NEWS2 performs barely above chance as a readmission predictor on NHS virtual wards

A retrospective observational cohort of 2,533 admissions across NHS virtual ward services found NEWS2 achieved an area under the receiver operating characteristic curve (AUC) of 0.55 (95% CI 0.51–0.60) for predicting unplanned hospital readmission — barely above chance, with an acceptable threshold typically set at 0.70 or above (Scott et al., PLoS One, April 2026; PMID 42048369). The AUC was computed across the whole cohort, not within individual pathways; frailty and OPAT patients together constituted 37% of virtual ward activity, alongside respiratory (41%), heart failure, and general pathways. The observed readmission rate was low (7%), which the authors note warrants cautious interpretation of the predictive estimates.

NEWS2 was developed and validated for acute inpatient populations; it captures physiological changes that signal acute deterioration in patients already receiving hospital-level monitoring at regular intervals. The physiological profile of clinically significant deterioration in patients managed at home — where measurement frequency is lower, patient co-operation is variable, and the antecedents of deterioration differ — was not the population in which NEWS2 validation evidence was generated. Its widespread adoption in virtual care settings has preceded rather than followed that validation.

This finding is the third independent convergence on the same principle in this evidence cycle [VF-P4]. Ceftriaxone MSSA bloodstream mortality (PMID 42112603) and TROMBODOM VTE risk assessment in hospital-at-home populations (PMID 42002440, where Padua and IMPROVE-VTE scores classified 46–75% of hospital-at-home patients as high risk against an observed 90-day VTE incidence of 1.2%) identify the same mechanism across separate domains: tools designed and validated in acute hospital populations carry real clinical risk when applied in community settings without systematic recalibration.

Virtual ward services expanding frailty and OPAT pathways should treat NEWS2 thresholds as orientation information rather than decisional criteria pending population-specific validation, and should prioritise commissioning community-derived predictive validity studies — including pathway-level analyses for frailty and OPAT cohorts — as a governance priority.