Stephan F Lanes1, Cynthia de Luise. 1. Boehringer Ingelheim Pharmaceuticals Inc., Ridgefield, Connecticut 06877-0368, USA. slanes@rdg.boehringer-ingelheim.com
Abstract
BACKGROUND AND OBJECTIVE: Automated database studies have become a cornerstone of drug safety assessment. To assess the reliability of automated data, we compared the hospitalisation and mortality rates among three similar studies of automated healthcare databases in North America. METHODS: Similar protocols were used to identify patients diagnosed with chronic obstructive pulmonary disease (COPD) who were treated with inhaled bronchodilators or inhaled corticosteroids in the Saskatchewan Health Database (SHD), the Kaiser Permanente Medical Care Program (KPMCP) of Northern California, and a proprietary automated insurance claims database available from i3 (formerly Ingenix). Automated data were used to compute incidence rates of total hospitalisation, cardiovascular (CV) hospitalisation and hospitalisation due to several specific types of CV outcomes. Record linkage with registries of vital statistics was used to identify deaths, obtain death certificates, and compute rates of total mortality, CV mortality and deaths due to certain CV outcomes. We compared rates in the i3 population with rates in the other two populations using age-adjusted rate ratio estimates and 95% CIs. RESULTS: The i3 cohort had approximately one-half the rates of total mortality, CV mortality and total hospitalisations, but twice the rate of CV hospitalisations, compared with each of the other two database cohorts. DISCUSSION: The unexpectedly higher rates of CV hospitalisations in the i3 population are inconsistent with its lower CV mortality, total mortality and total hospitalisation rates. This discrepancy is not readily explained by a higher prevalence of CV disease or procedures, random variation or confounding. Instead, high CV hospitalisation rates in the i3 population are consistent with a high rate of false-positive diagnoses recorded on insurance billing claims. CONCLUSION: These results underscore the importance of ensuring valid endpoints in automated claims databases.
BACKGROUND AND OBJECTIVE: Automated database studies have become a cornerstone of drug safety assessment. To assess the reliability of automated data, we compared the hospitalisation and mortality rates among three similar studies of automated healthcare databases in North America. METHODS: Similar protocols were used to identify patients diagnosed with chronic obstructive pulmonary disease (COPD) who were treated with inhaled bronchodilators or inhaled corticosteroids in the Saskatchewan Health Database (SHD), the Kaiser Permanente Medical Care Program (KPMCP) of Northern California, and a proprietary automated insurance claims database available from i3 (formerly Ingenix). Automated data were used to compute incidence rates of total hospitalisation, cardiovascular (CV) hospitalisation and hospitalisation due to several specific types of CV outcomes. Record linkage with registries of vital statistics was used to identify deaths, obtain death certificates, and compute rates of total mortality, CV mortality and deaths due to certain CV outcomes. We compared rates in the i3 population with rates in the other two populations using age-adjusted rate ratio estimates and 95% CIs. RESULTS: The i3 cohort had approximately one-half the rates of total mortality, CV mortality and total hospitalisations, but twice the rate of CV hospitalisations, compared with each of the other two database cohorts. DISCUSSION: The unexpectedly higher rates of CV hospitalisations in the i3 population are inconsistent with its lower CV mortality, total mortality and total hospitalisation rates. This discrepancy is not readily explained by a higher prevalence of CV disease or procedures, random variation or confounding. Instead, high CV hospitalisation rates in the i3 population are consistent with a high rate of false-positive diagnoses recorded on insurance billing claims. CONCLUSION: These results underscore the importance of ensuring valid endpoints in automated claims databases.
Authors: Bruce E Sands; Mei-Sheng Duh; Clorinda Cali; Anuli Ajene; Rhonda L Bohn; David Miller; J Alexander Cole; Suzanne F Cook; Alexander M Walker Journal: Pharmacoepidemiol Drug Saf Date: 2006-01 Impact factor: 2.890
Authors: Stephen Sidney; Michael Sorel; Charles P Quesenberry; Cynthia DeLuise; Stephan Lanes; Mark D Eisner Journal: Chest Date: 2005-10 Impact factor: 9.410
Authors: Suellen M Curkendall; Cynthia DeLuise; Judith K Jones; Stephan Lanes; Mary Rose Stang; Earl Goehring; Dewei She Journal: Ann Epidemiol Date: 2005-07-21 Impact factor: 3.797