Validation of Diagnosis Codes for Low Birth Weight and Small-for-Gestational Age in the Medicaid Analytic Extract Database

Background: The accuracy of low birth weight (LBW) and small for gestational age (SGA) in administrative healthcare records is crucial for perinatal studies but has few validity studies.

Methods: Using 1999-2010 MAX linked to birth certificates (BC), we identified mother-infant dyads (≥30 days enrollment after delivery, with valid gestational age (GA) and birth weight (BW)). LBW and SGA were identified based on ICD-9-CM codes. Infants with BW <10% of the U.S. reference were flagged as SGA. For LBW group diagnoses, we imputed birthweight using median, mean BW from BCs, and ICD code boundaries of infants in the same LBW group. We calculated the sensitivity, specificity, and positive/negative predictive values to assess performance.

Results: We identified 1,536,272 live births. All LBW groups had low SEs, high SPs, and NPVs, whereas PPVs varied. Among infants with SGA diagnoses based on GA/BW from the BC, SE of the SGA codes was 13.36%; SP 99.01%; PPV 67.37%. Combining imputation with LBW codes increased SE up to 22.09% (lower boundary) but decreased PPV to 41.53% (lower boundary).

Conclusions: ICD-9-CM codes from administrative healthcare records had low SE but high SP. Imputation based on GA and BW did not add much value to SGA identification.