The three-item patient-reported instrument for retrospective adherence in resource constrained settings: reliability, validity and potential utility

Introduction: Patient-report instruments remain the most feasible and sustainable approaches for monitoring medication adherence in clinical settings. However, questions of their reliability and validity persist. Studies suggest that the 3-item instrument for retrospective adherence (IRA) developed by Wilson et al. offers a viable option for clinically monitoring medication adherence. Here we report the reliability and validity of the IRA among patients recruited from community-based HIV clinics and antiretroviral therapy (ART) dispensaries in a resource constrained township in Cape Town, South Africa. Methods: Women (n = 794) and men (n = 228) receiving ART completed the IRA at three time points: (a) in a face-to-face administration at enrollment (b) in1-week phone interview and (c) 1-month phone interview. Participants also provided contemporaneous blood samples for HIV viral load testing as a clinical outcome and unannounced phone-based pill counts as an objective assessment of ART adherence. Results: The IRA was internally consistent and showed evidence of time stability. The IRA also demonstrated validity with respect to pill count adherence, correlates of adherence, and HIV viral load. Response operating curve analysis yielded an area under the curve of 0.646, using 75% adherence as the cut-off, with 0.637 sensitivity and 0.567 specificity. Conclusions: The IRA demonstrated reliability, construct validity and criterion validity in a resource constrained setting, supporting use of the IRA in research and clinical care.