A new method for identifying antibiotic-treated infections using automatedpharmacy records

For research purposes, there are few alternatives to costly surveillance fo r ascertaining infections in community populations. We propose a new approa ch based on antibiotic prescription fills in automated pharmacy records of the Group Health Cooperative of Puget Sound, in Seattle, Washington, to ide ntify treated infections in postmenopausal women. After excluding probable antimicrobial prophylaxis and chronic antibiotic use, four intervals betwee n antibiotic fills (30, 45, 60, and 90 days) were tested for their ability to detect new infections. Concordance with outpatient medical record review s was evaluated in 150 women. The sensitivity of the automated pharmacy rec ords using the four cutpoints for detecting new infections ranged from 88 t o 80%, from 30 to 90 days, respectively. Of the 81 women with no infection in the chart reviews, 75% also had no infection using the pharmacy method. Good agreement was found between the two methods for counts of infections p er person over the 2-year follow-up, with the 60-day cutpoint showing the g reatest overall agreement with chart reviews (kappa = 0.55). The pharmacy m ethod presented here offers a useful new approach for infection ascertainme nt for epidemiologic research. (C) 2000 Elsevier Science Inc. All rights re served.