Model-based patient adherence classification and intervention
Systems and methods for using predictive modeling to improve patient adherence to prescription medication regimens are provided. Training data may be generated using historical prescription adherence data associated with a patient population. A prescription adherence machine learning model may be trained using the training data, and the trained model may be applied to current prescription adherence data associated with a patient to predict a likelihood that a proportion of days that the patient will be covered by a prescribed medication will be below a threshold value over a calendar year. A patient risk score may be generated for the patient based at least in part on the predicted likelihood that the proportion of days covered by the prescribed medication will be below the threshold value. Based on the patient's patient risk score, the patient may be automatically contacted for intervention.
1. A computer-implemented method of using predictive modeling to improve patient adherence to prescription medication regimens, comprising:
generating, by a processor, training data using historical prescription adherence data and parameters of historical interventions to improve patient adherence to prescription medication regimens associated with a plurality of historical patients, and a corresponding proportion of days that each respective historical patient of the plurality of historical patients is covered by their respective prescribed medication over a calendar year,
wherein the historical prescription adherence data includes data indicating one or more of: (i) a number of days between a patient's expected first prescription fill date of the calendar year and the patent's first prescription fill date of the calendar year, (ii) a month of a patient's first prescription fill date of the calendar year, (iii) a number of days' supply associated with a patient's first prescription fill date of the calendar year, and/or (iv) a highest number of consecutive days that a patient was not covered by their prescribed medication during a previous calendar year, and
wherein the parameters of historical interventions include one or more of: (i) methods of contact that were used to schedule any interventions with a patient, or (ii) types of interventions that were attempted to be scheduled for a patient, and also include one or more of: (a) an indication of which interventions were successfully scheduled for the patient, (b) an indication of which interventions that were actually attended by the patient, or (c) a number of days, after an intervention actually attended by the patient, that the patient was covered by their prescribed medication;
training, by the processor, a prescription adherence machine learning model, using the training data, to predict a likelihood that a proportion of days that a patient, of a plurality of patients, will be covered by a prescribed medication over the calendar year will be below a threshold value, and parameters of interventions that are likely to be effective for the patient;
applying, by the processor, the trained prescription adherence machine learning model to prescription adherence data associated with the patient, to assign the patient to a risk profile group;
prioritizing, by the processor, each of the plurality of patients based on risk profile groups of each of the plurality of patients; and
automatically attempting, by the processor, to contact each of the plurality of patients for intervention in an order based on the prioritizing.
2. The computer-implemented method of claim 1 , wherein prescription adherence data further includes data indicating one or more of:
(i) a proportion of days, over a previous calendar year, that a patient was covered by the prescribed medication, and/or
(ii) a number of days between a patient's actual first prescription fill date of the previous calendar year and an end of the previous calendar year.
3. The computer-implemented method of claim 1 , further comprising:
selecting, by the processor, a mode of communication for attempting to contact the patient based on the risk profile group of the patient.
4. The computer-implemented method of claim 1 , further comprising:
selecting, by the processor, a type of intervention based on the risk profile group of the patient.
5. The computer-implemented method of claim 1 , wherein prioritizing each of the plurality of patients is further based on a number of failed contact attempts associated with each of the plurality of patients.
6. A computer system for using predictive modeling to improve patient adherence to prescription medication regimens, comprising:
one or more processors; and
a non-transitory program memory communicatively coupled to the one or more processors and storing executable instructions that, when executed by the one or more processors, cause the computer system to:
generate training data using historical prescription adherence data and parameters of historical interventions to improve patient adherence to prescription medication regimens associated with a plurality of historical patients, and a corresponding proportion of days that each respective historical patient of the plurality of historical patients is covered by their respective prescribed medication over a calendar year,
wherein the historical prescription adherence data includes data indicating one or more of: (i) a number of days between a patient's expected first prescription fill date of the calendar year and the patent's first prescription fill date of the calendar year, (ii) a month of a patient's first prescription fill date of the calendar year, (iii) a number of days' supply associated with a patient's first prescription fill date of the calendar year, and/or (iv) a highest number of consecutive days that a patient was not covered by their prescribed medication during a previous calendar year, and
wherein the parameters of historical interventions include one or more of: (i) methods of contact that were used to schedule any interventions with a patient, or (ii) types of interventions that were attempted to be scheduled for a patient, and also include one or more of: (a) an indication of which interventions were successfully scheduled for the patient, (b) an indication of which interventions that were actually attended by the patient, or (c) a number of days, after an intervention actually attended by the patient, that the patient was covered by their prescribed medication;
train a prescription adherence machine learning model, using the training data, to predict a likelihood that a proportion of days that a patient, of a plurality of patients, will be covered by a prescribed medication over the calendar year will be below a threshold value, and parameters of interventions that are likely to be effective for the patient;
apply the trained prescription adherence machine learning model to prescription adherence data associated with the patient, to assign the patient to a risk profile group;
prioritize each of the plurality of patients based on risk profile groups of each of the plurality of patients; and
automatically attempt to contact each of the plurality of patients for intervention in an order based on the prioritizing.
7. The computer system of claim 6 , wherein prescription adherence data includes data indicating one or more of:
(i) a proportion of days, over a previous calendar year, that a patient was covered by the prescribed medication, and/or
(ii) a number of days between a patient's actual first prescription fill date of the previous calendar year and an end of the previous calendar year.
8. The computer system of claim 6 , wherein the executable instructions further cause the computer system to:
select a mode of communication for attempting to contact the patient based on the risk profile group of the patient.
9. The computer system of claim 6 , wherein the executable instructions further cause the computer system to:
select a type of intervention based on the risk profile group of the patient.
10. The computer system of claim 6 , wherein the executable instructions further cause the computer system to prioritize each of the plurality of patients based on a number of failed contact attempts associated with each of the plurality of patients.
11. A tangible, non-transitory computer-readable medium storing executable instructions for using predictive modeling to improve patient adherence to prescription medication regimens that, when executed by at least one processor of a computer system, cause the computer system to:
generate training data using historical prescription adherence data and parameters of historical interventions to improve patient adherence to prescription medication regimens associated with a plurality of historical patients, and a corresponding proportion of days that each respective historical patient of the plurality of historical patients is covered by their respective prescribed medication over a calendar year,
wherein the historical prescription adherence data includes data indicating one or more of: (i) a number of days between a patient's expected first prescription fill date of the calendar year and the patent's first prescription fill date of the calendar year, (ii) a month of a patient's first prescription fill date of the calendar year, (iii) a number of days' supply associated with a patient's first prescription fill date of the calendar year, and/or (iv) a highest number of consecutive days that a patient was not covered by their prescribed medication during a previous calendar year, and
wherein the parameters of historical interventions include one or more of: (i) methods of contact that were used to schedule any interventions with a patient, or (ii) types of interventions that were attempted to be scheduled for a patient, and also include one or more of: (a) an indication of which interventions were successfully scheduled for the patient, (b) an indication of which interventions that were actually attended by the patient, or (c) a number of days, after an intervention actually attended by the patient, that the patient was covered by their prescribed medication; and
train a prescription adherence machine learning model using the training data, to predict a likelihood that a proportion of days that a patient, of a plurality of patients, will be covered by a prescribed medication over the calendar year will be below a threshold value, and parameters of interventions that are likely to be effective for the patient;
apply the trained prescription adherence machine learning model to prescription adherence data associated with the patient, to assign the patient to a risk profile group;
prioritize each of the plurality of patients based on risk profile groups of each of the plurality of patients; and
automatically attempt to contact each of the patients for intervention in an order based on the prioritizing.
12. The tangible, non-transitory computer-readable medium of claim 11 , wherein prescription adherence data includes data indicating one or more of:
(i) a proportion of days, over a previous calendar year, that a patient was covered by the prescribed medication, and/or
(ii) a number of days between a patient's actual first prescription fill date of the previous calendar year and an end of the previous calendar year.
13. The tangible, non-transitory computer-readable medium of claim 11 , further storing executable instructions that cause the computer system to:
select a mode of communication for attempting to contact the patient based on the risk profile group of the patient.
14. The tangible, non-transitory computer-readable medium of claim 11 , further storing executable instructions that cause the computer system to:
select a type of intervention based on the risk profile group of the patient.