Using trends in electronic recordings of vital signs to identify patients stable for transfer from acute hospitals

Abstract

Patients who are stable might not be required to remain in hospital. We aimed to create objective criteria to indicate stability based on vital signs. An index based on NEWS (NBI) was compared to a Patient Stability Index (PSI) algorithm created by random forest analysis. Data from the VITAL II study was used to train the algorithm and data from the VITAL III study to validate it. Failure rate of the algorithms was set close to the rate of readmission to UK hospitals at 15%. After a training period of two days the NBI identified stability with acceptable failure rates only after a further 96 hours with a subsequent release of 2143 bed days compared to the PSI which identified stability after only 12 hours leading to potential earlier release of 2652 bed days. Vital sign-based algorithms might be able to predict safe transfer from hospital and inform management of flow.