Computers and Adverse Drug Events
Nebeker JR, Hoffman JM, Weir CR, Bennett CL, Hurdle JF. High rates of adverse drug events in a highly computerized hospital. Arch Intern Med. 2005;165:1111-1116.
Adverse drug events account for a significant number of hospital admissions and the ensuing costs associated with these hospitalizations. Electronic endeavors, such as computerized physician order entry (CPOE), bar code systems, and electronic medical records attempt to reduce the preventable adverse drug events.
Nebeker, et al. attempted to assess the effects of the implementation of CPOE and other computerized medication systems on adverse drug events in a tertiary care Veterans Administration Medical Center. They used an observational study design whereby 937 out of 2,306 newly admitted patients from several hospital wards were randomly chosen and assigned to a pharmacist reviewer during a 20-week period.
They reviewed complete medical records of hospital stays every other day to document adverse drug events. Not only were traditional adverse drug events identified, but harm from overdoses and/or inappropriate dose reductions or discontinuations, as well as intolerable harm from dose titration, were documented as adverse drug events. The harms caused by the drugs were considered only if the drugs were started in the hospital.
Harms were classified based on prior literature and included standards for pharmacological typology, causality assessment, error type, event terminology, drug class, seriousness index, and medication error category indexing. Additional uncommon classifications were also used, including additional resource utilization. Consensus meetings were held weekly to confirm classification of adverse drug events. Of the admissions reviewed, 483 adverse drug events were identified of which 93% were drug reactions while 7% were due to over- or underdosing. Of the drug reactions, 90% were considered dose-dependent while 10% were considered to be idiosyncratic.
Two different indexing scales were used in classifying the harms. Using the LDS Hospital Scale, it was suggested that 91% of the adverse drug events caused moderate harm while 9% caused serious harm. Using the National Coordinating Council for Medication Error Reporting and Prevention indexing, it was suggested that 87% of the adverse drug events fell into category E (requiring treatment) and 4% into category F (requiring prolonged hospitalization). Twenty-seven percent of the total adverse drug events were thought to be due to errors, including execution and planning steps. Sixty-one percent of errors occurred with the ordering mechanism while 25% of the errors occurred in the monitoring process.
This study highlighted rates of adverse drug events five to 19 times higher than baseline. The authors explained this higher-than-expected rate in part by study elements, such as the use of clinical pharmacists as reviewers, iterative case reviews, and accessible electronic data that make adverse drug events more noticeable.
Weaknesses of this study included issues of comparability of CPOEs because there were significant feature differences among commercial software programs. In addition, this was an observational study lacking a control group. The authors felt that their study did not support the idea that the computerized patient record of the study institution had caused adverse drug events. Rather, the study supported the idea that the system increased the visibility of adverse drug events compared with a paper system. In addition, the authors recommended that the choice of CPOEs be carefully considered, with a focus on decision support features when integrated into a healthcare organization.