Abstract
BACKGROUND: Maternal near-miss is a serious public health concern in impoverished countries such as Ethiopia. Despite its huge burden, the prognostic predictive model of maternal near-miss has received little attention in research in the Ethiopian context. As a result, this study aimed to build and validate (internally) a clinical prediction model of maternal near-miss in Bahir Dar City, Northwest Ethiopia, in 2024. METHODS: A prospective follow-up study was conducted among 2110 randomly selected pregnant women in Bahir Dar city between May 1, 2023, and March 6, 2024. Pregnant women with gestational age less than 20 weeks were included in the cohort and followed up to 42 days after delivery. Data were extracted from antenatal care records and collected by an interview-administered questionnaire. The model was developed using the standard Cox regression model, and model fitness was checked using the Schoenfeld assumption test. After applying a stepwise elimination, a p-value of less than 0.15 was used to fit the reduced model. Both discrimination and calibration were used to assess the model's performance. The model was internally validated through the bootstrapping method. The clinical usefulness of the model was checked using decision curve analysis. A nomogram was used for the model presentation. RESULTS: Maternal near-miss incidence density rate was 1.94 per 1,000 woman-weeks. Maternal age, residence, decision-making power, intention to pregnancy, time of antenatal initiation, genital mutilation, history of cesarean section, middle upper arm circumference, systolic blood pressure, hemoglobin, and history of obstetric morbidity were identified as important predictors to predict maternal near-miss. The model demonstrated good discriminatory performance with a C-index of 0.82(95%CI: 0.80-0.85), and good calibration with close alignment with 45 degrees. A simplified risk score of 40 maximum points was developed. The model was presented using a nomogram. CONCLUSION: The maternal near-miss incidence density rate was high in the present study. Socio-demographic and clinical factors were key variables for predicting maternal near-miss. The model has good discrimination and calibration. The researchers recommend external validation in different settings to assess the model's generalizability before applying it to clinical settings.