Updating the International Early Warning Score with frailty and comparing to gestalt for prediction of 3-day critical illness and mortality in emergency department patients

The International Early Warning Score (IEWS) showed strong prediction of mortality in all ages, but its performance compared to clinical gestalt remains uncertain. Furthermore, it is unknown whether frailty improves prediction besides age. This study aimed to compare IEWS with gestalt, and assess whether updating with frailty improved prediction. This secondary analysis of a prospective study enrolled 774 adult ED patients transported by ambulance to Amsterdam University Medical Centre between March and October 2021. The primary outcome was the performance of IEWS (± frailty) and clinical gestalt of paramedics, nurses, and physicians in predicting critical illness and 28-day mortality. Critical illness included serious adverse events, ICU admission, or mortality within 72 h. Critical illness occurred in 14.1% (n = 109) and 28-day mortality was 7.1% (n = 55). Both gestalt and IEWS had low accuracy predicting mortality with substantial underestimation of risk in all patients and low clinical usefulness. Gestalt performed better than IEWS in terms of discrimination and calibration for critical illness, with AUROC for physicians' gestalt of 0.83 (0.80-0.86), of nurses' gestalt of 0.84 (0.81-0.87), and paramedics' gestalt of 0.78 (0.75-0.81) compared to AUROC for IEWS of 0.64 (0.60-0.69) and IEWS + frailty of 0.64 (0.60-0.69). However, gestalt was only clinically useful for patients in whom an acceptable risk threshold for critical illness was above 5%. In these high-risk patients who arrived by ambulance to an Academic Hospital, clinical gestalt performed better than IEWS plus or minus frailty but was only useful if a risk threshold above 5% was acceptable.