General Purpose Computer-Alert System in Intensive Care: Evaluation of Effect on Practitioner Performance

Abstract

We report the effect of an electronic alerting system that uses multiple data sources and requires minimal user interaction on doctors' behaviour in Intensive Care. Design: Observational study. Setting: University Hospital ICU. Participants: Medical staff. Intervention: Interview of medical staff, commenting on computerised expert system alerts generated without user prompting. Results: 289 alerts were generated from a list of 11 different alerts on 25 days during eight weeks. Overall, 29.06% of the alerts were considered useful. These alerts pertained to 41 of the 225 ICU patients involved (18.22%). They generated management changes in 84 of the 713 patient-days involved (11.78%). Conclusions: For minimal user-effort, this system generated one or more management changes in around one-fifth of patients at a ratio of one management change for every ten patient-days.