BACKGROUND: Birth defect registries using passive surveillance methods have suspect reliability in estimating prevalence. These systems are challenged by incomplete reporting and poor specificity of disease classification. Some passive systems “enhance” these methods through medical records review of selected conditions. Alaska has conducted medical records review for select conditions for over a decade. This information however, has never been utilized due to incomplete review. This study used this historical case confirmation information to create informative priors to estimate birth defect prevalence and establish a systematic ongoing process for sampling and review of reported conditions.
METHODS: We estimated the 5-year birth prevalence of 12 birth defects born during 2008:2012. This estimate uses a Bayesian approach to incorporate the confirmation probability [p(d|r) and the estimated missed cases probability by restricting the analysis to children diagnosed and seen by a medical provider before age 3 years [p(d|nr)] . From the known informative prior, and observed report probability, the estimated prevalence is calculated. We then applied this methodology to estimate the prevalence of microcephaly in Alaska by reviewing a sample of reported cases.
RESULTS: The probability of disease given a report ranged from a high of 86% to a low of 18%, indicating that for some conditions the reported ICD code is a poor representation of actual disease prevalence. This was especially apparent for conditions that shared ICD-9 codes (e.g. Gastroschisis and Omphalocele), and those that had extensive misclassification (e.g. Microcephaly). Based on reports the estimated prevalence of microcephaly in Alaska during 2007-2014 was 19 per 10,000 live births, after applying the informative priors the estimated prevalence is between 3 and 8 per 10,000 live births.
CONCLUSIONS: Birth defects contribute to a substantial amount of infant mortality and lifelong morbidity. Birth defects registries using passive surveillance methodologies have substantial challenges in estimating disease prevalence due to both reporting and coding issues. Enhanced surveillance using medical records review can be costly and challenging in large geographic states such as Alaska. This efficient methodology can account in part for reporting delays and correct coding issues to improve disease prevalence estimation. Due to the known limitations with the Alaska Birth Defects Registry, the estimates produced in the past have had little impact and often dismissed by providers and decision makers. These new estimates that have increased scientific rigor are much more well received and will likely lead to improved attention to these important issues.