Computerized Physician Order Entry Systems: The Coming of Age for Outpatient Medicine

I n this month's issue of PLoS Medicine, Steele et al. address a timely question in clinical medicine today—how best to use the potential of computers to change and improve physicians' prescribing behavior [1]. In evaluating the effectiveness of automated computerized alerts in a primary-care setting, the authors examined a system that had been developed to reduce the chances of a physician prescribing a medication that would lead to any of the following fi ve different drug-induced adverse events: hypokalemia, hyperkalemia, nephrotoxicity, thrombocytopenia, or hepatotoxicity. Specifi c rules triggered alerts that were delivered via computer to the prescribing physician whenever there existed a potential for one or more of these fi ve adverse events. These alerts presented suggestions for the physician during the prescription-writing process; faced with an alert, a physician could elect to proceed, revise, or stop a medication order, or order additional laboratory tests. Steele and colleagues used a " before and after " design to assess the impact of the intervention on physicians' prescribing behavior over a nine-month study period, during which 19,076 patients were seen. In the " before " phase, which lasted 17 weeks, the rules were switched on in the background without displaying any message to the providers. Since the rules were processing in the background, the providers did not receive any alerts recommending changes to their orders. Physicians' baseline ordering behavior was then compared to their ordering behavior in the " after " phase, during which alerts were presented to the provider. Adverse drug events were also assessed before and after the intervention by doing a random sample of chart reviews. During the six months after introduction of the intervention, almost 50% of all the medications prescribed triggered an assessment by the system. Alerts were presented to providers over 1,000 times (for about 6% of all the completed prescriptions). These alerts indicated a possible " lab value– prescription mismatch, " for instance, a lab test that was abnormal or absent, and the physician needed to account for this. After the intervention, there was a signifi cant increase in the percentage of time that physicians ordered a rule-associated laboratory test (39% at baseline versus 51% post-intervention), particularly if the relevant laboratory value was not present in the data fi les. There was also a signifi cant increase in the percentage of cases in which the provider stopped the ordering process and did not complete …