Predictors of medication adherence: fact or artifact.

Numerous studies have examined socio-demographic, psychosocial, and other factors as potential contributors to poor adherence. Variability exists in the strength and consistency of findings. We speculated that the method of measuring adherence might be a factor in the variability in identification of predictor variables. We examined the identification of predictors of adherence by method of measurement in two randomized, controlled trials of adherence interventions. Both studies used the Aardex Medication Event Monitor and the Morisky Self-Report Scale. Twenty-one days of baseline data from 698 subjects were examined in relation to measures of depression, functional status, perceived therapeutic efficacy, number of co-morbidities, and socio-demographic indices. Analysis included Spearman rho, Pearson r, and multiple logistic regression. Differences in the identification of predictors between adherence measurement methods were identified. These data support the hypothesis that different measurement methods yield different predictors of adherence.