forecast model for the 10 AM hourly interval, the forecast model being determined based on a subset of the patient's historical patient data associated with the 10 AM hourly interval (e.g., task 804). It should be noted here that, in one or more embodiments, the amount of active insulin or carbohydrate need not necessarily be calculated or input into the LSTM cell 1002 within the 10 AM interval, as active insulin, carbohydrate, and / or surrogate amounts may be obtained from a previous LSTM cell and scaled, reduced, discarded, or otherwise adjusted based on the model associated with the LSTM cell 1002. The average glucose value 1005 associated with the 10 AM interval (e.g., 115 mg / dL) is displayed in the forecast area 902 associated with the 10 AM hourly interval of the GUI display 900. The predicted 10 AM glucose value 1005 is also input into the 11 AM hourly interval LSTM cell 1004 along with the predicted patient behavior 1007 for the 11 AM hourly interval (e.g., the predicted amount of carbohydrates consumed, insulin delivered, exercise, medication, etc., by the patient between 11 AM and 12 PM). The LSTM cell 1004 calculates or otherwise determines an average glucose value 1009 associated with the 11 AM interval using a forecast model for the 11 AM hourly interval, which is determined based on a subset of the patient's historical patient data associated with the 11 AM hourly interval (e.g., task 804). The predicted glucose value 1009 for the 11 AM interval (e.g., 110 mg / dL) is displayed in the forecast area 902 associated with the 11 AM hourly interval of the GUI display 900 and is input into the 12 PM hourly interval LSTM cell 1006 for use in determining a predicted glucose value for the 12 PM interval (e.g., 100 mg / dL), and so on.

[0131] In one or more exemplary embodiments, the predicted glucose values in the glucose forecast area 902 are displayed or otherwise presented with visually distinguishable characteristics that indicate the relationship of the individual predicted glucose values relative to one or more threshold values. Figure 9In the illustrated embodiment of , predicted glucose values within a target range for a patient's glucose value (e.g., between 80 mg / dL and 140 mg / dL) are presented using a visually distinguishable characteristic that indicates that these values are normal, expected, or other acceptable values (e.g., green), wherein predicted glucose values outside the target range are presented using a different visually distinguishable characteristic that indicates that these values may be problematic or undesirable (e.g., red). In this regard, for the illustrated embodiment, the predicted glucose values associated with the 4 PM hourly interval and the 5 PM hourly interval in the glucose forecast area 902 can be presented in red to indicate that they are below a lower threshold of the patient's target glucose range, which lower threshold indicates that a possible hypoglycemic event is predicted for the patient at or around those times, while the predicted glucose values before 4 PM in the glucose forecast area 902 can be presented in green to indicate that the patient's glucose is forecast to be within the target range for the next 6 hours.

[0132] In one or more embodiments, glucose forecast area 902 is scrollable or can be otherwise adjusted to allow the patient or user to check the forecast glucose value for the future. For example, in one or more embodiments, glucose forecast area 902 is scrollable or can be otherwise adjusted to allow the patient or user to check the forecast glucose value for the next 24 hours. In addition, it should be noted that in some embodiments, the forecast glucose value in glucose forecast area 902 can be dynamically updated in real time in response to the patient's current sensor glucose measurement value, current operating background or other real-time behaviors or active changes performed by the patient.

[0133] Now refer to Figure 11 According to one or more embodiments, an ensemble prediction process 1100 may be performed to determine an ensemble prediction for a patient's physiological condition as a combination of predicted values determined using a plurality of different prediction models. Because different prediction models may utilize different input variables, different prediction ranges, and / or different formulas or techniques to determine future glucose values, an ensemble prediction of a patient's glucose level may better reflect potential changes in the patient's future glucose level than would be possible with reliance on any individual prediction model. In this regard, the predicted hourly glucose values determined according to the forecast process 800 may be weighted or otherwise combined with predicted glucose values for the patient determined using other prediction models to obtain an ensemble prediction of the patient's glucose level with respect to time that reflects the relative reliability or accuracy of the respective prediction models with respect to time.

[0134] The various tasks performed in conjunction with the ensemble prediction process 1100 may be performed by hardware, firmware, software executed by processing circuitry, or any combination thereof. Figure 1 and Figures 6 and 7 elements mentioned. In practice, portions of the ensemble prediction process 1100 may be performed by different elements of the patient data management system 100 or the infusion system 600 (e.g., the server 102, the one or more electronic devices 106, the infusion device 602, and / or the pump control system 620, 700). It will be understood that the ensemble prediction process 1100 may include any number of additional or alternative tasks, which need not be performed in the order illustrated and / or which may be performed simultaneously, and / or the ensemble prediction process 1100 may be incorporated into a more comprehensive program or process having additional functionality not described in detail herein. Furthermore, the elements described in the ensemble prediction process 1100 may be omitted from an actual implementation of the ensemble prediction process 1100 as long as the intended overall functionality remains unchanged. Figure 11 One or more tasks shown and described in the context of.

[0135] The ensemble prediction process 1100 begins by retrieving or otherwise obtaining historical data associated with a particular patient, and developing, training, or otherwise determining a plurality of different patient-specific glucose prediction models based on the patient's historical data (tasks 1102, 1104). In this regard, in addition to determining the glucose prediction model described above in the context of the prediction process 800, one or more additional models for the patient may be determined based on the patient's historical data 120 that predicts the patient's future glucose levels in different ways (e.g., using different algorithms or modeling techniques, etc.), based on different input variables with different levels of time granularity (e.g., in minutes, in hours, etc.), etc. It should be noted that, in practice, many different types of prediction models may be utilized, and the subject matter described herein is not intended to be limited to any particular type of model, technique, or method, or combination thereof, for predicting glucose levels.

[0136] For example, in one or more embodiments, in addition to a patient-specific neural network-based forecasting model, the patient's historical data is used to determine an autoregressive integrated moving average (ARIMA) model for predicting future glucose levels. In this regard, the ARIMA model predicts future glucose levels based on periodic patterns in the patient's historical data, which may be related to different events and / or operational contexts. In an exemplary embodiment, the ARIMA model is configured to determine the patient's predicted glucose values in future increments, which correspond to the sampling rate associated with the glucose sensing device 604 and / or the patient's historical sensor glucose measurement data. For example, in one embodiment, the glucose sensing device 604 provides new or updated sensor glucose measurements every 5 minutes, and the ARIMA model is configured to determine predicted glucose values at intervals of 5 minutes in the future. In one or more exemplary embodiments, machine learning or similar techniques are utilized to determine which combination of historical transport data, historical auxiliary measurement data, historical event log data, historical geographic location data, and other historical or contextual data is associated with or predictive of historical sensor glucose measurement data, and then, based on the set of input variables and a previous subset of historical sensor glucose measurement values, a corresponding ARIMA model is determined for calculating or predicting future sensor glucose measurement values. In this regard, the trajectory of the previous subset of historical sensor glucose measurement values is combined with concurrent or previous events or operational context that historically correlate with or predict changes in the patient's sensor glucose level, thereby influencing the predicted future sensor glucose measurement values determined using the model. In one embodiment, training of the autoregressive component of the ARIMA model attempts to determine the ability of the patient's glucose level to regress on its own, while the moving average component of the ARIMA model attempts to compensate for slow background drifts in the patient's glucose level.

[0137] In one or more embodiments, the patient's historical data is also used to determine a patient-specific physiological model for predicting future glucose levels, which is in an attempt to simulate the patient's pharmacodynamics and pharmacokinetics and compensate for differences within and between personnel. Similar to the ARIMA model, the patient-specific physiological model can be configured to determine the patient's predicted glucose value in the future increment corresponding to a sampling rate, and this sampling rate is associated with the historical sensor glucose measurement data of glucose sensing device 604 and / or the patient. That is, the patient-specific physiological model may determine the predicted glucose value in a manner different from the ARIMA model and / or based on input variables different from the input variables used by the ARIMA model. For example, in one or more embodiments, one or more patient-specific physiological parameters (e.g., glucose occurrence rate, insulin action, etc.) are determined for the patient based on the relationship between the patient's historical sensor glucose measurement data, historical meal data, historical delivery data, historical bolus data, etc. Based on the patient's current or recent sensor glucose measurement value, current insulin in body, recent meal data, etc., the physiological model utilizes the patient-specific physiological parameters to predict future sensor glucose measurement values. In this regard, the output of the patient-specific physiological parameters represents the patient's expected glucose level based on the patient's historical physiological response given the current amount of insulin in the body and / or the amount of carbohydrate to be metabolized by the patient.

[0138] Still refer to Figure 11 In an exemplary embodiment, after determining or otherwise obtaining a plurality of different patient-specific glucose prediction models, the ensemble prediction process 1100 proceeds to identify or otherwise obtain a current operating context for the patient and, based on the current operating context, calculate or otherwise determine reliability metrics associated with different patient-specific glucose prediction models for different prediction horizons before the current time of day (tasks 1106, 1108). In this regard, the current time of day, the current day of the week, the current geographic location, the current network address and / or network connection type, and / or other contextual data associated with the device 106, 602 associated with the patient are identified or otherwise obtained. Based on the current operating context, a subset of the patient's historical data 120 corresponding to the current operating context is obtained, and the subset is used to determine one or more accuracy or reliability metrics associated with different glucose prediction models. For example, if the current operating context indicates that it is 8 AM on a Wednesday and the patient is at home, previous subsets of the patient's historical data 120 that have associated timestamps on or around 8 AM on a Wednesday and / or have associated geographic locations that correspond to the patient's home geographic location can be obtained and then used to determine the reliability of different glucose prediction models.

[0139] For each prediction model, appropriate input variables are obtained from a relevant subset of the patient's historical data, and the calculated glucose values output by the model are then compared with the patient's historical sensor glucose measurements at the corresponding time to obtain a reliability metric for the model. For example, the mean absolute difference, standard deviation, or other statistical measure can be calculated by comparing a set of output values from the glucose prediction model corresponding to a prediction range after a certain time point (e.g., predicted glucose values four hours after an 8 AM reference point) with corresponding historical glucose measurements (e.g., patient sensor historical glucose measurements four hours after an 8 AM reference point on Wednesday). In one or more embodiments, a reliability metric is determined for hourly intervals, for example, by calculating the mean absolute difference within the first hour after the prediction time (e.g., using values corresponding to a time frame from 8 AM to 9 AM), the mean absolute difference within the second hour after the prediction time (e.g., using values corresponding to a time frame from 9 AM to 10 AM), etc. In this regard, the reliability metric associated with each particular prediction model can vary depending on the particular prediction range or time window before the current prediction time.

[0140] Based on the reliability metrics associated with the different prediction models, the ensemble prediction process 1100 calculates or otherwise determines weighting factors associated with the outputs of the different patient-specific glucose prediction models for the different prediction horizons prior to the current time of day, and then uses these weighting factors to calculate or otherwise determine the ensemble-predicted glucose values within the different prediction horizons as a weighted average of the outputs of the different patient-specific glucose prediction models (tasks 1110, 1112). In this regard, based on the relationship between the reliability metrics among the different patient-specific glucose prediction models for a particular prediction horizon, time window, or sampling time, weighting factors can be assigned to the different models accordingly to increase the influence of one or more more reliable models on the ensemble-predicted glucose values within the prediction horizon.

[0141] For example, if the reliability of the patient's ARIMA model for the second hour before the current forecast time (e.g., the 9 AM to 10 AM time frame) is fifty percent higher than the reliability of the patient's hourly forecast model, the predicted glucose value output by the ARIMA model at the second hour can be assigned a weighting factor that is fifty percent higher than the weighting factor assigned to the hourly forecast model. The ensemble forecast value for the forecast horizon (i.e., the second hour before the current time) can then be determined as a weighted average of the 5-minute predicted glucose values output by the ARIMA model within that time frame (e.g., the 9 AM value to the 10 AM value) and the hourly predicted glucose values output by the hourly forecast model (e.g., the predicted average glucose level within the 9 AM to 10 AM time window), resulting in a 5-minute ensemble forecast value that is composed of 60% of the ARIMA predicted glucose values for that particular 5-minute sampling time (e.g., the ARIMA predicted glucose value at 9:05 AM) and 40% of the hourly predicted glucose values within the forecast horizon. However, for the third hour before the current prediction time, the reliability of the patient's hourly forecast model may be fifty percent higher than the reliability of the predicted glucose value output by the ARIMA model in the third hour, resulting in a weighting factor assigned to the hourly forecast model that is fifty percent higher than the weighting factor assigned to the ARIMA model, thereby resulting in a 5-minute ensemble forecast value consisting of 40% of the ARIMA predicted glucose values at a specific 5-minute sampling time (e.g., the ARIMA predicted glucose values for 10:05 AM, 10:10 AM, etc.) and 60% of the hourly forecast glucose values within the time range (e.g., the hourly forecast glucose values within the 10 AM to 11 AM time period).

[0142] After determining the future aggregate glucose forecast, the aggregate forecast process 1100 proceeds to generate or otherwise provide a graphical representation of the aggregate forecast glucose value to the patient or other user (task 1114). Figure 12As depicted, a GUI display 1200 can be presented on the client electronic device 106 and / or infusion device 602 that includes a graphical representation 1202 of the patient's sensor glucose measurement data relative to a time before a marker 1206, or a similar graphical indication of the current time, followed by a graphical representation 1204 of the collective predicted glucose value relative to a time after the marker indicating the current time. In one or more embodiments, as the forecast horizon progresses further into the future before the current time, the reliability metric associated with the ARIMA model, physiological model, or other short-term forecast model decreases relative to the reliability metric associated with the hourly forecast model, such that when the patient or user scrolls, slides, or otherwise adjusts the GUI display 1200 to advance the forecast horizon associated with the displayed values, the graphical representation of the collective predicted glucose values converges toward the hourly predicted glucose level. In this regard, scrolling or adjusting the sensor glucose measurement display area 904 can result in updating the sensor glucose measurement display area 904 to present the GUI display 1200 that depicts the collective glucose forecast 1204 extending from the current time marker 1206 into the future.

[0143] In one or more exemplary embodiments, earlier portions of the ensemble glucose forecast 1204 are weighted more heavily toward forecast model outputs that are more reliable in the short term, while future portions of the ensemble glucose forecast 1204 are weighted more heavily toward forecast model outputs that have better long-term reliability. For example, the portion of the ensemble glucose forecast 1204 from 10 AM to 11 AM may be composed of 60% of the output of the patient's ARIMA model and 40% of the patient's hourly forecast model, while the subsequent portion of the ensemble glucose forecast 1204 from 11 AM to 12 PM may be composed of 40% of the output of the patient's ARIMA model and 60% of the patient's hourly forecast model, the portion of the ensemble glucose forecast 1204 from 12 PM to 1 PM may be composed of 30% of the output of the patient's ARIMA model and 70% of the patient's hourly forecast model, and so on.

[0144] Figure 13 An exemplary patient simulation process 1300 suitable for implementation in conjunction with the ensemble prediction process 1100 is depicted for simulating or otherwise predicting how different events or actions performed by a patient may affect the patient's glucose level in the future. The various tasks performed in conjunction with the patient simulation process 1300 may be performed by hardware, firmware, software executed by processing circuitry, or any combination thereof. For illustrative purposes, the following description relates to the process described above in conjunction with the ensemble prediction process 1100. Figure 1 and Figures 6 and 7elements mentioned. In practice, portions of the patient simulation process 1300 may be performed by different elements of the patient data management system 100 or the infusion system 600 (e.g., the server 102, the one or more electronic devices 106, the infusion device 602, and / or the pump control system 620, 700). It should be understood that the patient simulation process 1300 may include any number of additional or alternative tasks, which need not be performed in the order shown and / or which may be performed simultaneously, and / or the patient simulation process 1300 may be incorporated into a more comprehensive program or process having additional functionality not described in detail herein. Furthermore, the patient simulation process 1300 may be omitted from an actual implementation of the patient simulation process 1300 as long as the intended overall functionality remains unchanged. Figure 13 One or more tasks shown and described in the context of.

[0145] Patient simulation process 1300 begins with receiving or otherwise obtaining the user input (task 1302) of the future event, action or other activities of the indication patient. In this regard, the patient or another user can input or otherwise provide characterization, quantization or otherwise define the action or event that the patient anticipates, thinks or otherwise considers. For example, the user input can indicate the expected amount of the carbohydrate that the patient will consume, the expected amount of the exercise that the patient will perform, the expected bolus amount of the insulin that the patient will use etc. In addition, the user input can indicate the time of the day associated with the expected activity (for example, the expected meal time for the carbohydrate of future amount), the time window of interest or the duration, or other time information of the expected activity that the characterizes the patient. In some events, prediction engine 708 can automatically predict future action or event and the corresponding parameter or standard associated therewith based on the patient's historical measurement data, event log data, background data etc. In this type of embodiment, the patient or another user can input or otherwise provide confirmation to predicting future events, or otherwise adjust one or more characteristics associated with the future events of prediction (for example, adjusting the future time, quantity, duration, type or other characteristics associated with the event etc.).

[0146] The patient simulation process 1300 continues by calculating or otherwise determining adjusted weighting factors for combining the outputs from the patient's glucose prediction models into an ensemble prediction based on the current operating context and the input future activity information (task 1304). In this regard, similar to that described above in the context of the ensemble prediction process 1100, the patient simulation process 1300 calculates or otherwise determines reliability metrics associated with different patient-specific glucose prediction models for different prediction horizons into the future, based on the current time of day and other operating context, in a manner that also takes into account the input future activity information.

[0147] In one or more embodiments, when selecting subsets of the patient history data 120 for use in determining a reliability metric, the patient simulation process 1300 can exclude from the subset of the patient history data 120 corresponding to the current operational context any subset of the patient history data 120 that does not include one or more inputted future activities within the prediction horizon or otherwise within a threshold time period from the current time of day. For example, if the current operational context indicates that it is 8 AM on Wednesday and the patient is at home, and user input indicates that the patient intends to consume carbohydrates or otherwise experience a meal event within a threshold amount of time, only previous subsets of the patient history data 120 that have an associated timestamp on or around 8 AM on Wednesday and / or have an associated geographic location that corresponds to the patient's home geographic location that also have simultaneous or concurrent meals within the threshold amount of time after 8 AM are selected for analysis. In this regard, varying the data sets used to calculate the reliability metrics associated with different prediction models can result in prospectively adjusted reliability metric values associated with the respective prediction models that differ from normal reliability metric values that would otherwise be associated with the respective prediction models without consideration of future events. The patient simulation process 1300 then determines prospectively adjusted weighting factors based on the relationships between the adjusted reliability metrics across the different patient-specific glucose prediction models. In a similar manner as described above, the prospectively adjusted weighting factors can also vary relative to prediction times prior to the current time.

[0148] Still refer to Figure 13 After determining the weighting factors that are prospectively adjusted to account for the input future activity information, the patient simulation process 1300 calculates or otherwise determines an expected aggregate glucose forecast based on the input future activity information using the prospectively adjusted weighting factors (task 1306). In this regard, the input future activity information is provided as input to one or more patient glucose prediction models to thereby alter or influence the predicted glucose values output by the models in a manner that takes into account a specified future event corresponding to a future time. For example, if the user input indicates that the patient is likely to eat a meal at a particular time in the future, the input meal information will be provided as input to the LSTM cells of the patient's hourly forecast model that include or correspond to that future time, thereby influencing the predicted glucose values for that time interval and / or subsequent time intervals. As another example, if the user input indicates that the patient is likely to be administering a bolus of insulin at the current time, the input bolus amount may be provided to each of the patient's glucose prediction models in a manner that takes into account the insulin bolus amount upon initialization (e.g., by adding the input bolus amount to the current active insulin amount associated with the current time of day).

[0149] After each glucose prediction model for the patient is used to calculate a predicted glucose value that takes into account the expected patient activity, the collective prediction value is determined as a weighted average of the corresponding predicted glucose values output by the corresponding glucose prediction models using the prospectively adjusted weighting factors in a manner similar to that described above (e.g., task 1112). By prospectively adjusting the weighting factors and using future activity as input to the prediction models, the resulting collective prediction value will effectively simulate or project the patient's glucose level if the patient engages in the input activity. Therefore, the expected collective prediction may alternatively be referred to herein as the patient's simulated glucose level.

[0150] In an exemplary embodiment, the patient simulation process 1300 generates or otherwise provides a graphical representation of the patient's simulated glucose level or other feedback affected by the expected aggregate glucose forecast (task 1308). For example, in some embodiments, a line graph or graph of the patient's simulated glucose level can be presented in response to the input future activity information. In other embodiments, the simulated glucose values can be processed or otherwise analyzed to provide one or more recommendations to the patient (e.g., an indication of whether to engage in the input activity, etc.).

[0151] For example, now refer to Figure 14 ,refer to Figure 13In one or more embodiments, the patient simulation process 1300 is performed in conjunction with a session interaction with the patient, which may be supported by a session interaction application 712 at an infusion device 602 or other client device 106. In the illustrated embodiment, the patient manipulates or otherwise interacts with the client device 106, 602 to input that the patient wishes to view their simulated glucose levels for the next four hours if the patient also consumes 60 grams of carbohydrates and administers a bolus of 3 units of insulin. In response to receiving the user input, the session interaction application 712 may provide the input parameters to the prediction engine 708 for use in simulating the patient's glucose levels according to the patient simulation process 1300. In this regard, for the subsequent four hours following any instances in the patient's history of consuming carbohydrates and / or administering an insulin bolus at or around the current time of day, the patient simulation process 1300 determines a prospectively adjusted weighting factor for the patient's prediction model based on the corresponding reliability metrics associated with the model. The patient simulation process 1300 then inputs or otherwise provides 60 grams of carbohydrates and 3 units of insulin to each patient's prediction model to initialize the model, as if carbohydrates were being consumed and insulin was being delivered simultaneously or otherwise at the time the model was started. Thereafter, the patient's prediction model, initialized with 60 grams of carbohydrates and 3 units of insulin, is used to calculate the patient's predicted glucose value for the next four hours. The expected ensemble predicted glucose value for the next four hours is then determined as a weighted average of the predicted glucose values using the forward-looking adjusted weighting factor, which takes into account the simultaneous intake of carbohydrates and insulin.

[0152] After determining the expected set predicted glucose value for the patient, the prediction engine 708 can provide the expected set predicted glucose value to the conversation interaction application 712 for presentation to the patient in the context of the ongoing conversation interaction. In this regard, the conversation GUI display 1400, which includes a graphical representation 1402 of the user input, is updated to include a conversation response 1404 of the user input influenced by the simulated glucose value. In the illustrated embodiment, the conversation response 1404 includes a graphical representation 1406 of the patient's simulated glucose level (e.g., a line graph of the expected set predicted glucose value) for the next four hours following a marker 1408 indicating the current time of day.

[0153] Figure 15 Another exemplary GUI display 1500 is depicted that depicts a mergeable Figure 11 The ensemble prediction process 1100 and / or Figure 13The session interaction of the patient simulation process 1300 is performed. In the illustrated embodiment, the session interaction application 712 receives initial user input 1502 and analyzes the initial user input to determine whether the patient is interested in the prediction of his or her physiological condition. The session interaction application 712 generates or otherwise provides a session response 1504 that prompts the patient to enter or otherwise provide possible other parameters for the prediction range and / or the prediction to be performed. For example, in some embodiments, the session interaction application 712 can prompt the patient to provide input for any anticipated activities within the prediction range for proactively adjusting the prediction according to the patient simulation process 1300.

[0154] In response to receiving subsequent user input 1506 indicating that the patient is interested in a prediction for the next 12 hours, the conversational interaction application 712 commands, signals, or otherwise instructs the prediction engine 708 to predict the patient's glucose level for the next 12 hours. The prediction engine 708 performs the collective prediction process 1100 to calculate or otherwise determine the patient's collective predicted glucose value for the next 12 hours based on the patient's current or most recent glucose measurement, current active insulin, current operating context, etc., as described above. In the illustrated embodiment, the prediction engine 708 also utilizes a reliability metric (e.g., standard deviation, mean absolute difference, etc.) associated with the relevant prediction model to probabilistically determine the likelihood of one or more physiological events (e.g., hypoglycemic events, hyperglycemic events, and / or the like) within the prediction range based on the collective predicted glucose values. The prediction engine 708 provides the collective predicted glucose values and corresponding physiological event probabilities to the conversational interaction application 712, which generates a conversational response 1508 that provides feedback influenced by the patient's collective predicted value. For example, the illustrated session response 1508 includes a graphical representation of the probability of a hypoglycemic event relative to different intervals within the forecast range, as well as an indication of the probability of a hypoglycemic event based on the current time of day.

[0155] In the illustrated embodiment, the conversation response 1508 also prompts the patient, based on the predicted glucose level, whether the patient wishes to configure one or more settings at the device 106, 602. In response to receiving user input 1510 indicating a desire to configure a reminder, the conversation interaction application 712 can configure itself to provide a reminder at a time of day when the probability of a hypoglycemic event is highest based on the aggregate predicted value and the reliability metric, and then generate or otherwise provide a conversation response 1512 confirming or otherwise indicating that the reminder has been set.

[0156] Figure 16Another exemplary GUI display 1600 is depicted that depicts a conversational interaction that may incorporate one or more of the above-described processes 300, 800, 1100, and 1300. In the illustrated embodiment, the conversational interaction application 712 receives initial conversational user input 1602 and analyzes the initial user input to determine that the patient is interested in a prediction of his or her physiological condition in response to a future exercise event. The conversational interaction application 712 generates or otherwise provides a series of conversational responses 1604, 1608, and 1612 that prompt the patient for conversational input 1606, 1610, and 1614 and define desired attributes of the future exercise event, such as the desired type of event, the desired duration of the event, and the desired timing of the event. After defining the attributes of the future event, the prediction engine 708 performs a patient simulation process 1300 to calculate or otherwise determine the patient's expected glucose level after the event (e.g., at a time corresponding to the sum of the input timing 1614 of the event and the input duration 1610 of the event) based on the patient's current glucose measurement, current active insulin, and current operating context, where the attributes associated with the future event are input or otherwise provided to the patient's forecast and prediction model based on the expected time entered by the patient.

[0157] In some embodiments, in order to adjust the model weighting factors, reliability metrics, or other aspects of the patient simulation process 1300 to take into account expected patient activities, the query process 300 can be performed to obtain data or information characterizing the responses of other similar patients to expected future events. For example, the query process 300 can be performed to identify similar patients based on common links or edges between nodes within the logical database layer and obtain historical measurement data of the blood glucose response of these similar patients to the input activity type for the input duration at the input time of day (e.g., sensor glucose measurement data when a similar patient jogs at 10 AM and continues for the next 30 minutes). The average or typical physiological response of the similar patients can then be used to adjust or otherwise enhance the physiological prediction model of the individual patient, which is then used by the ensemble prediction process 1100 and / or the patient simulation process 1300 to obtain the expected ensemble glucose prediction, which is influenced by the patient's hourly glucose forecasts, which are taken into account for the input future exercise (e.g., by inputting exercise attributes into the 10 AM LSTM unit) and combined with the adjusted physiological prediction for the patient's glucose level.

[0158] As another example, a patient may interact conversationally with client device 106, 602 to obtain a prediction of what their sensor glucose level may be upon waking in the morning. Based on the patient's historical event log data, the patient's estimated sleep time and / or estimated wake time may be determined, and this may be used to adjust model weighting factors and provided as input to the patient prediction model to obtain an expected aggregate prediction of the patient's glucose level at or around the estimated wake time. In one or more embodiments, the expected aggregate glucose prediction is also used to generate one or more recommendations for the patient. For example, if the expected aggregate glucose prediction at the estimated wake time is outside the target glucose range, a query process 300 may be executed to identify actions performed by similar patients at or before bedtime, or otherwise associated with overnight periods that resulted in changes in the patient's glucose levels, that, if corresponding increases or decreases were made relative to the current patient, would result in the expected aggregate glucose prediction at the estimated wake time being within the target range. In this regard, the query process 300 may be used to identify a recommended amount of carbohydrates the patient should consume, a recommended amount of insulin the patient should administer, and / or a recommended amount of exercise the patient should perform before bedtime, in order to achieve the desired glucose level upon waking. If the expected aggregate glucose prediction at the estimated wake-up time is within the target glucose value, other suggestions that may improve the patient's glucose regulation (e.g., increasing the percentage of the day within the target glucose range, minimizing glucose excursion events, etc.) can be determined based on similar patients and provided to the patient, for example, a recommended duration of sleep, a recommended amount of carbohydrates for the next day, a recommended amount of exercise for the next day, etc.

[0159] refer to Figures 8 to 16 And refer to Figure 1 In some embodiments, one or more of the processes 800, 1100, and 1300 can be implemented in conjunction with the patient data management system 100 and adapted to utilize the graph data structures in the database 104 to improve the accuracy of modeling and outcome prediction. In this regard, weighted directed or causal links between nodes or entities can be used to identify predictive relationships and corresponding impacts on patient outcomes for improved modeling, while such relationships may otherwise be impossible or computationally impractical to determine using conventional databases that rely on tables lacking causal and / or probabilistic relationships between entities.

[0160] Expectant treatment management

[0161] Now refer to Figure 17According to one or more embodiments, the risk management process 1700 utilizes measurement data related to the patient's physiological condition in combination with the patient's medical record data in order to calculate or otherwise determine a metric that indicates the patient's risk of experiencing a particular condition. For example, the patient's sensor glucose measurement data, or a metric calculated based on that data, may be utilized in combination with a subset of the patient's medical record data in order to calculate or otherwise determine a metric that indicates the patient's risk of experiencing one or more acute diabetic crises (e.g., severe hypoglycemia, acute diabetic ketoacidosis, hyperosmolarity, etc.) and / or long-term complications. In an exemplary embodiment, the risk management process 1700 generates or otherwise provides a notification or recommendation to an end user (e.g., a patient, a patient's healthcare provider, a patient's care partner, etc.) regarding a condition that the patient is at risk for. For purposes of illustration, the subject matter may be described herein in the context of providing notifications or recommendations to a patient, however, it should be understood that the subject matter described herein is not limited to the type of end user to whom the notifications or recommendations are provided. In some embodiments, according to one or more of processes 1800 and / or processes 1900, one or more treatment recommendations are provided to the patient, as described below in Figures 18 and 19 In the context of Figure 17 In the illustrated embodiment, a value indicating a metric of a patient's risk level for a particular condition can be used to adjust, modify, or otherwise affect the delivery of fluid by an infusion device 602 associated with the patient and / or otherwise alter the patient's treatment.

[0162] The various tasks performed in conjunction with the risk management process 1700 may be performed by hardware, firmware, software executed by processing circuitry, or any combination thereof. Figure 1 and Figures 6 and 7 elements mentioned. In practice, portions of the risk management process 1700 may be performed by different elements of the patient data management system 100 or the infusion system 600 (e.g., the server 102, the one or more electronic devices 106, the infusion device 602, and / or the pump control system 620, 700). It will be understood that the risk management process 1700 may include any number of additional or alternative tasks, which need not be performed in the order shown and / or which may be performed simultaneously, and / or the risk management process 1700 may be incorporated into a more comprehensive program or process having additional functionality not described in detail herein. Furthermore, the risk management process 1700 may be omitted from an actual implementation of the risk management process 1700 so long as the intended overall functionality remains unchanged. Figure 17 One or more tasks shown and described in the context of.

[0163] In the illustrated embodiment, the risk management process 1700 begins by receiving or otherwise obtaining measurement data and medical record data for a patient population (tasks 1702, 1704). For example, the server 102 may retrieve a subset of historical patient data 120 for the patient population and a corresponding subset of electronic medical record data 122 for the patient population from the database 104. In one embodiment, the historical patient data 120 and electronic medical record data 122 are obtained for all common patients across the dataset. However, in other embodiments, the patient population may be customized for a particular demographic or combination of demographic attributes (e.g., by age, gender, income, etc.).

[0164] After obtaining the measurement data and medical record data for the patient population, the risk management process 1700 determines a risk model for a particular condition based on the relationships between the measurement data and the medical record data across the patient population (task 1706). In an exemplary embodiment, stepwise feature selection, such as recursive feature elimination, is performed to identify which fields or attributes of the patient measurement data and medical record data are most correlated with or predictive of the occurrence of the particular condition within the patient population.

[0165] For example, the server 102 may analyze historical sensor glucose measurement data for a patient population to identify which sensor glucose metrics (e.g., average sensor glucose measurement value, sensor glucose measurement standard deviation, overnight average sensor glucose measurement value, percentage of time sensor glucose measurement value is within range, percentage of time sensor glucose measurement value is above a hyperglycemic threshold, percentage of time sensor glucose measurement value is below a hypoglycemic threshold, etc.) for a subset of the patient population that predict or are associated with the occurrence of a particular medical diagnosis code within the electronic medical records of the subset of the patient population. In this regard, for a given medical diagnosis code of interest (e.g., hypoglycemia, diabetic ketoacidosis, hyperosmolarity, cardiovascular disease, etc.), the server 102 may perform stepwise feature selection across different sensor glucose measurement metrics associated with the patient population to identify or determine a subset of sensor glucose measurement metrics that are associated with or predict the occurrence of a diagnosis code for that medical condition within the electronic medical record data 122. Similarly, for a medical condition of interest, server 102 can analyze the electronic medical record data of a patient population by performing stepwise feature selection to identify which fields or attributes of the patient's medical record (e.g., age, gender, income, education, smoking, A1C value or other laboratory value, insulin status or other medication or therapy, other medical conditions, etc.) are associated with or predict the occurrence of the disease condition. It should be noted that in some embodiments, the operational context data of the patient population can also be analyzed to identify whether any specific operational environment (e.g., geographic location, temperature, humidity, and / or similar conditions) is associated with or predicts the occurrence of a particular medical condition.

[0166] After identifying sensor glucose measurement variables and medical record variables that are associated with or predict the occurrence of a medical condition, the server 102 calculates or otherwise determines an equation, function, or model for calculating the probability or likelihood of the occurrence of the medical condition of interest based on a predicted subset of the sensor glucose measurement variables and medical record variables. For example, a risk prediction model for cardiovascular disease can calculate the probability of a patient suffering from cardiovascular disease in the future based on the patient's average sensor glucose measurement value, the sensor glucose measurement standard deviation, the percentage of time that the patient's sensor glucose measurement value is outside the target range, the patient's age, and whether the patient receives insulin treatment. Depending on the embodiment, the risk prediction model can calculate the risk probability within a limited future prediction range (e.g., within the next 18 months, within the patient's life expectancy, etc.) or within an infinite or unbounded duration. After determining the risk prediction model for various medical conditions and / or patient populations, the server 102 can store or maintain the risk prediction model for different medical conditions in a database 104 associated with the overall statistical criteria of the patient population used for the corresponding model. In other embodiments, the server 102 can transmit or push the risk prediction model to one or more client electronic devices 106, 602. In this regard, in some embodiments, the server 102 may periodically update the risk prediction model (e.g., weekly, monthly, yearly, etc.) to reflect new or more recent data in the database 104.

[0167] Still refer to Figure 17 , the risk management process 1700 shown receives or otherwise obtains measurement data and medical record data of an individual patient, and applies one or more risk prediction models to the patient's measurement data and medical record data to determine the patient's individual risk of experiencing a condition associated with the corresponding risk prediction model (tasks 1708, 1710, 1712). In this regard, the risk prediction model can be used to periodically or continuously analyze the sensor glucose measurement data and electronic medical record data of the individual patient to determine whether the patient's risk of having a specific condition is higher than a threshold risk tolerance. In one or more exemplary embodiments, the risk of the individual patient for a specific condition is analyzed or otherwise determined at the client device 106, 602 associated with the patient. In this regard, the client device 106, 602 can download or otherwise retrieve a risk prediction model for a condition that is not or has not yet been diagnosed for its associated patient from the database 104 via the server 102. The overall demographic information and medical record data associated with the patient can be used to identify the patient group to which the patient belongs, and then, for medical conditions for which there is no diagnosis code present in the patient's medical record data, a risk prediction model associated with the identifier patient group is selected.

[0168] After obtaining a risk prediction model for a specific medical condition, the client device 106, 602 uses the current or most recent sensor glucose measurement data associated with the patient to calculate or otherwise determine one or more inputs to the risk prediction model. In addition, the client device 106, 602 can download or otherwise retrieve fields of the patient's medical record data from the database 104 via the server 102, which are also used as inputs to the risk prediction model. The client device 106, 602 then uses equations, formulas, or functions associated with the risk prediction model to calculate or otherwise determine an output value representing the probability of the patient developing or experiencing a medical condition associated with the risk prediction model based on the patient's most recent measurement data and medical record fields, i.e., the patient's risk score for the condition. In addition, in an embodiment in which the risk prediction model uses background information as input, the client device 106, 602 can obtain the current operating context via one or more sensing devices 650, 660 at the client device 106, 602, and input the current operating context into the risk prediction model.

[0169] In an exemplary embodiment, when a patient's risk score is greater than a notification threshold, the risk management process 1700 generates or otherwise provides a user notification indicating a potential risk to the patient (tasks 1714, 1716). For example, a user notification may be generated or otherwise provided at a client device 106, 602 that identifies a medical condition for which the patient may be at risk of experiencing or exhibiting risk. In some embodiments, the risk management process 1700 generates or otherwise provides treatment recommendations based on the medical condition. In this regard, the patient's measurement data and / or medical record data input to the risk prediction model may be analyzed to identify or otherwise determine whether any input variables can be modified to lower the patient's risk score and provide the patient with recommended remedial actions. For example, a GUI display may be generated at the client device 106, 602 that includes recommended actions the patient can take to lower their average sensor glucose measurement (e.g., exercise, dietary changes, etc.) when a higher sensor glucose measurement predicts a specific medical condition. As another example, the GUI display at the client device 106, 602 may include recommended treatment changes (e.g., changing the type of treatment, adding a new medication, etc.). In this regard, the risk management process 1700 may initiate the process 1800 described below to identify which therapy modifications should be recommended to the patient to achieve a desired reduction in the patient's risk score.

[0170] Still refer to Figure 17In one or more exemplary embodiments, the risk management process 1700 adjusts or modifies the delivery of fluid by the infusion device based, at least in part, on the patient's risk score for a particular medical condition (task 1718). In this regard, based on the patient's risk score and / or the magnitude of the medical condition, the command generation application 710 can adjust one or more delivery commands to compensate for the patient's risk. For example, when the patient's risk score indicates that the patient's risk of a severe hypoglycemic event is greater than a threshold probability, the command generation application 710 can reduce the delivery command to mitigate the risk of a hypoglycemic event. Thus, even if the patient's current sensor glucose measurement or a glucose level predicted based on previous measurements or trends is above a hypoglycemic threshold, or is otherwise expected to remain within a target glucose value range, the command generation application 710 can reduce insulin delivery (e.g., by narrowing or reducing the delivery command, increasing the patient's target glucose level, increasing the patient's insulin sensitivity factor, etc.) to proactively account for the increased risk of hypoglycemia. As another example, when a patient's risk score indicates that the patient's risk of diabetic ketoacidosis or other acute hyperglycemic event is greater than a threshold probability, the command generation application 710 can increase delivery commands, reduce the patient's insulin sensitivity factor, and / or reduce the patient's target glucose value to mitigate the risk by increasing insulin in the patient's body. Thus, even if the patient's current sensor glucose measurement or predicted glucose level is below the hyperglycemic threshold, or is otherwise expected to remain within the target glucose value range, the command generation application 710 can increase insulin delivery to proactively reduce the patient's risk of a hyperglycemic event. In some embodiments, the risk management process 1700 can dynamically determine the risk score in real time in response to new or updated sensor glucose measurements and stop modifying the delivery command once the patient's risk for a particular condition drops below a threshold.

[0171] Now refer to Figure 18In one or more exemplary embodiments, a lift recommendation process 1800 is performed to identify a treatment recommendation that is likely to have the most favorable impact on the patient's physiological condition based on the patient's historical data (e.g., measurement data, event log data, background data, etc.) and medical record data. In some embodiments, the lift recommendation process 1800 can identify which therapy change or intervention is likely to have the greatest impact on aspects of the individual's physiological condition. In other embodiments, lift metrics are used in conjunction with cost, compliance, patient burden, and / or other metrics to apply cost-effectiveness analysis or similar optimization techniques to identify the best treatment recommendation for the patient. It should be noted that although the terms lift, lift modeling, and variations thereof may be used for purposes of explanation, the present subject matter is not limited to lift modeling. Therefore, in the absence of explicit indication, lift modeling should be understood to include any kind of incremental modeling of the impact of specific events or actions on specific outcomes, including true referral modeling, net referral modeling, and variations thereof.

[0172] The various tasks performed in conjunction with the improvement suggestion process 1800 may be performed by hardware, firmware, software executed by processing circuitry, or any combination thereof. Figure 1 and Figures 6 and 7 elements mentioned. In practice, portions of the escalation suggestion process 1800 may be performed by different elements of the patient data management system 100 or the infusion system 600 (e.g., the server 102, the one or more electronic devices 106, the infusion device 602, and / or the pump control system 620, 700). It should be understood that the escalation suggestion process 1800 may include any number of additional or alternative tasks, which tasks need not be performed in the order shown, and / or which tasks may be performed simultaneously, and / or the escalation suggestion process 1800 may be incorporated into a more comprehensive program or process having additional functionality not described in detail herein. Furthermore, as long as the intended overall functionality remains unchanged, Figure 18 One or more of the tasks shown and described in the context of may be omitted from an actual implementation of the promotion suggestion process 1800.

[0173] The improvement suggestion process 1800 receives or otherwise obtains historical patient data and medical record data for a patient population and then analyzes the relationships between the historical patient data and the medical record data to identify different patient groups for use in modeling the effects of different therapeutic interventions on the patient's physiological condition (tasks 1802, 1804, 1806). For example, the server 102 may retrieve the historical patient data 120 and the electronic medical record data 122 from the database 104 and then utilize machine learning to identify patient groups where different therapeutic interventions or changes have statistically significant improvements in physiological patient aspects within the patient group, such as a reduction in A1C laboratory values, a reduction in glucose excursion events, an increase in the percentage of time that sensor glucose measurements are within a target range, etc. In this regard, the patient groups may be defined by common demographic attributes (e.g., gender, income, etc.), common medical diagnoses, common treatment regimens or types of treatment (e.g., monotherapy patients, dual therapy patients, etc.), common medications or prescriptions, and / or other medical record commonalities. For example, in addition to defining patient groups statistically overall (e.g., by age, location, race, gender, socioeconomic status, occupation, etc.), clustering techniques can be used to characterize or define patient groups to categorize similar patients using other available data sets (e.g., mood records, program interactions, personal goals, etc.), which can be tracked, monitored, or recorded by an application at the client device 106, for example.

[0174] After identifying different patient groups to model for different therapeutic interventions, the lift recommendation process 1800 determines a lift model for calculating the impact of the corresponding therapeutic intervention on each patient group (task 1808). In this regard, the server 102 identifies sensor glucose measurement variables, medical record variables, and / or operational context variables that are correlated with or predictive of an improvement in physiological condition, and then, based on the identified subset of variables, calculates or otherwise determines an equation, function, or model for calculating a possible improvement in physiological condition. For example, stepwise feature selection can be performed to identify which fields or attributes of the patient measurement data and medical record data are most correlated with or predictive of an A1C reduction in the patient group. A lift model for calculating an estimated A1C reduction for patients within the particular patient group can then be determined based on the relevant subset of sensor glucose measurement variables, medical record variables, and / or operational context variables. In this regard, for each different patient cohort identified for a different potential therapeutic intervention, the server 102 can determine a boost model for calculating a metric indicating the effect of the corresponding therapeutic intervention on the physiological condition of the corresponding cohort of patients based on a subset of sensor glucose measurement variables, medical record variables, and / or operating environment variables. The boost model determined by the server 102 can be stored or otherwise maintained in the database 104 associated with the corresponding combination of patient cohort attributes and therapeutic interventions. In some embodiments, the server 102 can push or otherwise transmit the boost model to the client device 106, 602 associated with the patient classified within the corresponding patient cohort. It should be noted that boost modeling is not limited to stepwise feature selection, and in other embodiments, random forest analysis, logistic regression, and / or other machine learning or artificial intelligence techniques can be used to generate the boost model.

[0175] Still refer to Figure 18To determine a treatment recommendation for an individual patient, the lift recommendation process 1800 receives or otherwise obtains historical observed patient data and medical record data for the individual patient and then identifies or otherwise obtains lift models associated with patient groups that include the patient or into which the patient is categorized based on the patient's overall demographics, medical records, etc. (tasks 1810, 1812, 1814). In other words, the patient's medical records, measurement data, event log data, and / or current operating environment can be utilized to identify which lift models in the database 104 may be most relevant to the individual patient being analyzed. Thereafter, the lift recommendation process 1800 calculates or otherwise determines an impact or lift metric associated with each respective therapeutic intervention for the patient based on the patient's measurement data and medical record data and the corresponding lift models associated with the different therapeutic interventions (task 1816). In this regard, for each potential therapeutic intervention, the lift recommendation process 1800 may calculate or otherwise determine an estimated A1C reduction or other estimate of the lift or impact associated with the corresponding therapeutic intervention for the patient based on the patient's medical records, measurement data, and / or current operating environment.

[0176] After determining the lift metrics for different possible therapeutic interventions for the patient, the lift suggestion process 1800 determines a therapeutic intervention recommendation based on the lift metrics and generates or otherwise provides an indication of the recommended therapeutic intervention to the patient (tasks 1818, 1820). For example, in one embodiment, the lift suggestion process 1800 identifies the therapeutic intervention with the greatest estimated impact or benefit (e.g., the greatest estimated A1C reduction) as the recommended therapeutic intervention for the patient. In other embodiments, the lift suggestion process 1800 performs a cost-effectiveness analysis or other optimization based on the estimated lift values associated with different potential therapeutic interventions and the costs associated with the respective possible therapeutic interventions to identify the best therapeutic intervention. For example, the lift suggestion process 1800 identifies the therapeutic intervention with the highest ratio of estimated lift value to cost as the recommended therapeutic intervention. In this regard, in some embodiments, the claims data 124 maintained in the database 104 can be used to calculate or otherwise determine an estimated cost associated with a particular therapeutic intervention, which in turn can be used to determine the relative impact or benefit of the therapeutic intervention (e.g., by dividing the lift value by the estimated cost).

[0177] In one or more embodiments, the escalation recommendation process 1800 identifies or otherwise determines the optimal therapeutic intervention based on: estimated escalation values associated with different potential therapeutic interventions, costs associated with the different potential therapeutic interventions, and estimated adherence metrics associated with the different potential therapeutic interventions. In this regard, some embodiments of the escalation recommendation process 1800 may calculate or otherwise determine an adherence metric that represents the likelihood that a patient will adhere to a particular therapeutic intervention, as described below in Figure 19 As described in more detail in the context of the adherence recommendation process 1900, in some embodiments, the improvement recommendation process 1800 can identify the following therapeutic interventions for recommendation as optimal: therapeutic interventions that do not have the highest estimated improvement value, but have relatively lower costs and / or higher adherence than one or more therapeutic interventions with higher estimated improvement values. Thus, patients can be advised of the most cost-effective and more likely successful treatments based on their likelihood of adherence to the recommended treatments.

[0178] Now refer to Figure 19 In one or more exemplary embodiments, the adherence recommendation process 1900 can be performed to identify treatment recommendations that an individual patient is most likely to adhere to or will otherwise result in the highest level of adherence. In this regard, in the exemplary embodiments described herein, adherence modeling is used to determine an adherence metric that represents the respective probability that a patient will adhere to a particular treatment regimen, for example, by taking a fully prescribed treatment regimen or engaging in some other action prescribed by the corresponding treatment regimen or indicating an attempt to achieve the corresponding treatment regimen (e.g., fulfilling a prescription within a threshold amount of time after being written). Using the adherence metric, therapeutic interventions that are likely to result in better adherence (and thus more likely to have a beneficial outcome relative to a treatment regimen that is less likely to have that level of adherence) can be recommended to the patient. For example, for a given patient, if the adherence metric value associated with an injectable insulin treatment regimen (e.g., 15%) is lower than the adherence metric value for an oral medication (e.g., 50%), then the oral medication is recommended because it is likely to provide a greater improvement in adherence given its associated adherence.

[0179] The various tasks performed in conjunction with the compliance recommendation process 1900 may be performed by hardware, firmware, software executed by processing circuitry, or any combination thereof. Figure 1 and Figures 6 and 7elements mentioned. In practice, portions of the compliance recommendation process 1900 may be performed by different elements of the patient data management system 100 or the infusion system 600 (e.g., the server 102, the one or more electronic devices 106, the infusion device 602, and / or the pump control system 620, 700). It should be understood that the compliance recommendation process 1900 may include any number of additional or alternative tasks, which tasks need not be performed in the order shown, and / or which tasks may be performed simultaneously, and / or the compliance recommendation process 1900 may be incorporated into a more comprehensive program or process having additional functionality not described in detail herein. Furthermore, as long as the intended overall functionality remains unchanged, Figure 19 One or more of the tasks shown and described in the context of may be omitted from an actual implementation of compliance suggestion process 1900.

[0180] The adherence suggestion process 1900 receives or otherwise obtains historical observational data, medical record data, and medical claims data for a patient population from a database (tasks 1902, 1904, and 1906). The adherence suggestion process 1900 calculates or otherwise determines adherence metrics for different therapeutic interventions or regimens based on the relationships between the historical observational data, medical record data, and medical claims data for the patient population (task 1908). For example, for each patient whose corresponding medical record and medical claims data are stored in database 104, server 102 may analyze the relationship between the patient's prescriptions and other treatment information from the patient's medical record data, as well as the number and / or frequency of the patient's medical claims corresponding to those prescriptions or treatments in the patient's claims data, to determine an adherence metric associated with the patient for the corresponding treatment based on the patient's claims data indicating a degree of adherence or alignment with the patient's prescribed treatment. In this regard, a patient whose claims data indicates that the prescription was fulfilled with a prescribed frequency or relatively small delay after writing the prescription may be assigned a relatively high adherence value, while a patient whose claims data indicates that the prescription was not fulfilled regularly or in a timely manner may be assigned a relatively low adherence value. Additionally, in some embodiments, event log data or other observed patient data may also be utilized when determining adherence metrics. For example, a patient's event log data may indicate when a patient took a prescribed medication and the corresponding dosage, which may then be compared with prescription information from the patient's medical record data to determine whether the patient's behavior adhered to the patient's prescribed treatment.

[0181] In the illustrated embodiment, after determining the adherence metrics associated with different therapies for different patients, the adherence suggestion process 1900 proceeds to analyze the relationships between the observed patient data, medical records, claims data, and adherence values to determine adherence models for calculating adherence metrics for the different therapies based on the observed data, medical record data, and claims data for the individual patients (task 1910). In this regard, for a subset of patients who share a particular treatment regimen, the server 102 identifies observed patient variables (e.g., sensor glucose measurement variables, meal, exercise, or other event log variables, operating environment variables, etc.), medical record variables (e.g., population demographics, medical conditions, and / or claims data variables (e.g., refill data for previous prescriptions, etc.)) that are associated with or predictive of a patient's adherence metric for the treatment regimen, and then calculates or otherwise determines an equation, function, or model for calculating a possible adherence metric value for the given patient based on the identified subset of variables associated with the prospective patient. For example, stepwise feature selection or other machine learning techniques can be performed to identify which fields or attributes of the historically observed patient data 120 and medical record data 122 are most correlated with or predictive of adherence metrics among patients prescribed a corresponding prescribed treatment regimen. A compliance model can then be determined based on the relevant subset of variables for calculating estimated adherence for patients not currently taking the treatment regimen. In this regard, for each different potential treatment regimen or intervention, the server 102 can determine an compliance model for calculating a metric indicative of likely adherence to the corresponding treatment regimen based on existing patients not prescribed the corresponding treatment regimen. The compliance model determined by the server 102 can be stored or otherwise maintained in the database 104 associated with the corresponding treatment regimen or intervention, or pushed or otherwise transmitted to the client device 106, 602.

[0182] Still refer to Figure 19 To determine a treatment recommendation for an individual patient, the adherence recommendation process 1900 receives or otherwise obtains observational data, medical record data, and claims data for the individual patient and then applies various adherence models to different treatment regimens not currently prescribed to the patient in order to calculate or otherwise determine adherence metrics for the different treatment regimens (tasks 1912, 1914). In this regard, the server 102 or client device 106, 602 obtains data associated with the patient of interest from the database 104, as well as recent measurement data and / or operational context information from the client device 106, 602 associated with the patient, and then utilizes the adherence model to estimate the patient's likely adherence to each of the different possible treatment regimens not currently prescribed for the patient.

[0183] In an exemplary embodiment, the adherence suggestion process 1900 determines a treatment recommendation for the patient based on the adherence metric values associated with different possible treatment regimens, and generates or otherwise provides an indication of the treatment recommendation to the patient or another user (e.g., a physician, a healthcare provider, etc.) (tasks 1916, 1918). In some embodiments, the adherence suggestion process 1900 selects or otherwise identifies the treatment regimen with the highest adherence metric value as the recommended treatment for the patient. In other embodiments, the adherence metric values associated with the different possible treatment regimens are considered in conjunction with the lift metric values and / or estimated costs associated with the different possible treatment regimens to identify the best treatment regimen, as described above in Figure 19 , described in the context of . For example, in an embodiment where the adherence metric represents a probability or percentage, the lift metric value associated with a possible treatment regimen for a patient of interest can be scaled or otherwise multiplied by the adherence metric value associated with the treatment regimen for the patient of interest to obtain a possible lift value for the patient that represents the possible benefit after taking into account adherence. In one embodiment, the recommended therapy can be selected as the treatment regimen with the highest possible lift value to cost ratio (e.g., the product of the lift metric value and the adherence probability divided by the estimated cost). A GUI display can be generated or otherwise provided at a client device 106, 602 that indicates the recommended treatment to the patient or other user of the client device 106, 602. In one embodiment, the GUI display can include a list of possible therapies that are categorized, prioritized, or otherwise sorted in a manner influenced by the adherence metric value such that the highest priority therapy corresponds to the treatment regimen recommended based on the adherence metric.

[0184] refer to Figures 17 to 19 It should be noted that in some embodiments, processes 1700, 1800, and 1900 may be combined Figure 1 The patient data management system 100 is implemented in accordance with the present invention and is adapted to utilize the graph data structure in the database 104 to improve the accuracy of modeling. In this regard, weighted directed or causal links between nodes or entities can be used to identify predictive relationships and corresponding impacts on patient outcomes for improved modeling. Additionally, shared links within or between logical database layers can be utilized to identify commonalities between patients that might not otherwise be readily identified using conventional databases that rely on tables lacking causal and / or probabilistic relationships between entities.

[0185] Infusion system integration

[0186] Figure 20An exemplary embodiment of an infusion system 2000 suitable for use with the above-described subject matter is depicted. For example, a computer 2008 (e.g., computing device 102) can communicate with and / or obtain data from various client electronic devices (e.g., electronic device 106), such as a fluid infusion device (or infusion pump) 2002 (e.g., infusion device 602), a sensing device 2004 (e.g., glucose sensing device 604), and a command and control device (CCD) 2006. The components of the infusion system 2000 may be implemented using different platforms, designs, and configurations, and Figure 20 The embodiments shown are not intended to be exhaustive or limiting. In practice, the infusion device 2002 and the sensing device 2004 are fixed to the desired location on the user's (or patient's) body, such as Figure 20 In this regard, the infusion device 2002 and the sensing device 2004 are Figure 20 The locations where the infusion system 2000 is secured to the user's body are provided merely as representative, non-limiting examples. The elements of the infusion system 2000 may be similar to those described in U.S. Patent No. 8,674,288, the subject matter of which is hereby incorporated by reference in its entirety.

[0187] exist Figure 20 In the exemplary embodiment of the invention, the infusion device 2002 is designed as a portable medical device suitable for infusing fluids, liquids, gels, or other pharmaceutical agents into the body of a user. In the exemplary embodiment, the infused fluid is insulin, but many other fluids can be administered by infusion, such as, but not limited to, HIV medications, medications for treating pulmonary hypertension, iron chelation drugs, painkillers, anti-cancer drugs, vitamins, hormones, etc. In some embodiments, the fluids can include nutritional supplements, dyes, tracking media, saline media, hydration media, etc.

[0188] The sensing device 2004 generally represents a component of the infusion system 2000 configured to sense, detect, measure, or otherwise quantify a condition of the user, and may include sensors, monitors, and the like for providing data indicative of the condition sensed, detected, measured, or otherwise monitored by the sensing device. In this regard, the sensing device 2004 may include electronics and enzymes responsive to a biological condition of the user, such as blood glucose level, and provide data indicative of the blood glucose level to the infusion device 2002, the CCD 2006, and / or the computer 2008. For example, the infusion device 2002, the CCD 2006, and / or the computer 2008 may include a display for presenting information or data to the user based on the sensor data received from the sensing device 2004, such as, for example, the user's current glucose level, a graph or chart of the user's glucose level over time, a device status indicator, an alarm message, and the like. In other embodiments, the infusion device 2002, the CCD 2006, and / or the computer 2008 may include electronics and software configured to analyze sensor data and operate the infusion device 2002 to deliver fluid to the user's body based on the sensor data and / or a preprogrammed delivery plan. Thus, in an exemplary embodiment, one or more of the infusion device 2002, the sensing device 2004, the CCD 2006, and / or the computer 2008 include a transmitter, a receiver, and / or other transceiver electronics that allow communication with other components of the infusion system 2000, such that the sensing device 2004 can transmit sensor data or monitor data to one or more of the infusion device 2002, the CCD 2006, and / or the computer 2008.

[0189] Still refer to Figure 20 In various embodiments, the sensing device 2004 can be affixed to the user's body or embedded in the user's body at a location remote from where the infusion set 2002 is affixed to the user's body. In various other embodiments, the sensing device 2004 can be incorporated into the infusion set 2002. In other embodiments, the sensing device 2004 can be separate and distinct from the infusion set 2002 and can be, for example, part of the CCD 2006. In such embodiments, the sensing device 2004 can be configured to receive a biological sample, analyte, etc. to measure a condition of the user.

[0190] In some embodiments, the CCD 2006 and / or computer 2008 may include electronics and other components configured to perform processing, deliver daily doses, and control the infusion device 2002 in a manner influenced by sensor data measured by and / or received from the sensing device 2004. By including control functionality in the CCD 2006 and / or computer 2008, the infusion device 2002 can be made with more simplified electronics. However, in other embodiments, the infusion device 2002 can include all control functionality and can operate without the CCD 2006 and / or computer 2008. In various embodiments, the CCD 2006 can be a portable electronic device. Additionally, in various embodiments, the infusion device 2002 and / or sensing device 2004 can be configured to transmit data to the CCD 2006 and / or computer 2008 for display or processing by the CCD 2006 and / or computer 2008.

[0191] In some embodiments, CCD 2006 and / or computer 2008 can provide information to the user that facilitates the user's subsequent use of infusion device 2002. For example, CCD 2006 can provide information to the user to allow the user to determine the rate or dosage of a drug to be administered to the user's body. In other embodiments, CCD 2006 can provide information to infusion device 2002 to autonomously control the rate or dosage of a drug administered to the user's body. In some embodiments, sensing device 2004 can be integrated into CCD 2006. Such embodiments can allow a user to assess their condition by, for example, providing a blood sample to sensing device 2004 to monitor the condition. In some embodiments, sensing device 2004 and CCD 2006 can be used to determine the glucose level in the user's blood and / or body fluids without using or requiring a wired or cable connection between infusion device 2002 and sensing device 2004 and / or CCD 2006.

[0192] In some embodiments, the sensing device 2004 and / or the infusion device 2002 are cooperatively configured to utilize a closed-loop system to deliver fluid to the user. Examples of sensing devices and / or infusion pumps utilizing closed-loop systems can be found in, but are not limited to, the following U.S. Patents: No. 6,088,608, No. 6,119,028, No. 6,589,229, No. 6,740,072, No. 6,827,702, No. 7,323,142, and No. 7,402,153, or U.S. Patent Application Publication No. 2014 / 0066889, all of which are incorporated herein by reference in their entirety. In such embodiments, the sensing device 2004 is configured to sense or measure a condition of the user, such as blood sugar level. The infusion device 2002 is configured to deliver fluid in response to a condition sensed by the sensing device 2004. In turn, the sensing device 2004 continues to sense or otherwise quantify the user's current condition, thereby allowing the infusion device 2002 to continue delivering fluid indefinitely in response to the condition currently (or most recently) sensed by the sensing device 2004. In some embodiments, the sensing device 2004 and / or infusion device 2002 can be configured to utilize the closed-loop system only during a portion of the day, such as only when the user is asleep or awake.

[0193] Figures 21 to 23 An exemplary embodiment of a fluid infusion device 2100 (or alternatively, an infusion pump) suitable for use in an infusion system is shown, for example, as Figure 6 The infusion device 602 in the infusion system 600 or Figure 20 The infusion device 2002 in the infusion system 2000 of FIG. The fluid infusion device 2100 is a portable medical device designed to be carried or worn by a patient (or user), and the fluid infusion device 2100 can utilize any number of conventional features, components, elements, and properties of existing fluid infusion devices, such as, for example, some of the features, components, elements, and / or properties described in U.S. Patent Nos. 6,485,465 and 7,621,893. It should be understood that Figures 21 to 23 Certain aspects of the infusion device 2100 are shown in a simplified manner; in practice, the infusion device 2100 may include additional elements, features, or components not shown or described in detail herein.

[0194] like Figures 21 to 22As shown, the exemplary embodiment of the fluid infusion device 2100 includes a housing 2102 adapted to receive a reservoir 2105 containing a fluid. An opening 2120 in the housing 2102 accommodates an accessory 2123 (or cap) for the reservoir 2105, wherein the accessory 2123 is configured to cooperate with or otherwise connect to a conduit 2121 of an infusion set 2125 to provide a fluid path to / from the user's body. In this way, fluid communication is established from the interior of the reservoir 2105 to the user via the conduit 2121. The illustrated fluid infusion device 2100 includes a human-machine interface (HMI) 2130 (or user interface) comprising elements 2132, 2134 that can be manipulated by the user to administer a bolus of fluid (e.g., insulin), change treatment settings, change user preferences, select display features, and the like. The infusion device also includes a display element 2126 such as a liquid crystal display (LCD) or another suitable display element, which can be used to present various types of information or data to the user, such as, but not limited to: the patient's current glucose level; the time; a graph or chart of the patient's glucose level relative to time; a device status indicator, etc.

[0195] The housing 2102 is formed of a substantially rigid material having a hollow interior space 2114 suitable for allowing an electronic component 2104, a sliding member (or slider) 2106, a drive system 2108, a sensor assembly 2110 and a drive system cover member 2112 in addition to the reservoir 2105 to be disposed therein, wherein the contents of the housing 2102 are enclosed by a housing cover member 2116. The opening 2120, the slide 2106, and the drive system 2108 are coaxially aligned in an axial direction (indicated by arrow 2118), whereby the drive system 2108 facilitates linear displacement of the slide 2106 in the axial direction 2118 to dispense fluid from the reservoir 2105 (after the reservoir 2105 has been inserted into the opening 2120), wherein the sensor assembly 2110 is configured to measure an axial force (e.g., a force aligned with the axial direction 2118) applied to the sensor assembly 2110 in response to operation of the drive system 2108 to displace the slide 2106. In various embodiments, the sensor assembly 2110 can be used to detect one or more of the following: slowing, preventing, or otherwise reducing a blockage in the fluid path of fluid delivery from the reservoir 2105 to the user's body; when the reservoir 2105 is empty; when the slide 2106 is properly seated with the reservoir 2105; when a dose of fluid has been delivered; when the infusion pump 2100 is subjected to shock or vibration; or when the infusion pump 2100 requires maintenance.

[0196] Depending on the embodiment, the reservoir 2105 containing the fluid may be implemented as a syringe, a vial, a cartridge, a bag, etc. In certain embodiments, the infused fluid is insulin, but many other fluids may be administered by infusion, such as, but not limited to, HIV medications, medications for treating pulmonary hypertension, iron chelation drugs, pain medications, anti-cancer treatments, vitamins, hormones, etc. Figures 22 to 23 As best shown, the reservoir 2105 generally includes a reservoir barrel 2119 that contains fluid and is concentrically and / or coaxially aligned (e.g., in an axial direction 2118) with the slider 2106 when the reservoir 2105 is inserted into the infusion pump 2100. The end of the reservoir 2105 proximate the opening 2120 may include or otherwise cooperate with a fitting 2123 that secures the reservoir 2105 in the housing 2102 and prevents displacement of the reservoir 2105 in the axial direction 2118 relative to the housing 2102 after the reservoir 2105 is inserted into the housing 2102. As described above, the fitting 2123 extends from (or through) the opening 2120 of the housing 2102 and cooperates with the tubing 2121 to establish fluid communication from the interior of the reservoir 2105 (e.g., the reservoir barrel 2119) to the user via the tubing 2121 and the infusion set 2125. The opposite end of the reservoir 2105 from the slider 2106 includes a plunger 2117 (or stopper) positioned to push fluid from the interior of the barrel 2119 of the reservoir 2105 along a fluid path through the conduit 2121 to the user. The slider 2106 is configured to mechanically couple or otherwise engage with the plunger 2117, thereby becoming seated with the plunger 2117 and / or the reservoir 2105. When the drive system 2108 is operated to displace the slider 2106 in the axial direction 2118 toward the opening 2120 in the housing 2102, fluid is forced out of the reservoir 2105 via the conduit 2121.

[0197] exist Figures 22 to 23In the illustrated embodiment, the drive system 2108 includes a motor assembly 2107 and a drive screw 2109. The motor assembly 2107 includes a motor coupled to a drive train component of the drive system 2108 that is configured to convert the rotational motor motion into translational displacement of the slider 2106 in an axial direction 2118, and thereby engage and displace the plunger 2117 of the reservoir 2105 in the axial direction 2118. In some embodiments, the motor assembly 2107 can also be powered to translate the slider 2106 in an opposite direction (e.g., opposite to the direction 2118) to retract and / or remove the slider from the reservoir 2105 to allow replacement of the reservoir 2105. In an exemplary embodiment, the motor assembly 2107 includes a brushless direct current (BLDC) motor having one or more permanent magnets mounted, attached, or otherwise disposed on its rotor. However, the subject matter described herein is not necessarily limited to use with a BLDC motor, and in alternative embodiments, the motor may be implemented as a solenoid motor, an AC motor, a stepper motor, a piezoelectric track drive, a shape memory actuator drive, an electrochemical gas cell, a thermally driven gas cell, a bimetallic actuator, etc. The drive train components may include one or more lead screws, cams, pawls, jacks, pulleys, pawls, clamps, gears, nuts, slides, bearings, levers, beams, stops, plungers, slides, brackets, guides, bearings, supports, bellows, caps, diaphragms, bags, heaters, etc. In this regard, while the illustrated embodiment of the infusion pump utilizes a coaxially aligned drive train, the motor may be offset relative to the longitudinal axis of the reservoir 2105 or arranged in another non-coaxial manner.

[0198] like Figure 23As best shown, the drive screw 2109 mates with threads 2302 on the interior of the slider 2106. When the motor assembly 2107 is powered and operated, the drive screw 2109 rotates and forces the slider 2106 to translate in an axial direction 2118. In an exemplary embodiment, the infusion pump 2100 includes a sleeve 2111 to prevent the slider 2106 from rotating when the drive screw 2109 of the drive system 2108 rotates. Thus, rotation of the drive screw 2109 causes the slider 2106 to extend or retract relative to the drive motor assembly 2107. When the fluid infusion device is assembled and operational, the slider 2106 contacts the plunger 2117 to engage the reservoir 2105 and control the delivery of fluid from the infusion pump 2100. In an exemplary embodiment, a shoulder portion 2115 of the slider 2106 contacts or otherwise engages the plunger 2117 to displace the plunger 2117 in the axial direction 2118. In an alternative embodiment, the slider 2106 can include a threaded tip 2113 that can be removably engaged with internal threads 2304 on the plunger 2117 of the reservoir 2105, as described in detail in US Patent Nos. 6,248,093 and 6,485,465, which are incorporated herein by reference.

[0199] like Figure 22 As shown, the electronic assembly 2104 includes a control electronic device 2124 coupled to a display element 2126, wherein the housing 2102 includes a transparent window portion 2128 aligned with the display element 2126 to allow a user to view the display 2126 when the electronic assembly 2104 is disposed within the interior 2114 of the housing 2102. The control electronic device 2124 generally represents the hardware, firmware, processing logic and / or software (or a combination thereof) configured to control the operation of the motor assembly 2107 and / or the drive system 2108. Whether such functionality is implemented as hardware, firmware, a state machine or software depends on the specific application and design constraints imposed on the embodiment. Concepts similar to those described herein can implement such functionality in a manner suitable for each specific application, but such specific implementation decisions should not be interpreted as being limited or restrictive. In an exemplary embodiment, the control electronic device 2124 includes one or more programmable controllers that can be programmed to control the operation of the infusion pump 2100.

[0200] The motor assembly 2107 includes one or more electrical leads 2136 adapted to be electrically coupled to the electronics assembly 2104 to establish communication between the control electronics 2124 and the motor assembly 2107. In response to a command signal from the control electronics 2124 that operates a motor driver (e.g., a power converter) to adjust the amount of power supplied to the motor from the power source, the motor actuates the drive train components of the drive system 2108 to displace the slider 2106 in the axial direction 2118, forcing the fluid to flow out of the reservoir 2105 along the fluid path (including the tubing 2121 and the infusion set), thereby administering a dose of the fluid contained in the reservoir 2105 to the user's body. Preferably, the power source is implemented as one or more batteries housed within the housing 2102. Alternatively, the power source may be a solar panel, a capacitor, AC or DC power supplied via a power cord, or the like. In some embodiments, the control electronics 2124 may operate the motor assembly 2107 and / or the motor of the drive system 2108 in a stepwise manner, typically on an intermittent basis; administering separate, precise doses of fluid to the user according to a programmed delivery profile.

[0201] refer to Figures 21 to 23As described above, the user interface 2130 includes HMI elements such as buttons 2132 and arrow keys 2134 formed on a graphical keypad overlay 2131 that overlies a keypad assembly 2133, which includes features corresponding to the buttons 2132, arrow keys 2134, or other user interface items indicated by the graphical keypad overlay 2131. When assembled, the keypad assembly 2133 is coupled to the control electronics 2124, thereby allowing a user to manipulate the HMI elements 2132, 2134 to interact with the control electronics 2124 and control the operation of the infusion pump 2100, such as to administer a bolus of insulin, change therapy settings, change user preferences, select display features, set or disable alarms and reminders, etc. In this regard, the control electronics 2124 maintains and / or provides information to the display 2126 regarding program parameters, delivery profiles, pump operation, alarms, warnings, status, etc., which can be adjusted using the HMI elements 2132, 2134. In various embodiments, the HMI elements 2132, 2134 can be implemented as physical objects (e.g., buttons, knobs, joysticks, etc.) or virtual objects (e.g., using touch sensing and / or proximity sensing technology). For example, in some embodiments, the display 2126 can be implemented as a touch screen or touch-sensitive display, and in such embodiments, the features and / or functionality of the HMI elements 2132, 2134 can be integrated into the display 2126, and the HMI 2130 may not be present. In some embodiments, the electronic assembly 2104 can also include an alarm generation element that is coupled to the control electronics 2124 and is suitably configured to generate one or more types of feedback, such as, but not limited to, auditory feedback, visual feedback, tactile (physical) feedback, etc.

[0202] refer to Figures 22 to 23According to one or more embodiments, the sensor assembly 2110 includes a backplate structure 2150 and a loading element 2160. The loading element 2160 is disposed between the cover member 2112 and a beam structure 2170, which includes one or more beams having sensing elements disposed thereon that are affected by a compressive force applied to the sensor assembly 2110 that causes the one or more beams to deflect, as described in more detail in U.S. Patent No. 8,474,332, which is incorporated herein by reference. In an exemplary embodiment, the backplate structure 2150 is attached, adhered, mounted, or otherwise mechanically coupled to the bottom surface 2138 of the drive system 2108 such that the backplate structure 2150 is located between the bottom surface 2138 of the drive system 2108 and the housing cover 2116. The drive system cover member 2112 is contoured to accommodate and match the bottom of the sensor assembly 2110 and the drive system 2108. A drive system capping member 2112 can be attached to the interior of the housing 2102 to prevent the sensor assembly 2110 from being displaced in a direction opposite to the direction of the force provided by the drive system 2108 (e.g., opposite to direction 2118). Thus, the sensor assembly 2110 is positioned between the motor assembly 2107 and the capping member 2112 and is secured by the capping member, which prevents the sensor assembly 2110 from being displaced in a downward direction opposite to the direction of arrow 2118, such that when the drive system 2108 and / or the motor assembly 2107 are operated to displace the slider 2106 in an axial direction 2118 opposite to the pressure of the fluid in the reservoir 2105, the sensor assembly 2110 is subjected to a reactive compressive force. Under normal operating conditions, the compressive force applied to the sensor assembly 2110 is related to the pressure of the fluid in the reservoir 2105. As shown, the electrical lead 2140 is suitable for electrically coupling the sensing element of the sensor assembly 2110 to the electronic assembly 2104 to establish communication with the control electronic device 2124, wherein the control electronic device 2124 is configured to measure, receive or otherwise obtain an electrical signal from the sensing element of the sensor assembly 2110, which is indicative of the force applied by the drive system 2108 in the axial direction 2118.

[0203] Figure 24An exemplary embodiment of a patient monitoring system 2400 suitable for use with the subject matter described herein is depicted. The patient monitoring system 2400 includes a medical device 2402 communicatively coupled to a sensing element 2404 that is inserted into the body of a patient or worn by the patient to obtain measurement data indicative of a physiological condition in the patient's body, such as a sensed glucose level. The medical device 2402 is communicatively coupled to a client device 2406 via a communication network 2410, wherein the client device 2406 is communicatively coupled to a remote device 2414 via another communication network 2412. In this regard, the client device 2406 can serve as an intermediary for uploading or otherwise providing measurement data from the medical device 2402 (e.g., the server 102) to the remote device 2414. It should be understood that for purposes of illustration, Figure 24 A simplified representation of a patient monitoring system 2400 is shown and is not intended to limit the subject matter described herein in any way.

[0204] In an exemplary embodiment, client device 2406 is implemented as a mobile phone, smartphone, tablet computer, or other similar mobile electronic device; however, in other embodiments, client device 2406 can be implemented as any type of electronic device capable of communicating with medical device 2402 via network 2410, such as a laptop or notebook computer, desktop computer, etc. In an exemplary embodiment, network 2410 is implemented as a Bluetooth network, a ZigBee network, or other suitable personal area network. That is, in other embodiments, network 2410 can be implemented as a wireless ad hoc network, a wireless local area network (WLAN), or a local area network (LAN). Client device 2406 includes or is coupled to a display device (such as a monitor, screen, or another conventional electronic display) that can graphically present data and / or information related to the patient's physiological condition. Client device 2406 also includes or is otherwise associated with a user input device (such as a keyboard, mouse, touch screen, etc.) that can receive input data and / or other information from a user of client device 2406.

[0205] In an exemplary embodiment, a user (such as a patient, the patient's doctor, or another healthcare provider) manipulates a client device 2406 to execute a client application 2408 that supports communication with a medical device 2402 via a network 2410. In this regard, the client application 2408 supports establishing a communication session with the medical device 2402 over the network 2410, and receiving data and / or information from the medical device 2402 via the communication session. The medical device 2402 can similarly execute or otherwise implement a corresponding application or process that supports establishing a communication session with the client application 2408. The client application 2408 generally represents a software module or another feature generated or otherwise implemented by the client device 2406 to support the processes described herein. Thus, the client device 2406 generally includes a processing system and a data storage element (or memory) capable of storing programming instructions for execution by the processing system, which, when read and executed, causes the processing system to create, generate, or otherwise facilitate the client application 2408 and perform or otherwise support the processes, tasks, operations, and / or functions described herein. Depending on the embodiment, the processing system can be implemented using any suitable processing system and / or device, such as one or more processors, central processing units (CPUs), controllers, microprocessors, microcontrollers, processing cores and / or other hardware computing resources configured to support the operation of the processing system described herein. Similarly, the data storage element or memory can be implemented as random access memory (RAM), read-only memory (ROM), flash memory, magnetic or optical mass storage, or any other suitable non-transitory short-term or long-term data storage or other computer-readable media and / or any suitable combination thereof.

[0206] In one or more embodiments, the client device 2406 and the medical device 2402 establish an association (or pairing) with each other via the network 2410 to support subsequent establishment of a point-to-point or peer-to-peer communication session between the medical device 2402 and the client device 2406 via the network 2410. For example, according to one embodiment, the network 2410 is implemented as a Bluetooth network, wherein the medical device 2402 and the client device 2406 are paired with each other by performing a discovery process or other suitable pairing process (e.g., by obtaining and storing each other's network identification information). The pairing information obtained during the discovery process allows either the medical device 2402 or the client device 2406 to initiate establishment of a secure communication session via the network 2410.

[0207] In one or more exemplary embodiments, the client application 2408 is further configured to store or otherwise maintain the address and / or other identifying information of the remote device 2414 on a second network 2412. In this regard, the second network 2412 can be physically and / or logically distinct from the network 2410, such as the Internet, a cellular network, a wide area network (WAN), or the like. The remote device 2414 generally represents a server or other computing device configured to receive and analyze or otherwise monitor measurement data, event log data, and other possible information obtained for a patient associated with the medical device 2402. In an exemplary embodiment, the remote device 2414 is coupled to a database 2416 (e.g., database 104), which is configured to store or otherwise maintain data associated with individual patients. In practice, the remote device 2414 can reside in a location physically distinct and / or separate from the medical device 2402 and the client device 2406, such as at a facility owned and / or operated by or otherwise affiliated with the manufacturer of the medical device 2402. For purposes of illustration, but not limitation, remote device 2414 may optionally be referred to herein as a server.

[0208] still Figure 24 , sensing element 2404 generally represents a component of patient monitoring system 2400 configured to generate, produce, or otherwise output one or more electrical signals indicative of a physiological condition sensed, measured, or otherwise quantified by sensing element 2404. In this regard, the physiological condition of the user will affect the characteristics of the electrical signal output by sensing element 2404 such that the characteristics of the output signal correspond to or are otherwise correlated with the physiological condition to which sensing element 2404 is sensitive. In an exemplary embodiment, sensing element 2404 is implemented as an interstitial glucose sensing element that is inserted into a location on the patient's body and generates an output electrical signal having an electrical current (or voltage) associated therewith that is correlated to an interstitial fluid glucose level sensed or otherwise measured in the patient's body by sensing element 2404.

[0209] The medical device 2402 generally represents a component of the patient monitoring system 2400 that is communicatively coupled to the output of the sensing element 2404 to receive or otherwise obtain measurement data samples (e.g., measured glucose and characteristic impedance values) from the sensing element 2404, store or otherwise maintain the measurement data samples, and upload or otherwise transmit the measurement data to the server 2414 via the client device 2406. In one or more embodiments, the medical device 2402 is implemented as an infusion device 602, 2002 that is configured to deliver a fluid, such as insulin, to the patient's body. That is, in other embodiments, the medical device 2402 can be a stand-alone sensing or monitoring device that is separate and independent from the infusion device (e.g., sensing device 604, 2004). It should be noted that although Figure 24 Medical device 2402 and sensing element 2404 are depicted as separate components, but in practice, medical device 2402 and sensing element 2404 may be integrated or otherwise combined to provide a unitary device wearable by a patient.

[0210] In an exemplary embodiment, medical device 2402 includes a control module 2422, a data storage element 2424 (or memory) and a communication interface 2426. The control module 2422 generally represents the hardware, circuitry, logic, firmware and / or one or more other components of the medical device 2402, which are coupled to the sensing element 2404 to receive the electrical signal output by the sensing element 2404 and perform or otherwise support the various additional tasks, operations, functions and / or processes described herein. Depending on the embodiment, the control module 2422 can be implemented or realized using a general-purpose processor, microprocessor, controller, microcontroller, state machine, content addressable memory, application specific integrated circuit, field programmable gate array, any suitable programmable logic device, discrete gate or transistor logic, discrete hardware components or any combination thereof designed to perform the functions described herein. In some embodiments, the control module 2422 includes an analog-to-digital converter (ADC) or another similar sampling structure that samples the output electrical signal received from the sensing element 2404 or otherwise converts it into a corresponding digital measurement data value. In other embodiments, the sensing element 2404 may incorporate an ADC and output digital measurements.

[0211] The communication interface 2426 generally represents the hardware, circuitry, logic, firmware, and / or other components of the medical device 2402 that are coupled to the control module 2422 for outputting data and / or information from / to the medical device 2402 to / from the client device 2406. For example, the communication interface 2426 may include or otherwise be coupled to one or more transceiver modules capable of supporting wireless communication between the medical device 2402 and the client device 2406. In an exemplary embodiment, the communication interface 2426 is implemented as a Bluetooth transceiver or adapter that is configured to support Bluetooth Low Energy (BLE) communications.

[0212] In an exemplary embodiment, the remote device 2414 receives measurement data values associated with a particular patient (e.g., sensor glucose measurements, acceleration measurements, etc.) obtained using the sensing element 2404 from the client device 2406, and the remote device 2414 stores or otherwise maintains the historical measurement data in a database 2416 associated with the patient (e.g., using one or more unique patient identifiers). Additionally, the remote device 2414 may also receive meal data or other event log data that may be input or otherwise provided by the patient (e.g., via the client application 2408) from or via the client device 2406, and store or otherwise maintain the historical meal data and other historical event or activity data associated with the patient in the database 2416. In this regard, the meal data includes, for example, a time or timestamp associated with a particular meal event, a meal type or other information indicating the composition or nutritional characteristics of the meal, and an indication of the size associated with the meal. In an exemplary embodiment, remote device 2414 also receives historical fluid delivery data, which corresponds to the basis or bolus dose of the fluid delivered to the patient by infusion device 602,2002. For example, client application 2408 can communicate with infusion device 602,2002 to obtain insulin delivery dose and corresponding timestamp from infusion device 602,2002, and then upload the insulin delivery data to remote device 2414 for storage associated with a specific patient. Remote device 2414 can also receive geographic location data and possible other background data associated with device 2402,2406 from client device 2406 and / or client application 2408, and store or otherwise maintain historical operation background data associated with a specific patient. In this regard, one or more of device 2402,2406 can include a global positioning system (GPS) receiver, or can output in real time or otherwise provide similar modules, components or circuits that characterize the data of the geographical location of respective device 2402,2406.

[0213] As described above, in one or more exemplary embodiments, remote device 2414 utilizes machine learning to determine which combination of variables, fields or attributes of historical observation patient data is associated with or predicts the occurrence of a specific event, activity, or metric of a specific patient, and then determines the corresponding equation, function or model for calculating the value of the parameter of interest based on the set of input variables. Therefore, the resulting model can characterize one or more specific combinations of current (or recent) sensor glucose measurement data, auxiliary measurement data, delivery data, geographic location, patient behavior or activity, etc., or map the specific combination to a value representing the current probability or possibility of a specific event or activity or the current value of the parameter of interest. It should be noted that since the physiological response of each patient may be different from that of other groups, when modeling is performed by patient, the subset of input variables predicted or associated with a specific patient may be different from that of other users. In addition, in such embodiments, based on the different correlations between specific input variables and the historical data of the specific patient, the relative weights applied to the corresponding variables of the prediction subset can also be different from those of other patients who may have a common prediction subset. It should be noted that the remote device 2414 may utilize any number of different machine learning techniques to determine which input variables predict the patient of current interest, such as artificial neural networks, genetic programming, support vector machines, Bayesian networks, probabilistic machine learning models or other combinations of Bayesian techniques, fuzzy logic, heuristic derivations, and the like.

[0214] Overview of Diabetes Data Management System

[0215] Figure 25 A computing device 2500 suitable for use as part of a diabetes data management system is shown, incorporating one or more of the processes described above. In some embodiments, the diabetes data management system (DDMS) may be called the Medtronic MiniMed CARELINK TM A DDMS is a medical data management system or medical data management system (MDMS). The DDMS can be built on a server or multiple servers that can be accessed by users or healthcare professionals through a communication network via the Internet or the World Wide Web. Some models of DDMS described as MDMS are described in U.S. Patent Application Publication Nos. 2006 / 0031094 and 2013 / 0338630, which are incorporated herein by reference in their entirety.

[0216] Although the embodiments are described with respect to monitoring the medical or biological condition of subjects with diabetes, the systems and processes herein are applicable to monitoring the medical or biological condition of subjects with heart disease, cancer, HIV, subjects with other diseases, infections, or manageable conditions, or various combinations thereof.

[0217] In an embodiment of the present invention, the DDMS can be installed on a computing device in a healthcare provider's office (e.g., a doctor's office, a nurse's office, a clinic, an emergency room, or an urgent care unit). Healthcare providers may be reluctant to use a system where their confidential patient data is stored on a computing device (e.g., a server on the Internet).

[0218] The DDMS can be installed on a computing device 2500. The computing device 2500 can be coupled to a display 2533. In some embodiments, the computing device 2500 can be located in a physical device separate from the display (e.g., in a personal computer, a microcomputer, etc.). In some embodiments, the computing device 2500 can be located in a single physical housing or a device with a display 2533 (e.g., a laptop computer with the display 2533 integrated into the computing device). In an embodiment of the present invention, the computing device 2500 hosting the DDMS can be, but is not limited to, a desktop computer, a laptop computer, a server, a network computer, a personal digital assistant (PDA), a portable phone including computer functions, a pager with a large visual display, an insulin pump including a display, a glucose sensor including a display, a glucose meter including a display, and / or a combination insulin pump / glucose sensor with a display. The computing device can also be an insulin pump connected to a display, a glucose meter connected to a display, or a glucose sensor connected to a display. The computing device 2500 can also be a server located on the Internet, which can be accessed through a browser installed on a laptop computer, a desktop computer, a network computer, or a PDA. Computing device 2500 may also be a server located in a doctor's office that can be accessed via a browser installed on a portable computing device (e.g., a laptop, PDA, network computer, portable phone) that has wireless capabilities and can communicate via wireless communication protocols (e.g., Bluetooth and IEEE 802.11 protocols).

[0219] exist Figure 25 In the illustrated embodiment, the data management system 2516 includes a set of related software modules or layers that specialize in handling different tasks. The system software includes a device communication layer 2524, a data parsing layer 2526, a database layer 2528, a database storage device 2529, a reporting layer 2530, a graphics display layer 2531, and a user interface layer 2532. The diabetes data management system can be connected to multiple subject support devices 2512 ( Figure 2525 (two of which are shown in FIG). Although different reference numerals refer to multiple layers (e.g., device communication layer, data parsing layer, database layer), each layer may include a single software module or multiple software modules. For example, the device communication layer 2524 may include multiple interactive software modules, libraries, etc. In an embodiment of the present invention, the data management system 2516 can be installed on a non-volatile storage area of the computing device 2500 (such as a flash memory, a hard disk, a mobile hard disk, a DVD-RW, or a CD-RW memory). If the data management system 2516 is selected or started, the system 2516 can be loaded into a volatile storage device (such as a memory of DRAM, SRAM, RAM, or DDRAM) for execution.

[0220] The device communication layer 2524 is responsible for interacting with at least one, and in other embodiments, multiple different types of subject support devices 2512 (e.g., blood glucose meters, glucose sensors / monitors, or infusion pumps). In one embodiment, the device communication layer 2524 can be configured to communicate with a single type of subject support device 2512. However, in a more comprehensive embodiment, the device communication layer 2524 is configured to communicate with multiple different types of subject support devices 2512 (e.g., devices manufactured by multiple different manufacturers), multiple different models from a specific manufacturer, and / or multiple different devices that provide different functions (e.g., infusion functions, sensing functions, metering functions, communication functions, user interface functions, or a combination thereof). By providing the ability to interact with multiple different types of subject support devices 2512, the diabetes data management system 2516 can collect data from a significantly greater number of discrete sources. Such embodiments can provide expanded and improved data analysis capabilities by including a greater number of subjects and subject groups in the form of statistics or other analysis that can benefit from a greater amount of sample data and / or a greater diversity of sample data, thereby improving the ability to determine appropriate treatment parameters, diagnoses, etc.

[0221] The device communication layer 2524 allows the DDMS 2516 to receive information from and transmit information to each subject support device 2512 in the system 2516. Depending on the implementation and context of use, the types of information that can be transmitted between the system 2516 and the device 2512 may include, but are not limited to, data, programs, updated software, educational materials, warning messages, notifications, device settings, treatment parameters, and the like. The device communication layer 2524 may include appropriate routines for detecting the type of subject support device 2512 communicating with the system 2516 and implementing appropriate communication protocols for that type of device 2512. Alternatively or additionally, the subject support device 2512 may transmit information in packets or other data arrangements, wherein the communication includes a preamble or other portion of device identification information that identifies the type of subject support device. Alternatively or additionally, the subject support device 2512 may include a suitable user-operable interface to allow the user to input information corresponding to the type of subject support device being used (e.g., by selecting a selectable icon or text or other device identifier). This information may be transmitted to the system 2516 via a network connection. In another embodiment, system 2516 may detect the type of subject support device 2512 with which it is communicating in the manner described above, and may then send a message requesting the user to verify that system 2516 correctly detected the type of subject support device being used by the user. For systems 2516 capable of communicating with multiple different types of subject support devices 2512, device communication layer 2524 may be capable of implementing multiple different communication protocols and selecting a protocol appropriate for the detected type of subject support device.

[0222] The data parsing layer 2526 is responsible for verifying the integrity of the received device data and correctly entering it into the database 2529. A cyclic redundancy check (CRC) process can be used to check the integrity of the received data. Alternatively or in addition, the data can be received in packets or other data arrangements, wherein the preamble or other portion of the data includes device type identification information. Such a preamble or other portion of the received data may also include a device serial number or other identification information that can be used to verify the authenticity of the received information. In such an embodiment, the system 2516 can compare the received identification information with pre-stored information to assess whether the received information is from a valid source.

[0223] The database layer 2528 may include a centralized database repository that is responsible for warehousing and archiving stored data in an organized format for later access and retrieval. The database layer 2528 works in conjunction with one or more data storage devices 2529 that are suitable for storing and providing access to data in the manner described herein. Such data storage devices 2529 may include, for example, one or more hard disks, optical disks, magnetic tapes, digital libraries, or other suitable digital or analog storage media and associated drive devices, drive arrays, and the like.

[0224] Data can be stored and archived for various purposes, depending on the implementation and use environment. Information about specific subjects and patient support devices can be stored and archived and can be used by these specific subjects, their authorized healthcare providers, and / or authorized healthcare payment entities to analyze the subjects' conditions. In addition, certain information about groups of subjects or groups of subject support devices can be used more generally by healthcare providers, subjects, personnel of entities managing system 2516, or other entities to analyze group data or other forms of aggregated data.

[0225] Embodiments of the database layer 2528 and other components of the system 2516 may employ appropriate data security measures to protect the subject's personal medical information while also allowing non-personal medical information to be more generally available for analysis. Embodiments may be configured to comply with appropriate government regulations, industry standards, policies, and the like, including but not limited to the Health Insurance Portability and Accountability Act of 1996 (HIPAA).

[0226] The database layer 2528 can be configured to limit each user's access to the type of information that has been pre-authorized for that user. For example, a subject may be allowed to access his or her personal medical information (using a personal identifier) stored by the database layer 2528, but not be allowed to access the personal medical information (using a personal identifier) of other subjects. Similarly, the subject's authorized healthcare provider or payment entity may be provided with access rights to some or all of the subject's personal medical information (using an individual identifier) stored by the database layer 2528, but not be allowed to access the personal information of others. Moreover, an operator or administrator user (on a separate computer that communicates with the computing device 2500) may be provided with access rights to some or all of the subject's information, depending on the role of the operator or administrator. On the other hand, the subject, healthcare provider, operator, administrator, or other entity may be authorized to access general information of unidentified individuals, groups, or consortiums (without personal identifiers) stored in the data storage device 2529 by the database layer 2528.

[0227] In an exemplary embodiment, database 2529 stores uploaded patient measurement data (e.g., sensor glucose measurement results and characteristic impedance values) and event log data, which includes event records created during the monitoring period corresponding to the measurement data. In an embodiment of the present invention, database layer 2528 can also store preference profiles. In database layer 2528, for example, each user can store information about specific parameters corresponding to the user. For example, these parameters may include target blood glucose or sensor glucose levels, the type of device used by the user (insulin pump, glucose sensor, blood glucose meter, etc.), and may be stored in records, files, or memory locations in one or more data storage devices 2529 in the database layer. Preference profiles may include various threshold values, monitoring cycle values, priority criteria, screening criteria, and / or other user-specific values for parameters displayed on display 2533 or support device 2512 in a personalized or patient-specific manner.

[0228] The DDMS 2516 can measure, analyze, and track a user's blood glucose (BG) or sensor glucose (SG) measurements (or readings). In embodiments of the present invention, the medical data management system can measure, track, or analyze both a user's BG and SG readings. Thus, while certain reports may only reference or describe BG or SG, reports can monitor and display the results of either or both glucose readings.

[0229] The reporting layer 2530 may include a reporting wizard that extracts data from selected locations in the database 2529 and generates report information based on the desired parameters of interest. The reporting layer 2530 may be configured to generate a plurality of different types of reports, each report having different information and / or displaying information in a different format (layout or style), wherein the type of report can be selected by the user. A plurality of pre-set types of reports (with predefined types of content and formats) may be available and selectable to the user. At least some of the pre-set types of reports may be common industry standard report types that many healthcare providers should be familiar with. In the exemplary embodiment described herein, the reporting layer 2530 also facilitates the generation of snapshot reports including snapshot GUI displays.

[0230] In an embodiment of the present invention, database layer 2528 can calculate values for various medical information that will be displayed on reports generated by report layer 2530. For example, database layer 2528 can calculate average blood glucose or sensor glucose readings for a specified time range. In an embodiment of the present invention, report layer 2530 can calculate values for medical or physical information that will be displayed on reports. For example, a user can select parameters that report layer 2530 then uses to generate medical information values corresponding to the selected parameters. In other embodiments of the present invention, a user can select a parameter profile that previously existed in database layer 2528.

[0231] Alternatively or in addition, a report wizard can allow a user to design a custom report. For example, a report wizard can allow a user to define and input parameters (e.g., parameters specifying the type of content data, the time period of such data, the report format, etc.), and can select data from a database and arrange the data into a printable or displayable layout based on the user-defined parameters. In another embodiment, the report wizard can interact with or provide data for use by other programs available to the user (e.g., common report generation, formatting, or statistical analysis programs). In this way, the user can import data from system 2516 into other reporting tools familiar to the user. The report layer 2530 can generate reports in a displayable form that allows the user to view the report on a standard display device, a printable form that allows the user to print the report on a standard printer, or other suitable forms that are convenient for user access. The embodiment can operate with conventional file format solutions to simplify storage, printing, and transmission functions, including but not limited to PDF, JPEG, etc. For example, the user can select the report type and parameters for the report, and the report layer 2530 can create the report in PDF format. A PDF plug-in can be activated to help create the report and allow the user to view the report. Under these operating conditions, the user can use the PDF plug-in to print the report. In certain embodiments implementing security measures, e.g., to comply with government regulations, industry standards, or policies that restrict the communication of a subject's personal information, some or all reports may be generated via a form (or using appropriate software controls) to prohibit printing or electronic transmission (e.g., in a non-printable and / or unusable format). In other embodiments, the system 2516 may allow a user generating a report to designate the report as non-printable and / or non-transferable, whereupon the system 2516 will provide the report in a format that prohibits printing and / or electronic transmission.

[0232] The reporting layer 2530 may transmit the selected report to the graphic display layer 2531. The graphic display layer 2531 receives information about the selected report and converts the data into a format that can be displayed or shown on the display 2533.

[0233] In an embodiment of the present invention, the report layer 2530 can store a plurality of user parameters. For example, the report layer 2530 can store carbohydrate unit type, blood glucose exercise or sensor glucose reading, carbohydrate conversion factor, and time range for a particular type of report. These examples are intended to be illustrative and not limiting.

[0234] Data analysis and presentation of the reported information can be used to develop and support diagnostic and treatment parameters. When the information in the report is related to an individual subject, the diagnostic and treatment parameters can be used to assess the subject's health status and relative physical and mental health, assess the subject's compliance with treatment, and develop or modify the subject's therapy, and assess the subject's behavior that affects his / her treatment. If the information in the report is related to a subject group or data set, the diagnostic and treatment parameters can be used to assess the health status and relative physical and mental health of a subject group with a similar medical condition, such as, but not limited to, diabetic subjects, heart disease subjects, diabetic subjects with a specific type of diabetes or heart disease, subjects of a specific age, sex, or other overall statistical group, subjects with symptoms that affect treatment decisions (such as, but not limited to, pregnancy, obesity, hypoglycemic awareness disorder, learning disabilities, limited self-care ability, various levels of insulin resistance, combinations thereof, etc.).

[0235] The user interface layer 2532 supports interaction with end users, such as for user login and data access, software navigation, data entry, user selection of desired report types, and display of selected information. The user may also input parameters to be used in the selected report through the user interface layer 2532. Examples of users include, but are not limited to, healthcare providers, healthcare payers, system operators or administrators, researchers, commercial entities, healthcare institutions and organizations, etc., depending on the services provided by the system and on the embodiment of the invention. More comprehensive embodiments are capable of interacting with some or all of the above types of users, where different types of users may have access to different services or data or different levels of services or data.

[0236] In an exemplary embodiment, the user interface layer 2532 provides one or more websites that can be accessed by users via the Internet. The user interface layer may include, or operate in conjunction with, at least one (or more) suitable web servers to provide the websites via the Internet and allow global access from Internet-connected computers using standard Internet browser software. The one or more websites may be accessed by a variety of users, including but not limited to subjects, healthcare providers, researchers, commercial entities, healthcare delivery organizations and organizations, payer entities, pharmaceutical partners or other sources of drugs or medical devices, and / or support personnel or other personnel operating the system 2516, depending on the embodiment used.

[0237] In another exemplary embodiment, when DDMS 2516 is hosted on a computing device 2500, a user interface layer 2532 provides a user with multiple menus for navigating within the DDMS. These menus can be created using any menu format, including but not limited to HTML, XML, or Active Server Pages. A user can access DDMS 2516 to perform one or more of a variety of tasks, such as accessing general information available on the website for all subjects or groups of subjects. The user interface layer 2532 of DDMS 2516 can allow the user to access specific information or generate reports regarding the subject's medical condition or one or more of the subject's medical devices 2512, transfer data or other information from the subject's one or more support devices 2512 to the system 2516, transfer data, programs, program updates, or other information from the system 2516 to the subject's one or more support devices 2512, manually enter information into the system 2516, participate in a remote consultation with a healthcare provider, or modify custom settings in the subject's support device and / or the subject's DDMS / MDMS data file.

[0238] System 2516 can provide different users and different types or groups of users with access to different optional resources or activities (including access to different information items and services) so that each user can have a customized experience and / or each type or group of users (e.g., all users, diabetic users, cardiac users, healthcare provider users, or paying users, etc.) can have a different set of information items or services available on the system. System 2516 may include or employ one or more suitable resource configuration programs or systems to allocate appropriate resources to each user or user type based on a predefined authorization plan. Resource configuration systems are well known to be associated with configuring electronic office resources (email, licensed software programs, sensitive data, etc.) in an office environment (e.g., a local area network LAN in an office, company, or enterprise). In an exemplary embodiment, such a resource configuration system is suitable for controlling access to medical information and services on DDMS2516 based on the type of user and / or the identity of the user.

[0239] Upon successfully entering user identification information and a password, the user may be provided with access to secure, personalized information stored on the DDMS 2516. For example, the user may be provided with access to a secure, personalized location in the DDMS 2516 that has been assigned to the subject. This personalized location may be referred to as a personalized screen, home screen, main menu, personalized page, etc. The personalized location may provide the subject with a personalized home screen, including selectable icons or menu items for selecting optional activities, including, for example, options for transferring device data from the subject's support device 2512 to the system 2516, manually entering additional data into the system 2516, modifying the subject's customized settings, and / or viewing and printing reports. Reports may include data specific to the subject's condition, including, but not limited to, data obtained from one or more of the subject's support devices 2512, manually entered data, data from a medical library or other networked treatment management system, data from the subject or a group of subjects, etc. When a report includes subject-specific information and subject identification information, the report may be generated from some or all of the subject's data stored in a secure storage area employed by the database layer 2528 (e.g., storage device 2529).

[0240] The user may select an option for transferring (sending) device data to the medical data management system 2516. If the system 2516 receives a user request to transfer device data to the system, the system 2516 may provide the user with step-by-step instructions on how to transfer data from one or more support devices 2512 of a subject. For example, the DDMS 2516 may have multiple different stored instruction sets for instructing the user on how to download data from different types of subject support devices, where each instruction set relates to a specific type of subject support device (e.g., pump, sensor, meter, etc.), a specific manufacturer version of a type of subject support device, etc. The registration information received from the user during registration may include information about the type of one or more subject support devices 2512 used by the subject. The system 2516 uses this information to select one or more stored instruction sets associated with one or more support devices 2512 of a particular subject for display to the user.

[0241] Other activities or resources available to the user on system 2516 may include options for manually entering information into DDMS / MDMS 2516. For example, from a user's personalized menu or location, the user may select an option for manually entering additional information into system 2516.

[0242] Additional optional activities or resources may be available to the user of DDMS 2516. For example, from the user's personalized menu, the user may select an option to receive data, software, software updates, treatment recommendations, or other information from system 2516 on one or more support devices 2512 of the subject. If system 2516 receives a request from the user to receive data, software, software updates, treatment recommendations, or other information, system 2516 may provide the user with a list or other arrangement of a plurality of selectable icons or other indicia representing available data, software, software updates, or other information available to the user.

[0243] On the medical data management system 2516, the user can use additional optional activities or resources, including, for example, options for customizing or otherwise further personalizing the user's personalized location or menu. Specifically, from the user's personalized location, the user can select options for customizing parameters for the user. In addition, the user can create a profile for customizable parameters. When the system 2516 receives such a request from the user, the system 2516 can provide the user with a list or other arrangement of multiple optional icons or other markings representing parameters that can be modified to accommodate the user's preferences. When the user selects one or more icons or other markings, the system 2516 can receive the user's request and make the requested modification.

[0244] In one or more exemplary embodiments, for an individual patient in the DDMS, the computing device 2500 of the DDMS is configured to analyze the patient's historical measurement data, historical delivery data, historical event log data, and any other historical or contextual data associated with the patient maintained in the database layer 2528 to support one or more processes described herein. In this regard, machine learning, artificial intelligence, or similar mathematical modeling of the patient's physiological behavior or response can be performed at the computing device 2500 to facilitate patient-specific correlations or predictions. The resulting model can be used at the computing device 2500 or another device 2512 of the DDMS to analyze current measurement data, delivery data, and event log data related to the patient, as well as current contextual data, to determine, in real time, predictions or other possible events, behaviors, or outcomes related to the patient.

[0245] Additionally, the following exemplary embodiments are provided, which are numbered for ease of reference:

[0246] Embodiment 1: A method for monitoring a patient's physiological condition, the method comprising: obtaining, at a computing device, current measurement data of the patient's physiological condition provided by a sensing device; obtaining, at the computing device, user input indicating one or more future events associated with the patient; and in response to the user input: using one or more prediction models associated with the patient, determining a prediction of the patient's future physiological condition based at least in part on the current measurement data and the one or more future events; and displaying, at the computing device, a graphical representation of the prediction on a display device.

[0247] Example 2: A method according to Example 1, wherein determining a prediction includes: for each corresponding prediction model in a plurality of different prediction models, determining a weighting factor associated with the corresponding prediction model based at least in part on one or more future events; using each of a plurality of different prediction models associated with the patient, determining a plurality of prediction values indicating a future physiological condition based at least in part on current measurement data and one or more future events, wherein each of the plurality of prediction values is associated with a corresponding prediction model in the plurality of different prediction models; and determining the prediction of the patient's physiological condition as a weighted average of the corresponding prediction values in the plurality of prediction values and the weighting factors associated with the corresponding prediction models.

[0248] Example 3: A method according to Example 1, wherein determining the prediction includes: using a first prediction model to determine a first plurality of prediction values of the patient's future physiological condition based at least in part on current measurement data; using a second prediction model different from the first prediction model to determine a second plurality of prediction values of the patient's future physiological condition based at least in part on the current measurement data; for each of the first prediction model and the second prediction model, determining a weighting factor associated with the corresponding prediction model based at least in part on one or more future events; and determining a collective prediction of the patient's physiological condition relative to future time based at least in part on the first plurality of prediction values, the second plurality of prediction values, and the weighting factors associated with the corresponding first prediction model and the second prediction model, wherein the graphical representation of the prediction includes a graphical representation of the collective prediction relative to time.

[0249] Embodiment 4: A method according to embodiment 3, wherein: based on the relationship between a first reliability metric associated with the first prediction model and a second reliability metric associated with the second prediction model, the weighting factor varies relative to the amount of time in the future; and the first reliability metric and the second reliability metric are affected by one or more future events.

[0250] Embodiment 5: The method of embodiment 3, wherein determining the first plurality of predicted values comprises determining hourly predicted values of the physiological condition using an hourly prediction model associated with the patient based at least in part on current measurement data and one or more future events.

[0251] Example 6: A method according to Example 5, wherein the hourly forecast model comprises: a neural network including a plurality of cells; each cell corresponds to a corresponding hourly interval; and at least one of the plurality of cells is configured to output an average value of the physiological condition during the corresponding hourly interval in the future based at least in part on a subset of one or more future events predicted to occur within the corresponding hourly interval in the future.

[0252] Embodiment 7: The method according to embodiment 6 also includes obtaining background measurement data from a second sensing device, wherein determining the hourly forecast value includes determining the hourly forecast value based at least in part on the current measurement data, one or more future events and the background measurement data.

[0253] Example 8: A method according to Example 5, wherein determining the second plurality of predicted values includes determining predicted sample values of the physiological condition using at least one of an autoregressive integrated moving average model determined based on historical data associated with the patient, or a physiological model determined based on historical data associated with the patient.

[0254] Embodiment 9: The method of embodiment 8, wherein the ensemble forecast comprises a weighted average of an hourly forecast value weighted using a first weighting factor and a second plurality of forecast values weighted using a second weighting factor.

[0255] Example 10: A method according to Example 1, wherein: obtaining current measurement data includes receiving sensor glucose measurement data from a glucose sensing device; obtaining user input includes receiving at least one of the following via a user interface at a computing device: an expected amount of carbohydrates to be consumed by the patient, an expected amount of exercise to be performed by the patient, and an expected bolus amount of insulin to be administered; determining a prediction includes using multiple prediction models associated with the patient to determine an aggregate prediction of the patient's future glucose levels based at least in part on the sensor glucose measurement data and one or more future events; and displaying a graphical representation of the prediction includes displaying a graphical representation of the aggregate prediction relative to time.

[0256] Example 11: A method according to Example 10, wherein determining the ensemble prediction includes: using a prediction model associated with the patient to determine a first plurality of prediction values of the patient's future glucose levels based at least in part on sensor glucose measurement data; using a second prediction model to determine a second plurality of prediction values of the patient's future glucose levels based at least in part on the sensor glucose measurement data; determining a first weighting factor associated with the prediction model based at least in part on one or more future events; determining a second weighting factor associated with the second prediction model based at least in part on one or more future events; and using the first weighting factor and the second weighting factor, determining the ensemble prediction as a weighted average of the first plurality of prediction values and the second plurality of prediction values.

[0257] Example 12: The method according to Example 1 also includes providing a graphical user interface display on a display device that prompts a patient for conversational interaction, wherein: obtaining user input includes receiving conversational input from the patient indicating one or more future events; and displaying a graphical representation of the prediction includes providing a graphical representation of the prediction within the graphical user interface display in response to the conversational input.

[0258] Embodiment 13: A computer-readable medium having instructions stored thereon, the instructions being executable by a processing system of a computing device to perform the method according to embodiment 1.

[0259] Example 14: An electronic device comprising: a communication interface for receiving current measurement data of a patient's physiological condition from a sensing device; a display device having a graphical user interface display presented thereon; a user interface for obtaining user input indicating one or more future events; and a control system coupled to the communication interface, the display device, and the user interface to use one or more predictive models associated with the patient to determine a prediction of the patient's future physiological condition based at least in part on the current measurement data and one or more future events, and to display a graphical representation of the prediction within the graphical user interface display of the display device.

[0260] Embodiment 15: An electronic device according to embodiment 14, wherein: the prediction includes a weighted average of a first plurality of prediction values determined using a first prediction model and a second plurality of prediction values determined using a second prediction model; and the weighting factors associated with the corresponding first prediction model and second prediction model vary relative to future time based at least in part on the relationship between one or more future events and a first reliability metric associated with the first prediction model and a second reliability metric associated with the second prediction model.

[0261] Embodiment 16: The electronic device of Embodiment 15, wherein the first plurality of predicted values comprises predicted hourly averages of the patient's physiological condition.

[0262] Example 17: An electronic device according to Example 14, wherein: the current measurement data includes sensor glucose measurement data from a glucose sensing device; the one or more future events include at least one of: an expected amount of carbohydrates to be consumed by the patient, an expected amount of exercise to be performed by the patient, and an expected bolus amount of insulin to be administered; and the prediction includes a simulated glucose level determined using multiple prediction models.

[0263] Embodiment 18: An electronic device according to embodiment 14, wherein a graphical user interface display prompts a conversational interaction, wherein: the user input includes a conversational input from the patient indicating one or more future events; and in response to the conversational input, a graphical representation of the prediction is provided within the graphical user interface display.

[0264] Example 19: A method for monitoring a patient's glucose level, the method comprising: obtaining, at a computing device, current sensor glucose measurement data of the patient from a glucose sensing device; obtaining, at the computing device, user input indicating one or more future events for the patient via a user interface; determining, at the computing device, a future simulated glucose level of the patient based at least in part on the current sensor glucose measurement data and the one or more future events using a plurality of different predictive models associated with the patient, wherein the plurality of different predictive models include an hourly forecast model associated with the patient; and displaying, on a display device associated with the computing device, a graphical representation of the simulated glucose level relative to future time.

[0265] Example 20: The method according to Example 19 also includes determining a weighting factor associated with a corresponding prediction model in a plurality of different prediction models based at least in part on one or more future events, wherein determining the simulated glucose level includes using the weighting factor to determine a weighted average of a plurality of predicted glucose values output by the plurality of different prediction models.

[0266] Example 21: A method for monitoring a patient's physiological condition, the method comprising: obtaining current measurement data of the patient's physiological condition from a sensing device; using a first prediction model to determine a first plurality of prediction values of the patient's future physiological condition based at least in part on the current measurement data; using a second prediction model different from the first prediction model to determine a second plurality of prediction values of the patient's future physiological condition based at least in part on the current measurement data; determining a collective prediction of the patient's physiological condition relative to a future time based at least in part on the first plurality of prediction values, the second plurality of prediction values, and weighting factors associated with the corresponding first prediction model and second prediction model, wherein the weighting factor varies relative to the future time based on a relationship between a first reliability metric associated with the first prediction model and a second reliability metric associated with the second prediction model; and displaying a graphical representation of the collective prediction of the patient's physiological condition relative to the future time on a display device.

[0267] Embodiment 22: The method according to embodiment 21 further includes displaying a graphical representation of current measurement data relative to time at a graphical user interface display on a display device, wherein displaying a graphical indication of the aggregate forecast includes displaying the graphical representation of the aggregate forecast in response to user input to adjust the graphical user interface display to view a future time.

[0268] Embodiment 23: The method of Embodiment 21, wherein determining a first plurality of predicted values comprises determining hourly predicted values for the physiological condition using an hourly prediction model associated with the patient.

[0269] Embodiment 24: The method of Embodiment 23, wherein determining the second plurality of predicted values comprises determining predicted sample values of the physiological condition using an autoregressive integrated moving average model determined based on historical data associated with the patient.

[0270] Embodiment 25: The method of Embodiment 23, wherein determining a second plurality of predicted values comprises determining predicted sample values of the physiological condition using a physiological model determined based on historical data associated with the patient.

[0271] Embodiment 26: The method of embodiment 23, wherein a first one of the weighting factors associated with the hourly forecast model increases relative to the future amount of time, and a second one of the weighting factors associated with the second forecast model decreases relative to the future amount of time.

[0272] Embodiment 27: The method of embodiment 26, wherein the ensemble forecast comprises a weighted average of an hourly forecast value weighted using a first weighting factor and a second plurality of forecast values weighted using a second weighting factor.

[0273] Embodiment 28: The method of embodiment 21 further comprising identifying a current operating context; and determining the weighting factor with respect to time based at least in part on the current operating context.

[0274] Embodiment 29: The method of embodiment 21, wherein determining the ensemble prediction comprises determining a weighted average of the first plurality of prediction values and the second plurality of prediction values using a weighting factor.

[0275] Example 30: A method according to Example 29, wherein: determining a first plurality of prediction values includes using an hourly prediction model associated with the patient to determine hourly prediction values of the physiological condition; determining a second plurality of prediction values includes using an autoregressive integrated moving average model associated with the patient and one of the physiological models associated with the patient to determine predicted sample values of the physiological condition; and the ensemble prediction includes a weighted average of the hourly prediction values and the second plurality of prediction values.

[0276] Embodiment 31: A computer-readable medium having instructions stored thereon, the instructions being executable by a processing system coupled to a display device to perform the method according to embodiment 21.

[0277] Example 32: A method for monitoring a patient's physiological condition, the method comprising: obtaining current measurement data of the patient's physiological condition from a sensing device; using a plurality of different prediction models associated with the patient, determining a plurality of prediction values indicating the physiological condition at a future time based at least in part on the current measurement data, wherein each of the plurality of prediction values is associated with a corresponding prediction model among the plurality of different prediction models; for each prediction model among the plurality of different prediction models, determining a reliability metric associated with the corresponding prediction model based at least in part on a relationship between the future time and the current time; for each corresponding prediction model among the plurality of different prediction models, determining a weighting factor associated with the corresponding prediction model based at least in part on the reliability metric associated with the corresponding prediction model; determining a collective prediction value of the patient's physiological condition as a weighted average of the corresponding prediction values among the plurality of prediction values and the weighting factors associated with the corresponding prediction models; and displaying a graphical indication of the collective prediction value of the patient's physiological condition associated with the future time.

[0278] Embodiment 33: The method of embodiment 32, wherein the weighting factors associated with the respective prediction models vary with respect to time.

[0279] Example 34: According to the method described in Example 32, multiple different prediction models include an hourly forecast model, wherein determining the collective prediction value includes determining the predicted hourly average value of the physiological condition associated with the hourly forecast model, and a weighted average of corresponding prediction values among multiple prediction values associated with one or more different prediction models.

[0280] Example 35: A method according to Example 32, wherein determining a reliability metric includes: determining a prediction range corresponding to a future time before a current time of day; obtaining historical data associated with the patient for the prediction range, the historical data including historical measurement data of physiological conditions associated with the prediction range; and for each corresponding prediction model in a plurality of different prediction models, determining a reference prediction for the prediction range based on the historical data using the corresponding prediction model; and determining a reliability metric associated with the corresponding prediction model based on a difference between the reference prediction and the historical measurement data.

[0281] Embodiment 36: The method of embodiment 35, wherein determining the reliability metric comprises determining a mean absolute difference associated with the respective prediction models for the prediction horizon.

[0282] Embodiment 37: An electronic device comprising: a communication interface for receiving current measurement data of a patient's physiological condition from a sensing device; a display device having a graphical user interface display including a graphical representation of the current measurement data; a user interface for obtaining user input for adjusting the graphical user interface display to view the future; and a control system coupled to the communication interface, the display device, and the user interface to determine, using a first prediction model, a first plurality of prediction values of the patient's future physiological condition based at least in part on the current measurement data; to determine, using a second prediction model different from the first prediction model, a second plurality of prediction values of the patient's future physiological condition based at least in part on the current measurement data; to determine, based at least in part on the first plurality of prediction values and the second plurality of prediction values, a collective prediction of the patient's physiological condition relative to a future time; and to display, on the graphical user interface display, a graphical representation of the collective prediction in response to the user input.

[0283] Embodiment 38: The electronic device of embodiment 37, wherein: the ensemble prediction comprises a weighted average of the first plurality of prediction values and the second plurality of prediction values; and based on a relationship between a first reliability metric associated with the first prediction model and a second reliability metric associated with the second prediction model, weighting factors associated with the respective first prediction model and second prediction model vary relative to the future time.

[0284] Embodiment 39: The electronic device of Embodiment 38, wherein the first plurality of predicted values comprises predicted hourly averages of the patient's physiological condition.

[0285] Embodiment 40: An electronic device according to embodiment 39, wherein a first one of the weighting factors associated with the hourly forecast model increases as the amount of future time increases, and a second one of the weighting factors associated with the second forecast model decreases as the amount of future time increases.

[0286] Example 41: A method for monitoring a patient's physiological condition, the method comprising: obtaining current measurement data of the patient's physiological condition from a sensing device; predicting one or more events that may affect the patient's physiological condition at one or more different times in the future based at least in part on historical event data associated with the patient; using a predictive model associated with the patient, determining a plurality of predicted values of the patient's physiological condition associated with a plurality of different time periods in the future based at least in part on the current measurement data and the one or more events; and displaying the plurality of predicted values relative to the plurality of different time periods in the future on a display device.

[0287] Embodiment 42: The method of Embodiment 41 further comprising determining a prognostic model associated with the patient based at least in part on a relationship between historical measurement data of the patient's physiological condition and historical event data associated with the patient.

[0288] Example 43: A method according to Example 42, wherein: the prediction model includes a neural network, which includes multiple cells; and each cell corresponds to a corresponding time period among multiple different time periods, and is configured to output an average value of the physiological condition during the corresponding time period in the future based at least in part on a subset of one or more future events predicted to occur in the corresponding time period in the future.

[0289] Example 44: A method according to Example 42, wherein determining the forecast model includes determining, for each hourly interval of a plurality of hourly intervals, a corresponding long short-term memory (LSTM) unit based at least in part on a relationship between a corresponding subset of historical measurement data corresponding to the corresponding hourly interval and a corresponding subset of historical event data corresponding to the corresponding hourly interval, the unit being configured to output an average value of the physiological condition during the corresponding hourly interval of the plurality of hourly intervals.

[0290] Example 45: A method according to Example 44, wherein determining multiple forecast values includes, for each corresponding hourly interval in the future, using a corresponding LSTM unit associated with the corresponding hourly interval, calculating a corresponding hourly average forecast value associated with the corresponding hourly interval based at least in part on a subset of one or more events predicted to occur within the corresponding hourly interval.

[0291] Embodiment 46: The method of Embodiment 41, further comprising obtaining context measurement data from a second sensing device, wherein predicting one or more events comprises predicting the one or more events based at least in part on the context measurement data.

[0292] Embodiment 47: The method according to embodiment 41 also includes obtaining background measurement data from a second sensing device, wherein determining multiple forecast values includes using a forecast model to determine multiple forecast values based at least in part on current measurement data, one or more events and the background measurement data.

[0293] Example 48: A method according to Example 41, wherein the current measurement data includes sensor glucose measurement data, wherein: predicting one or more events includes predicting one or more of meals, exercise, insulin delivery, and drug dosage at one or more different times in the future; and determining multiple prediction values includes determining the patient's predicted hourly average glucose levels a...

Claims

1. A method for monitoring a patient's physiological condition, the method comprising: obtaining current measurement data of the patient's physiological condition from a sensing device; predicting one or more events that are likely to affect the patient's physiological condition at one or more different times in the future based at least in part on historical event data associated with the patient; determining, using a prognostic model associated with the patient, a plurality of predicted values of the patient's physiological condition associated with a plurality of different time periods in the future based at least in part on the current measurement data and the one or more events; as well as The plurality of predicted values relative to the plurality of different time periods in the future are displayed on a display device.

2. The method of claim 1 , further comprising determining a prognostic model associated with the patient based at least in part on a relationship between historical measurement data of the patient's physiological condition and historical event data associated with the patient.

3. The method according to claim 2, wherein: The prediction model includes a neural network including a plurality of cells; and Each cell corresponds to a respective time period of a plurality of different time periods and is configured to output an average value of the physiological condition during the respective future time period based at least in part on a subset of one or more future events predicted to occur within the respective future time period.

4. The method of claim 2 , wherein determining the prediction model comprises determining, for each hourly interval in a plurality of hourly intervals, a corresponding long short-term memory (LSTM) unit based at least in part on a relationship between a corresponding subset of historical measurement data corresponding to the corresponding hourly interval and a corresponding subset of historical event data corresponding to the corresponding hourly interval, the unit being configured to output an average value of the physiological condition during the corresponding hourly interval in the plurality of hourly intervals.

5. The method of claim 4 , wherein determining the plurality of predicted values comprises, for each respective hourly interval in the future, calculating, using a respective LSTM unit associated with the respective hourly interval, a respective hourly average forecast value associated with the respective hourly interval based at least in part on a subset of one or more events predicted to occur within the respective hourly interval. 6 . The method of claim 1 , further comprising obtaining context measurement data from a second sensing device, wherein predicting the one or more events comprises predicting one or more events based at least in part on the context measurement data.

7. The method of claim 1 , further comprising obtaining background measurement data from a second sensing device, wherein determining the plurality of predicted values comprises using the forecast model to determine the plurality of forecast values based at least in part on the current measurement data, the one or more events, and the background measurement data.

8. The method of claim 1 , wherein the current measurement data comprises sensor glucose measurement data, wherein: Predicting the one or more events includes predicting one or more of meals, exercise, insulin delivery, and medication dosage at one or more different times in the future; and determining the plurality of predicted values includes determining predicted hourly average glucose levels of the patient associated with a plurality of hourly intervals in the future based at least in part on the sensor glucose measurement data and the predicted one or more of meals, exercise, insulin delivery, and medication dosage at the one or more different times in the future.

9. The method of claim 1, wherein displaying the plurality of predicted values comprises displaying a plurality of predicted hourly average glucose levels for the patient for a plurality of hourly intervals in the future.

10. A computer readable medium having instructions stored thereon, the instructions being executable by a processing system coupled to a display device to perform the method of claim 1.

Images