Maternal mortality prediction using WHO near miss criteria and Maternal severity models: Evidence from a Tertiary Care Study in India

Background Maternal mortality remains a critical public health issue, particularly in low- and middle-income countries (LMICS). Over the past two decades, there has been a significant global reduction in maternal mortality. As maternal deaths continue to decline, measuring maternal morbidity—including the near-miss ratio, mortality index, severe maternal outcome ratio, and maternal near-miss mortality ratio—is essential for assessing the quality of obstetric care. This study was conducted to validate the performance of the WHO MNM criteria and compare predictive models that integrate the Maternal Severity Score and Maternal Severity Index to predict maternal mortality in a tertiary care setting. Methods A prospective observational study was conducted at KAHER’s Dr. Prabhakar Kore Hospital and Medical Research Centre, Belagavi, India (1st February 2024–31st January 2025). Pregnant women fulfilling the WHO maternal near-miss criteria were enrolled, and diagnostic accuracy tests for overall WHO criteria and organ dysfunction severity markers were performed. Pearson’s correlation coefficient was used to determine the association between MSS and MSI. Two binary logistic regression models to predict the probability of maternal death were developed and compared via the area under the receiver operating curve (AUROC), with additional assessment via Nagelkerke R² and the Hosmer–Lemeshow goodness-of-fit test. Results Out of the 295 women identified with the WHO maternal near miss criteria, 191 fulfilled the criteria for maternal near misses, and 15 resulted in maternal death. The severe maternal outcome ratio (SMOR) was 51.7, the MNM ratio was 47.9 per 1,000 live births, and the mortality index was 7.27%. indicating that a majority of women with life-threatening conditions survived with better quality of care. The diagnostic accuracy of the WHO near-miss criteria showed good sensitivity (100%) and high specificity (93.11%) and improved significantly, up to 95.43%, when the organ dysfunction subset was used. The number of cases with severity markers per thousand deliveries ranged from 0.49–24.82. The correlation between the Maternal Severity Score and the Maternal Severity Index was strong (R = 0.805, p < 0.001). supporting the internal validity of severity assessment tools, the multivariate logistic regression model that included additional clinical parameters performed better, achieving an AUROC of 0.939, indicating excellent discriminatory ability for predicting maternal mortality and underscoring the clinical utility of the enhanced model.