Variations in patient compliance with common long-term drugs.

An indirect measure of adherence to drug regimens was developed from pharmacy records of prescription refills. This measure was validated by demonstrations of significant relationships between compliance with hydrochlorothiazide and propranolol and expected physiologic effects of these drugs (decreased blood pressures and decreased pulse rates, respectively). The measure was used to survey compliance with 20 common drugs prescribed to 419 outpatients for long-term administration on regular schedules. Eight of the study drugs showed very low levels of compliance, suggesting "take-as-needed" (PRN) use, in spite of medical record, pharmacy record, and label instructions for regular administration. For the whole survey population, 12 truly "non-PRN" drugs showed showed a considerable gradient in mean compliance rates, ranging from digoxin (76.6 per cent) to hydralazine (50.4 per cent). Within-patient differences in mean compliance with different drugs were shown (e.g., hydrochlorothiazide 61.2 per cent versus potassium chloride 41.2 per cent mean compliance, p < 0.001). Correlations between compliance rates with different drugs in a single regimen were sometimes strikingly low. We conclude that 1) chart and prescription directions for regular administration are not sufficient to identify "non-PRN" drugs in compliance studies; 2) when compliance is an outcome measure in health care trials, drug mix is an important confounding variable; 3)in outpatient clinical settings, it may be appropriate to be inherently more suspicious of noncompliance with some drugs than others; and 4) patients may comply at different rates with individual drug components of a single regimen.