Validity of ART Adherence Self-Reports in Adults

Mandisarisa, John

Introduction:The absence of a gold standard to measure adherence results in studies employing adherence methods, alone or in combination. Objective measures used in research, although impractical for most clinical settings, remain sensitive than patient self-report for detecting non-adherence although they remain difficult to apply in clinical practice due to complexity and cost implications. Patient self reports, whilst simple and easy-to-apply, still lack adequate reliability and validity in some settings. The purpose of this research was to evaluate the applicability, reliability and validity of simple self-reported adherence in Zimbabwe.

Methods:A 12-month longitudinal cohort tracking of adherence measurement methods at three month intervals with comparison of levels of agreement (Kappa test) between two self-report methods and CD4 count change (CD4) was used. The pill-count method was used at 12 month of follow-up to validate both the self-reported and CD4 adherence. Survey data was used to explain possible causes of observed variability between methods.

Results: Moderate agreements between seven–day recall (7DR) and thirty–day recall (30DR) at three months (46.6%, p = 0.000); six months (46.2%, p = 0.001) and nine months (60.8%, p = 0.006). Excellent to moderate agreement between pill count (PC) and thirty – day recall at 12 months of follow up (agreement = 99.03%; p = 0.000); 7DR (64.57%; p = 0.452) and CD4 (52.87%; p = 0.186). Moderate to fair agreement between CD4 and 30DR (53.53%, p = 0.184) and 7DR (46.95%, p=0.905). Good agreement between CD4 and 30DR (64%, p = 0.000); and 7DR (37%, p = 0.030) at six months. Fair agreement between CD4 and 30DR (47%, p=) and 7DR (43%, p=) at nine months. Moderate agreement between CD4 and both 30DR (53%, p=0.17) and 7DR (54%, p=0.64) at twelve months.

Conclusions: CD4 count not a significant predictor of adherence in treatment experienced patients. Self reported methods useful in predicting adherence but need constant periodic validation with reliable methods like pill counts.