Non-invasive blood pressure measurement

Abstract

A method of adaptively arranging non-invasive blood pressure measurement time intervals based on calculating a risk of a patient suffering from predefined one or more adverse clinical events (e.g., within a predefined time window) using a risk model and also determining time intervals based on a clinician's risk assessment of the patient.

Description

Technical Field

[0001] The invention relates to the field of non-invasive blood pressure monitoring. Background Art

[0002] US2018042493A1 relates to a device and method for providing a control signal for a blood pressure measuring device. In addition, it also relates to a system for monitoring a patient.

[0003] WO2021062292A1 relates to risk-based patient monitoring systems and methods.

[0004] Arterial blood pressure (ABP) is a key physiological parameter relevant to medical diagnosis, prevention, and treatment guidance.

[0005] The gold standard for blood pressure measurement is invasive measurement. This uses an intra-arterial cannula. This provides the most accurate measurement and allows for continuous measurement. However, it requires trained medical personnel and is therefore usually only applied in the highest emergency situations, where very close monitoring and real-time alerts are required.

[0006] Most blood pressure measurements are performed non-invasively. The most common non-invasive blood pressure measurement (NIBP) technique for patient monitoring is standard cuff-based ABP measurement using the automatic oscillometric method. This provides intermittent measurements. Typically, the intervals are set to fixed or standardized time intervals, depending on the patient's condition. For example, in an ICU setting, a typical interval might be every 15 minutes. However, this leaves a significant risk that the patient's condition could suddenly change during a certain interval.

[0007] For example, Figure 1 An example of a patient's real-time systolic blood pressure (y-axis) over time is shown (trace 22), compared to blood pressure measurements recorded by a measuring device at fixed intervals (trace 24). The interval in this example is 15 minutes. As can be seen from this example, if there is a sudden drop in blood pressure during the measurement interval, this may be missed for a considerable period of time until the next measurement is taken. For example, in the time interval from about 16:15 to about 16:30, the patient's real-time blood pressure drops significantly, but this occurs within the measurement interval, so that about 15 minutes have passed before the clinician can resolve the change in condition.

[0008] In the prior art, to account for the fact that the appropriate measurement interval varies depending on the patient's condition, the physician supervising the patient will manually adjust the measurement interval based on their judgment of the patient's severity and risk of rapid deterioration. Clinicians use standard protocols combined with medical experience to set the optimal measurement interval.

[0009] It would be beneficial to provide a method that can more objectively determine the optimal measurement time interval for a patient in view of the patient's condition. Summary of the invention

[0010] The invention is defined by the claims.

[0011] According to an example of one aspect of the present invention, a computer-implemented method is provided. The method includes receiving biometric data for a patient. The method also includes applying a patient risk model to calculate a risk parameter p that indicates a risk level of at least one predefined adverse clinical event, wherein at least one input to the model is the biometric data. The method also includes obtaining a patient risk parameter A defined by a clinician. The method also includes: calculating a measurement time interval T for successive non-invasive blood pressure measurements for the patient based on A and p and according to the following relationship:

[0012]

[0013] Here, α is a predefined parameter, preferably having a value greater than zero.

[0014] The method further comprises controlling the non-invasive blood pressure measurement device to collect blood pressure measurement results at a frequency defined by the determined time interval T.

[0015] Thus, embodiments of the present invention provide a method for adaptively configuring the timing intervals between automatically triggered non-invasive blood pressure measurements based on a risk model, the method taking into account the current value of at least one biological parameter of the patient and advantageously also taking into account the clinician's assessment of risk. Thus, the clinician's guiding medical opinion about the situation still has an impact on the timing, but at the same time the time intervals can be set in a partially objective manner based on the risk model (configured to calculate the risk of at least one specific serious adverse event (such as hypotension or hypertension)).

[0016] The defined relationship for T holds the product of the model-based probability p for a critical event and the time between blood pressure measurements as a constant.

[0017] With respect to the clinician-defined patient risk parameter A, this may represent the clinician's assessment of the risk level of at least one predefined adverse clinical event. In other words, it may represent the clinician's (subjective) judgment about the patient's risk level. In other words, it may represent the clinician's independent assessment of the patient's risk.

[0018] With respect to the variable parameter α, this allows the weighting of machine-derived risk assessments versus human-derived risk assessments to be tuned according to preference or according to context.

[0019] In some embodiments, the value of α may be obtained or retrieved as a step of the method. For example, the value of α may be obtained or retrieved from a memory or a register or any other data source.

[0020] The value of α may be user defined. The value of α may be received from a user interface, or the method may include receiving or determining the value of α (e.g. in a setup phase or at least as a preliminary step), and optionally storing the value of α in a memory or register for subsequent retrieval during the execution of the remainder of the method (i.e. during the calculation time interval T).

[0021] In some embodiments, the method may include calculating or determining a value of α. For example, it may be determined using a predefined lookup table or mapping that relates different patient or clinician states or categories (e.g., patient urgency levels) to different α values. The state or category of the patient or clinician may be collected or retrieved from a data store or other data source. For example, for patients with higher urgency, it may be preferred to set a lower α value, which will have the effect of reducing the time interval between blood pressure measurements when (for p values ​​varying between 0 and 1) the risk parameter p increases during the time period when the clinician updates his own risk assessment A. Therefore, in this case, the higher the α value, the more caution is needed. By way of further example, for more experienced clinicians who trust their own assessments, the value of α may be set lower than for less experienced clinicians. This will have the effect (assuming that the value of p varies between 0 and 1) that a higher relative weight is given to the parameter A defined by the clinician compared to the case where the α value is lower.

[0022] It should be appreciated that a patient may often be at risk for more than one specific adverse clinical event, and therefore, in some embodiments, a model may be designed or trained to output a combined risk level for a patient to suffer any one of a set of different adverse events. Additionally or alternatively, the model may output a risk associated with only one adverse clinical event, but wherein the method includes running multiple versions of the model, each version configured to predict the probability of a different type of adverse event, and selecting the determined probability parameters based on predefined criteria or based on input from a clinician.

[0023] For example, the at least one adverse or critical event may include the patient having low blood pressure and / or high blood pressure; the patient being in a state of shock, cardiac arrest, or any cardiac event.

[0024] In some embodiments, the model is such that p is the probability of an event occurring within a predefined time hop.

[0025] In some embodiments, the biometric data includes real-time sensor data, or data derived therefrom.

[0026] In some embodiments, the biometric data includes historical biometric data of the patient retrieved from a data repository. For example, this may include heart rate or other vital signs, previous blood pressure measurements, pathology tests, or any other biological parameters.

[0027] In some embodiments, the risk parameter p is repeatedly recalculated during a monitoring session for a patient. The biometric data is preferably updated for each recalculation.

[0028] In some embodiments, the risk model is a Bayesian model.

[0029] In some embodiments, the Bayesian model is a personalized risk model for the patient and / or the clinician treating the patient. In some embodiments, the Bayesian model is pre-configured based on prior information about one or more of the following: patient history; severity / urgency of the patient's condition; training level of the clinician treating the patient; and a measure of the physician's responsiveness to changes in the condition.

[0030] In some embodiments, the method further comprises repeatedly adjusting or updating the risk model based on a patient monitoring database comprising monitoring data of a plurality of patients. Thus, a self-learning model can be implemented that can learn over time, for example via retrospective analysis of monitored patients. This can be based on monitoring data obtained at the same medical institution or department, and / or can be based on monitoring data obtained more widely, for example, wide population monitoring data across a plurality of different medical institutions.

[0031] In some embodiments, the biometric data includes data from one or more of: an electrocardiogram (ECG) sensing device; a photoplethysmography (PPG) sensing device; a carbon dioxide measurement device; and a bioimpedance measurement device.

[0032] The method may be implemented in the form of software.

[0033] Therefore, another aspect of the method is a computer program product comprising code units, which are configured to, when executed on a processor, cause the processor to perform the method according to any embodiment or example in the present disclosure.

[0034] The present invention can also be implemented in hardware form. Therefore, another aspect of the method is a processing unit comprising: an input / output; and one or more processors adapted to: receive biometric data for a patient at the input / output; apply a patient risk model to calculate a risk parameter p indicating a risk level of at least one predefined adverse clinical event, wherein at least one input to the model is biometric data; obtain a patient risk parameter A defined by a clinician; calculate a measurement time interval T for successive non-invasive blood pressure measurements for the patient based on A and p and according to the following relationship:

[0035]

[0036] wherein α is a defined (variable) parameter; and the non-invasive blood pressure measurement device is controlled via generating a control signal for output at the input / output to acquire blood pressure measurement results at a frequency defined by the determined time interval T.

[0037] Another aspect of the present invention is a system, comprising: a processing unit as described above or according to any embodiment described in the present disclosure; and a non-invasive blood pressure measurement device operatively coupled to the processing unit.

[0038] In some embodiments, the system further comprises one or more bio-parameter sensing devices for collecting biometric data, wherein the one or more bio-parameter sensing devices are operatively coupled to the processing unit.

[0039] In some embodiments, the one or more bio-parameter sensing devices include a PPG sensor integrated into a portion of the non-invasive blood pressure measurement device.

[0040] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiment(s) described hereinafter. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] For a better understanding of the invention, and to show more clearly how it may be put into practice, reference will now be made, by way of example only, to the accompanying drawings, in which:

[0042] Figure 1 A comparison of a patient's continuous blood pressure with intermittent blood pressure measurements as a function of time is illustrated;

[0043] Figure 2 Outlines the steps of an example method according to one or more embodiments of the present invention; and

[0044] Figure 3 Components of an example processing device and a process flow according to one or more embodiments of the present invention are schematically outlined. DETAILED DESCRIPTION

[0045] The present invention will be described with reference to the accompanying drawings.

[0046] It should be understood that the detailed description and specific examples, although indicating exemplary embodiments of the device, system and method, are intended only for illustrative purposes and are not intended to limit the scope of the invention. These and other features, aspects and advantages of the device, system and method of the present invention will be better understood based on the following description, the appended claims and the accompanying drawings. It should be understood that the drawings are only schematic and are not drawn to scale. It should also be understood that throughout the drawings, the same reference numerals are used to represent the same or similar parts.

[0047] The present invention provides a method for adaptively scheduling non-invasive blood pressure measurements, wherein the time intervals are determined based on using a risk model to calculate the patient's risk of suffering one or more predefined adverse clinical events (e.g., within a predefined time window) and also based on a clinician's risk assessment of the patient.

[0048] Blood pressure is typically measured via intermittent cuff non-invasive blood pressure (NIBP) measurements. Currently, medical personnel must decide which blood pressure monitoring interval is appropriate based on an assessment of the patient's hemodynamic stability.

[0049] Because a patient's health status may change over time, the NIBP measurement interval would ideally need to be continually adjusted to reflect changes in the patient's risk level. However, in reality, this is not always done, which can lead to several adverse consequences, including:

[0050] Missing critical events (e.g., periods of hypotension and hypertension);

[0051] Unnecessary NIBP measurement affects patient comfort and causes sleep disturbance;

[0052] An inaccurate representation of a patient's response to a treatment (eg, medication) leading to subsequent suboptimal management decisions.

[0053] References have been discussed above Figure 1 Examples of NIBP trend monitoring during unstable and stable phases are discussed. Figure 1 The diagram shows that if NIBP monitoring is performed using intermittent measurements with a fixed measurement interval (line 24), many events with large BP changes (line 22) will be missed. The measurement interval selected by the clinician is not appropriate for this patient. It is worth noting that Figure 1 The graph shown shows a blood pressure trend line according to a zero-order hold representation (eg, representing a discrete-time blood pressure signal as a continuous-time signal by holding each sample value for one sampling interval).

[0054] Blood pressure can be used as an indicator of a variety of adverse medical events in a patient. Two adverse events that clinicians typically focus on when monitoring patients are periods of hypertension and hypotension. Hypertensive events are periods of very high blood pressure, typically understood as systolic blood pressure above about 140 mm / Hg. Hypotensive events are periods of very low blood pressure, typically understood as systolic blood pressure below about 90 mm / Hg.

[0055] Depending on the stage of care, if the duration of the hypotensive period is experienced, the patient's likelihood of death or long-term health problems may increase significantly. For example, even if the hypotensive period is only experienced for a short time, the postoperative care stage is particularly at great risk for myocardial infarction and death due to the resulting hypoperfusion (ischemia) of the heart and brain. For example, refer to the following article: McEvoy, M et al, Perioperative Quality Initiative consensus statement on postoperative blood pressure, risk and outcomes for elective surgery, British Journal of Anaesthesia, Volume 122, Issue 5, 2019, Pages 575-586. The article explains, with reference to the data presented on page 578, that every 10-minute episode of hypotension on postoperative day 0 is associated with a 3% increased risk of myocardial infarction (MI) and death, while any episode of hypotension on postoperative days 1-4 is associated with almost double the risk of MI and death.

[0056] If the patient is in the hypertensive or hypotensive stage, the clinician must act quickly through medication or other interventions to restore the patient to a normal hemodynamic state.

[0057] A hypotensive event occurs when the patient's actual blood pressure value falls below a certain predefined critical threshold. The duration of a hypotensive event can be defined as the interval (in minutes) from when the actual blood pressure falls below the critical blood pressure threshold to when it rises above the predefined critical threshold again.

[0058] The severity of hypotension is related to its depth and duration. Depth refers to the lowest blood pressure value reached during a hypotensive episode. By generalizing this concept, a depth-duration dose metric (DDD) can be defined for each patient's hypotensive episode in a general form as follows:

[0059]

[0060] Among them, t1 is the start time of the period, t2 is the end time of the period, BP 阈值 is the critical blood pressure threshold, and BP测量 is the actual blood pressure value of the patient over time. You can refer to the following paper that outlines the definition of DDD: DWSpaite et al., "Association of Out-of-Hospital Hypotension Depth and Duration With Traumatic Brain Injury Mortality," Ann. Emerg. Med., vol. 70, no. 4, pp. 522-530. e1, 2017.

[0061] Regarding the critical hypotension threshold, there are many ways to define this. Regardless of the definition, studies have shown that in different populations (e.g., noncardiac surgery patients and patients with creative brain injury), the patient's prognosis depends largely on the duration and depth of the hypotensive episode (i.e., the extent of the drop in blood pressure, such as the lowest value). This is because the damage is accumulated during the continuous hypotensive episode. For example, by quantifying the hypotensive dose (the integral of the depth of hypotension during the exposure time) in 7521 patients with traumatic brain injury, one study (ibid) showed that hypotension was associated with a 20% increase in mortality. In addition, there is also evidence that there is a significant association between the degree and duration of hypotension during and after surgery on the one hand and myocardial damage and acute kidney injury (AKI) on the other hand. For example, refer to the following paper: EJMascha, et al, "Intraoperative Mean Arterial Pressure Variability and 30-day Mortality in Patients Having Noncardiac Surgery," Anesthesiology, vol. 123, no. 1, pp. 79–91, 2015.

[0062] More evidence shows that postoperative hypotension events are associated with poor prognosis. One study showed that SBP < 90 mmHg was the most common reason (25%) for the activation of the emergency team in the ward after orthopedic or general surgery. The study is detailed in the following paper: Mohammed Iddrisu, et al, "Frequency, nature and timing of clinical deterioration in the early postoperative period," J. Clin. Nurs., vol. 27, no. 19–20, pp. 3544–3553, 2018.

[0063] A further study collected and studied data from 90 hospitals in the United Kingdom, Australia, and New Zealand and found that the most common antecedent to a patient's deterioration was a systolic blood pressure (SBP) < 90 mmHg. This study is detailed in the following paper: J. Kause, et al, "A comparison of Antecedents to Cardiac Arrests, Deaths and Emergency Intensive care Admissions in Australia and New Zealand, and the United Kingdom-The ACADEMIA study," Resuscitation, vol. 62, no. 3, pp. 275–282, 2004.

[0064] As a consequence of these problems, patient outcomes are affected, such as longer hospital stays, higher morbidity, and higher mortality. Therefore, improved blood pressure management will help improve patient outcomes and reduce complications, and thereby increase hospital efficiency.

[0065] Embodiments of the present invention provide a system and method for estimating and automatically configuring optimal / appropriate blood pressure measurement intervals that are customized to the patient's risk profile and physician's opinion.

[0066] Figure 1 The steps of an example method according to one or more embodiments are summarized in block diagram form. These steps will be described in summary form and then further explained in the form of an example embodiment.

[0067] The method includes receiving 12 biometric data for a patient.

[0068] The method further comprises applying 14 a patient risk model to calculate a risk parameter p indicating a risk level of at least one predefined adverse clinical event, wherein at least one input to the model is biometric data.

[0069] The method further comprises obtaining 16 a clinician-defined patient risk parameter A.

[0070] The method further comprises calculating 18 a measurement time interval T for successive non-invasive blood pressure measurements of the patient based on A and p and according to the following relationship:

[0071]

[0072] Here, α is a predefined parameter.

[0073] The method further comprises controlling 20 the non-invasive blood pressure measurement device to acquire blood pressure measurements at a frequency defined by the determined time interval T.

[0074] As mentioned above, the method may also be embodied in hardware form, for example in the form of a processing unit configured to perform the method according to any example or embodiment described in this document or according to any claim of this application.

[0075] To further aid understanding, Figure 2 A schematic diagram of an example processing unit 32 configured to perform a method according to one or more embodiments of the invention is presented and also schematically illustrates the processing flow in more detail.

[0076] The processing unit 32 includes an input / output unit 34 and one or more processors 36. The one or more processors are configured to perform the following steps:

[0077] Receiving biometric data of the patient at the input / output 34, such as from one or more biometric devices or sensors 44;

[0078] applying a patient risk model to calculate a risk parameter p indicating a risk level of at least one predefined adverse clinical event, wherein at least one input to said model is said biometric data;

[0079] obtaining a clinician-defined patient risk parameter A, such as based on input received at the input / output 34 from the user interface 46;

[0080] The measurement time interval T of successive non-invasive blood pressure measurements for the patient is calculated based on A and p according to the following relationship:

[0081]

[0082] where α is a predefined parameter; and

[0083] The non-invasive blood pressure measurement device 42 is controlled to acquire blood pressure measurements at a frequency defined by the determined time interval T via the generation of a control signal for output at the input / output section 34 .

[0084] Another aspect of the invention is a system 30 comprising a processing unit 32 . Figure 2A processing unit 32 is shown operating in the context of such a system. The system 30 may also include a non-invasive blood pressure measurement device 42, such as a cuff-based oscillometric blood pressure measurement device, operatively coupled to the processing unit. The system 30 may also include one or more bio-parameter sensing devices 44 for collecting biometric data, operatively coupled to the processing unit 32. For example, the one or more bio-parameter sensing devices include a PPG sensor. The PPG sensor can be used to collect pulse rate or heart rate and blood oxygen concentration. It is also possible to infer the patient's breathing rate from the PPG sensor signal. In some examples, the PPG sensor can be advantageously integrated into a portion of the non-invasive blood pressure measurement device.

[0085] The system 30 may also include a user interface 46, for example for receiving user input indicating a clinician-defined patient risk parameter A, and / or for displaying a representation of a determined blood pressure measurement interval, a current value of a calculated risk parameter, one or more biometric results of the patient, or any combination thereof.

[0086] To further clarify, a brief example workflow of the method will now be outlined.

[0087] Patients in a medical setting are monitored using a set of one or more measuring devices 42 (e.g., sensors), each of which is configured to measure at least one biological parameter, such as a physiological parameter. Data output from the one or more measuring devices is received at a processing unit 32. The data forms a biometric data set for the patient. In some examples, the processing unit may be a processing unit of a patient monitoring system. In addition to or in lieu of real-time data from the measuring device, the biometric data set may also include data retrieved from a data repository, for example, the data may include historical values ​​of one or more biological parameters, or variables related to the patient's medical history. It may include parameters that cannot be measured with sensors, such as pathology test results.

[0088] The processing unit 32 is also used to obtain an indication of a risk parameter A defined by a clinician. In some examples, the parameter may be obtained based on user input received from the user interface device 46. For example, the clinician responsible for monitoring the patient may input the parameter. In another example, the parameter A may be determined for the patient based on a lookup table. For example, the clinician may have determined in advance a set of relationships between different patient state classifications and associated values ​​of the parameter A, and these may be encoded in a stored lookup table.

[0089] The processing unit 32 is also adapted to retrieve a patient risk model from a data repository. The patient risk model is adapted to receive input values ​​of one or more variables corresponding to one or more corresponding biometric measurements. These may include some or all of the values ​​forming the received biometric data set. The patient risk model is configured to generate as an output a value of a parameter p that indicates at least one predefined adverse risk level, such as hypotension, hypertension, or a cardiac event. The value of p may be a probability value and may take a value between 0 and 1. The model may be a generic model for any patient or a personalized model for the patient, as will be explained in more detail later. In some cases, the model may also be additionally or alternatively customized to the clinician monitoring the patient, as will be explained in more detail later.

[0090] After retrieving the model, the processing unit 32 applies the model by providing the relevant biometric measurements as input to the model, thereby generating an output risk parameter value p.

[0091] Then, the processing unit 32 firstly The value of is used to calculate the time interval T of consecutive non-invasive blood pressure measurements for the patient, and then based on the relationship To calculate the value of T, where α is a further variable parameter.

[0092] In some embodiments, the formula To simply determine T.

[0093] In some embodiments, T may be determined based on evaluating the following formula: Wherein, β may be, for example, a constant, or β may be, for example, a predefined transfer function. For example, β may be a function of any one or all of A, p, and α.

[0094] The purpose of parameter β is to The value of can be transferred or converted to a value in the correct unit suitable for T. The parameter β can be predefined to provide the required ratio of T and A specific mapping between .

[0095] The parameter α provides a further degree of freedom in the calculation to optimize the calculation of the time interval. This parameter effectively defines the relative weight to be applied to the computer-determined risk parameter p compared to the clinician-defined risk parameter A. It will be appreciated that as the value of α increases, the scaling of the value pα as a function of p increases in steepness. The effect is that a higher value of α means that a greater relative weight is given to the computer-defined risk parameter p compared to the clinician-defined risk parameter A.

[0096] In some examples, parameter α may be freely determined by the user. In some examples, it can be determined using a predefined lookup table or mapping, which relates different patient or clinician states or categories to different α values. For example, for patients with higher urgency, it may be preferred to set a lower α value, which will have the effect of reducing the time interval between blood pressure measurements when (for p values ​​varying between 0 and 1) the risk parameter p increases during the time period when the clinician updates his own risk assessment A. Therefore, in this case, the higher the α value, the more cautious it is. For further example, for more experienced clinicians who trust their own assessments, the value of α may be set lower than that of less experienced clinicians. This will have the effect (assuming that the value of p varies between 0 and 1), that is, giving a higher relative weight to the parameter A defined by the clinician compared to the case where the α value is lower.

[0097] The processing unit 32 may store in a memory a record of the newly calculated value of the calculation interval time T. This may be updated each time T is recalculated. The calculated value of T may be recorded over time. T may be recalculated repeatedly during a monitoring session for a patient. The risk parameter p may be recalculated repeatedly during a monitoring session for a patient, wherein the biometric data is updated for each recalculation. Using the newly calculated value of p, the processing unit may recalculate T.

[0098] The processing unit 32 is further configured to control the NIBP measurement device to acquire BP measurements at a frequency defined by the determined time interval T. The actual control of the NIBP measurement device may be implemented by the processing unit 32 itself, or by a subsystem operatively coupled to the processing unit. In the latter case, the step of controlling the NIBP measurement device to acquire BP measurements at a frequency defined by the determined time interval may comprise simply transmitting a control command to the NIBP measurement device control subsystem to cause the subsystem to acquire measurements at the calculated intervals T. Alternatively, if the processing unit is adapted to coordinate the control of the NIBP measurement device itself, it may simply update the value of the parameter T in a local register, and wherein the processing unit runs a control routine or program for triggering blood pressure measurements at time intervals determined by the current value of T in the register.

[0099] As described above, the value of T is preferably recalculated repeatedly. With respect to the timing of the recalculation, this may be performed at regular fixed time intervals of seconds, or may be performed at intervals set by the clinician, or may be performed at automatically determined or triggered intervals. For example, one approach may be to recalculate T each time a new blood pressure measurement is triggered. In other words, the value of T itself is updated at a frequency determined by the current value of T, so that at the end of each measurement interval, a new blood pressure measurement is acquired and, at the same time, a new value of the interval time T is calculated. To calculate a new value for T, each of the one or more biometric measurements may be sampled and these values ​​used as input to rerunning the patient risk model to generate a new value for the parameter p, which is then calculated using the relationship outlined above.

[0100] It is worth noting that the patient risk model is naturally trained or programmed to model the patient risk for a specific adverse clinical event. Therefore, in some embodiments, there may be multiple versions of the model, each version being configured to predict the probability of a different type of adverse event.

[0101] In some embodiments, multiple such versions of the risk model may be run to derive a set of initial risk parameters p i , each indicating the likelihood of a corresponding adverse clinical event occurring. These initial risk parameters may then be further processed to derive an overall risk parameter p, which indicates the risk of one or more adverse clinical events. In some embodiments, only one of the initial risk parameters pi may be selected, corresponding to the adverse event that is most likely to occur or has the most serious consequences for the patient. The selection may be based on user input at a user interface. For example, a user interface displays a set of calculated initial risk parameters, along with an indication of each corresponding adverse clinical event, and the user may select one to be used as an overall patient risk parameter. In another example, some or all of the parameters in the set of initial risk parameters may be combined to form an overall patient risk parameter p. For example, the probabilities may be simply added to obtain the cumulative probability of any corresponding adverse clinical event occurring. Alternatively, an average value may also be calculated.

[0102] Further details regarding one or more of the features briefly summarized above will now be described.

[0103] With regard to the biometric data obtained as part of the method, preferably it is received continuously or in real time from one or more measurement devices or sensors used to monitor the patient. In this way, the patient risk model can be provided with up-to-date information for the patient. An alternative approach is to use archived biometric data for the patient.

[0104] By way of non-limiting example, biometric data may include one or more of the following:

[0105] ECG data;

[0106] Photoplethysmography (reflection or transmission) data;

[0107] Pulse oximetry / blood oxygen saturation data;

[0108] Capnography data;

[0109] Bioimpedance measurement data;

[0110] Any other vital signs;

[0111] Pathological examination results.

[0112] In some examples, a photoplethysmogram (PPG) sensor may be provided as an auxiliary sensor, separate from the blood pressure measurement device, such as a finger PPG sensor. In some examples, the PPG sensor may be provided to be integrated in the blood pressure measurement device, such as in a blood pressure measurement cuff, and arranged so that when the cuff is worn on a body part to measure blood pressure, the PPG sensor optical region is in optical communication with the patient's skin surface.

[0113] These data may be received in real time from one or more bio-parameter sensors. Additionally or alternatively, the biometric data may include historical biometric data of the patient retrieved from a data repository.

[0114] With regard to the non-invasive blood pressure (NIBP) measurement device controlled as part of the method, this preferably includes a cuff-based measurement device. In use, the cuff may be applied to the patient's arm or wrist. A typical example of such a device operates as follows: by actively pressurizing the cuff by a pump or pressure reservoir, and then measuring the cuff pressure using a pressure sensor. There are also other methods of operating a cuff-based NIBP measurement device, and these methods are known to those skilled in the art. Any other type of NIBP measurement device may also be used as a substitute for the cuff system. NIBP measurement devices are well known in the art, and those skilled in the art will understand how to implement this feature.

[0115] The guiding insight behind the inventors' proposed approach is that for a constant value of a clinician-defined parameter A, the product of the probability p for a critical event (perhaps adjusted by a power law) and the NIBP interval time T is maintained constant. Thus, in any interval in which A remains constant (i.e., the clinician does not provide any updated assessment of the patient's status or risk by themselves), the patient's risk level calculated by the model increases, resulting in a shortening of the NIBP interval.

[0116] As described above, according to this principle, in some embodiments, the measurement value time interval T can be determined as follows:

[0117]

[0118] T is the recommended measurement interval. A is a physician-defined parameter that represents the physician's assessment of the patient's risk level for a specific (predefined) adverse or serious event (e.g., hypotension, hypertension) during the future monitoring period. It is expected that the physician will choose this based on their judgment of the situation, using their knowledge and experience to assess the risk level. Parameter p is the probability of the same predefined adverse event occurring within a defined time window calculated based on the patient risk model. The exponent / power parameter α can be adjusted to provide more or less weight to the computer risk parameter p compared to the clinician risk parameter A.

[0119] In some examples, a calibration multiplier or factor β can be added to the above formula to provide an estimate of The correct mapping between the resulting numerical values ​​and the desired time units of T. For example, the formula can then take In some examples, β may be a single-valued constant. In some examples, β may be a function of any or all of A, p, and α, such that it provides The transfer function between the value of and the timing value T for the appropriate measurement interval. Throughout this disclosure, reference to the use of It may be understood to optionally include a coefficient or transfer function β as a further multiplier.

[0120] Alternatively, the calibration may have been incorporated into the permissible range of values ​​for A, p and α. For example, each of these parameters may take values ​​within a respective range, and / or within a respective interval within the range, which is configured such that The evaluation of automatically produces T values ​​with the correct units.

[0121] For example, in a clinical scenario, a typical interval is estimated to be 15 minutes. Assume that the hospital policy allows a risk level of 1 event per 1,000 patients, A = 0.001. The probability of an event occurring per minute is p = 6.67*10 -5 / minute, resulting in a recommended interval time T = 15 minutes.

[0122] For illustration only, the following table shows other different example combinations and their resulting interval times. The parameter α has been set to α=1.

[0123]

[0124] The effect of this formula is that an increased risk level for patients who are likely to suffer a critical event results in a decrease in the measurement interval.

[0125] Parameter A may be specified by the specific clinician treating the patient and represents the clinician's assessment of the patient's risk level. Additionally or alternatively, parameter A may be set based on standardized guidelines for a specific department of the hospital or in addition to general guidelines, for example, based on a lookup table that associates different classifications of patient conditions with different values ​​of parameter A.

[0126] The time-dependent estimation of the critical event probability p allows the interval time T to be adjusted in real time.

[0127] In some embodiments, in addition to determining the value of T, the method may further comprise determining an estimate of the uncertainty of T. For this purpose, an uncertainty parameter ΔT may be calculated.

[0128] A simple way to estimate uncertainty is to define uncertainty interval values ​​ΔA and Δp. Δp can be estimated by a Bayesian model. The Bayesian model inherently outputs a probability distribution, as well as an uncertainty dp, and this uncertainty can be used for Δp. The value of ΔA can be predefined, for example, using a lookup table, or defined according to hospital policy based on patient condition or other factors.

[0129] Using ΔA and Δp, the estimated uncertainty parameter, ΔT for T can be calculated as follows:

[0130]

[0131] The calculated value of the uncertainty parameter ΔT can be presented to the user (clinician) on a display screen of the user interface together with the current calculated value of T. The user can be provided with the option of whether to accept T as the current measurement interval via the user interface. The value of the uncertainty parameter provides additional information that helps the user make a decision. If the uncertainty is high, the user (clinician) may decide to act cautiously and set the interval to a shorter time than the current value of T.

[0132] In some examples, the user may be provided with a set of three preset options for the measurement interval time: T, T+ΔT, or T-ΔT. For example, if the uncertainty is relatively high, the user may select T-ΔT.

[0133] In the above equation, the value of ΔA can be set to zero or estimated based on historical data or guidelines.

[0134] Part of the method is to use the patient risk model to calculate a risk parameter p, the risk parameter p being indicative of at least one predefined adverse clinical event. Details regarding the patient risk model will now be discussed in further detail.

[0135] The patient risk model is configured to receive inputs during operation, the inputs including at least: values for each of one or more biometric parameters for the patient (as described above). The patient risk model is configured to generate an output value p, which indicates the probability of a predefined adverse clinical event occurring within a specific time interval.

[0136] A particularly suitable method for estimating such a probability is to use a Bayesian statistical model.

[0137] A statistical model using the Bayesian principle can determine the posterior probability distribution of a specific outcome based on one or more input variables, where the model has been constructed based on one or more prior probability distributions.

[0138] Using the Bayesian principle in this way to derive a risk estimate based on the current values of one or more parameters and the prior distributions of these parameters is well known to those skilled in the art.

[0139] For example, a Bayesian model can be constructed to provide a value of p, which indicates the estimated risk that blood pressure drops below a value x for a duration of y minutes, where x and y are set defined values corresponding to the hypotensive event of interest. For example, x might be 90 mm / Hg and y = 15 minutes. For this scenario, the prior distribution can be determined in advance based on the variation of the measured biometric parameter d(t). If multiple biometric parameters are to be measured and used as inputs to the model, the prior distribution of p can be calculated as a function of the multi-dimensional space of d(t), where d(t) is a vector representing the set of all biometric parameters (and / or other parameters) that will be used as inputs to the model.

[0140] Using the standard Bayesian theorem, the probability p that blood pressure (BP) drops below x and persists for y minutes can be expressed as:

[0141]

[0142] where,

[0143] K is a context parameter, such as the patient's condition; d(t)|BP < x mmHg for y min, K represents the probability distribution of obtaining the measurement dataset d(t) given that the condition [BP < x mmHg for y minutes; K] is true;

[0144] p(BP < x mmHg for y minutes, K) is the prior probability distribution for the condition [BP < x mmHg for y minutes; K] for any d(t); and

[0145] p(d(t)) is the prior probability distribution of obtaining d(t) for any blood pressure measurement.

[0146] Using standard Bayesian notation, the equation for T described above can be written as:

[0147]

[0148] As described above, the probability p may preferably be estimated iteratively (eg, continuously or quasi-continuously) based on collected real-time biometric data and / or available historical data.

[0149] Using Bayesian methods, a variety of different factors and variables can be considered as long as appropriate prior distributions are obtained. For example, in addition to the measured biometric data d(t), the following parameters can be factored into the probability p output from the model:

[0150] the level of training of the physician responsible for the patient;

[0151] How quickly the physician can respond within a given time span;

[0152] Patient history;

[0153] Patient severity.

[0154] Each of these variables can be encoded according to its own standardized score metric, so that the prior distribution of p can be measured by the value of each variable.

[0155] By taking these various factors into account, the psychological process of humans considering new data can be effectively simulated.

[0156] In some embodiments, the method may also include updating the patient risk model based on monitoring data for patients in the medical setting. For example, the monitoring data may include blood pressure measurement data and biometric data, such that the relationship between these parameters may be used to train the model. Thus, the model may learn over time via retrospective analysis of monitored patients.

[0157] Furthermore, changes in clinician training levels can be monitored and reflected in updates to the prior information used to train the model.

[0158] To update the model, the method may further include a model update phase, which may be run once or at intervals that are the same as or different from any (optional) update intervals of the parameters p and T. For example, the update phase may include executing an algorithm for collecting and collating relevant updated model prior distributions, and updating the model using the updated prior distributions.

[0159] It is worth noting that although the above embodiments use a Bayesian model for the patient risk model, the risk model may also be implemented using other types of models. For example, a machine learning model may be used in a further example, which includes one or more machine learning algorithms, and the one or more machine learning algorithms are trained to receive at least the patient's biometric data as input and generate a risk parameter p as output, and the risk parameter p indicates the risk level of at least one predefined adverse clinical event. Suitable machine learning algorithms include, but are not limited to, a multivariate regression model or an artificial neural network such as a convolutional neural network.

[0160] It is worth noting that, although the above-described embodiments relate to blood pressure measurement intervals, the concepts of the present invention are more broadly applicable to determining time intervals for any other biological measurement results (e.g., another physiological parameter measurement, such as another vital sign measurement). The method is most applicable to measurements that cannot or should not preferably be performed continuously.

[0161] As mentioned above, the present invention can be embodied in software form. Therefore, another aspect of the present invention is a computer program product comprising code units, which are configured to, when running on a processor, cause the processor to perform a method according to any example or embodiment of the present invention described in this document or according to any claim of this patent application.

[0162] It should be noted that for any embodiment of the present invention, the method can be performed by a processing unit located in a variety of different locations. For example, the method can be performed by a processing unit of a bedside patient monitoring device, or by a patient monitoring system, or by a separate computing system, or by a cloud computing system.

[0163] The embodiments of the present invention described above employ a processing unit. A processing unit may generally include a single processor or multiple processors. It may be located within a single containing device, structure or unit, or it may be distributed between multiple different devices, structures or units. Therefore, reference to a processing unit being adapted or configured to perform a specific step or task may correspond to the step or task being performed by any one or more of a plurality of processing components, either individually or in combination. Those skilled in the art will understand how to implement such a distributed processing device. A processing unit may include a communication module or input / output unit for receiving data and outputting data to other components.

[0164] The one or more processors of the processing unit can be implemented in a variety of ways, such as software and / or hardware, to perform the various functions required. The processor usually adopts one or more microprocessors, which can be programmed using software (e.g., microcode) to perform the required functions. The processor can be implemented as a combination of dedicated hardware for performing some functions and one or more programmed microprocessors and associated circuits for performing other functions.

[0165] Examples of circuits that may be used in various embodiments of the present disclosure include, but are not limited to, conventional microprocessors, application specific integrated circuits (ASICs), and field programmable gate arrays (FPGAs).

[0166] In various implementations, the processor may be associated with one or more storage media, such as volatile and non-volatile computer memory, such as RAM, PROM, EPROM, and EEPROM. The storage media may be encoded with one or more programs that perform the desired functions when run on one or more processors and / or controllers. The various storage media may be fixed within the processor or controller, or may be transportable so that one or more programs stored thereon can be loaded into the processor.

[0167] Those skilled in the art can understand and implement variations to the disclosed embodiments when practicing the claimed invention by studying the drawings, the disclosure and the claims. In the claims, the word "comprising" does not exclude other elements or steps, and the word "one" or "an" does not exclude a plurality.

[0168] A single processor or other unit may fulfill the functions of several items recited in the claims.

[0169] The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

[0170] The computer program may be stored / distributed on suitable media such as optical storage media or solid-state media provided together with or as part of other hardware, but may also be distributed in other forms such as via the Internet or other wired or wireless telecommunication systems.

[0171] If the term "suitable for" is used in the claims or the specification, it should be noted that the term "suitable for" is intended to be equivalent to the term "configured to".

[0172] Any reference signs in the claims should not be construed as limiting the scope.

Claims

1. A computer-implemented method comprising: receiving (12) biometric data for a patient; applying (14) a patient risk model to calculate a risk parameter p, the risk parameter p being indicative of a risk level of at least one predefined adverse clinical event, wherein at least one input to the model is the biometric data; Obtaining (16) a patient risk parameter A defined by a clinician; Based on A and p and according to the following relationship, the measurement time interval T for consecutive non-invasive blood pressure measurements of the patient is calculated (18): where α is a predefined parameter; and The non-invasive blood pressure measurement device is controlled (20) to collect blood pressure measurement results at a frequency defined by the determined time interval T.

2. The method according to claim 1, wherein: At least one adverse or critical event comprises hypotension or hypertension in said patient.

3. The method according to claim 1 or 2, wherein: p is the probability of the event occurring within a defined time interval.

4. The method according to any one of claims 1 to 3, wherein: The biometric data includes real-time sensor data, or data derived from the real-time sensor data.

5. The method according to any one of claims 1 to 4, wherein: The biometric data includes historical biometric data for the patient retrieved from a data repository.

6. The method according to any one of claims 1 to 5, wherein: The risk parameter p is repeatedly recalculated during a monitoring session for the patient, wherein the biometric data are updated for each recalculation result.

7. The method according to any one of claims 1 to 6, wherein: The risk model is a Bayesian model.

8. The method according to claim 7, wherein: The Bayesian model is a personalized risk model for the patient and / or the clinician treating the patient, and is pre-configured based on prior information including one or more of the following: Patient history; the severity of the patient's condition; the training level of the clinicians treating the patients; A measure of how quickly a physician responds to changes in a patient's condition.

9. The method according to any one of claims 1-8, further comprising iteratively adjusting or updating the risk model based on a patient monitoring database comprising monitoring data for a plurality of patients.

10. The method according to any one of claims 1 to 9, wherein: The biometric data includes data from one or more of the following: Electrocardiogram (ECG) sensing device; Photoplethysmography (PPG) sensing device; Capnography devices; and Bioimpedance measurement device.

11. A computer program product comprising code means configured to, when executed on a processor, cause the processor to perform the method according to any one of claims 1-10.

12. A processing unit (32), comprising: Input / output (34); as well as One or more processors (36) adapted to: receiving biometric data for a patient at the input / output; applying a patient risk model to calculate a risk parameter p indicating a risk level of at least one predefined adverse clinical event, wherein at least one input to the model is the biometric data; Obtain the patient risk parameter A defined by the clinician; The measurement time interval T between successive non-invasive blood pressure measurements for a patient is calculated based on A and p according to the following relationship: where α is a predefined parameter; and The non-invasive blood pressure measurement device is controlled to acquire blood pressure measurement results at a frequency defined by the determined time interval T by generating a control signal for output at the input / output section.

13. A system (30) comprising: The processing unit (32) according to claim 12; as well as A non-invasive blood pressure measurement device (42) is operatively coupled to the processing unit.

14. The system according to claim 13, further comprising one or more bio-parameter sensing devices (44) for collecting biometric data, the one or more bio-parameter sensing devices being operatively coupled to the processing unit.

15. The system of claim 14, wherein: The one or more bio-parameter sensing devices include a PPG sensor integrated in a portion of a non-invasive blood pressure measurement device.

Images