If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
The obstetric comorbidity index (OCI), a weighted aggregation of maternal comorbidities defined by ICD-10 code, has been validated in predicting non-transfusion severe maternal morbidity (ntSMM). We sought to improve OCI using clinical features that are automatically extractable during labor and delivery.
Study Design
Retrospective analysis performed of all women delivering in a four-hospital system from January 2016 through January 2020. Patients were excluded if pre-delivery Hematocrit (Hct) was unknown. Demographic & clinical data were extracted using an automated query from the hospital electronic data warehouse. OCI score was calculated per patient using the weights reported in Leonard et al, 2020, denoting likelihood of ntSMM. Logistic regression was used to predict ntSMM using OCI score and combinations of clinical features. An ‘Augmented OCI’ score was also calculated in which clinical informatic data including pre-delivery Hct < 31.5, Gestation Weeks < 37, and BMI>40 were swapped in for OCI indicators: Anemia, Obesity, and Preterm. Model accuracy was assessed using AUC.
Results
Of 60,034 deliveries, 4,754 (7.9%) were excluded due to missing Hct records. Characteristics of the 55,280 remaining deliveries are shown in Table 1, segmented by whether patients had ntSMM. Two sets of pre-delivery clinical features were defined: L&D-OCI denoting features redundant with OCI indicators, and L&D-additional, denoting additional features not in OCI that might predict ntSMM. Logistic regression models were run using combinations of these feature sets and OCI or Augmented OCI (Table 2). Models including OCI plus extra features (AUC=0.88) and Augmented OCI plus extra features (AUC=0.88) outperformed OCI alone (AUC=0.85). An Augmented OCI model (AUC=0.85) performed similarly to OCI. A model including only automatically extractible features (AUC=0.8) also performed well.
Conclusion
We found that augmenting OCI using automatically extractable L&D features surpasses OCI alone in predicting ntSMM, and that L&D features alone are predictive of ntSMM, suggesting that a clinical decision support tool for real-time ntSMM risk prediction is attainable.