Background: Drug interactions have been shown to be preventable by computerized prescription entry and screening only in hospitals and not in community-based practice.
Methods: We retrospectively evaluated the effect of online prescription screening in community pharmacies and physician offices of one health maintenance organization, phased in during 3 consecutive 6-month periods in 1998 to 1999 (period I, system active only in 40% of pharmacies; period II, system active in 90% of pharmacies and 50% of physician offices; period III, system active in 95% of pharmacies and 90% of physician practices), on rates of prescriptions with-, patient exposure to-, and physician prescribing of-potential drug interactions.
Findings: Cumulative data included 775,186 patients given at least one prescription, by one or more of 5504 physicians, whose prescriptions were dispensed at 572 pharmacies. Dispensing of drug interaction prescriptions was reduced by 21.1% and by 67.5% in periods II and III compared with period I (odds ratio, 0.79; 95% confidence limit, 0.75-0.83 and odds ratio, 0.28; 95% confidence limit, 0.26-0.30, respectively). Patient exposure decreased only in those receiving 3 to 7 concurrent drugs (odds ratio, 0.80; 95% confidence limit, 0.71-0.90) with no reductions for patients who were given 2 drugs or 8 or more drugs. Only 19% to 25% of physicians wrote prescriptions for drugs that interact, but 85% of these repeated this pattern after being alerted. The proportion of prescriptions of drugs that interact that originated with a single prescriber, as opposed to 2 prescribers, decreased during the 3 periods from 0.81 to 0.74 and 0.69 (P <.001).
Interpretation: Computerized prescription entry and drug interaction screening in the community caused a 62.8% reduction in pharmacy-dispensed prescriptions with severe drug interactions and a 20% reduction in patient exposure to prescriptions with severe drug interactions; this reduction was negated by polypharmacy of 8 or more drugs. The effect of interaction alerts on physician prescribing patterns was limited.