CPOE Cuts Medication Errors By 48%, Says Study

Ordering medications through a computerized provider order entry (CPOE) system decreases the likelihood of medication errors by 48%, according to a new paper in the Journal of the American Medical Informatics Association. Projected nationwide, this means that CPOE averted about 17.4 million errors in 2008, the year of the study. However, the authors note, "It is unclear whether this translates into reduced harm for patients."

The study included a meta-analysis of nine papers that compared the medical error rates in hospitals before and after their adoption of CPOE. In addition, the researchers used national data on CPOE prevalence, medication ordering and medication errors for 2008, or data that were deemed reflective of that year.

The researchers chose 2008 as the study period partly because there was sufficient data in the literature for that year, said Lauren Olsho, a co-author and an economist at Abt Associates in Cambridge, Mass., in an interview with InformationWeek Healthcare. Also, she noted, it predated the HITECH Act of 2009, which created the EHR incentive program. That led to a big increase in EHR and CPOE adoption, so 2008 was considered a good "baseline" year.

In 2008, 34% of U.S. acute-care hospitals had CPOE systems, but the extent to which clinicians used them varied markedly. Of the nearly 1.76 billion medications ordered in hospitals in 2008, about 26% were processed using CPOE.

[ EHRs and health management systems offer best ROI, research says. Read more at Health IT Proves Economic Mettle, Research Says. ]

At the rate of CPOE adoption and implementation in 2008, the study said, medication errors were reduced by 12.5% nationally, meaning there were 17.4 million fewer errors than there would have been without CPOE. If all hospitals adopted CPOE and if the implementation level remained around 60%, the researchers added, up to 51 million medication errors a year could be averted.

But there's a big asterisk next to these findings: The researchers originally analyzed 10 studies that were rigorously done and that used the same methods to measure medication errors before and after CPOE adoption. But in the peer review process, they dropped a large study that showed a significant increase in medication errors, largely because of mistakes associated with CPOE. (Another paper also showed a slight increase in the error rate, but it was not significant.) If that study's results had been factored into the meta-analysis, the national estimate of 12.5% in error reduction would have fallen to 9.2%.

This paper was excluded from the analysis because it used voluntary reporting of medication errors -- a method that has been shown to be unreliable.

On balance, Olsho noted, her team did establish that CPOE resulted in a large reduction in reported errors. These errors did not necessarily result in any harm to patients, she said. For example, a small mistake in dosage might not hurt a patient, or a pharmacist might catch a wrong medication on an order before it was filled. So while the literature shows an association between ordering errors and adverse drug events, she said, this study does not quantify that.

In addition, Olsho pointed out, the study doesn't show whether the benefits of ordering drugs through CPOE outweigh the risks. "You can't take the results of our study and compare them to the recent concerns about safety issues without assigning a subjective value to those safety issues."

A number of studies and reports, including one from the Institute of Medicine, have raised the alarm about safety issues associated with CPOE and other aspects of health IT. The Office of the National Coordinator for Health IT last December issued a proposal to address these concerns, and industry reaction has been positive. But the data on whether CPOE reduces the risk of patient harm is still insufficient to reach a solid conclusion.

As large healthcare providers test the limits, many smaller groups question the value. Also in the new, all-digital Big Data Analytics issue of InformationWeek Healthcare: Ask these six questions about natural language processing before you buy. (Free with registration.)