Computer-decision support for predicting and managing non-adherence to treatment

1. A method for developing a predictive model configured for predicting the likelihood of non-adherence for an individual target subject, the method comprising:

receiving, by a data communication controller, information for a target subject, from a system that logs adherence to attendance of the target subject at periodic visits or online sessions or from third-party agencies;

generating, based on at least the information for the target subject, a timeseries of historical activity, wherein random noise comprising Gaussian noise is added to the timeseries;

analyzing, the timeseries by computing a statistical relationship between the timeseries and an outcome of a singular spectrum analysis (SSA) computation of attendance information from the target subject to produce an analysis result, wherein the SSA computation is performed with column and row projector centering;

calculating a possibility of non-adherence for the target subject based on the analysis result and comparison of the result to a decision threshold; and

communicating an alert to a case manager assigned to the target subject where the possibility of non-adherence exceeds the decision threshold.

2. The method of claim 1 , wherein the method further comprises:

comparing the possibility of non-adherence with a previously determined possibility of non-adherence for the target subject where the possibility of non-adherence does not exceed the decision threshold; and

communicating the alert to the case manager where a discrepancy between the possibility of non-adherence and the previously determined possibility of non-adherence indicates an increase in the possibility of non-adherence.

3. The method of claim 1 , wherein the method further comprises:

comparing the possibility of non-adherence with a previously determined possibility of non-adherence for the target subject where the possibility of non-adherence does not exceed the decision threshold; and

sending a notification to the case manager where no discrepancy exists between the possibility of non-adherence and the previously determined possibility of non-adherence.

4. The method of claim 1 , further comprising sending an alert to the case manager if the expected log status of adherent attendance or appearance or participation is not received at a scheduled interval.

5. The method of claim 1 , wherein the system logs adherence to attendance of the target subject from the third-party agencies, and the third-party agencies comprise at least one of a health services provider, a law enforcement agency, a school, a social services agency, a religious services organization, a club, or a society.

6. The method of claim 1 , wherein the system logs adherence to attendance of the target subject from the third-party agencies, and the information for the target subject from the third-party agencies comprises at least one of following: past health records, criminal records, employment status, school enrollment, housing status, participation status in health coaching, training or rehabilitation programs, and substance use.

7. The method of claim 1 , wherein the timeseries is comprised of binomial, multinomial, or ordinal values, and wherein at least one value denotes non-adherence, absconding, relapse, or recidivism.

8. The method of claim 1 , wherein the time dimension of the timeseries is one of (a) continuous actual date-time; (b) coarsened continuous date-time; (c) transformed continuous time; and (d) discrete serial occurrences indexed in sequence of occurrence.

9. One or more computer storage media storing computer-useable instruction that, when implemented on a computing device, cause the computing device to perform operations, the operations comprising:

receiving session log data, wherein the session log data comprises at least an indication of a target subject's performance of prescribed program events;

generating, based on at least the session log data, a timeseries of historical activity, wherein random noise comprising Gaussian noise is added to the timeseries;

analyzing the timeseries by computing a statistical relationship between the timeseries and an outcome of a singular spectrum analysis (SSA) computation to produce an analysis result wherein the SSA computation is performed with column- and row-projector centering;

generating a possibility value for the target subject based on the analysis result;

determining the possibility value satisfies a decision threshold; and

in response to the possibility value exceeding the decision threshold, sending an alert to a case manager assigned to the target subject, wherein the alert comprises at least an indication of a high possibility of relapse, recidivism, non-adherence, or absconding for the target subject.

10. The computer storage media of claim 9 , wherein analyzing the timeseries further comprises a Markov dynamic programming computation, a CUSUM computation, Bayesian quickest detection computation, a Lorden's test, a Page's test, a hierarchical divisive estimation computation, or a group-fused Lasso computation.

11. The computer storage media of claim 10 , wherein the input variables comprise the timeseries and at least one of a decision threshold (P); an offset; a noise amplitude; a window length (L); a base series length (B); and a neighborhood size.

12. The computer storage media of claim 11 , wherein the input variables comprise the window length L and L is a fraction of a length of the timeseries (Λ), or the input variables comprise the base series length B and B is a fraction of a length of the timeseries (Λ).

13. The computer storage media of claim 12 wherein the fraction of Λ is between 0.2*Λ and 0.5*Λ.

14. The computer storage media of claim 11 , wherein the input variables comprise the neighborhood size, and the neighborhood size is a fraction of L between 0.2*L and 0.5*L.

15. The computer storage media of claim 10 , wherein the SSA computation produces a heterogeneity matrix (Hankel matrix) and Hankel matrix factor vector.

16. The computer storage media of claim 10 , wherein the notification further comprises a recommendation for at least one of:

altering the target subject's prescribed program events;

reserving resources, for the target subject, in a care facility associated with treatment and/or management of the consequences of potential non-adherence;

ordering increased monitoring of the target subject; and

ordering increased testing of the individual; and/or ordering prescriptions for the patient.

17. A system for predicative intervention, the system comprising:

a target monitoring device for determining adherence information for a target subject, the adherence information associated with prescribed program events;

one or more processors;

computer readable media having computer-executable instructions stored thereon which, when executed by the one or more processors, cause the one or more processors to:

receive adherence information for the target subject;

generate, based on the adherence information for the target subject, a timeseries of historical performance of prescribed program events, wherein random noise comprising Gaussian noise is added to the timeseries;

generate, based on the timeseries, a possibility value by computing a statistical relationship between a set of input variables and an outcome of a singular spectrum analysis (SSA) computation, wherein the SSA computation is performed with column- and row-projector centering;

compare the possibility value with a decision threshold;

in response to the possibility value satisfying the decision threshold, send an alert to a case manager assigned to the target subject, wherein the alert comprises at least an indication of a high possibility of relapse, recidivism, non-adherence, or absconding for the target subject.

18. The system of claim 17 , wherein the timeseries is comprised of binomial, multinomial, or ordinal values, and wherein at least one value denotes non-adherence, absconding, relapse, or recidivism.