The reliability of passively collected AIDS surveillance data in Massachusetts

PURPOSE: To determine the reliability of demographic, clinical, and treatment information collected during routine AIDS surveillance. METHODS: Information from the medical records of a random sample of 212 AIDS cases reported to the Massachusetts Department of Public Health between November 1993 and November 1994 was compared with that from the original case reports. We assessed levels of agreement by calculating overall percent agreement and kappa statistics with 95% confidence limits. We used an intraclass correlation coefficient to compare the CD4+ lymphocyte count reported from the two sources. RESULTS: There was excellent agreement for gender (K = 0.97) and race (K = 0.87). Agreement was lower for transmission mode (K = 0.73), CD4+ cell count (ICC = 0.76) and category of AIDS case definition (K = 0.59). There was poor agreement for use of antiretrovirals (K = 0.23), use of prophylaxis for Pneumocystis carinii pneumonia (K = 0.12) and vital status (K = 0.22). The month and year of diagnosis agreed in only 55% of cases reviewed. CONCLUSIONS: Routine AIDS surveillance data was reliable for demographic variables, but less reliable for information about clinical events, laboratory findings, or treatment. Future AIDS surveillance efforts should improve the collection of these data by using sources where this information is more reliable. (C) 2003 Elsevier Science Inc. All rights reserved.