Background:

Medication discrepancies, defined as unexplained differences in documented medication regimens across different sites of care, are prevalent and an important contributor to adverse drug events among hospitalized and recently discharged patients. In November 2011, we implemented a Web‐based discharge module, including a new tool to reconcile medications at discharge. The goal of this study was to evaluate the impact of this tool on medication discrepancies.

Methods:

The new tool presents the preadmission medication list (PAML) and current inpatient medications side by side, sorted by class, with differences highlighted. Clinicians are required to indicate for each medication whether or not it will be continued at discharge. Clinical decision support is then run on the preliminary discharge list, which populates a third list to the right of the other 2. When finished, clinicians move to a screen that explicitly documents how the discharge medication list differs from the PAML. Using this comparison, the tool produces patient discharge instructions, which explain these differences in plain English, for example, which medications are new, stopped, or changed. To evaluate the impact of this tool on medication safety, we identified unintentional medication discrepancies. Study pharmacists took a “gold standard” medication history, compared that list with admission and discharge orders, and recorded unexplained differences. The primary outcome was the number of medication discrepancies per patient in a pre/post analysis (i.e., before and after implementation of the discharge ordering module), compared using Poisson regression. We distinguished discrepancies because of “medication history errors” from those due to “reconciliation errors.”

Results:

We evaluated 25 patients preimplementation and 90 patients postimplementation. The total number of discrepancies per patient dropped from 5.48 to 4.13 (RR, 0.75; 95% CI, 0.63–1.1; P = 0.20). Surprisingly, the tool had a greater impact on discrepancies from history errors (RR, 0.65; 95% CI, 0.47–1.2) than on reconciliation errors at discharge (RR, 0.81; 95% CI, 0.62–1.4).

Conclusions:

A new tool to assist with medication reconciliation and ordering at discharge, as part of a broader Web‐based discharge module, was associated with a trend toward a reduction in unintentional medication discrepancies. The lack of significant results is likely a result of the small preimplementation sample size and possibly a learning curve in using the new tool. The unexpectedly large impact on history errors may be attributed to an increased incentive to take an accurate medication history, which facilitated reconciliation at discharge and the production of clear patient instructions on how discharge medications differ preadmission. This tool is being used to inform the design of future vendor‐based medication reconciliation software.