View source for Medication errors: a prospective cohort study of hand-written and computerised physician order entry in the intensive care unit

This is a review of Shulman's 2005 article, Medication errors: a prospective cohort study of hand-written and computerised physician order entry in the intensive care unit. <ref name="shulman 2005">Rob Shulman, Mervyn Singer,John Goldstone, and Geoff Bellingan. Medication errors: a prospective cohort study of hand-written and computerised physician order entry in the intensive care unit. Crit Care. 2005; 9(5): R516–R521. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1297620/</ref>

Several articles have reported on the implementation of [[CPOE]] systems within health care facilities, although few (see [[Unexpected increased mortality after implementation of a commercially sold computerized physician order entry system| Han et al 2005]], and [[Cordero L, Kuehn L, Kumar RR, Mekhjian HS. Impact of computerized physician order entry on clinical practice in a newborn intensive care unit. J Perinatol. 2004 Feb; 24(2): 88-93| Cordero et al 2004]]) have focused on implementation within an [http://en.wikipedia.org/wiki/Intensive_Care_Unit Intensive Care Unit (ICU)]. Given higher adverse drug event rates within medical ICUs, with over 50% of the total errors detected occurring during the prescribing stage of a medication order [http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=pubmed&cmd=Retrieve&dopt=AbstractPlus&list_uids=7791255&query_hl=11&itool=pubmed_docsum (Bates 1995)], implementation of a CPOE system within an ICU may have a greater impact on medication error reduction.  This study sought to compare the impact of implementation of a CPOE system lacking active decision support capabilities with handwritten medication orders on the frequency and type of medication errors, and potential adverse outcomes within an ICU setting.

== Study design ==

Medication orders were reviewed during two 4-5 day periods (combined for analysis) before CPOE implementation, during which orders were handwritten, and during four 2-5 day periods after implementation of the QS 5.6 Clinical Information System (GE HealthCare, Annapolis MD) at the University College Hospitals London ICU.  This system had access to the on-line formulary, IV guidelines, and to information regarding drug interactions, contraindications, and side effects, but all available information was passive; no active clinical decision support capability was available.  An ICU clinical pharmacist prospectively reviewed all orders and recorded each medication error, defined as a “prescription decision or writing process that resulted in unintended significant reduction in probability of treatment being timely and effective, or an unintended significant increase in risk of harm”.  The clinical pharmacist and Clinical Director of the ICU reviewed all errors, classified them according to type and patient outcome (both actual and potential), and ranked them in severity (minor, moderate, major).  The remaining ICU medical and nursing staff were blinded to the study.  Patient acuity of illness (APACHE) scores were also collected during all study periods.  The data were analyzed using the Chi Square test for error rate and outcome comparisons and learning effect trends, together with the Student T-test for comparison of acuity scores.

== Results ==

There was a significant overall reduction of medication errors after CPOE implementation, with 69 errors from 1036 orders (6.7%) pre- vs. 117 errors from 2429 orders (4.8%) post-implementation (p<0.04).  The rates of errors varied over the time periods, with a trend to decline over time after CPOE implementation (p<0.001).  Patient acuity scores were not significantly different over the study periods.  The frequency of different types of medication errors varied before and after introduction of the CPOE system. The greatest change in error frequency after CPOE implementation was a reduction in orders missing key details such as dose, units, and frequency.  In contrast, the CPOE time periods had a greater proportion of dosing calculation errors, errors of omission of a required drug given the condition of the patient, and, surprisingly, missing prescriber’s signature.  Other types of errors (e.g. incorrect route, wrong drug) were essentially unchanged.

Most errors for both groups were minor and non-intercepted.  During the CPOE time periods there were 4 moderate/ 0 major non-intercepted, and 15 moderate/ 3 major intercepted errors. During the handwritten order period there were 0 moderate/ 0 major non-intercepted, and 19 moderate/ 0 major intercepted errors.  Three of the major error outcomes during the CPOE period were unique to computerized order entry, and included lack of activation of an order for an anticonvulsant, a pull down menu choice error for drug units (mg/kg vs. mg), and an incorrect choice of drug from a drug list. There was no difference between intercepted vs. non-intercepted errors between the two groups.

== Comment ==

There are several issues with the study design and analysis that influence the interpretation of the results.  The observation periods were not uniform in length between the control (handwritten orders) and CPOE time periods; errors may have been missed during the shorter CPOE observation periods, and the analysis did not account for variable length of observation, or the combination of both control time periods into one.  In addition, the junior medical staff changed during the study periods.  Thus, any difference in error rates may be related to the change in physicians (given the short observation periods) rather than the CPOE system. Likewise, a trend in reduction of errors was seen after CPOE implementation, yet observation periods for the control periods were inadequate to exclude the possibility that the trend was simply short term variation in error rate.  The error outcome classification was unblinded and may have introduced bias.  In fact, 2 CPOE non-intercepted errors were listed as moderate, yet from their descriptions appeared to be major in severity.  Finally, the study was not powered for subgroup analysis to assess differences in specific types of errors or potential outcomes between the two groups.
  
This study illustrates the difficulty in achieving meaningful medical error reduction in real-life scenarios that match those reported from pilot trials at academic centers, as well as the need for active clinical decision support to achieve a greater reduction in errors.  Simply following the National Health Service Plan’s mandate to implement an electronic patient record system did not guarantee automatic, significant improvement in patient care and safety. The study found a reduction in overall errors after implementation of CPOE, derived mostly from a reduction in minor errors related to incomplete orders.  Unfortunately, there was no apparent improvement in errors of moderate to major significance; in fact, the absolute percentage of errors of this magnitude were greater in the CPOE study periods, although true statistical analysis is limited by sample size and number of subgroups. One should also note that potential adverse outcomes were included together with actual (non-intercepted) adverse outcomes in the analysis. In addition, CPOE implementation brought CPOE-specific errors related to pull down menus and incomplete order activation.  


== References ==
<references/>

[[Category: OHSU-F-06]]
[[Category: Reviews]]
[[Category: CPOE]]