Methods and systems to improve patient medication adherence

A system for estimating an adherence risk score for a medication taken by a patient, the system comprising:

• a network interface configured to:
• receive from a server, a request for an adherence risk score for a patient and medication combination, the adherence risk score representing a risk that the patient will not be adherent with taking the medication as prescribed during a time period, and
• provide the adherence risk score to the server;
• a data transformer configured to:
• obtain pharmacy records for the patient and medication combination,
• obtain updated medication adherence outcomes for a respective each patient of a plurality of patients representing whether they were adherent with their prescription for the medication, and
• obtain updated data representing updated pharmacy records for the medication for a plurality of patients;
• a machine learning model, wherein the machine learning model is a statistical machine learning model verified during training by computing a sum of squares metric to determine when performance of the machine learning model is no longer improving, the machine learning model configured to:
• process the pharmacy records to determine the adherence risk score, and
• process the updated data to determine updated adherence risk scores for a respective each patient and medication combination;
• a comparer configured to compare the updated medication adherence outcomes and the updated adherence risk scores for the respective each patient and medication combination for the plurality of patients;
• and
• the machine learning model further configured to:
• update the machine learning model based upon the comparison of the updated medication adherence outcomes and the respective updated adherence risk scores for the respective each patient and medication combination, and
• process the pharmacy records with the updated machine learning model upon receiving the request for the adherence risk score for the patient and medication combination.

A non-transitory, machine-readable storage medium comprising instructions that, when executed by one or more processors, cause a machine to:

• receive, via a network interface from a server, a request for an adherence risk score for a patient and medication combination, the adherence risk score representing a risk that a patient will not be adherent with taking a medication as prescribed during a time period;
• obtain pharmacy records for the patient and medication combination;
• process the pharmacy records with a machine learning model to determine the adherence risk score,
• wherein the machine learning model is a statistical machine learning model verified during training by computing a sum of squares metric to determine when performance of the machine learning model is no longer improving;
• provide, via the network interface, the adherence risk score to the server;
• obtain:
• updated medication adherence outcomes for a respective each patient of a plurality of patients representing whether they were adherent with their prescription for the medication, and
• updated data representing updated pharmacy records for the medication for a plurality of patients;
• process the updated data with the machine learning model to determine updated adherence risk scores for a respective each patient and medication combination;
• compare the updated medication adherence outcomes and the updated adherence risk scores for the respective each patient and medication combination for the plurality of patients;
• update the machine learning model based upon the comparison of the updated medication adherence outcomes and the respective updated adherence risk scores for the respective each patient and medication combination;
• and
• process the pharmacy records with the updated machine learning model upon receiving the request for the adherence risk score for the patient and medication combination.