We compared the Zentus composite with reconstructed NEWS2, partial SOFA and lactate on the same patients in four cohorts. It scored higher in the three adult cohorts and not in the paediatric one.

The composite is one score per patient-hour for any deterioration event, read beside the scores a unit already uses.

AUROC for any deterioration event

Zentus compositeBedside score
Bedside scores reconstructed from charted values; not identical to formal NEWS2 or SOFA. Patient-hours: internal 1,176,991; NWICU 649,553; Zigong 16,834; PICDB 171,258.

The paediatric exception

In PICDB the composite does not lead: lactate matches it on AUROC, and lactate and partial SOFA both beat it on AUPRC. The composite was developed on adult cohorts and is for adults; children need recalibration and separate validation.

The internal threshold at other sites

A threshold chosen on internal validation data gave 80% sensitivity on the internal test set. The same threshold gave 57% on NWICU, 50% on PICDB and 34% on Zigong. A pilot therefore logs against thresholds fixed in advance, and any alerting threshold is set on the site's own silent log.

ScopeThat the composite changes clinical decisions or outcomes. It shows ranking performance on retrospective data only.