(680) Comorbidity Score for the Prediction of Severe Maternal Morbidity in a High-Risk Population

Comorbidity scores predict severe maternal morbidity (SMM) and are used to identify and develop interventions for patients at risk of SMM. Most comorbidity scores are developed from claims databases and are representative of low-risk populations (risk of SMM of 1.2% and < 5% prevalence of obesity, diabetes, substance use and chronic hypertension [CHTN]). We sought to assess the predictive ability of a previously developed comorbidity score for SMM in a high-risk population.


Study Design: Retrospective cohort study of patients receiving prenatal care and delivering at a single US tertiary center between 2015-2018. The primary exposure was a comorbidity score developed by Bateman et al from a US national cohort of 854,823 pregnancies solely insured by Medicaid delivering in 2000-2007 (c-statistic 0.66, 95% CI 0.65-0.67). The primary outcome was a SMM composite as defined by the Center for Disease Control, using ICD-10 codes extracted from the electronic medical record. Using published weights for comorbid conditions, a summary (Bateman) comorbidity score was generated for each participant. Logistic regression models were then used to generate a receiver operating curve for the association of this comorbidity score with SMM.

Results: Of 16835 included patients, 1706 (10%) had a SMM event (699 [4%] excluding blood transfusion). Blood product transfusion (1508 [9%]) was the most common SMM component. Overall, obesity (40%), substance use (18%), preeclampsia (15%), and CHTN (8%) were the most prevalent comorbidities. Patients who developed SMM (versus those who did not) were more likely to have comorbidities such as preeclampsia, cardiac disease, and CHTN (Table). Among our cohort, the published comorbidity score was poorly predictive of SMM (AUC 0.55, 95% CI 0.53-0.57, Figure).

Conclusion:

In our high-risk population, the existing comorbidity score poorly predicts SMM. Developing individual population-based comorbidity scores that incorporate pregnancy and labor characteristics may improve SMM prediction in high-risk populations.