In search of strategies to improve compliance, we assessed medication-
taking behavior among 19 ambulatory hypertensives, using both pill cou
nts and electronic monitoring. The regimen consisted of one pill each
day for less-than-or-equal-to 63 weeks with return visits at 1 to 4 we
ek intervals. The study population was 79% male, 68% white, and 16% bl
ack with mean (+/- SEM) age 58.4 +/- 2.5 years. Only 51% of the interv
isit intervals displayed greater-than-or-equal-to 80% of vial openings
within the desirable range (24 +/- 6 h). Pill counts detected only 2%
of suboptimal interdosing intervals identified by electronic monitori
ng. Early changes in compliance correlated well with later changes (r
= 0.83, P = .002). A few of the subjects exhibited a large deviation f
rom the prescription, uncorrectable with drugs having a long duration
of action. We conclude that (a) pill counts tend to overestimate patie
nts' compliance rates, (b) changes in medication-taking behavior early
in therapy may predict subsequent compliance rates; and (c) prolongin
g drug action may compensate for some imperfect medication-taking beha
vior.