(766) Differences in severe maternal morbidity and comorbidity indices in two North Carolina health systems

Our objective was to define variation in the pre-delivery comorbidities and occurrence of severe maternal morbidity (SMM) in two large hospital systems in North Carolina. 

Study Design:

We used data from the electronic medical records of two large hospital systems to identify birth hospitalizations from 2016 to 2020. We compared rates of SMM and non-transfusion SMM between patients delivered in our health system and to national rates generated from delivery hospitalizations in the 2018 National Inpatient Sample (NIS). We similarly compared the distributions of OB Comorbidity Index (OB-CMI; Leonard 2020 version) using weighting designed to predict SMM, and non-Transfusion SMM, across the health systems and births in the 2018 NIS sample. Results were compared using chi-squared and Kruskal-Wallis tests as appropriate.

Results:

In total, 82,271 patients had births across 16 hospitals in our systems. While rates of SMM were 1.8% for both systems (p=0.89), non-Transfusion SMM rates differed significantly from 0.90% to 1.73% (p < 0.001; Figure 1). Rates of SMM were comparable to 2018 national rates, however rates of non-transfusion SMM were higher in both systems. Similarly, the median OB-CMI score for predicting SMM differed from 7 [25^th percentile: 0, 75^th percentile: 20] to 15 [25^th percentile: 2, 75^th percentile: 31] between the two health systems (p < 0.001; Figure 2), with similar results when calculating OB-CMI scores using weights to predict non-transfusion SMM. Both health systems had higher median OB-CMI scores when compared to the national sample (Figure 2).

Conclusion:

Among two large health systems in North Carolina, both of which have significant referral volume for high levels of care, there was large variation in burden of comorbid conditions and non-transfusion SMM among birthing patients. These findings suggest the importance of tailoring SMM and non-transfusion SMM targets to individual systems patient populations and characteristics and that SMM prediction models developed to fit to each systems’ characteristics may have superior performance.