Electronic prescribing can improve medication safety
Clinical decision-support capabilities key
A University of Washington study has found that potential prescribing errors frequently occur but don't generally reach the patient or cause harm. The researchers say the most severe errors may be reduced by implementing an electronic prescribing system with clinical decision-support capabilities.
According to lead study author Emily Beth Devine, PharmD, MBA, BCPS, FASHP, in the University of Washington's Department of Pharmacy Pharmaceutical Outcomes Research and Policy Program, four Institute of Medicine reports have established patient safety as a national priority for research, identified patient safety as one of six U.S. health care system dimensions needing improvement, described how an information technology infrastructure is central to the needed improvement, and begun to outline a comprehensive plan for implementing and standardizing the infrastructure to support health care delivery and improve patient safety.
Electronic prescribing with clinical decision-support capabilities has emerged as a part of the infrastructure and as a viable option for improving patient safety by decreasing medication errors, Devine says.
The research, reported in the American Journal of Health-System Pharmacy, was conducted at The Everett Clinic, which is owned and managed by 250 physicians who provide primary and specialty care delivered in 60 clinics in 13 geographic locations in the state of Washington. The care delivery system includes two ambulatory care surgery centers, a cancer center, comprehensive laboratory and imaging services, three community pharmacies, and a hospitalist team that admits to the local community hospital. The clinic's physicians write more than 2.5 million prescriptions a year. Everett Clinic contracts with 18 separate health plans, each of which has its own drug formulary.
Devine tells Drug Formulary Review that the clinic was interested in building an electronic prescribing system to add to its electronic medical record in hopes that it would improve medication safety. They started brainstorming electronic prescribing systems in 2002 and implemented their project in July of 2003, she says.
The research study called for a retrospective review of prescriptions written in one internal medicine clinic, which was the first entity to receive the electronic prescribing system, within the larger health system. Prescriptions were written between March 1 and July 15, 2002, 12 months before implementation of the electronic prescribing system and well before the prescribers knew of the study.
Three data sources
The three data sources reviewed were the handwritten prescription, the electronic medical record, and the prescription as it was entered into the pharmacy computer system. The study scope was limited to evaluating prescriptions for medication errors and potential errors that could be attributed to having occurred during the prescribing process.
Reviewing each prescription as it was entered into the pharmacy computer system enabled assessment of errors and potential errors arising from discrepancies between each prescription as written and each prescription as entered into the pharmacy computer system. With that information, prescription legibility from the perspective of the pharmacist was assessed. Errors attributed to the prescribing process and errors attributed to the order-entry process were determined by reviewing notes made by the pharmacist on each prescription. (Pharmacists make clarifying notes on prescriptions after discussion with a prescriber.)
In the absence of notes, if a researcher identified a discrepancy between a prescription as written and the prescription as entered into the pharmacy system, the discrepancy was attributed to a prescribing error such as illegibility. The presence of notes indicated to the researcher that the pharmacist had clarified the prescription with the prescriber and thus any discrepancy between the prescription as written and as entered into the pharmacy system was attributed to order-entry error. Capturing order-entry errors was outside the study scope.
Three clinically trained pharmacists evaluated the data—one was a primary care specialist, the second was in geriatric pharmacotherapy, and the third was in general clinical practice. By using the electronic medical record as one of the data sources, the researchers said they were able to evaluate each prescription and error in the clinical context, taking into consideration patient comorbidities and concomitant medications.
Medication errors were found in 386 (27.4%) of the 1,411 prescriptions evaluated. Some 77 prescriptions contained more than one error, bringing the total to 463, or an error rate of 32.8%.
Evaluate nine severity levels
The study used a scale with nine levels of severity ranging from level A (circumstances or events that have the capacity to cause error) to level I (error occurred, reached patient, contributed to or resulted in death). By far, Devine says, the largest proportion of errors was attributed to those in severity level A. No errors were categorized as level B. Thirty-two errors (6.9%) reached the patient and had the potential to cause harm (levels C through I). Only one error caused patient harm (level E—"error occurred, reached patient, may have contributed to or resulted in temporary harm, required intervention") and no errors were categorized at levels more severe than level E.
The characteristic most frequently identified was missing information, but only three of those 122 errors reached the patient and none caused harm. Incorrect directions and illegible prescriptions also occurred with a frequency of greater than 10%, but caused no harm. Several administrative errors were documented, all involving the number of prescription refills authorized and all were level A.
Some 21% (97 of the 463) of errors were characterized as clinical errors. The most frequently characterized clinical error was a contraindication of a drug for a patient at least 65 years old. Of the clinical errors, the more severe errors were most often characterized as drug-disease interactions, with three at level D and one at level E, followed by a lack of appropriate laboratory monitoring (two level D).
The researchers reported that more than 90% of the errors identified were not actual errors but rather potential errors. While 6.9% of the errors reached patients, meaning they had the potential to cause harm, only 0.2% actually did cause harm, an error rate of two in every 1,000 prescriptions written.
Devine tells DFR the findings were consistent with other studies that have shown that relatively few errors actually lead to patient problems. While the researchers had suspected that the error characteristics most likely to occur would be illegibility and allergy and drug-drug interactions, illegibility actually ranked third in the list of non-clinical characteristics that caused an error, behind missing information and wrong directions. Errors involving the Beers criteria (used to determine appropriateness of prescriptions for those at least 65 years old), drug-disease interactions, and lack of appropriate laboratory monitoring occurred more frequently than did errors involving allergies and drug-drug interactions.
Of the errors that had the potential to cause harm, only one of the non-clinical characteristics required monitoring to prevent harm (level D)—a dose that was too high. The characteristics of errors that required monitoring to prevent harm (level D) and the characteristic of the one error that caused harm (level E) were clinical errors.
E-prescribing can help non-clinical errors
The researchers said they believe that it is the non-clinical errors that may be affected by a basic electronic prescribing system (illegibility, missing information, wrong dose), while those defined as clinical errors may be affected only when more sophisticated levels of clinical decision-support programming are added. Devine said the Everett Clinic has used the study results to prioritize development of clinical decision-support programming that will provide prescriber alerts for drug-disease interactions and guidelines for appropriate laboratory monitoring to prevent harm.
An electronic prescribing system with clinical decision support, she says, will provide patient-specific or drug-specific information to prescribers at the point of care. She says such systems are most useful when they actively present information to prescribers as they go through the prescription ordering process, rather than passively waiting to be asked for information.
[Editor's note: Contact Dr. Devine at (206) 221-5760 or e-mail firstname.lastname@example.org.]
- Devine EB, Wilson-Norton JL, Lawless NM, et al. Characterization of prescribing errors in an internal medicine clinic. Am J Health-Syst Pharm 2007; 64:1062-1070.