Systematic approaches for compliance problem detection and intervention are needed if the benefits of prescribed drug therapy in chronic disease management are to be optimized. As with all measures of compliance, computer algorithms based on refill patterns have advantages and disadvantages. They are unobtrusive and easily determined, but they measure the timeliness of prescription refills, not actual drug-taking. Computer-generated algorithms for assessing compliance based on refill patterns should be used by practitioners with caution, because they are not only markers for potential drug taking compliance problems, but also for discrepancies between the medical chart, pharmacy records and verbal advice given to the patient. Because patients may obtain refills before depleting their supply, compliance rates using this methodology are best determined across several refills. In particular, we urge caution in applying them over time periods of less than 60 days. Longer minimum time periods further decrease the likelihood of "false positives" but limit the number of patients for whom a compliance measure can be computed. For the health professional (eg, the pharmacist) responsible for monitoring drug-taking compliance of patients, the message seems clear: when reviewing computer-generated noncompliance "flags," the first task is to fully explore the possibility of discrepancies in drug records before initiating compliance-related interventions.