The BPD estimator peaked at day 14—but discrimination is not calibration

A Saudi external validation found the NICHD BPD estimator's composite-outcome AUC was 0.762 on day 14 and 0.642 on day 28, with changing case sets across time. Moderate discrimination without reported calibration or impact analysis is not a bedside treatment threshold.

THE HOT TAKE

The useful signal is not that day 14 wins. It is that a familiar calculator changed performance with timing and case mix—and discrimination alone cannot justify a treatment trigger.

Abuzaid and colleagues externally evaluated the 2022 NICHD BPD Outcome Estimator in one level III NICU in eastern Saudi Arabia. The retrospective cohort covered February 2021 through January 2024 and targeted infants born at 23+0 to 28+6 weeks with stated birth-weight eligibility of 501–1250 g. Of 270 infants enrolled, 29 were excluded and 241 entered the analysis. The prespecified outcomes were grade 2–3 bronchopulmonary dysplasia and the composite of death or grade 2–3 BPD at 36 weeks' postmenstrual age, or discharge if earlier.