E-Prescriptions Just as Error-Prone as Paper Scripts

July 1, 2011 — Government and the healthcare industry have placed big bets on digital technology, and electronic prescribing in particular, for the sake of patient safety, but a new study reports that the error rate with computer-generated prescriptions in physician offices roughly matches that for paper scripts: about 1 in 10.

However, results from the study, published online June 29 in the Journal of the American Medical Informatics Association, are not as damning as they may initially appear. Error rates varied widely depending on the type of e-prescribing software used, with some programs outperforming pen and paper. In addition, software improvements could eliminate more than 80% of the mistakes, most of them involving omitted information.

In 2010, an estimated 190,000 physicians were electronically prescribing, the technical term for transmitting scripts directly to a pharmacy computer, according to a pharmacy industry group called Surescripts. That number does not include physicians who create a prescription with computer software and then either fax it to the pharmacy or give patients a printout.

Since 2009, the federal government has been paying hundreds of millions of dollars in Medicare bonuses to physicians and other clinicians who electronically prescribe. The government operates an even pricier incentive program for electronic health records, and e-prescribing is one of the prerequisites for earning a 6-figure bonus.

The new study study examined nearly 3900 computer-generated prescriptions received by a pharmacy chain in 2008 in Florida, Massachusetts, and Arizona, regardless of whether they were faxed or electronically transmitted to pharmacies or were printed out. Of those prescriptions, 11.7% contained at least 1 error. Researchers did not ascertain whether errors were corrected by the pharmacy chain or whether they led to an actual adverse drug event. Lead author Karen Nanji, MD, MPH, writes that the 11.7% figure is "consistent with the literature on manual handwritten prescription error rates."

Roughly one third of the errors represented potential adverse drug events, none of them life-threatening.

Software Improvements Must Be Physician-Friendly

Omitted information such as drug dose, duration, and frequency accounted for almost 61% of the errors detected by the authors. The rest of the errors stemmed from unclear, conflicting, or clinically incorrect information.

Software improvements, Dr. Nanji and coauthors write, could eliminate the vast majority of these mistakes. E-prescribing programs can incorporate so-called forcing functions that would prevent physicians from completing a prescription unless they enter required information, including complete drug names and proper abbreviations. Likewise, decision-support tools can issue alerts about a wrong drug dose or frequency. However, the authors note, physicians may rebel against e-prescribing software if antierror safeguards make it too slow or annoying to use.

Some e-prescribing programs included in the study appeared to give users a technological edge. The error rate associated with one such program was only 5.1% compared with a whopping 37.5% for another. However, the study did not assess whether the root cause was system design or how well or poorly the systems were implemented in physician offices. Training physicians and staff on new software systems, the authors note, is often given short shrift.

The study was supported by the federal Agency for Healthcare Research and Quality and the Harvard Risk Management Foundation. The authors have disclosed no relevant financial relationships.

J Am Med Inf Assn. Published online June 29, 2011. Abstract