Computerized physician order entry (CPOE) systems with decision support provide advice, such as recommended drug dosages, and warnings, for example, about patient drug allergies or drug-drug interactions, as physicians use the computer software to order patient medications. One of the major reasons for implementing CPOE systems is to prevent serious errors that can cause patient harm. However, simulation of CPOE as it is used by physicians at 62 U.S. hospitals detected only 53 percent of the medication orders that would have likely resulted in fatalities and, in individual hospitals, 10-82 percent of orders that would have caused serious injuries. In the hospitals studied, the overall scores varied by as much as 40 to 65 percent among hospitals using CPOE software from the same vendor. The six top-performing hospitals used software from six different vendors (one homegrown solution and five commercial products).
In a statistical model, choice of vendor was significant in predicting performance, but only accounted for 27 percent of the total variation in performance. Hospital teaching status also was a significant predictor of CPOE performance, but only accounted for 10 percent of the observed variation in hospital performance. The hospitals were much more likely to use basic decision support than advanced decision tools that require more configuration and customization. However, the top-performing hospitals implemented advanced clinical decision support, as well as basic tools, thereby demonstrating what is possible with current systems.
The simulation tool mimics what happens when a physician writes a medication order for a real patient but uses fictitious patients and orders created for the test. The test orders would cause serious harm if the medication actually reached the patient and represent the categories of adverse drug events (excessive dose, wrong route, contraindicated based on patient age, diagnosis, or renal function) research shows contribute the most adverse drug events. For each assessment, a participating physician enters about 50 test orders for 10-12 test patients into the local electronic health record via its CPOE and notes any guidance provided by the software's decision report module.
In the study, the researchers calculated an overall score (percentage of test orders identified) as well as scores for each category of adverse drug event for all hospitals in aggregate. The researchers then compared hospitals based on hospital size (bed number), teaching status, and whether the hospital was part of a multihospital health system. The assessment tool employed in the study was funded in part by the Agency for Healthcare Research and Quality (Contract No. 290-04-0016).
More details are in "Mixed results in the safety performance of computerized physician order entry," by Jane Metzger, B.A., Emily Welebob, R.N., M.S., David W. Bates, M.D., M.Sc., and others in the April 2010 Health Affairs 29(4), pp. 655-663.