Impact of a computer-generated alert system prompting review of antibiotic use in hospitals

The aim of this study was to measure the impact on antibiotic use of a computer-generated alert prompting post-prescription review and direct counselling in hospital wards. A computer-generated alert on new prescriptions of 15 antibiotics was reviewed weekly by an infectious disease physician for 41 weeks. During the first 6 months of the study, criteria selected for potential intervention were: (i) a planned duration of treatment of >= 10 days; (ii) discordance between the spectrum of the prescribed antibiotic and available microbiological results; or (iii) prescriptions of broad-spectrum beta-lactams, fluoroquinolones, glycopeptides or linezolid. During the following 5 months, the alert was restricted to any prescription of the 15 antibiotics in the 9 wards where overall antibiotic use had not decreased in the past year. We analysed 2385 prescriptions, 932 (39%) of which generated an alert for potential intervention. Among the latter, 482 (51.7%) prescriptions prompted direct counselling, mainly for shortening the planned duration of therapy (18.9%), withdrawing antibiotics (16.2%) or streamlining therapy (15.5%). The attending physicians' compliance with the recommendations was 80%. The overall median (interquartile range) days of therapy prescribed by the attending physicians was reduced from an initial duration of 8 (7-14) to 7 (6-11) days (P < 0.0001), resulting in 26.5% less antibiotic days prescribed. The time required for the intervention was 6 h per week. This computer-prompted post-prescription review led physicians to modify one half of the antibiotic courses initially prescribed and was well accepted by the majority, although they had not requested counselling.