Population Improvement Bias Observed in Estimates of the Impact of Antenatal Steroids to Outcomes in Preterm Birth

Objective To examine the hypothesis that increasing rates and differential uptake of antenatal steroids would bias estimation of impact of antenatal steroids on neonatal death and severe (grade III-IV) intraventricular hemorrhage (IVH). Study design The study population included infants born between 24 and 28 weeks of gestational age in the California Perinatal Quality Care Collaborative. Outcomes were in-hospital mortality and severe IVH. Mixed multivariable logistic regression models estimated the effect of antenatal steroid exposure, one model accounting for individual risk factors as fixed effects, and a second model incorporating a predicted probability factor estimating overall risk status for each time period. Results The study cohort included 28 252 infants. Antenatal steroid exposure increased from 80.1% in 2005 to 90.3% in 2016, severe IVH decreased from 14.5% to 9.0%, and mortality decreased from 12.8% to 9.1%. When stratified by group, 3-year observed outcomes improved significantly in infants exposed to antenatal steroids (12.5%-8.6% for IVH, 11.5%-8.8% for death) but not in those not exposed (20.7%-19.1% and 16.6%-15.5%, respectively). Women not receiving antenatal steroids had greater risk profile (such as no prenatal care) and greater predicted probability for severe IVH and mortality. Both outcomes exhibited little change (P > .05) over time for the group without antenatal steroids. In contrast, in women receiving antenatal steroids, observed and adjusted rates for both outcomes decreased (P Conclusions As the population's proportion of antenatal steroid use increased, the observed positive effect of antenatal steroids also increased. This apparent increase may be designated as the “population improvement bias.”