HK40008528B - Method and device for risk-based control-to-range - Google Patents

Description

Risk-based scope control methods and devices

Cross-references to related applications

This application claims priority to U.S. Patent Application Serial No. 15/170468, filed June 1, 2016, entitled “RISK-BASED CONTROL-TO-RANGE,” the entire contents of which are incorporated herein by reference.

Technical Field

The present invention generally relates to processing glucose data measured from individuals with diabetes, and more specifically, to controlling adjustments to the provisional basal rate based on the risk associated with the glucose status of individuals with diabetes.

Background Technology

Many people have type 1 or type 2 diabetes, in which the body does not properly regulate blood sugar levels. Continuous glucose monitoring (CGM) allows for continuous measurement of interstitial glucose levels in patients with diabetes, such as every few minutes. The timing and dosage of insulin administered to the patient can be determined based on the measurements recorded by the CGM device. The glucose readings from the CGM device are displayed to the patient, who can inject insulin or consume meals to help control glucose levels. An insulin pump can deliver precise insulin doses according to a programmable schedule that can be adjusted by the patient or healthcare provider.

A risk metric can be derived from glucose data to assess the risk to diabetic patients based on detected glucose levels. For example, known risk metrics include the risk function proposed in the following article: Kovatchev, B. P. et al., Symmetrization of the blood glucose measurement scale and its applications, Diabetes Care, 1997, 20, 1655-1658. The Kovatchev risk function is defined by an equation where g is the blood glucose concentration (in milligrams per deciliter or mg/dl) and h(g) is the corresponding penalty value. The Kovatchev function provides a static penalty (i.e., risk) value because the penalty depends only on the glucose level. The minimum (zero) risk occurs at 112.5 mg/dl. The risk of hypoglycemia increases significantly faster than the risk of hyperglycemia when glucose levels approach hypoglycemia.

The Kovatchev hazard function fails to account for the rate of change in glucose levels and the uncertainties associated with the measured glucose level. For example, the risk associated with a patient at 100 mg/dL and a rapidly decreasing blood glucose level may be greater than the risk associated with a patient at 100 mg/dL with a constant rate of change in glucose. Furthermore, the measured glucose results may be inaccurate due to sensor noise, sensor malfunction, or sensor detachment.

Various methods have been developed to control glucose levels in diabetic patients based on CGM glucose data. One method for limiting hypoglycemic conditions involves an insulin pump shutdown algorithm that completely shuts off basal insulin if CGM glucose levels drop below a low glucose threshold (such as 50 to 70 mg/dL), and then resumes basal insulin several hours later. However, this on/off approach disadvantageously requires an adverse condition of crossing the low glucose threshold to occur before action is taken. Furthermore, this method does not account for the rate at which glucose crosses the threshold, which can be problematic for patients with high glucose variability (e.g., children, active individuals, etc.).

Another approach is to alert the patient to a predicted hypoglycemia, after which the patient consumes a certain amount of carbohydrates and waits for a predetermined period. If the system still predicts hypoglycemia, the patient repeats this cycle until the system no longer predicts hypoglycemia. However, this method assumes that the patient can immediately consume carbohydrates when alerted to a predicted hypoglycemia. Furthermore, patients may overcorrect by consuming too many carbohydrates, potentially leading to weight gain or causing glucose levels to tend towards hyperglycemia.

Therefore, some embodiments of this disclosure provide a prediction method for adjusting the therapeutic basal rate based on the cumulative hazard value of a return path generated according to a glucose state distribution surrounding the estimated glucose state by mapping the estimated risk of glucose state to an adjustment of the basal rate. The risk associated with glucose state is based on blood glucose level, rate of change of blood glucose level, and standard deviation of blood glucose level and rate of change. Furthermore, some embodiments provide adjustments for the calculated risk for glucose state in response to dietary increases, insulin increases, and/or other events such as exercise, glucagon availability, and stress responses that may affect the risk of hypoglycemia or hyperglycemia.

Summary of the Invention

In one embodiment, a method is provided for determining basal rate adjustment of insulin based on the risk associated with the glucose status of a person with diabetes. The method includes receiving a signal representing at least one glucose measurement result by at least one computing device. The method further includes detecting the person's glucose status by the at least one computing device based on the signal, the detected glucose status including the person's glucose level and the rate of change of the glucose level. Furthermore, the method includes determining a current risk measure associated with the detected glucose status by the at least one computing device based on a target glucose status, the target glucose status being stored in memory accessible by the at least one computing device, the current risk measure indicating the risk of at least one of hypoglycemia and hyperglycemia in the person. A return path is determined based on the transition from the current glucose status to the target glucose status, the return path including at least one intermediate glucose value associated with the return to the target glucose status. Additionally, a cumulative hazard value for the return path is determined, the cumulative hazard value including the sum of hazard values for at least one glucose value on the return path, each hazard value indicating a hazard associated with a corresponding intermediate glucose value. Furthermore, a current risk measure is determined based on a weighted average of the cumulative hazard values of the return path, the return path being generated according to a glucose status distribution surrounding the detected glucose status. The method further includes identifying a reference glucose state and a reference risk metric associated with the reference glucose state by the at least one computing device; and calculating an adjustment to the baseline rate of the treatment delivery device by the at least one computing device based on the current risk metric associated with the detected glucose state and the reference risk metric associated with the reference glucose level.

In another embodiment, a blood glucose management device is provided, configured to determine a basal rate adjustment based on a risk associated with the glucose status of a person with diabetes. The device includes a non-transitory computer-readable medium storing executable instructions; and at least one processing device configured to execute the executable instructions such that, when executed by the at least one processing device, the executable instructions cause the at least one processing device to receive a signal representing at least one glucose measurement result. The executable instructions also cause the at least one processing device to detect the person's glucose status based on the signal, the detected glucose status including the person's glucose level and the rate of change of glucose level. Additionally, the executable instructions cause the at least one processing device to determine a current risk measure associated with the detected glucose status based on a target glucose status stored in memory accessible by the at least one computing device, the current risk measure indicating a risk of at least one of hypoglycemia and hyperglycemia in the person. A return path is determined based on the transition from the current glucose status to the target glucose status, the return path including at least one intermediate glucose value associated with the return to the target glucose status. A cumulative hazard value is determined for the return path, the cumulative hazard value comprising the sum of hazard values for the at least one glucose value along the return path, each hazard value indicating a hazard associated with a corresponding intermediate glucose value. A current risk metric is determined based on a weighted average of the cumulative hazard values for the return path, the return path being generated based on the glucose state distribution surrounding the detected glucose state. The executable instructions further cause the at least one processing device to identify a reference glucose state and a reference risk metric associated with the reference glucose state. Finally, the executable instructions further cause the at least one processing device to calculate an adjustment to the basal rate of the treatment delivery device based on the current risk metric associated with the detected glucose state and the reference risk metric associated with the reference glucose level.

Attached Figure Description

The embodiments illustrated in the accompanying drawings are illustrative and exemplary in nature and are not intended to limit the invention as defined by the claims. The following detailed description of the illustrative embodiments can be understood when read in conjunction with the following drawings, wherein similar structures are indicated by similar reference numerals, and wherein:

Figure 1 illustrates a continuous glucose monitoring (CGM) system according to one or more embodiments shown and described herein;

Figure 2 illustrates an exemplary blood glucose management device, treatment delivery device, and glucose sensor of the CGM system of Figure 2. The blood glucose management device includes a dosage calculator module, range control logic, hazard analysis logic, recursive filter, and basal rate adjustment logic.

Figure 3 illustrates a graph of an exemplary CGM trace and the adjusted maximum allowable glucose after a dietary event;

Figure 4 illustrates the graph of the periodic update of the base rate;

Figure 5 illustrates a graph of the danger function with exemplary hyperglycemic aggression and hyperglycemic shift adjustment;

Figure 6 illustrates a graph of the risk function for hypoglycemia shift caused by exercise or glucagon availability;

Figure 7 illustrates a graphical representation of an exemplary return path to a target glucose level;

Figure 8A illustrates a hypoglycemia risk surface, which has an array of sample locations corresponding to the distribution of glucose states.

Figure 8B illustrates an exemplary return path for the glucose status highlighted in Figure 8A;

Figure 9 illustrates the graphs for providing continuous and incremental base multipliers;

Figure 10A illustrates the base rate adjustment plot;

Figure 10B illustrates the basal rate adjustment plot of Figure 10A, showing the hyperglycemia shift caused by recent dietary or corrective increases; and.

Detailed Implementation

The embodiments described herein generally relate to methods and systems for determining basal rate adjustments of insulin in a continuous glucose monitoring system for a person with diabetes, and more specifically, to methods and systems for determining basal rate adjustments of insulin based on risk associated with the glucose status of a person with diabetes. For the purposes of this disclosure, "measured glucose result" refers to a person's glucose level as measured by a glucose sensor; "actual glucose level" or "true glucose measurement result" refers to a person's actual glucose level.

Referring to Figure 1, an exemplary continuous glucose monitoring (CGM) system 10 is illustrated for monitoring glucose levels in a person with diabetes (PWD) 11. Specifically, the CGM system 10 is operable to collect measured glucose values at predetermined, adjustable intervals, such as every minute, five minutes, or at other suitable intervals. The CGM system 10 illustratively includes a glucose sensor 16 having a needle or probe 18 inserted under the skin 12 of the person. The tip of the needle 18 is positioned in an interstitial fluid 14 (such as blood or another bodily fluid) such that the measurement obtained by the glucose sensor 16 is based on the glucose level in the interstitial fluid 14. The glucose sensor 16 is positioned adjacent to the person's abdomen or at another suitable location. Furthermore, the glucose sensor 16 can be periodically calibrated to improve its accuracy. This periodic calibration can help correct for sensor drift due to sensor degradation and changes in the physiological condition of the sensor insertion site. The glucose sensor 16 may also include other components, including but not limited to a wireless transmitter 20 and an antenna 22. The glucose sensor 16 can alternatively use other suitable devices (such as, for example, non-invasive devices, such as infrared sensors) to obtain measurement results. During measurement, the glucose sensor 16 transmits the measured glucose value to a computing device 26, illustratively a blood glucose (bG) management device 26, via a communication link 24. The bG management device 26 can also be configured to store multiple glucose measurements received from the glucose sensor 16 over a period of time in a memory 39.

The CGM system 10 also includes a treatment delivery device 31, illustratively an insulin infusion pump 31, for delivering treatment (e.g., insulin) to a person. The insulin pump 31 communicates with a management device 26 via a communication link 35, and the management device 26 is capable of transmitting dosing and basal rate information to the insulin pump 31. The insulin pump 31 includes a catheter 33 with a needle inserted into the skin 12 of a person 11 for insulin injection. The insulin pump 31 is illustratively positioned adjacent to the person's abdomen or in another suitable location. Similar to the glucose sensor 16, the infusion pump 31 also includes a wireless transmitter and antenna for communicating with the management device 26. The insulin pump 31 is operable to deliver basal insulin (e.g., small doses of insulin released continuously or repeatedly at the basal rate) and dosing insulin (e.g., spurt doses of insulin, such as, for example, around meal events). Dosing insulin can be delivered in response to user input triggered by the user or in response to a command from the management device 26. Similarly, the basal rate of basal insulin is set based on user input or in response to a command from the management device 26. The infusion pump 31 may include a display for showing pump data and a user interface for providing user control. In an alternative embodiment, the insulin pump 31 and glucose sensor 16 may be provided as a single device worn by the patient, and at least a portion of the logic provided by a processor or microcontroller may reside on that single device. Supplemental insulin doses may also be injected in other ways, such as manually by the user via a needle.

In one embodiment, this CGM system 10 is referred to as an artificial pancreas system, which provides closed-loop or semi-closed-loop therapy to a patient to approximate or mimic the natural function of a healthy pancreas. In this system, an insulin dose is calculated based on CGM readings from a glucose sensor 16, and this insulin dose is automatically delivered to the patient based on the CGM readings. For example, if the CGM indicates that the user has high blood sugar levels or suffers from hyperglycemia, the system can calculate and automatically deliver the insulin dose required to lower the user's blood sugar level below a threshold level or to a target level. Alternatively, the system can automatically suggest changes to the treatment, such as increasing the basal insulin rate or boosting the delivery dose, but may require the user to accept the suggested changes before delivery. If CGM data indicates that a user has low blood glucose levels or suffers from hypoglycemia, the system may, for example, automatically lower the basal rate alone or in any desired combination or sequence, suggest that the user lower the basal rate, automatically deliver a certain amount of substance (such as, for example, a hormone (glucagon)) or suggest that the user initiate the delivery of a certain amount of substance (such as, for example, a hormone (glucagon)) to increase blood glucose concentration, suggest that the user consume carbohydrates and/or automatically take other actions and/or make other suggestions that may be appropriate to resolve the hypoglycemic condition. In some embodiments, multiple drugs may be used in such a system, such as a first drug that lowers blood glucose levels, such as insulin, and a second drug that raises blood glucose levels, such as glucagon.

Communication links 24 and 35 are illustratively wireless, such as radio frequency (“RF”) or other suitable wireless frequencies, wherein data and control are transmitted between sensor 16, treatment delivery device 31, and management device 26 via electromagnetic waves. Bluetooth® is an exemplary type of wireless RF communication system that uses a frequency of approximately 2.4 GHz. Another exemplary type of wireless communication scheme uses infrared light, such as systems supported by Infrared Data Association® (IrDA®). Other suitable types of wireless communication may be provided. Furthermore, each communication link 24 and 35 can facilitate communication between multiple devices, such as glucose sensor 16, computing device 26, insulin pump 31, and other suitable devices or systems. Alternatively, wired links, such as wired Ethernet links, may be provided between the devices of system 10. Other suitable public or proprietary wired or wireless links may be used.

Figure 2 illustrates an exemplary management device 26 of the CGM system 10 of Figure 2. The management device 26 includes at least one microprocessor or microcontroller 32 that executes software and/or firmware code stored in a memory 39 of the management device 26. The software/firmware code contains instructions that, when executed by the microcontroller 32 of the management device 26, cause the management device 26 to perform the functions described herein. The management device 26 may alternatively include one or more application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), hardwired logic, or combinations thereof. While the management device 26 is illustratively a glucose monitor 26, other suitable management devices 26 may be provided, such as, for example, desktop computers, laptop computers, computer servers, personal data assistants (“PDAs”), smartphones, cellular devices, tablet computers, infusion pumps, integrated devices including a glucose measurement engine and a PDA or mobile phone, etc. Although the management device 26 is illustrated as a single management device 26, multiple computing devices may be used together to perform the functions of the management device 26 described herein.

Memory 39 is any suitable computer-readable medium accessible by microcontroller 32. Memory 39 may be a single storage device or multiple storage devices, may be located internally or externally to management device 26, and may include both volatile and non-volatile media. Furthermore, memory 39 may include one or both of removable and non-removable media. Exemplary memory 39 includes random access memory (RAM), read-only memory (ROM), electrically erasable programmable ROM (EEPROM), flash memory, CD-ROM, digital versatile disc (DVD) or other optical disc storage devices, magnetic storage devices, or any other suitable medium configured to store data and accessible by management device 26.

The microcontroller 32 may also include additional programming to allow it to learn user preferences and/or user characteristics and/or user history data. This information can be used to implement changes in usage, recommendations based on detected trends (such as weight gain or loss). The microcontroller 32 may also include programming to allow the device 26 to generate reports (such as reports based on user history, compliance, trends, and/or other such data). Additionally, embodiments of the insulin infusion pump 31 of this disclosure may include a “power-off” or “pause” function for suspending one or more functions of the device 26, such as suspending the delivery protocol, and/or for powering off the device 26 or its delivery mechanism. In some embodiments, two or more microcontrollers 32 may be used for controller functions of the insulin infusion pump 31, including high-power and low-power controllers, for maintaining programming and pumping functions in low-power modes to conserve battery life.

The management device 26 also includes a communication device 41 operatively coupled to the microcontroller 32. The communication device 41 includes any suitable wireless and/or wired communication modules operable to transmit and receive data and control between the device 26 and the glucose sensor 16 and the insulin pump 31 via communication links 24 and 35. In one embodiment, the communication device 41 includes an antenna 30 (FIG. 1) for wirelessly receiving and/or transmitting data via communication links 24 and 35. The management device 26 stores measured glucose results and other data received from the glucose sensor 16 and/or the insulin pump 31 via the communication device 41 in a memory 39.

The management device 26 includes one or more user input devices 34 for receiving user input. The input devices 34 may include buttons, switches, mouse pointers, keyboards, touchscreens, or any other suitable input device. The display 28 is operatively coupled to the microcontroller 32 and may include any suitable display or monitor technology (e.g., liquid crystal display, etc.) configured to display information provided by the microcontroller 32 to the user. The microcontroller 32 is configured to transmit to the display 28 information relating to the detected glucose status of a person, the associated risks of the glucose status, and basal rate and dosage information. The glucose status may include an estimated glucose level and an estimated rate of change of the glucose level, as well as an estimate of the quality or uncertainty of the estimated glucose level. Furthermore, the displayed information may include warnings, alerts, etc., regarding whether the estimated or predicted glucose level of the person is hypoglycemic or hyperglycemic. For example, a warning may be issued if the person's glucose level drops below (or is predicted to drop below) a predetermined hypoglycemic threshold, such as 50 to 70 milligrams of glucose per deciliter of blood (mg/dL). The management device 26 can also be configured to transmit information or warnings to a person in a tactile manner, such as through vibration, for example.

In one embodiment, the management device 26 communicates with a remote computing device (not shown), such as at a caregiver's facility or a location accessible to caregivers, and data (e.g., glucose data or other physiological information) is transferred between them. In this embodiment, the management device 26 and the remote device are configured to transfer physiological information via a data connection (e.g., physical transfer via the Internet, cellular communication, or storage devices such as floppy disks, USB keys, compressed disks, or other portable storage devices).

The microcontroller 32 also includes range control logic 44. The range control system reduces the likelihood of hypoglycemic or hyperglycemic events by adjusting the insulin dose only when the glucose level of PWD11 is close to the low or high glucose threshold.

The microcontroller 32 includes hazard analysis logic 40, which calculates a target return path from multiple initial glucose states to a target glucose state based on cumulative hazard values. The target glucose state is illustratively defined as the optimal or ideal glucose state, which has no associated hazards or risks, such as a glucose level of 112.5 mg/dL and a glucose change rate of zero, but any suitable target glucose state can be identified. Each target return path includes multiple intermediate glucose states encountered during the transition from the initial glucose state to the target glucose state. Cumulative penalty values associated with the target return path are stored in memory 76, which can be used as a lookup table. The calculation of the cumulative penalty values is discussed below.

In some embodiments, inaccurate glucose measurements may be caused by malfunctions and/or noise associated with glucose sensor 24. Therefore, hazard analysis logic 40 also analyzes the probability of accurate detection of the glucose state provided by glucose sensor 24. Hazard analysis logic 40 can use any suitable probabilistic analysis tool to determine the probability of accurate glucose measurement, such as a hidden Markov model. Based on the determined probability of accuracy, hazard analysis logic 40 uses recursive filter 42 to estimate the person's glucose level and rate of change in glucose. Specifically, recursive filter 42 (such as, for example, a Kalman filter) weights the detected glucose states (including glucose level and rate of change) using the determined probability of glucose sensor accuracy. Based on the probability of glucose sensor accuracy, recursive filter 42 calculates an uncertainty measure of the estimated glucose state. The uncertainty measure indicates the quality of the estimated glucose state. For a range of detected glucose states, the uncertainty may differ for each state.

The microcontroller 32 in Figure 2 also includes a dosing calculator module 48, which calculates recommended dosing and the user's maximum permissible glucose level, which can be displayed to the user via display 28. The management device 26 maintains a record of the user's historical data accumulated over time up to the current time in memory 39. Historical data includes blood glucose history, prescription data, previous dosing recommendations, previous dosings, previous basal rates, glucose sensitivity factors for the user's sensitivity to insulin and carbohydrates, blood glucose responses to previous dosing and dietary events, other user health and medical data, and timestamps for each event and data record. Historical data includes information entered via user input 34 from patient records, such as dietary events, carbohydrate consumption, dosing delivery confirmations, medications, exercise events, stress response periods, physiological events, manual insulin injections, and other health events. The dosing calculator module 48 uses historical data to more accurately and efficiently determine recommended insulin dosings and/or carbohydrate amounts.

The dosage calculator module 48 determines a user-specific recommended dosage, such as an insulin correction dosage or a dietary dosage, based on the current glucose status, historical data, and user input. The recommended dietary dosage (e.g., carbohydrate amount) may be in response to a detected or predicted hypoglycemic condition. The recommended insulin correction dosage may be in response to a detected glucose level exceeding the maximum permissible glucose level. The actual amount of carbohydrates consumed and the actual amount of insulin administered can be confirmed by the user as information entered via user input 34 and recorded in memory 39 along with other historical data. The recommended dosage may be displayed on display 28.

Referring to Figure 3, an exemplary CGM trace 100 is illustrated, where the x-axis represents time in minutes and the y-axis represents glucose in mg/dL. The CGM trace 100 includes a series of detected glucose levels measured over a period of time. In the illustrated embodiment, the CGM trace 100 represents filtered glucose levels, i.e., glucose levels estimated based on measured glucose levels weighted with accurate sensor probability. The most recently estimated glucose level 110 has an associated negative rate of change, indicated by arrow 112. The dosing calculator module 48 determines a target glucose level 102 and a target range of glucose levels indicated by upper glucose limit 104 and lower glucose limit 106. For illustrative purposes, the target glucose level 102 is 110 mg/dL, the upper glucose limit 104 is 140 mg/dL, and the lower glucose limit 106 is 80 mg/dL, but other suitable values may be provided. The dosing calculator module 48 may determine the target glucose level 102 and the limits 104, 106 based at least in part on user history data described herein. Management device 26 uses trend glucose data from CGM trace 100 to recommend corrective actions to move blood glucose toward target glucose level 102. Target glucose level 102 in Figure 3 corresponds to the maximum allowable glucose before time _t1 and after time _t2 (i.e., when there are no recent dietary or corrective increases). Between time _t1 and _t2 , the maximum allowable glucose is adjusted based on dietary event 114 or other appropriate events.

At time _t1 , a meal event 114 occurs when the user consumes a meal and inputs carbohydrate data indicating the amount of carbohydrates consumed with the meal into management device 26. In some cases, an insulin bolus is administered around meal event 114 to offset the anticipated increase in glucose levels caused by the meal. Dosage calculator module 48 determines the predicted glucose level rise and the duration of the glucose rise based on the consumed carbohydrates, the insulin correction bolus (if administered), and the user's historical data related to glucose fluctuations following the meal and insulin injection. Based on the predicted glucose rise, dosage calculator module 48 determines an allowable rise value 124, an offset time value 126, and an action time value 122. The allowable rise value 124 may be based on other events, such as, for example, glucagon injection, exercise, sleep, driving, or time of day.

The allowable rise 124 is the amount by which a patient's glucose level can increase relative to the target glucose level 102 as a result of carbohydrate intake and insulin dosing. In some embodiments, the allowable rise 124 is a combination of a corrected Δglucose value 130 resulting from insulin dosing and a dietary rise 132 resulting from a dietary event 114. The corrected Δglucose value 130 is the difference between the current glucose level and the target glucose level 102 at the time of insulin dosing, allowing time for the glucose level to follow the insulin's decline. As illustrated, the allowable rise 124 is constant for a first predetermined time period after diet and insulin administration (i.e., offset time 126) (see line 118), and then decreases linearly after offset time 126 (see ramp 120). The total time for which diet and insulin doses affect a patient's bG level is the time of action 122. Figure 3 illustrates a trapezoidal graph 116 of the allowable rise 124 that accounts for the effects of insulin dose and dietary events.

The maximum allowable glucose level is increased based on the allowable rise value 124 and follows plot 116 of Figure 3. Therefore, the dosing calculator module 48 expands the range of allowable glucose levels over the duration of action time 122 following the dietary event according to plot 116. The allowable rise value 124 illustratively has an initial height of 50 mg/dL, but the allowable rise value 124 may have other suitable heights based on dietary size, insulin levels, and typical user responses to dosing increases from historical data. In some embodiments, the dietary rise value 132 is fixed for dietary events exceeding a threshold amount of carbohydrates. As an example, depending on the user, dietary size, and insulin dosing, the offset time 126 is approximately 2 hours, and the action time 122 is approximately 3 to 5 hours.

Referring again to Figure 2, the management device 26 also includes basal rate adjustment logic 50, which is operable to calculate and adjust the basal rate based on the current glucose state and the risk associated with the current glucose state. The management device 26 transmits the adjustment to the basal rate in a control signal via communication link 35 to the insulin pump 31, and the insulin pump 31 adjusts the current insulin basal rate based on this adjustment. Alternatively, the adjusted basal rate can be displayed to the user, and the user can manually adjust the basal rate of the insulin pump 31. In one or more embodiments, the adjustment is a percentage reduction of the initial, unadjusted, or nominal basal rate based on the risk of hyperglycemia, or a percentage increase of the initial, unadjusted, or nominal basal rate based on the risk of hypoglycemia.

Basal rate adjustment logic 50 determines whether the basal rate needs to be adjusted. If the adjusted basal rate is appropriate, basal rate adjustment logic 50 calculates the adjusted basal rate, and management device 26 sends a control signal to insulin pump 31 to cause insulin pump 31 to deliver insulin according to the adjusted basal rate. Alternatively, management device 26 may display the adjusted basal rate to the user to prompt the user to manually adjust insulin pump 31. In some embodiments, manual control of insulin pump 31 by the user may take precedence over the implementation of the adjusted basal rate.

The basal rate multiplier adjustment is determined based on glucose measurement results. In one or more embodiments, the basal rate multiplier is changed at fixed intervals (e.g., 15 minutes). When calculating the new basal rate multiplier, glucose values and the rate of change in glucose are used to predict the glucose value at the midpoint of the next fixed interval. Figure 4 shows an example of a fixed interval with length d, so that at time _t1 , the glucose value and trend are used to predict the value at that time. The values are then used to calculate the basal rate multiplier that will be used between times _t1 and _t2 .

Determining the base rate multiplier to be achieved begins with estimating the current glucose state. The complete glucose state includes the glucose level, the rate of change of glucose, and a covariance matrix indicating the diffusion of the glucose level and the rate of change of glucose. These values are provided by recursive filter 42. If the sensor noise is approximately constant, the glucose state can be simplified to just glucose and the rate of change.

When determining adjustments to the basal rate, it is assumed that the CGM controller receives measurements every minute (or other periodic intervals), but communicates with the insulin pump at a lower frequency. Once the temporary basal rate (TBR) for a period is transmitted to the pump, the algorithm waits at least d minutes before another TBR command is sent. In at least one embodiment, d equals 15 minutes, such that the TBR is updated on a periodic 15-minute basis. In further embodiments, d equals, for example, 10 minutes, 5 minutes, 2 minutes, or 1 minute. It will be understood that d can be adjusted based on the individual needs of the PwD.

As previously described, the microcontroller 32 includes hazard analysis logic 40, which calculates a target return path from multiple initial glucose states to a target glucose state based on a cumulative hazard value. Figures 5 and 6 illustrate an exemplary hazard function 80 used to calculate a hazard value for a given glucose level, which is ultimately used to determine the cumulative hazard value. The hazard function 80 is defined by the following equation:

(2)

(3)

Where is the blood glucose value (mg/dl) shown on the x-axis, is the corresponding hazard value shown on the y-axis, is the hyperglycemia shift, is the hypoglycemia shift, _hMAX is the maximum hazard, _hMIN is the minimum hazard, _αhyper is the hyperglycemia control aggression, and α, β, and c are process variables. In the illustrated embodiment, the variables α, β, and c are defined as follows: α = 1.509, β = 5.381, and c = 1.084. is a glucose value above which no other incremental hazard higher than _hMAX is calculated, and similarly, a glucose value below which no other incremental hazard higher than _hMIN is calculated. Test cases are generated for hazard functions for the hyperglycemia range () and the hypoglycemia range (). The function determines whether _hMAX , _hMIN , or is implemented as the final hazard value for the tested blood glucose value.

In determining _hMAX and _hMIN , excessive positive or negative hazard values for extreme blood glucose levels are implemented and prevented, respectively. In one or more embodiments, it is set to 600 mg/dL, and _hMAX is associated with it. Similarly, in one or more embodiments, it is set to 10 mg/dL, and _hMIN is associated with it. Thus, if it exceeds or falls below, the hazard value associated with the blood glucose level is prevented from exceeding the range defined by _hMAX and _hMIN .

Patients with diabetes exhibit varying degrees of insulin sensitivity. Therefore, the parameter _αhyper serves to adjust the aggressiveness of the hyperglycemia hazard function to account for these differences in insulin sensitivity. Referring to Figure 5, the nominal hazard function 80 and the hazard function 82 with reduced _αhyper are shown.

Referring to Figure 5, the hazard function 80 (positive hazard value) in the hyperglycemia region is shifted to account for recent dietary or corrective bolus increases. The over-shifted hazard function 84 illustrates the shift of the hazard function after a previous dietary or corrective bolus increase.

Referring to Figure 6, the hazard function is shifted to account for factors such as recent exercise, glucagon availability, or overcorrection escalation. For safety, the hyperglycemia hazard region associated with increased insulin never shifts to the left. When glucagon is present, the hypoglycemia hazard region shifts to the left by 86 because glucagon accounts for a portion of the hypoglycemia risk. In this case, the hyperglycemia hazard does not shift because insulin administration should not be increased due to glucagon. For example, in the case of exercise, the hypoglycemia hazard increases, and the curve shifts to the right by 88. In this case, the entire hazard curve shifts.

The cumulative hazard value of the return path from the current glucose state to the target glucose state is calculated by summing the hazard values of glucose values along the path between the current glucose state and the target glucose state. This path is constrained by limiting the maximum permissible glucose acceleration. Furthermore, it is assumed that the target has a zero rate of change, because once the target glucose state is reached, it is expected to remain at the target glucose state and not oscillate above or below it.

The fastest path is the return path with the least risk between the glucose state and the target. This return path uses the maximum permissible glucose acceleration (both positive and negative) to return to the target glucose state. The closed-form solution generated for the return path consists of a time interval with one extreme of the permissible glucose acceleration and then the opposite extreme.

If you are using a positive hypoglycemia shift, then the hypoglycemia shift must be added to the target glucose to obtain the shifted glucose target. This is necessary for correctly shifting the risk of hypoglycemia, because the glucose target represents the blood glucose level at which the danger shifts from positive (hyperglycemia) to negative (hypoglycemia). Adjusting the target glucose to the shifted glucose target is defined by the following equation:

(4)

Where is the shifted glucose target, is the nominal glucose target, and is the hypoglycemic shift. The maximum function in Equation 4 prevents negative hypoglycemic shifts from being added to the target glucose, and instead uses zero hypoglycemic shifts, resulting in and being equal.

As an initial problem, the general form of the return path must be determined. The return path can have an initial positive glucose acceleration followed by a negative glucose acceleration, or it can have an initial negative glucose acceleration followed by a positive glucose acceleration. The general form of the return path can be determined by solving which of the equations 5 and 6 given below returns a real solution.

(5)

(6)

in

(7)

(8)

(9)

(10)

is the rate of change of glucose level, is the maximum positive glucose acceleration, is the maximum negative glucose acceleration, and is the shifted glucose target from Equation 4. If Equation 5 returns a real number, and and are both greater than or equal to zero, then the return path first utilizes positive acceleration and then negative acceleration. Conversely, if Equation 6 returns a real number, and and are both greater than or equal to zero, then the return path first utilizes negative acceleration and then positive acceleration.

Once the general form of the return path is determined, the cumulative hazard value of the return path can be calculated. When the return path initially utilizes positive acceleration, the cumulative hazard value is defined by the following equation:

(11)

Furthermore, when the return path first utilizes negative acceleration, the cumulative danger value is defined by the following equation:

(12).

It should be understood that return paths encountering more extreme glucose values will tend to have higher cumulative hazard values because the hazard value at each time point is higher, as shown in Figures 5 and 6. For example, at the same rate of glucose change, a blood glucose value of 225 mg/dL will have a higher hazard value than a blood glucose value of 120 mg/dL. Furthermore, paths that take longer to return to the target glucose state will tend to have higher hazard values. Due to the initial rate of glucose change or extreme glucose values, paths may require longer return times to the target glucose state. Referring to Figure 7, exemplary return paths are provided for a wide range of initial glucose values with an initial rate of change of zero. The time to the target glucose state in Figure 7 ranges from approximately 20 minutes to almost 180 minutes. This amplifies the differences in cumulative hazard values for each initial glucose state. Calculating cumulative hazard values allows for comparison of glucose states with different glucose values and rates of change. If the rate of glucose change is more extreme, glucose values closer to the target glucose value generally have a higher hazard value than glucose values further away.

The cumulative hazard value provides the risk for a specific return path from the current glucose state to the target glucose state. However, there are uncertainties in the CGM blood glucose measurement results from the glucose sensor 16. Therefore, the actual blood glucose measurement result may differ from the blood glucose determined by the glucose sensor 16, and the specifically calculated cumulative hazard value may be inaccurate regarding the actual return path. To account for the variability of the actual return path, a current risk measure is determined, which includes changes in the CGM blood glucose measurement result.

To calculate the current risk measure, the predicted glucose state at the midpoint of the CTR period is first determined. In various embodiments, the midpoint of the CTR period is the true midpoint (1/2 of the CTR period), 1/4 of the CTR period, 1/3 of the CTR period, 2/3 of the CTR period, or 3/4 of the CTR period. In embodiments, the CTR is typically updated every 15 minutes, resulting in the midpoint being 7.5 minutes within the 15-minute sampling interval. For short time ranges, linear prediction performs as well or better than more complex models, so linear prediction is used for simplicity. When determining the predicted blood glucose level at the midpoint of the 15-minute sampling interval, it is assumed that the rate of change of glucose level remains constant within the 7.5-minute window. Therefore, the predicted glucose level is defined by the following equation:

(13)

Where is the initial measured blood glucose level, is the initial rate of change of glucose level, and is the prediction time measured from the beginning of the CTR cycle. Therefore, the predicted glucose state is .

Subsequently, the glucose state distribution surrounding the predicted glucose state is determined. Similarly, the glucose state distribution surrounding the current glucose state can also be determined. Samples for the glucose state distribution are selected based on the standard deviation of the distribution in the direction of glucose. The generation of glucose state distribution samples is defined by the following equation:

Where is the distribution of glucose values, is the distribution of the rate of change of glucose, is the glucose value for the current risk measure, is the rate of change of glucose level for the current risk measure, is the standard deviation of , is the standard deviation of , k is the divisor of , and n is the divisor of . It will be understood that if a glucose state distribution is desired for the current glucose state or the predicted glucose state, respectively, then can represent the current glucose level or the predicted glucose level. Equations 14 and 15 provide sample distributions ranging within two standard deviations of and . In at least one embodiment, the sample value for is selected by dividing the range defined by the two standard deviations by 10, and the sample value for is selected by dividing the range defined by the two standard deviations by 8, such that k = 10 and n = 8, respectively. Other sampling ranges and frequencies, such as 3 standard deviations, can also be used.

The current risk metric is determined based on a weighted average of the cumulative hazard values of the return paths generated from each sampled glucose state. Specifically, the risk is calculated by determining the weighted average of the cumulative hazard values at each combination of points and weighting them using a multivariate exponential function. The current risk metric is defined by the following equation:

(16)

This includes the current risk measurement.

(17)

This represents the distribution of glucose values, specifically the distribution of the rate of change of glucose levels determined based on the distribution of glucose states surrounding the detected glucose state. It also represents the cumulative hazard value for the return path at each glucose state. Finally, it represents the glucose value for the current risk measure, and the rate of change of glucose levels for the current risk measure.

(18)

The standard deviation is _____, and the standard deviation is _____. The weighting of cumulative hazard values results in the sample closest to the measured glucose state receiving the maximum weight in the final current risk metric calculation.

Referring to Figures 8A and 8B, the determination of the current risk measure is visually displayed. Figure 8A illustrates the 99 glucose states generated in an 11×9 matrix covering the hypoglycemia risk surface when k = 10 and n = 8. The return paths for the nine highlighted samples from Figure 8A are also highlighted in Figure 8B. The weighted average of the cumulative hazard values of the return paths for the entire grouping of the 99 glucose states provides the current risk measure.

The final base multiplier for each CTR period is determined using the current risk metric. First, the current risk metric is converted to a base multiplier value between 0 and TBR _MAX . TBR _MAX is the maximum percentage for the provisional base rate (TBR). In at least one embodiment, TBR _MAX defaults to 250%. In further embodiments, TBR _MAX is below or above 250% and is adjusted to regulate control and determination for low-adverse individuals. The base multiplier value is defined by the following equation:

(19)

Where is the basal multiplier, r is the current risk measure, and is the reference risk measure. In one or more embodiments, the reference risk measure is the glucose state associated with complete basal shut-off. For example, complete basal shut-off can occur at 70 mg/dL, such that no basal insulin is provided when the blood glucose level is below 70 mg/dL. The basal multiplier can be provided as a continuous function as the current risk measure changes. However, before the adjusted basal rate is provided to the treatment delivery device 31, it is converted to the nearest TBR increment (TBR _inc ) to provide the incremental basal rate multiplier (BM _inc ). The incremental basal rate multiplier is defined by the following equation:

(20)

Referring to Figure 9, an exemplary continuous base multiplier and an incremental base rate multiplier with TBR _inc of 10% and the implemented floor function are illustrated.

In another embodiment, a basal multiplier (BM _bolus ) greater than a threshold is delivered as a single bolus. The threshold can be 100%, 110%, or 130%. In these cases, the additional insulin ( _ITBR ) to be delivered in the next period d minutes later is calculated using the expected basal rate (I _BasalRate ) for that period and the duration of that period (d). This additional insulin is then delivered as a single bolus, and the basal rate multiplier is set to the threshold (BM _bolus ).

(twenty one).

As previously mentioned, if PwD has undergone a recent dietary or corrective spurt, a shift is applied to the hyperglycemic side of the hazard function 80. This reduces the calculated risk of hyperglycemia because insulin present in the subcutaneous layer accounts for a portion of the hyperglycemic risk. Referring to Figures 10A and 10B, the shift in basal rate adjustment resulting from the initial shift applied to the hyperglycemic side of the hazard function 80 is illustrated. Figure 10A provides an exemplary basal rate adjustment distribution where a curve passing through the rate of change at approximately 115 mg/dL glucose and 0 mg/dL/min divides the basal rate into above and below 100%; the lower curve represents the basal rate below 100%, and the upper curve represents the basal rate above 100%. Similarly, Figure 10B provides an exemplary basal rate adjustment distribution with the addition of a hyperglycemic shift. A single curve passing through the rate of change at approximately 140 mg/dL glucose and 0 mg/dL/min divides the basal rate into above and below 100%.

For some PwDs, the maximum permissible TBR (TBR _MAX ) should be set to a value below 250% or the default setting for TBR _MAX . These individuals are characterized by a large glucose correction equivalent ( _Gbr ) with their basal rate. This is calculated by multiplying the hourly basal rate (BR) by insulin sensitivity (IS). For example, an individual with a nominal basal rate of 0.9 IU/hr and insulin sensitivity of 50 mg/dl/IU would have a glucose correction equivalent of 45 mg/dl. PwDs with Gbr above a threshold ( _GbrT ) can benefit from a reduced TBR _MAX . In one or more embodiments, _GbrT is set to 150 mg/dl. It will be understood that _GbrT can be set to a value higher or lower than 150 mg/dl as needed for a particular PwD case. The temporary basal rate limit (TBR _limit ) used to provide a reduced TBR _MAX is defined by the following equation:

(twenty two).

Similar to the incremental base rate multiplier, the temporary base rate limit can be incremented to the nearest TBR increment. _{The TBR limit} increments to the nearest TBR increment as defined by the following equation:

(twenty three).

Glucose-corrected equivalents were calculated for 30 simulated PwDs. Simulated subjects numbered 21 and 24 exhibited oscillating behavior when insulin sensitivity increased. In this case, for subject 24, the basal rate increased to 1.5-fold, causing hypoglycemia, and the CTR algorithm was activated to mitigate this effect. Simulations were repeated with different values ranging from 125% to 250% of the maximum permissible TBR. Lower maximum permissible TBR values had lower oscillation values, demonstrating the benefit of achieving the TBR _limit for PwDs with _GbrT above _GbrT .

For further and alternative descriptions of the methods used to determine base rate adjustments, see U.S. Patent Application Serial No. 14/229016, filed March 28, 2015, entitled "System and Method for Adjusting Therapy Based on Risk Associated with a Sugar State," the entire disclosure of which is incorporated herein by reference. For further descriptions of the calculation of the target return path and the calculation of the risk metric, see U.S. Patent Application Serial No. 13/645198, filed October 4, 2012, entitled "System and Method for Assessing Risk Associated with a Sugar State," the entire disclosure of which is incorporated herein by reference. For a further description of the probabilistic analysis tools, recursive filters, uncertainty calculations, and other probability and risk analysis functions of the computing device 66, see U.S. Patent Application Serial No. 12/693701, filed January 26, 2010, entitled “Methods and Systems for Processing Glucose Data Measured from a Person Having Diabetes”, and U.S. Patent Application Serial No. 12/818795, filed June 18, 2010, entitled “Insulin Optimization Systems and Testing Methods with Adjusted Exit Criterion Accounting for System Noise Associated with Biomarkers,” the entire disclosures of which are incorporated herein by reference. For a further description of the addition calculator module 88, see U.S. Patent Application Serial No. 13/593557, filed August 24, 2012, entitled “Handheld Diabetes Management Device with Bolus Calculator”, and U.S. Patent Application Serial No. 13/593575, filed August 24, 2012, entitled “Insulin Pump and Methods for Operating the Insulin Pump”, the entire disclosure of which is incorporated herein by reference.

It should now be understood that the methods and systems described herein can be used to estimate glucose levels in individuals with diabetes and to adjust glucose levels using range control algorithms. Furthermore, the methods and systems described herein can also be used to determine adjustments to the basal rate of insulin administered to PwD. The methods described herein can be stored on a computer-readable medium having computer-executable instructions for performing the methods. Such computer-readable media may include a compact disk, hard disk drive, thumb drive, random access memory, dynamic random access memory, flash memory, etc.

It should be noted that the term "configuration" in this document refers to the components of this disclosure in a particular manner. "Configuration" is a structural description that describes a component as exhibiting a particular property or functioning in a particular way, as opposed to a description of its intended use. More specifically, references to the manner in which a component is "configured" indicate the existing physical state of the component and are therefore considered explicit descriptions of the component's structural characteristics.

While specific embodiments and aspects of the invention have been illustrated and described herein, various other changes and modifications may be made without departing from the spirit and scope of the invention. Furthermore, although various inventive aspects have been described herein, these aspects are not necessarily used in combination. Therefore, it is intended that the appended claims cover all such changes and modifications within the scope of the invention.

Claims (20)

1. A method for determining basal insulin rate adjustment based on risk associated with glucose status in a person with diabetes, the method comprising: The signal representing at least one glucose measurement result is received by at least one computing device; The glucose state of the person is detected by the at least one computing device based on the signal, and the detected glucose state includes the glucose level of the person and the rate of change of the glucose level; The at least one computing device determines a current risk measure associated with a detected glucose state based on a target glucose state stored in memory accessible by the at least one computing device. The current risk measure indicates the risk of at least one of hypoglycemia and hyperglycemia in the person. Specifically, a return path is determined based on the transition from the current glucose state to the target glucose state, and the return path includes at least one intermediate glucose value associated with the return to the target glucose state. The cumulative hazard value for the return path is determined, comprising the sum of hazard values for the at least one glucose value along the return path, with each hazard value indicating a hazard associated with a corresponding intermediate glucose value. The current risk measure is calculated as follows: the weighted average of the cumulative hazard values of the return path generated at each point combination in the glucose state distribution G _{_S} around the detected glucose state and weighted by using a multivariate exponential function, where G_{_S} is the distribution of glucose values and g_S is the distribution of glucose change rate, and g_{_S} is the glucose level of the detected glucose state and g_S is the rate of change of glucose level of the detected glucose state. The at least one computing device identifies a reference glucose state and a reference risk measure associated with the reference glucose state; and The at least one computing device calculates an adjustment to the baseline rate of the treatment delivery device based on a current risk metric associated with the detected glucose state and a reference risk metric associated with a reference glucose level. 2. The method of claim 1, wherein the calculation includes a percentage reduction in mapping the current risk measure to the base rate based on the reference risk measure. 3. The method of claim 2, wherein the reference glucose state includes a glucose level corresponding to a hypoglycemic condition. 4. The method according to claim 1, further comprising: displaying graphical data to a user on a graphical user interface, the graphical data representing the calculated adjustment to the base rate. 5. The method of claim 1, further comprising: transmitting a control signal to instruct the treatment delivery device to adjust the basal rate based on a calculated adjustment. 6. The method of claim 5, wherein the therapeutic delivery device comprises an insulin pump for delivering insulin to a person with diabetes, and the therapeutic delivery device communicates with the at least one computing device to receive an adjustment of the calculated basal rate. 7. The method of claim 1, wherein the at least one computing device determines a danger value for each danger value of the at least one glucose value on the return path according to the following equation. h(g) _hyper =max(α _hyper ·α(log(max(g-Δg _hyper -max(Δg _hypo ,0),1)) ^c -β),0), h(g) _hypo =min(α(log(max(g-Δg _hypo ,1))c-β),0), and Where g is the glucose value for the current risk measure, _Δghyper is the hyperglycemia shift, _Δghyper is the hypoglycemia shift, _hMAX is the maximum risk, _gMAX is the glucose value above which no other incremental risk higher than _hMAX can be calculated, _hMIN is the minimum risk, _gMIN is the glucose value below which no other incremental risk higher than _hMIN can be calculated, _αhyper is the hyperglycemia control aggression, and α, β, and c are process variables. 8. The method of claim 7, further comprising: before determining the cumulative hazard value of the return path, having the at least one computing device identify the person's shifted glucose target according to the following equation to account for the risk of positively shifted hypoglycemia, and... Wherein is the shifted glucose target, g _{_t} is the nominal glucose target, and Δg_{_hypo} is the hypoglycemic shift. 9. The method of claim 8, wherein, if it is a real number, the cumulative danger value of the return path is determined by the at least one computing device according to the following equation: in It is the rate of change of glucose level for the current risk measure, is the maximum positive glucose acceleration, t is the time of the return path, and is the maximum negative glucose acceleration. 10. The method of claim 8, wherein, if it is a real number, the cumulative danger value of the return path is determined by the at least one computing device according to the following equation: in It is the rate of change of glucose level for the current risk measure, is the maximum positive glucose acceleration, t is the time of the return path, and is the maximum negative glucose acceleration. 11. The method of claim 1, wherein the glucose state distribution is determined by the at least one computing device according to the following equation: and Where G _{_S} is the distribution of glucose values, σ_g is the distribution of the rate of change of glucose, g is the glucose value for the current risk measure, σ_g is the standard deviation of g, σ_{_g} is the standard deviation of σ_g, k is the divisor of G_{_S}, and n is the divisor of σ_g. 12. The method of claim 11, wherein k = 10 and n = 8. 13. The method of claim 1, wherein the current risk metric is determined by the at least one computing device according to the following equation: Where r is the current risk measure, and _G is the distribution of glucose values, which is the distribution of the rate of change of glucose determined based on the distribution of glucose states around the detected glucose state. G is the cumulative hazard value of the return path for each glucose state, g is the glucose value for the current risk measure, and σg is the standard deviation of g, where _σg is the standard deviation of g. 14. The method of claim 2, wherein the base multiplier value is determined by the at least one computing device according to the following equation: Where BM(r) is the base multiplier, r is the current risk measure, and r _0% is the reference risk measure. 15. The method of claim 14, wherein r0 _% is a risk measure of glucose state associated with complete basal shutdown. 16. The method of claim 15, wherein the temporary basal rate for delivery to the treatment delivery device is determined by the at least one computing device according to the following equation: Where TBR _inc is the size of the temporary base rate multiplier adjustment increment, and TBR _MAX is the maximum temporary base rate multiplier. 17. The method of claim 16, wherein the at least one computing device determines a temporary basal rate multiplier limit taking into account the person's insulin sensitivity according to the following equation: Where TBR _limit is the temporary basal rate multiplier limit, _GbrT is the glucose-corrected equivalent threshold, BR is the nominal basal rate, and IS is the insulin sensitivity of the person. 18. The method of claim 17, wherein the TBR _MAX is 250% and the _GbrT is 150 mg/dl. 19. The method of claim 7, wherein Δg _hyper and Δg _hypo are adjusted based on the detection of overcorrection increment. 20. A blood glucose management device configured to determine basal rate adjustment based on a risk associated with the glucose status of a person with diabetes, the device comprising: a non-transitory computer-readable medium storing executable instructions; and at least one processing device configured to execute the executable instructions such that, when executed by the at least one processing device, the executable instructions cause the at least one processing device to: Receive a signal representing at least one glucose measurement result; The glucose state of the person is detected based on the signal, and the detected glucose state includes the person's glucose level and the rate of change of the glucose level; A current risk metric is determined based on a target glucose state and associated with the detected glucose state, the target glucose state being stored in memory accessible by the at least one computing device. This current risk metric indicates the risk of at least one of hypoglycemia and hyperglycemia in the individual. Specifically, a return path is determined based on the transition from the current glucose state to the target glucose state, and the return path includes at least one intermediate glucose value associated with the return to the target glucose state. The cumulative hazard value for the return path is determined, comprising the sum of hazard values for the at least one glucose value along the return path, with each hazard value indicating a hazard associated with a corresponding intermediate glucose value. The current risk metric is determined based on the following: a weighted average of the cumulative hazard values of the return paths generated at each point combination in the distribution of glucose states around the detected glucose state and weighted using a multivariate exponential function. Identify reference glucose status and reference risk measures associated with reference glucose status; and The adjustment of the baseline rate for the treatment delivery device is calculated based on the current risk metric associated with the detected glucose status and the reference risk metric associated with the reference glucose level.