Role of computerized physician order entry systems in facilitating medication errors

From Clinfowiki
Revision as of 01:23, 18 November 2015 by Mbouie-hattix (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Conventional wisdom is in favor of the implementation of CPOE systems to reduce prescribing errors. A substantial body of literature advocates for these systems based on studies of medical errors and institutional reports (such as from the Institute of Medicine). Koppel and colleagues present a contrarian view based on qualitative user interviews and quantitative user questionnaires. The setting of the study was the University of Pennsylvania, a large 750 bed academic practice using the Eclipsys CPOE system. While interviews were with a variety of users including attending physicians, pharmacists, nurses and IT managers, the results focus on the findings from the resident physicians. The researchers conducted one-on-one interviews and focus groups with these house staff physicians but also performed content expert interviews and even shadowed users on rounds. After these data were digested, written questionnaires were developed and administered to the residents.


Results: Questionnaires were returned by 88% of the 291 house staff beginning in 2002. Combining the interview and questionnaire research techniques, 22 “previously unexplored medication-error sources” were identified that CPOE had facilitated. These errors were divided into 2 sources: information errors resulting from the failure to integrate disparate hospital systems and human-machine interface flaws. Examples of the interface flaws included:


  • Pharmacy drug database doses in CPOE based on purchased pill size not clinical guidelines for effective doses resulting in dose range errors
  • New or modified medication orders not connected to discontinuing an old dose
  • Procedure associated orders not discontinued when the procedure was cancelled
  • Requirements for antibiotic approvals resulting in gaps of treatment

Examples of the human-machine interface problems were

  • Selecting the incorrect patient for drug orders because of name screen presentation issues
  • Incorrect selection of medication because of the need to view multiple screens
  • Failure to log off by a previous ordering physician resulting in orders on incorrect patients
  • System unavailability resulting in errors when CPOE returned because of lost or un-entered orders
  • Time related errors when orders entered just after midnight


Comments This largely qualitative study identified situations that increased the probability of errors with CPOE without specifying actual error outcome rates. Pharmacists reviewed all CPOE orders and rejected 4% of them so no definite adverse outcomes were identified in this study. It was unclear where the users or the system was in the cycle of repeated iterations of error detection and system improvement. The authors recommend a focus on organizational not solely technological work. They advise a plan for “continuous revisions” of CPOE by pursing error causation. While this study did not enumerate specific patient adverse outcomes, the contrarian nature of its finding coupled with publication in a high profile journal gained widespread commentary. An editorial in the same issue of JAMA by Wears noted that “these results are disappointing but should not be surprising.”