Background: Hospitalized elderly patients are at risk for delirium. Despite the widely recognized importance of avoiding potentially inappropriate medications on the BEERS list, hospitalized patients often receive high-risk medications such as benzodiazepines for insomnia. This quality improvement (QI) project aimed to reduce the proportion of at-risk inpatients (defined as any hospitalized patient at age 65 or more) who receive any of the selected high-risk medications for insomnia by 20% within 12 months.
Methods: This QI project was conducted at a university-based quaternary medical center in Milwaukee between April 2017 and April 2018. To minimize adverse impacts on our providers’ workflow, the scope of our intervention was limited to zolpidem, diphenhydramine, and commonly used benzodiazepines (hereafter referred to as target medications). As our electronic medical record (EMR) was unable to accurately capture whether a medication was ordered or administered for insomnia, we embedded a section of “indication” in the target medications’ order screen for at-risk inpatients, which required providers to choose one or more reasons prior to placing an order. Our change hypothesis was that this intervention would lead to behavioral changes by making providers reconsider the indication and appropriateness when ordering a target medication. The outcome measure was the proportion of at-risk patients who received any of the target medications between 8pm and 4am with the assumption that nighttime use more likely suggests insomnia as the ordering indication. The process measure was the proportion of at-risk patients who had any of the target medications ordered. The balancing measure was informal feedback from frontline providers. P-chart was used to detect special variations in these measures over time. Our EMR-based intervention went live on February 26, 2018.
Results: After the implementation, we did not observe any significant change in our outcome or process measures per respective p-chart. The data however showed the following results: On average, at least one of the target medications was (1) given at nighttime in 17%, (2) ordered in more than 35% and (3) given in 25% of at-risk patients. Our intervention enabled more accurate data collection for the proportion of at-risk patients who received any of the target medications for insomnia. The proportion of such patients increased from 6% to nearly 10% suggesting that our EMR was missing nearly 4% of at-risk patients prior to the intervention. Our intervention did not seem overly burdensome to frontline providers per informal feedback.
Conclusions: Our EMR-based intervention to mandate providers to indicate reasons for high-risk medication use in hospitalized elderly patients did not reduce high-risk medication use. The intervention however enabled more accurate data collection and can help frame future improvements to optimize pharmacotherapy for insomnia in at-risk patients.