Using Previous Medication Adherence to Predict Future Adherence.

Area of Science:

• Health Services Research
• Pharmacoepidemiology
• Public Health

Background:

• Medication nonadherence is a significant public health challenge.
• Identifying patients at risk for nonadherence can optimize interventions and study designs.

Purpose of the Study:

• To evaluate various measures of prior patient medication adherence.
• To assess their effectiveness in predicting future statin adherence.
• Utilized a large U.S. administrative claims database.

Main Methods:

• Cohort of new statin initiators identified.
• Previous adherence to chronic preventive medications measured (365-day baseline).
• Metrics included proportion of days covered (PDC) and fill patterns.
• Statin adherence measured one year post-initiation (high adherence defined as PDC ≥ 80%).
• Logistic regression and modified Poisson models used for prediction and association analysis.

Main Results:

• Models incorporating patient demographics, comorbidities, and medication use achieved a c-statistic of 0.665.
• Previous medication adherence measures, particularly maximum PDC, yielded c-statistics up to 0.666.
• Adding mean PDC to the combined model improved the c-statistic to 0.695.
• Patients with prior mean PDC < 25% were half as likely to be highly adherent to statins (RR=0.49).

Conclusions:

• Measures of previous medication adherence enhance the prediction of future statin adherence.
• These adherence metrics outperform traditional baseline clinical measures in claims-based studies.
• Prior adherence data is valuable for identifying patients likely to adhere to statin therapy.