Mortality risk ranking after medical emergency team review: External validation and redevelopment of predictive models

Objective This study aimed to validate the REMAIN in-hospital mortality model and redevelop a risk ranking model (COMFORTER) using point-of-care measures of acuity, dependency, and patient factors. Design Multicentre cohort study. Setting Four Australian metropolitan hospitals. Participants A total of 1,937 adults who remained on a general ward following MET review across four hospitals during 2024. Interventions REMAIN was externally validated using its original model. COMFORTER was developed and internally validated using routinely collected measures of acute illness severity, care dependency, and patient characteristics. Model performance was assessed by discrimination (ability to identify patients at highest risk of death) and calibration (agreement between predicted and observed mortality). COMFORTER was compared with REMAIN and the National Early Warning Score-2 (NEWS). Main outcome measures In-hospital mortality. Results In-hospital mortality was 20.7% (n = 400). REMAIN accurately identified patients at higher risk of death (area under the curve 0.80, 95% confidence interval 0.75–0.84); however, mortality estimates varied between hospitals. COMFORTER used fewer variables and improved risk ranking compared with REMAIN and NEWS (area under the curve 0.84, 95% confidence interval 0.76–0.89, versus 0.80 and 0.75, respectively). Absolute mortality risk estimates remained inconsistent across hospitals. Conclusions COMFORTER improved post-MET risk stratification compared with REMAIN and NEWS using fewer routinely available variables. The model consistently identified patients at higher risk despite variation in absolute risk estimates across hospitals. The model may support team communication and decision-making for prioritising post-MET care.