To identify all published studies evaluating computerized physician order entry (CPOE) in the inpatient setting and uniformly classify these studies on outcome measure and study design.

All studies that evaluated the effect of CPOE on outcomes pertaining to the medication process in inpatients were electronically searched in MEDLINE^® (1966 to August 2006), EMBASE^® (1980 to August 2006) and the Cochrane library. In addition, the bibliographies of retrieved articles were manually searched. Articles were selected if one of their main objectives was CPOE evaluation in an inpatient setting.

Identified titles and abstracts were independently screened by three reviewers to determine eligibility for further review.

We found 67 articles, which included articles on CPOE evaluation on some outcome at the time of ordering. Most papers evaluated multiple outcome measures. The outcome measures were clustered in the following categories: adherence (n=22); alerts and appropriateness of alerts (n=7); safety (n=21); time (n=7); costs and (organizational) efficiency (n=23); and satisfaction, usage and usability (n=10). Most studies used a before–after design (n=35) followed by observational studies (n=24) and randomized controlled trials (n=8).

The impact of CPOE systems was especially positive in the category adherence to guidelines, but also to some extent in alerts and appropriateness of alerts; costs and organizational efficiency; and satisfaction and usability. Although on average, there seems to be a positive effect of CPOE on safety, studies tended to be non-randomized and were focused on medication error rates, not powered to detect a difference in adverse drug event rates. Some recent studies suggested that errors, adverse drug events (ADEs) and even mortality increased after CPOE implementation. Only in the category time the impact has been shown to be negative, but this only refers to the physician's time, not the net time. Except for safety, on the whole spectrum of outcomes, results of RCT studies were in line with non-RCT study results.