Clinical decision support
Computerized clinical decision support (CDS) aims to aid healthcare professionals in improving patient care. CDS interventions in antimicrobial stewardship (ASP) and infection prevention (IP) can streamline initiatives, improve antibiotic prescribing, and reduce healthcare-associated infections. Reference Rittmann and Stevens1,Reference Singh, Claeys and Advani2 However, thoughtful development of CDS interventions is essential for efficacy and minimizing unintended consequences. Reference Rittmann and Stevens1,Reference Sutton, Pincock, Baumgart, Sadowski, Fedorak and Kroeker3,Reference Wright and Robicsek4 Utilizing established frameworks like the “Five Rights” of CDS and “Ten Commandments of Effective Clinical Decision Support” can maximize intervention utility. Reference Osheroff, Teich and Levick5–Reference Bates, Kuperman and Wang7 This article provides ASP and IP professionals examples for developing, designing, and assessing CDS interventions through real-world application involving a vancomycin Best Practice Advisory (BPA) and a Clostridiodes difficile (C diff) order panel.
The CDS “five rights” in action
The “Five rights” of CDS framework, based on the five rights of medication use, is a structured method for enhancing interventions to improve patient care outcomes Reference Osheroff, Teich and Levick5,Reference Sirajuddin, Osheroff, Sittig, Chuo, Velasco and Collins6 . Utilizing this framework, we assessed and enhanced our vancomycin BPA (Figure 1). This real-time interruptive pop-up alert acts as an “antibiotic timeout”, triggering primary clinicians to reassess empiric vancomycin orders 48–72 hours post-administration. This evaluation is outlined below:
Our BPA design provided a clear reason for triggering, criteria for continued vancomycin use, and a link to institutional guidelines. No changes were implemented.
The patient target audience was those on vancomycin for ≥ 48 hours without a microbiologic indication, however 45% of BPA alerts fired for patients with microbiologic indication. Additionally, 10% of alerts fired for consultants/non-primary providers instead of primary clinicians.
We suppressed the BPA for patients with positive sterile site cultures for methicillin resistant Staphylococcus aureus within 72 hours. The BPA was also removed from ICU units where 50% of non-target prescriber interactions occurred.
A BPA was the appropriate intervention format, as using interruptive alerts for this purpose has been supported by CDS literature. Reference Shojania, Yokoe, Platt, Fiskio, Ma’luf and Bates8 However, the original BPA design required providers to exit before they could discontinue the order. We added a ‘discontinue vancomycin’ button directly to the BPA to streamline the process.
An alert within the electronic health record (EHR) was the appropriate channel, given this was the same application used to order vancomycin. No changes were made.
While the alert timing during vancomycin ordering seemed mostly appropriate, 37% of alerts were deferred, mainly in 6 ICU units where interruptions are common. Since alternative ASP interventions were in place, the BPA was removed from these units.
“Ten commandments” in action
The “Ten Commandments” of effective CDS aim to reduce lag time and errors in evidence-based implementation Reference Bates, Kuperman and Wang7 . We received a proposed design for an order panel, a collection of commonly grouped orders for user selection, with the goal to reduce inappropriate ordering of C diff by assessing alternative causes of diarrhea (Figure 1). We used the “Ten Commandments” to revise this design proposal, outlined below:
The initial proposed design requested the provider manually click through several questions/answers, reducing provider efficiency. To streamline, a dynamic component auto-answered questions using discrete EHR data.
An order panel was deemed appropriate to provide CDS at time of ordering, but the original design had too many questions. Based on user feedback, the order panel was simplified to three questions where no extra information was required beyond automated components: (1) Is the patient on laxatives? (2) Does the patient have a feeding tube? (3) Is the patient on hyper-osmolar medications?
The order panel was designed to display active laxative/hyper-osmolar medications, providing users with the exact alternative cause of diarrhea. Careful design was used to reduce end user actions to one or two clicks only.
Rather than requesting the provider to stop ordering C diff, a nursing order was suggested to hold active laxatives, if present. For hyper-osmolar medications, the order panel suggested placing a pharmacy consult to use alternatives.
Requested metrics on order panel utilization revealed poor use due to a separate standalone historical C. difficile test order. This was eliminated to encourage order panel uptake.
Ownership of the order panel was assigned to the requesting operational team to ensure timely updates with guideline changes.
Conclusion
This report highlights the value of employing established frameworks to enhance the efficiency of CDS interventions for ASP/IP programs. ASP/IP best practices can sometimes create tension by supporting the development of barriers for providers to make it harder to do the wrong thing, but CDS literature suggests this may backfire. Reference Bates, Kuperman and Wang7,Reference Westbrook, Coiera and Dunsmuir9 Our real-world examples illustrate that optimal design of CDS tools can both increase efficiency and align with best practices, making it easier for providers to do the right thing.
Acknowledgments
None.
Financial support
None.
Competing interests
None.