Background:

Computerized provider order entry (CPOE) systems can be configured to generate warnings when medications are ordered, which could promote safer patient care. Many CPOE systems generate numerous warnings, however, which may lead to alert–fatigue and overriding of potentially important warnings. Understanding the factors associated with accepting a warning may assist in tailoring these systems to better guide medication ordering.

Methods:

A retrospective study was conducted of all medication orders that generated a single warning in April 2010 on a medical unit at an academic medical center. 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 are listed on the same screen, in the same font, without distinction for severity. Medication orders that generated more than one warning were excluded. Descriptive data about patients, providers, warnings and medications was collected, including inclusion on the Institute for Safe Medication Practices (ISMP) list of High–Alert Medications. Unpaired t–tests and Chi–square tests were used to evaluate associations between acceptance of warnings and variables. Data was analyzed using Stata 12 (College Station, TX). The study was approved by the medical center’s institutional review board.

Results:

1240 medication orders for 287 patients were studied; 119 patients were men (41%); the median age was 67 (range 22–96 years). Patients had been hospitalized from 0 to 51 days at the time the order was entered (median 4 days). 43% of orders were placed by attendings, 1% by fellows, 29% by residents, 25% by nurse practitioners or physician assistants, and 1% by medical students. 64% were entered by physician or midlevel members of the hospitalist service. 900 orders (73%) were entered on a weekday. Similar percentages were entered in each of four 6–hour time periods per day. 31% were for parenteral medications; 30% were for medications on the ISMP list. 7% of warnings were for allergies, 42% for duplicates, and 51% for interactions. Only 5% of warnings were accepted. In univariate analyses, allergy warnings were accepted more often than duplicate or interaction warnings (14% vs. 7% vs. 3%, P < 0.0001). Warnings for parenteral medications were accepted more often than those for non–parenteral medications (8% vs. 4%, P = 0.002).

Conclusions:

Most medication warnings were not accepted. Providers were more attentive to warnings for parenteral medications and to allergy warnings. Developers of CPOE systems should do more to distinguish between warning types and adjust warning appearance depending on the potential severity of consequences.