Electronic Methods in Assessing Adherence to Medical Regimens
Studies of adherence to medical regimen have permeated the medical literature for more than two decades. These studies have shown widely varying rates of adherence to pharmacological therapies, (Dolce, Crisp, Manzella, Richards, Hardin, & Bailey, 1991; Dunbar-Jacob, Dwyer, & Dunning, 1991; Kruse, 1991) that, taken together, suggest that approximately half of patients on prescribed medications fail to adhere sufficiently to achieve therapeutic benefit (Haynes, McKibbon, & Kanani, 1996). Yet the methods of measuring adherence have varied substantially across these studies, with differences in reported adherence arising from the measurement methods. For example, in the Aspirin Myocardial Infarction Study, a multicenter, randomized clinical trial evaluating the efficacy of aspirin on the prevention of myocardial infarction, adherence was assessed using three procedures: (a) platelet aggregation, (b) urinary salicy late levels, and (c) pill count. The proportion of subjects judged to be adherent varied across these measurement methods by 16 percentage points in the drug group and 19 percentage points in the placebo group (Mattson & Friedman, 1984). Pill counts classified the largest proportion of patients as adherent in both the drug and placebo groups (97%). Platelet
aggregation placed 94% of the drug group and 82% of the placebo group in the good adherer category, whereas urine salicy late levels classifled 83% of the drug group and 97% of the placebo group as good adherers. Combining measures further reduced the proportion of patients classified as good adherers: 81% of the drug group and 78% of the placebo group were judged adherent on all three measures. Thus, there is likely to be considerable classification error attributable to measurement method alone when one is examining adherence.