A randomized-controlled trial of computerized alerts to reduce unapproved medication abbreviation use

Abstract

Abbreviation use is a preventable cause of medication errors. The objective of this study was to test whether computerized alerts designed to reduce medication abbreviations and embedded within an electronic progress note program could reduce these abbreviations in the non-computer-assisted handwritten notes of physicians. Fifty-nine physicians were randomized to one of three groups: a forced correction alert group; an auto-correction alert group; or a group that received no alerts. Over time, physicians in all groups significantly reduced their use of these abbreviations in their handwritten notes. Physicians exposed to the forced correction alert showed the greatest reductions in use when compared to controls (p=0.02) and the auto-correction alert group (p=0.0005). Knowledge of unapproved abbreviations was measured before and after the intervention and did not improve (p=0.81). This work demonstrates the effects that alert systems can have on physician behavior in a non-computerized environment and in the absence of knowledge.

Figure 1

Examples of computerized alert screens used in the intervention. (Top) Example of alert with forced functionality (‘hard stop’). (Bottom) Example of alert with an auto-correction feature.

Figure 2

Study design overview.

Figure 3

Comparisons of estimated percentages of unapproved abbreviations in handwritten notes across different groups. Error rate for Quarter 1 (baseline) was estimated using the raw data; error rates for Quarters 2–4 were estimated using a pooled logistic regression model which includes group indicator, indicator of follow-up time, and their interaction terms (group×follow-up time), and specifies autoregression working correlation matrix. p Value for comparisons of error rates across the three groups at baseline is 0.54. p Values for trend test within each group are 0.004, <0.0001, and 0.04 for control, hard stop, and auto-correction group respectively. p Value for comparisons of the error rate between hard stop and control groups is 0.02; 0.21 between auto-correction and control groups, and p=0.0005 between auto-correction and hard stop groups.