The validity of drug exposure measurement based on pharmacy records wa
s investigated taking into account completeness of data, drug complian
ce, and different methods of drug exposure measurement in pharmacy rec
ords. Data on prescription drug use were collected from home inventori
es and community pharmacies in a survey on drug use and compliance in
115 elderly people. To compare drug exposure in pharmacy records with
exposure in the home inventory, three different methods for exposure m
easurement in pharmacy records were used. Two employed a fixed time wi
ndow of 30 and 90 days, respectively, and the third method was based o
n the calculated duration of use of a prescription (''legend time'').
Drug exposure in the home inventory was taken as the gold standard and
sensitivity, specificity, and positive predictive Values of the diffe
rent methods were calculated for the most frequently used drugs and dr
ug categories. The specificity and positive predictive value of all th
ree methods was generally high (0.93-1.00 and 0.67-1.00, respectively)
. The 90-day fixed method and the legend time method generally showed
high sensitivity (range: 0.67-1.00 and 0.63-0.83, respectively) for dr
ugs that were used on a chronic basis, while the 30-day fixed method h
ad poor sensitivity (range: 0.29-0.69). Drugs that were used according
to the home inventory but not according to the pharmacy records metho
ds could be almost completely retrieved in the pharmacy records of a o
ne-year period showing that these records were virtually complete with
regard to prescription drugs. We conclude that computerized pharmacy
records can be a reliable source of the true drug exposure as estimate
d in a home inventory, when adequate attention is paid to the definiti
on of the exposure time-window and when these records are comprehensiv
e with regard to prescription drugs. (C) 1997 Elsevier Science Inc.