The utility of adding retrospective medication profiling to computerized provider order entry in an ambulatory care population

From Clinfowiki
Jump to: navigation, search

The utility of adding retrospective medication profiling to computerized provider order entry in an ambulatory care population.

Glassman PA, Belperio P, Lanto A, Simon B, Valuck R, Sayers J, Lee M. J Am Med Inform Assoc 2007;14:424–431


Will adverse drug events (ADEs) be reduced when a retrospective drug utilization program is added to a computerized physician order entry (CPOE) prescribing system?

Background and Rationale

ADEs are a major clinical problem. It has long been hoped that a CPOE system might reduce ADEs by prospectively identifying potentially harmful drug-allergy and drug-drug interactions. While promising, published results to date have been quite modest, demonstrating that ADEs do not universally decrease to a significant degree with CPOE implementation. The investigators, from a VA hospital in Southern California, explored whether or not the use of a computerized standalone program called RationalMed, retrospectively used to extract prescription data from the VA EHR system (VISTA), would produce a meaningful decrease in the incidence of ADEs over and above the presumed reduction achieved by the use of the VA CPRS system, with its embedded clinical decision support (CDS).


The RationalMed system was used to create a series of medication profiles for a seven-month period of time for a cohort of 913 patients. Each profile had at least one prescribing error (a “conflict”), and these were generated by a rules engine that sought drug-drug and drug-disease interactions. The patient profiles were randomized to either a control group (“Usual Care”) or intervention group (“Provider Feedback”). The providers associated with the patients in the Provider Feedback group received letters and emails notifying them of the identified conflicts. A clinical pharmacist then did a comprehensive chart review seeking any suspected ADEs and assessing causality to a drug, as well as assessing potential preventability. This pharmacist assigned a numerical score on an Adverse Drug Reaction Probability Scale (ADRP) and compared scores between the two groups.

Main Results

There were 571 suspected ADEs in the 913 patients, occurring in 371 patients (41%). No statistically significant differences in the mean ADRP scores were identified between the Usual Care and Provider Feedback groups. If anything, the ADE incidence per patient was higher in the Provider Feedback group, although it did not achieve statistical significance. There was no difference in ADE preventability between the two groups.


This study disappointingly showed that the intervention studied (application of an automated retrospective drug review and communication of the results to the provider) had no measurable benefit in reducing ADEs when compared to a control group. Furthermore, even with the VISTA CPRS CPOE used for prescribing, with its 2000 embedded drug-drug interaction rules, there was still a 41% incidence of a potential ADE as determined by the use of the RationalMed system and the manual application of the ADRP score by the pharmacist. One of the limitations of the study was that the ADRP score and the assessment of ADE severity and preventability were determined subjectively by the pharmacist, so their validation and reproducibility could not be verified. Nonetheless, the paper demonstrates that merely the existence of a robust CPOE system does automatically result in fewer drug errors, and that such improvement will likely only be seen if the functionality of CDS is improved at the point of care and user-related issues are addressed concurrently.

Robert S. Miller, M.D.