Demographic differences and potential bias from automated occupation coding among mothers of babies born with or without cleft lip and/or cleft palate in the Texas Birth Defects Registry.

ABSTRACT

OBJECTIVE: To compare maternal demographics based on occupation coding status and evaluate potential bias by excluding manually coded occupations. METHODS: This case-control study assessed cases with clefts obtained from the Texas Birth Defects Registry. The NIOSH Industry and Occupation Computerized Coding System automatically coded occupations, with manual coding for unclassified cases. Maternal demographics were tabulated by occupation coding status (manual vs. automatic). Logistic regression examined associations between major occupation groups and clefts. RESULTS: Automatic coding covered over 90% of all mothers. Building, grounds cleaning, and maintenance occupations, and office and administrative support occupations were significantly associated with cleft lip with or without cleft palate, even after excluding manually coded occupations. CONCLUSION: We found consistent associations before and after excluding manually coded data for most comparisons, suggesting that machine learning can facilitate occupation-related birth defects research.