Systems, devices, and methods relating to medication dose guidance
Systems, devices and methods are provided for determining a medication dose for a patient or user. The dose determination can account for recent and/or historical analyte levels of the patient or user. The dose determination can also take into account other information about the patient or user, such as physiological information, dietary information, activity, and/or behavior. Many different dose determination embodiments are set forth, pertaining to a wide array of different aspects of the system or environment in which the embodiments can be implemented.
1 . A system for parameterizing a patient's medication dosing practice for configuring dose guidance settings for the treatment of diabetes, the system comprising:
an on-body unit configured to be worn on a skin surface of the patient, the on-body unit comprising:
an analyte sensor configured to be in contact with interstitial fluid of the patient and monitor analyte levels of the user; and
sensor electronics coupled to the analyte sensor and configured to wirelessly transmit analyte data of the patient;
a medication delivery device configured to administer medication doses and wirelessly transmit medication dose data of the patient;
a display configured to visually present information; and
one or more processors in communication with the on-body unit, the medication delivery device, and the display, the one or more processors coupled to a memory storing instructions that when executed by the one or more processors, cause the system to:
receive analyte data of the patient from the on-body unit and medication dose data of the patient from the medication delivery device over an analysis period;
detect a meal time using a first model based on the analyte data and the medication dose data, wherein the first model is a machine learning model;
classify medication doses received by the patient over the analysis period based on a feature matrix correlating a set of classification features and the meal time to each of the medication doses, wherein the set of classification features comprises a probability of a meal starting within a defined interval prior to a medication time of the medication dose;
group each of the classified medication doses in one of a set of mealtime groups;
generate dose parameters for the patient at least in part by applying the analyte data and the medication dose data for each of the groups of the set of mealtime groups to a second model, wherein the second model is configured to output the generated dose parameters, and wherein the dose parameters include a fixed dose medication amount;
store the dose parameters for configuring dose guidance settings;
adjust a fixed dose medication amount of the medication delivery device for each corresponding mealtime group based on the stored dose parameters to minimize a time out of a target analyte range of the patient; and
output, on the display, the adjusted fixed dose medication amount for each corresponding mealtime group.
2 . The system of claim 1 , wherein the memory holds further instructions for correlating the classification features from the group comprising: a medication time for each dose, a time-filtered analyte value, a rate of change of the analyte value closest to the time of medication, a left Area-Under-Curve (AUC) indicating an integrated difference between analyte values and the analyte value closest to the time of medication over an interval prior to the medication time, a right AUC indicating an integrated difference between analyte values and the analyte value closest to the time of medication over an interval after the medication time, time elapsed between medication times, a most probable interval of time elapsed since the most recent meal, probability of a meal starting within a defined interval after the medication time, and a most probable interval of time until the next meal.
3 . The system of claim 1 , wherein the memory holds further instructions to classify the medication doses at least in part by estimating a time for each meal eaten by the patient during the analysis period.
4 . The system of claim 3 , wherein the memory holds further instructions for estimating the time for each meal further at least in part by generating the feature matrix based on the analyte data and the medication dose data, wherein the feature matrix correlates a set of analyte data features to each of distinct regions classed as rising, fall-preceding, and falling.
5 . The system of claim 4 , wherein the memory holds further instructions for estimating the time for each meal at least in part by generating estimated mealtimes based on the feature matrix, using an algorithm.
6 . The system of claim 4 , wherein the memory holds further instructions for selecting the set of analyte data features from the group comprising: a maximal analyte rate of change, a maximal analyte acceleration, an analyte value at the maximal analyte acceleration point, a duration of the region, a height of the region, a maximal deceleration, an average rate of the change in the region, and a time of the maximal analyte acceleration.
7 . The system of claim 1 , wherein the memory holds further instructions to group each of the classified medication doses into the set of mealtime groups comprising breakfast, lunch, and dinner.
8 . The system of claim 7 , wherein the memory holds further instructions for the grouping at least in part by a clustering analysis.
9 . The system of claim 1 , wherein the memory holds the analyte data comprising an indicator of glucose level and the medication dose data comprising the medication doses comprising insulin.
10 . The system of claim 1 , wherein the memory holds further instructions for applying the analyte data and the medication dose data for each of the mealtime groups to the second model at least in part by fitting data pairs to the second model, wherein the data pairs each comprise a pre-meal glucose level and a corresponding meal dose amount.
11 . The system of claim 10 , wherein the memory holds further instructions for selecting the second model for fitting data pairs from a linear model with zero slope, a linear model with non-zero slope, a piecewise model with joins at a single point, or a non-linear model that approximates the piecewise model.
12 . The system of claim 10 , wherein the memory holds further instructions for fitting data pairs at least in part by minimizing a residual sum of squares.
13 . The system of claim 10 , wherein the memory holds further instructions for fitting data pairs at least in part by evaluating each second model with Akaike Information Criterion (AIC) and choosing a second model having a minimum AIC value.
14 . The system of claim 13 , wherein the dose parameters further comprise an analyte level and a correction factor from a chosen second model for each group.
15 . The system of claim 14 , wherein the memory holds further instructions for combining data from multiple mealtime groups to form a combined group, and choosing a best-fitting one of the second models and a correction factor for the combined group.
16 . The system of claim 13 , wherein the memory holds further instructions for comparing the AIC value of a chosen second model to a threshold, and requesting user input if the AIC value exceeds the threshold.
17 . The system of claim 1 , wherein the machine learning model comprises a pre-trained machine learning model based on one or more classification algorithms.
18 . The system of claim 1 , wherein the machine learning model is based on tree building rules and a feature threshold for each feature in each tree.
19 . The system of claim 1 , wherein the classified medication doses comprise a meal dose, a pre-meal correction dose, a post-meal correction dose, a basal dose, a missed meal dose, or an ambiguous dose.
20 . The system of claim 1 , wherein the memory holds further instructions for the classifying at least in part by extracting one or more features of the analyte data and the medication dose data.
21 . The system of claim 1 , wherein rows of the feature matrix correspond to each of the medication doses over the analysis period and each row comprises a feature vector for a corresponding medication dose.
22 . The system of claim 21 , wherein each feature vector is based at least in part on a corresponding segment of analyte data in a time range relative to the corresponding medication dose.