Background:

Implementation of computerized provider order entry (CPOE) with drug–drug and drug–allergy checking is an important patient safety measure. The most commonly used CPOE systems, however, generate numerous warnings, which may lead to “alert fatigue.” Understanding the epidemiology of medication warnings is warranted when considering ways to increase provider attention to potentially valuable warnings.

Methods:

A retrospective epidemiologic study was conducted of all admissions to non–pediatric units at an academic medical center during the 7 month period from October 2009 to April 2010. The hospital has been using Meditech for CPOE since 2003, with a drug interaction database from First DataBank to which few customizations have been made. All warnings for each medication order are listed on a single screen. Data was collected regarding admission, medication order, patient, provider and warning characteristics. The study was approved by the medical center’s institutional review board.

Results:

During the 7 month study period there were 11,817 admissions to non–pediatric units, for which 596,684 medication orders were placed. 83,680 medication orders were placed that generated warnings (14% of all medication orders) during 9900 of the admissions (84% of admissions). 46% of the admissions that had medication warnings were for men; the mean age was 59 (range 0 to 104 years). The Progressive Care Unit, a specialized unit with pulmonary and cardiac patients, had the highest rate of medication orders generating warnings per admission with 19, followed by the Cardiac Intensive Care Unit with 15, the Surgical Intensive Care Unit with 12, and the Medical Intensive Care Unit with 11. Labor and Delivery had the lowest rate of medication orders generating warnings per admission (1). 1090 distinct medications generated warnings; warnings generated by five medications (hydromorphone, metoprolol, furosemide, potassium chloride and aspirin) accounted for 22% of all warnings. 58% of orders generating warnings were entered by residents, 18% by physician assistants or advanced practice nurses, 17% by attendings, and 4% by nurses. A total of 86,831 warnings were generated, from 1 to 17 per medication order; the median number of warnings per medication was 1. 167 of the warnings were for potential adverse drug reactions, 5,438 were allergy warnings, 51,769 were duplicate warnings, and 29,456 were interaction warnings. 4,151 medication orders were erased in response to the warnings (5%).

Conclusions:

Medication warnings affected a majority of hospital admissions, particularly ICU admissions, but these warnings led to changes in ordering in only a small fraction of patients. Focusing attention on the most commonly–ordered medications in the units with the highest rates of medication warning generation may be prudent when considering steps to increase the utility of warnings.