[Alerts system associated with computerized physician order entry: analysis and identification of improvement points]

To make a quantitative analysis of the alerts associated with a computerized physician order entry system and identify opportunities to improve the system.

A retrospective observational study in a general hospital with 750 beds, 500 of which have a computerized physician order entry system installed. The frequency per type and medication of 525,691 alerts produced for a year in the prescription of drug treatments to 15,466 patients was analysed, entering these on a database. The system includes three categories of alert relating to the drug, the characteristics of the patient and the hospital medicine policy. By means of a failure mode and effects analysis, opportunities for improving the system were identified and corrective measures were suggested.

It has been observed that from the total of 1,084 drugs, 20 of them produce 34% of alerts. The ten most frequently active ingredients involved are: potassium chloride, acenocumarol, imipenem, lorazepam, diazepam, mycophenolate, enoxaparin, tacrolimus, calcium carbonate and cyclosporine. The most frequent alerts generated during electronic prescription are associated with duplicated therapy (35.4%), renal failure (27.6%) and risk due to advanced age (17.2%), with these groups accounting for 80.2% of the total. The excess of alerts and information provided by the alerts were identified as priority improvement points.

The system produced excessive alerts which led to the risk of them being ignored and reducing the capacity to prevent adverse drug events. Modifications are required for the design of the alert system, which also needs to be continuously updated.