Validation of Insurance Billing Codes for Monitoring Antenatal Screening

Abstract

Background: Prevalence statistics for pregnancy complications identified through screening such as gestational diabetes usually assume universal screening. However, rates of screening completion in pregnancy are not available in many birth registries or hospital databases. We validated screening-test completion by comparing public insurance laboratory and radiology billing records with medical records at three hospitals in British Columbia, Canada. Methods: We abstracted a random sample of 140 delivery medical records (2014–2019), and successfully linked 127 to valid provincial insurance billings and maternal–newborn registry data. We compared billing records for gestational diabetes screening, any ultrasound before 14 weeks gestational age, and Group B streptococcus screening during each pregnancy to the gold standard of medical records by calculating sensitivity and specificity, positive predictive value, negative predictive value, and prevalence with 95% confidence intervals (CIs). Results: Gestational diabetes screening (screened vs. unscreened) in billing records had a high sensitivity (98% [95% CI = 93, 100]) and specificity (>99% [95% CI = 86, 100]). The use of specific glucose screening approaches (two-step vs. one-step) were also well characterized by billing data. Other tests showed high sensitivity (ultrasound 97% [95% CI = 92, 99]; Group B streptococcus 96% [95% CI = 89, 99]) but lower negative predictive values (ultrasound 64% [95% CI = 33, 99]; Group B streptococcus 70% [95% CI = 40, 89]). Lower negative predictive values were due to the high prevalence of these screening tests in our sample. Conclusions: Laboratory and radiology insurance billing codes accurately identified those who completed routine antenatal screening tests with relatively low false-positive rates.