External Validation of the fullPIERS Risk Prediction Model in a U.S. Cohort of Individuals with Preeclampsia.

Objective: To externally validate the fullPIERS risk prediction model in a cohort of pregnant individuals with preeclampsia in the United States.

Study design: This was a retrospective study of individuals with preeclampsia who delivered at 22 weeks or greater from January 1, 2010, to December 31, 2020. The primary outcome was a composite of maternal mortality or other serious complications of preeclampsia occurring within 48 hours of admission. We calculated the probability of the composite outcome using the fullPIERS prediction model based on data available within 12 hours of admission including, gestational age, chest pain or dyspnea, serum creatinine levels, platelet count, aspartate transaminase levels, and oxygen saturation. We assessed the model performance using the area under the curve (AUC) of the receiver operating characteristic curve. The optimal cutoff point was determined using Liu's method. A calibration plot was used to evaluate the model's goodness-of-fit.

Results: Among 1,510 individuals with preeclampsia, 82 (5.4%) experienced the composite outcome within 48 hours. The fullPIERS model achieved an AUC of 0.80 (95% confidence interval: 0.75-0.86). The predicted probability for individuals with the composite outcome (median: 18.8%; interquartile range: 2.9-59.1) was significantly higher than those without the outcome (median: 0.9%; interquartile range: 0.4-2.7). The optimal cutoff point of 5.5% yielded a sensitivity of 70.7% (95% CI: 59.6-80.3), a specificity of 85% (95% CI: 82.7-86.5), a positive likelihood ratio of 4.6 (95% CI: 3.8-5.5), and an odds ratio of 13.3 (95% CI: 8.1-21.8). The calibration plot indicated that the model underestimated risk when the predicted probability was below 1% and overestimated risk when the predicted probability exceeded 5%.

Conclusion: The fullPIERS model demonstrated good discrimination in this U.S. cohort of individuals with preeclampsia, suggesting it may be a useful tool for healthcare providers to identify individuals at risk for severe complications.