Method and apparatus for predicting, encouraging, and intervening to improve patient medication adherence

1. A system for predictively intervening with a user to improve medication adherence, the system comprising:

a display;

a video capture device;

one or more processors; and

one or more non-transitory, computer-readable storage media storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:

obtaining a set of training data associated with a group of patients;

splitting the group of patients into a plurality of sub-groups using unsupervised learning or cluster analysis;

training a corresponding machine learning model for each sub-group of the plurality of sub-groups based on training data of patients included in the sub-group;

outputting, to the display, one or more instructions related to proper medication administration;

obtaining one or more video sequences of the user administering medication in accordance with the one or more instructions, the one or more video sequences captured by the video capture device;

determining that the user in the one or more video sequences does not properly administer the medication;

based on determining that the user does not properly administer the medication, outputting, to the display, additional instructions to the user in near-real time;

assigning the user to a first sub-group of the plurality of sub-groups based on a similarity of the user to patients included in the first sub-group;

inputting data of the user, including a determination that the user did not properly administer the medication in response to the one or more instructions, into a first machine learning model corresponding to the first sub-group;

obtaining, as an output of the first machine learning model, for each future day of a plurality of future days, a corresponding prediction of whether the user will properly administer the medication on the future day, wherein, for at least one future day of the plurality of future days, the corresponding prediction indicates that the user will not properly administer the medication, and wherein, for at least one other future day of the plurality of future days, the corresponding prediction indicates that the user will properly administer the medication;

selecting a future time point that is prior to the at least one future day where the corresponding prediction indicates that the user will not properly administer the medication; and

at the future time point, outputting, to the display, additional information, wherein the additional information comprises an encouragement to improve medication adherence.

2. The system of claim 1 , wherein, for each future day of the plurality of future days, the prediction of whether the user will properly administer the medication on the future day comprises a binary determination of whether the user is predicted to properly administer the medication on the future day.

3. The system of claim 1 , wherein, for each future day of the plurality of future days, the prediction of whether the user will properly administer the medication on the future day comprises a predicted probability that the user properly administers the medication on the future day.

4. The system of claim 1 , wherein the output of the first machine learning model comprises a predicted lateness of the user in properly administering a dose of the medication.

5. The system of claim 1 , wherein the operations comprise:

determining a plurality of category scores for the user, each category score indicating behavior of the user in a respective category of behavior; and

determining a plurality of weighted combinations of the plurality of category scores, each weighted combination being based on a different weighting of the plurality of category scores,

wherein each weighted combination indicates past user medication adherence or future user medication adherence.

6. The system of claim 5 , wherein the operations comprise:

providing one or more weighted combinations of the plurality of weighted combinations to one or more different healthcare providers selected from the group of: nurse, doctor, family member, insurance company, hospital, and health system, wherein each weighted combination emphasizes a different aspect of user behavior.

7. The system of claim 5 , wherein the operations comprise:

determining a first weighted combination of the plurality of weighted combinations that best fits historical medication administration data of the user.

8. The system of claim 1 , wherein the operations comprise:

selecting, based on one or more of a disease state of the user, a population type of the first sub-group, socioeconomic data of the user, or historical data, the future time point as a time point at which a reminder to the user is predicted to be more effective compared to other possible future reminder times.

9. A method for intervening with a user to improve medication adherence in accordance with a medication adherence monitoring apparatus comprising

a display; and

a video capture device, the method comprising:

obtaining a set of training data associated with a group of patients;

splitting the group of patients into a plurality of sub-groups using unsupervised learning or cluster analysis;

training a corresponding machine learning model for each sub-group of the plurality of sub-groups based on training data of patients included in the sub-group;

providing, via the display, one or more instructions related to proper medication administration at a predetermined medication administration time;

obtaining one or more video sequences of the user administering medication in accordance with the one or more instructions, the one or more video sequences captured by the video capture device;

determining that the user in the one or more video sequences does not properly administer the medication in response to the one or more instructions;

based on determining that the user does not properly administer the medication, providing additional instructions to the user in near-real time via the display;

assigning the user to a first sub-group of the plurality of sub-groups based on a similarity of the user to patients included in the first sub-group;

inputting data of the user, including a determination that the user did not properly administer the medication in response to the one or more instructions, into a first machine learning model corresponding to the first sub-group;

obtaining, as an output of the first machine learning model, for each future day of a plurality of future days, a corresponding prediction of whether the user will properly administer the medication on the future day, wherein, for at least one future day of the plurality of future days, the corresponding prediction indicates that the user will not properly administer the medication, and wherein, for at least one other future day of the plurality of future days, the corresponding prediction indicates that the user will properly administer the medication;

selecting a future time point that is prior to the at least one future day where the corresponding prediction indicates that the user will not properly administer the medication; and

at the future time point, outputting, to the display, additional information, wherein the additional information comprises an encouragement to improve medication adherence.

10. The method of claim 9 , wherein, for each future day of the plurality of future days, the prediction of whether the user will properly administer the medication on the future day comprises a binary determination of whether the user is predicted to properly administer the medication on the future day.

11. The method of claim 9 , for each future day of the plurality of future days, the prediction of whether the user will properly administer the medication on the future day comprises a predicted probability that the user properly administers the medication on the future day.

12. The method of claim 9 , wherein the output of the first machine learning model comprises a predicted lateness of the user in properly administering a dose of the medication.

13. The method of claim 9 , comprising:

determining a plurality of category scores for the user, each category score indicating behavior of the user in a respective category of behavior; and

determining a plurality of weighted combinations of the plurality of category scores, each weighted combination being based on a different weighting of the plurality of category scores,

wherein each weighted combination indicates past user medication adherence or future user medication adherence.

14. The method of claim 13 , comprising:

providing one or more weighted combinations of the plurality of weighted combinations to one or more different healthcare providers selected from the group of: nurse, doctor, family member, insurance company, hospital, and health system, wherein each weighted combination emphasizes a different aspect of user behavior.

15. The method of claim 13 , comprising ranking the user relative to other users based on a first weighted combination of the plurality of weighted combinations.

16. The method of claim 15 , comprising employing a rank of the user to allocate resources for medication adherence intervention.

17. The method of claim 13 , comprising:

determining a first weighted combination of the plurality of weighted combinations that best fits historical medication administration data of the user.

18. One or more non-transitory, computer-readable media storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:

obtaining a set of training data associated with a group of patients;

splitting the group of patients into a plurality of sub-groups using unsupervised learning or cluster analysis;

training a corresponding machine learning model for each sub-group of the plurality of sub-groups based on training data of patients included in the sub-group;

outputting, to a display, one or more instructions related to proper medication administration;

obtaining one or more video sequences of a user administering medication in accordance with the one or more instructions;

determining that the user in the one or more video sequences does not properly administer the medication;

based on determining that the user does not properly administer the medication, outputting, to the display, additional instructions to the user in near-real time;

assigning the user to a first sub-group of the plurality of sub-groups based on a similarity of the user to patients included in the first sub-group;

inputting data of the user, including a determination that the user did not properly administer the medication in response to the one or more instructions, into a first machine learning model corresponding to the first sub-group;

obtaining, as an output of the first machine learning model, for each future day of a plurality of future days, a corresponding prediction of whether the user will properly administer the medication on the future day, wherein, for at least one future day of the plurality of future days, the corresponding prediction indicates that the user will not properly administer the medication, and wherein, for at least one other future day of the plurality of future days, the corresponding prediction indicates that the user will properly administer the medication;

selecting a future time point that is prior to the at least one future day where the corresponding prediction indicates that the user will not properly administer the medication; and

at the future time point, outputting, to the display, additional information, wherein the additional information comprises an encouragement to improve medication adherence.