Predicting complications in hypertensive disorders of pregnancy: external validation of a prognostic model for adverse perinatal outcomes

BACKGROUND: Prediction models can be used as simple evidence-based tools to identify fetuses at risk of perinatal death. Payne et al developed a prognostic model for perinatal death in women with hypertensive disorders of pregnancy, a leading cause of maternal/fetal morbidity and mortality. OBJECTIVE: This study aimed to externally validate the predictive performance of this model in pregnant women with hypertensive disorders of pregnancy admitted between 26 and 34 weeks of gestation in Ghana. STUDY DESIGN: The perinatal model was applied in the SPOT (Severe Pre-eclampsia adverse Outcome Triage) study, a cohort of women with hypertensive disorders of pregnancy admitted between 26 and 34 weeks of gestation to referral facilities in Ghana. Predictive performance was assessed by calibration (calibration-in-the-large coefficient and calibration slope) and discrimination (based on the c-statistic). RESULTS: Of the 543 women included in the validation analysis, 87 (16%) experienced perinatal death from delivery until hospital discharge. Predictive performance of the model was poor. The calibration-in-the-large coefficient was 1.12 (95% confidence interval, 0.87–1.36, 0 for good calibration), calibration slope was 0.08 (95% confidence interval, −0.21 to 0.36, 1 for good calibration), and c-statistic was 0.52 (95% confidence interval, 0.44–0.59). CONCLUSION: This perinatal prediction model performed poorly in this cohort in Ghana. Possible reasons include differences in case mix, clinical management strategies, or data collection procedures between development and validation settings; suboptimal modeling strategies at development; or omission of important predictors. Given the burden of perinatal mortality and importance of risk stratification, new prediction model development and validation is recommended.