Drug utilization review (DUR) programs are conducted to improve the effectiveness of drug use and are instituted in a majority of managed care institutions. DUR programs rely on pharmacoepidemiologic studies to assess drug safety and efficacy. For diseases with long latency and/or low incidence, retrospective study designs are used to evaluate the drug- disease association. These types of studies require respondents to recall prior drug use. Except for drugs such as estrogens and oral contraceptives, little attention has been focused on the validity of patient histories of prior drug exposures. Yet, clinicians are well aware of patients' inability to remember the drugs they are currently taking, not to mention those they took previously. The level of actual recall accuracy has serious implications for case-control studies and for the effectiveness of DUR programs. We believe that factors such as age, chronicity of drug use and patient characteristics such as sex and current health status influence a person's ability to remember drugs s/he has taken in the past. To test these hypotheses, 480 persons enrolled in the Group Health Cooperative of Puget Sound Health Maintenance Organization for 12 or more years will be asked to complete a telephone interview of prior non- steroidal anti-inflammatory drug (NSAID) and estrogen exposures. The results will be compared to the computerized pharmacy database maintained by Group Health which lists virtually all prescription drugs dispensed to its members (a gold standard). Of particular interest is the extent to which chronicity of drug use and recall period (number of years between last use and questionnaire date) influence recall. Sensitivity, specificity, and kappa statistics will be used to study agreement. To study the determinants of recall, chi-square statistics and trend tests will be used on the ground data, and log-linear modeling will be used to assess all factors simultaneously. This study would evaluate exposure misclassification which impacts the validity of research findings used for the development of DUR programs.
