Presented at: (poster) 10th annual meeting of the National Birth Defects Prevention Network, February 4–7 2007, San Antonio TX; (oral) 40th annual meeting of the Society for Epidemiologic Research, June 19–22 2007, Boston MA.
Using registry data to suggest which birth defects may be more susceptible to artifactual clusters and trends†
Article first published online: 7 NOV 2007
Copyright © 2007 Wiley-Liss, Inc.
Birth Defects Research Part A: Clinical and Molecular Teratology
Special Issue: Special Issue: 2007 Congenital Malformations Surveillance Report, Part I
Volume 79, Issue 11, pages 798–805, November 2007
How to Cite
Langlois, P. H. and Scheuerle, A. (2007), Using registry data to suggest which birth defects may be more susceptible to artifactual clusters and trends. Birth Defects Research Part A: Clinical and Molecular Teratology, 79: 798–805. doi: 10.1002/bdra.20407
- Issue published online: 7 NOV 2007
- Article first published online: 7 NOV 2007
- Manuscript Accepted: 6 AUG 2007
- Manuscript Revised: 20 JUL 2007
- Manuscript Received: 2 APR 2007
- Centers for Disease Control and Prevention. Grant Number: U50/CCU613232
- diagnostic methods;
- observer variation;
- time trends;
- prevalence studies
Some birth defects appear to be more susceptible than others to artifactual prevalence variability over time or geographically. This article uses an empirical approach to try to identify them.
Assumption: Variation in clinical practice and other artifactual sources of variability impact observed variation in prevalence of mild cases more than severe cases for a given birth defect. Approach: Data were examined from Texas Birth Defects Registry deliveries from 1999–2003. For each of 312 delivery hospitals, birth prevalence for mild cases was calculated for birth defect X. The 5th percentile was subtracted from the 95th percentile to measure spread in the frequency distribution of all hospitals. That was repeated for severe cases. The ratio of the mild:severe spread was calculated for 49 defects, and the defects ranked into quintiles. That was repeated using birth prevalence based on county, and using isolated cases. The percentages of severe cases were calculated and also ranked into quintiles. A sensitivity analysis and simulation were conducted.
Forty-nine birth defects were ranked from those least susceptible to differences in mild:severe prevalence variability (e.g., anencephaly, hypoplastic left heart syndrome) to most susceptible (e.g., atrial septal defect, fetal alcohol syndrome). Resulting quintile ranks based on the three measures were highly correlated, whether based on all cases or isolated cases.
This empirical approach may be helpful for a number of public health applications. Birth defects and other health outcomes more susceptible to prevalence variability may be more likely to exhibit artifactual trends or clusters. Birth Defects Research (Part A) 79:798–805, 2007. © 2007 Wiley-Liss, Inc.