Improving Medication Guidance Needs to be a Top Priority for Hospitals, Health Systems

Consider this excerpt from a study published in Medical Care: “Perceived poor specificity of drug alerts may be an important obstacle to efficient utilization of information and may impede the ability of such alerts to improve patient safety.”¹ Incredibly, that study is more than 20 years old. The problem of irrelevant medication alerts and medication over-alerting in clinical decision support (CDS) systems integrated within electronic health records (EHR) has stubbornly remained and even intensified since that article was written. A 2020 systematic review found that large proportions of alert overrides were appropriate across guidance for drug duplication, drug allergy, and formulary.²

The growing population of older patients dealing with multiple chronic conditions continues to be a contributing factor to clinician wellbeing and a challenge for informaticians. By 2050, this demographic is projected to expand by over 91%.³ With each new chronic disease, there arises a need for additional medications, blood tests, and potential adjustments to existing medications. This introduces fresh health risks and leads to new clinical guidelines that prescribers must take into account. Consequently, physicians, who are already grappling with unprecedented levels of stress and burnout, are bombarded with more and more CDS alerts.⁴

With each new chronic disease, there arises a need for additional medications, blood tests, and potential adjustments to existing medications. Image Credit: © TensorSpark - stock.adobe.com

Health systems, however, have had success over the years reducing the volume of medication alerts by improving their relevancy. The advancement and refinement of data analytics and clinical guidance technologies developed in recent years have made it easier to identify and customize alerts so that they are “specific” to the patient—just as the decades-old article referenced above recommended. Finding and implementing these systems now is crucial for provider organizations before the medication CDS alert burden becomes unbearable for prescribers, pharmacists, and other clinicians, thereby increasing the risks to patients’ safety and health.

Staying Focused on Patient Safety Goals

Despite its challenges, CDS guidance can, in fact, protect patients. A 2017 study authored by investigators at Brigham and Women’s Hospital found that even though clinicians overrode 73.3% of alerts, 40% were inappropriately dismissed, which increased the risk of an adverse event (AE).4 A more recent study of dose change and avoid medication alerts associated with renal insufficiency found providers overrode 100% of these medication-related CDS alerts, although 12.5% of dose change notifications were appropriate and 29.6% of avoid medication alerts were appropriate.5 During the study period, investigators identified 5 AEs, 4 of which were due to inappropriately overridden alerts.

Clearly, clinicians need medication decision support alerts—just not as many as they receive today. Zeroing in on patient-specificity and the current clinical context could take a significant chunk out of the burden of over-alerting.

I’m a pediatrician and former chief medical information officer at Lehigh Valley Health Network, a 10-hospital health system in Pennsylvania, where we began taking a hard look at medication alert relevancy back in 2001. We developed order sets, expert rules, and alerts, and we were quite proud of how far we had taken our systems. During a pharmacogenomics (PGx) project more than a decade later, we saw the potential to truly tailor medication-related alerts to make them more patient-specific, relevant, and actionable for the provider at the optimal point in the workflow.

The creation of medication alerts should adhere to the “Five Rights of Clinical Decision Support”—the right information, delivered to the right person, in the right intervention format, through the right channel, at the right time in the workflow. The Five Rights has become a guiding principle in effective CDS and was integrated into the PGx project, as well as a book I co-authored on the topic in 2012.⁶

Customizing medication alerts internally was time- and resource-intensive, just as our pharmacy colleagues had warned us would be the case. To accelerate the process, we implemented First Databank (FDB) Targeted Medication Warnings from FDB, given that FDB was already a trusted partner in our health system.

At the same time, we formed a multidisciplinary subcommittee of our clinical decision support committee to study the alert override challenge and develop a strategic plan. Having clinical pharmacists who understood our EHR and our workflows was a critical piece. The subcommittee then identified and prioritized medication alerts to be customized for patient-specific clinical scenarios to eliminate the noise of too many nuisance alerts while still delivering important warnings when applicable. Since physicians, nurses, pharmacists, and other clinicians all face numerous alerts, it was important for this subcommittee to include all stakeholders.

Starting With Low-Hanging Fruit

High blood potassium (hyperkalemia) alerts were the first to be optimized since they were frequently triggered and overridden 81% of the time. These warnings were usually linked to all medications that impact potassium levels, such as diuretics, often regardless of blood potassium levels. Alerts also fired across a broad range of serum potassium results.

To improve specificity, we customized the alert to fire only when a recent blood-potassium level of the patient exceeded a more narrowly defined threshold. At first, the alert and override volume remained the same, but when we fine-tuned the threshold, we began to experience a significant decrease in alert and override volume.

Over 6 months, providers received only 4590 targeted hyperkalemia alerts compared to 15,057 traditional alerts, a 70% reduction. Likewise, providers cut their alert overrides in half, overriding the targeted medication alerts only 38% of the time versus traditional alerts in 81% of instances.

Overall, acceptance of alerts jumped from less than 20% to more than 60%, because the alerts were more specific and relevant. We also restructured the alert’s appearance, making it more noticeable, easier to read, and with actionable buttons embedded while actively tailoring the alerts to conform to the Five Rights of CDS.

Getting More Personal With CDS

We’ve reached a point with clinical decision support technology where the software is much more flexible and capable of intelligently adjusting to different clinical scenarios. The programs are aware of the clinical context around medication orders and can silence or downgrade alerts that do not protect safety or improve outcomes given the patient’s factors and the care being delivered at that moment.

While remedying medication over-alerting may not seem like a top priority for health systems now, it is a problem that will likely only grow in the coming years considering the Baby Boomer generation (all of whom will be 65 or older by 2030) has a greater burden of multiple chronic conditions than any previous generation.^7,8 Moreover, addressing over-alerting now will likely improve the working experience for clinicians, which is one of the most urgent challenges for provider organizations today.

Reducing clinicians’ workload and increasing their face-to-face time with patients through optimized medication CDS is an investment that will pay dividends for a long time to come.

About the Author

Donald Levick, MD, MBA, is a practicing pediatrician and the former chief medical information officer at Lehigh Valley Health Network.

REFERENCES

1. Glassman PA, Simon B, Belperio P, Lanto A. Improving recognition of drug interactions: benefits and barriers to using automated drug alerts. Med Care. 2002;40(12):1161-1171. doi:10.1097/00005650-200212000-00004
2. Poly TN, Islam MM, Yang HC, Li YJ. Appropriateness of Overridden Alerts in Computerized Physician Order Entry: Systematic Review. JMIR Med Inform. 2020;8(7):e15653. doi:10.2196/15653
3. Ansah JP, Chiu CT. Projecting the chronic disease burden among the adult population in the United States using a multi-state population model. Front Public Health. 2023;10:1082183. Published 2023 Jan 13. doi:10.3389/fpubh.2022.1082183
4. Nanji KC, Seger DL, Slight SP, et al. Medication-related clinical decision support alert overrides in inpatients. J Am Med Inform Assoc. 2018;25(5):476-481. doi:10.1093/jamia/ocx115
5. Shah SN, Amato MG, Garlo KG, Seger DL, Bates DW. Renal medication-related clinical decision support (CDS) alerts and overrides in the inpatient setting following implementation of a commercial electronic health record: implications for designing more effective alerts. J Am Med Inform Assoc. 2021;28(6):1081-1087. doi:10.1093/jamia/ocaa222
6. Osheroff JA, Teich JM, Levick D, et al. Improving outcomes with clinical decision support: an implementer's guide. CRC Press. 2012. doi:10.4324/9781498757461
7. Knickman JR, Snell EK. The 2030 problem: caring for aging baby boomers. Health Serv Res. 2002;37(4):849-884. doi:10.1034/j.1600-0560.2002.56.x
8. Bishop NJ, Haas SA, Quiñones AR. Cohort Trends in the Burden of Multiple Chronic Conditions Among Aging U.S. Adults. J Gerontol B Psychol Sci Soc Sci. 2022;77(10):1867-1879. doi:10.1093/geronb/gbac070