Comparative performance of obstetric comorbidity indices within categories of race and ethnicity: an external validation study.

BACKGROUND: Existing obstetric comorbidity adjustment indices were created without explicitly accounting for sociodemographic diversity in the development populations, which could lead to imprecise estimates if these indices are applied to populations different from the ones in which they were developed. The objective of this study was to validate two obstetric comorbidity indices (one using severe maternal morbidity [SMM] and one using end-organ injury or mortality) within categories of race/ethnicity. METHODS: Delivery hospitalizations from the State Inpatient Databases for Florida, Maryland, Kentucky, Washington (2015-2018) and New York (2015-2016) were analyzed. Outcomes were modeled using logistic regression by category of race/ethnicity and overall, with each model having its respective index value as the covariate. Discrimination and calibration were assessed. RESULTS: There were 1 604 203 delivery hospitalizations, among which 1.6% experienced SMM and 0.4% had SMM excluding blood transfusions. Maternal end-organ injury or mortality was identified in 0.5% of cases. For the entire patient population, the area under the receiver operating curve (AUROC) was 0.72 (95% CI 0.71 to 0.72) and 0.75 (95% CI 0.75 to 0.76) for SMM and non-transfusion SMM, respectively. The AUROC for maternal end-organ injury or death was 0.65 (95% CI 0.65 to 0.66). All scores exhibited poor calibration across racial/ethnic groups. There was no substantial variation within categories of race/ethnicity in terms of index performance. CONCLUSION: Users of these indices should consider performance data in totality when choosing a measure for obstetric comorbidity adjustment. There were no marked differences in model performance observed across race/ethnicity groups within each index.

Patient race and racial composition of delivery unit associated with disparities in severe maternal morbidity: a multistate analysis 2007-2014.

BACKGROUND: High Black-serving delivery units and high hospital safety-net burden have been associated with poorer patient outcomes. We examine these hospital-level factors and their association with severe maternal morbidity (SMM), independently and as effect modifiers of patient-level factors. METHODS: Using the 2007-2014 State Inpatient Databases (Florida, New York, California, Maryland, Kentucky), we analyzed delivery hospitalizations. We constructed generalized linear mixed models with patient- and hospital-level variables (Black-serving delivery units: high: top 5th percentile; medium: 5th-25th percentile; low: bottom 75th percentile; hospital safety-net burden status defined by insurance status) and report adjusted odds ratios (aOR) and 99% confidence intervals (CI). We repeated our mixed models with stratification and interaction analysis. RESULTS: 6 879 332 delivery hospitalizations were included in the analysis. Deliveries at high (aOR 1.83; 99% CI 1.34 to2.50) or medium (aOR 1.27; 99% CI 1.10 to 1.46) Black-serving delivery units were more likely to have SMM than deliveries at low Black-serving delivery units. Hospital safety-net burden was not significantly associated with SMM. In stratified models by hospital category, deliveries of Black women were associated with an increase in SMM compared with deliveries of White women in all hospital categories. In interaction models, Black women giving birth in high Black-serving delivery units had more than twice the odds of White women in low Black-serving delivery units of experiencing SMM (aOR 2.42; 99% CI 1.90 to 3.08). CONCLUSION: The patient racial/ethnic composition of the delivery unit is associated with adjusted-odds of SMM, both independently and interactively with individual patient race.