LOOP

Lower respiratory tract infections (LRTIs) are among the most commonly treated conditions in the ED and are a leading cause of sepsis and death worldwide. Clinical decision-making for this population is complicated by nonspecific and overlapping disease presentations (e.g. cough, fever) and diagnosis (e.g. pneumonia, bronchitis). This makes treatment and disposition decisions are highly variable.

Using LRTIs as a specific scope, the CDEM team wanted to take a critical look at feedback systems for post-encounter data that could support physician decision making for an LRTI diagnosis.

The result of this longitudinal project is a dashboard, LOOP, that is intended to be an educational tool and support learning in the practice. Before, the physician would carry the burden of following up on the patient. Now, the dashboard alerts the physician about specific outcomes for patients they cared for and democratizes the process, creating a link between care and outcome.

Scoping the Project

Alongside the Whiting School of Engineering, members from Johns Hopkins School of Medicine led interviews around physician’s practice, specifically what actions they take around diagnosis, treatment and disposition. Themes teased out in the interviews created a picture of a physician's care practice and what best practices are held among physicians.

Alongside Becky Slogeris, I worked with the core team to scope the initial project for the first phase of the research, which gave us our design challenge, "How might we create a

closed feedback loop with post-encounter data to better support physician decision making?"

Left to right: PI Dr. Jeremiah Hinson, Becky Slogeris and students tour the trauma center at the Bayview Campus of Johns Hopkins in downtown Baltimore.

Design Research

(In partnership with the Center for Social Design)

The explore our design challenge, I co-faciliated a student studio and worked with students to employ different design methodologies to better understand clinician environments. Students enrolled in the studio shadowed doctors at two different campuses — Howard County General Hospital and Johns Hopkins Hospital. We collected data about the outcomes doctors were interested in learning more about, as well as noted behaviors around digital spaces.

Our guiding questions:

• What information is important and constructive for physicians?

• What information and feedback do physicians currently base their decision making on?

• How might we design incoming feedback and support systems to ensure the use of what we create?

From the information and experiences shared in synthesis and prototyping, we began to map a hierarchy of the dashboard. Through this process, the team was able to refine ideas using best practices in UX design and desired objectives expressed across the multiple phases of research.

Key insights which shaped our process:

• Physicians generally wanted to hear from their peers rather than internal QI teams.

• While they didn't have a preference for feedback (good or bad), they did want the feedback to center around patients.

• Most of the examples which physicians described were anecdotal in nature, creating an interesting ground for more storytelling based visualizations.

Above: Pictured are the iterations of the dashboard from early prototypes from students and physicians to the final product.

Continuing work | Development

Post- student studio, I pivoted to executing the mockups generated in the studio. Through the six-month development process, I centered common values in the work developed during the studio:

• Value the Physician’s Practice Understand that physicians are making the best choices they can given the circumstances. Be mindful of the demands on physicians and design to support their work.

• Capture Curiosity; Encourage Action Use encounter feedback as a way to share and generate knowledge. Build on this knowledge to inspire action in practice and care behaviors.

• Prioritize Clear & Simple Delivery Generate designs which are intuitive, relevant, and easily accessible.

• Be Agile to Need Have the design be customizable for providers’ unique needs and embrace iterations throughout the process.

I worked in tandem with core data engineers on the CDEM team to ground and visualize data pulled from EPIC (the electronic health record). I worked with clinicians at every major juncture of the development process to better scope the narrative physicians would bring to the data presented, such as, timeframes which would be crucial for each outcome and the interplay between handoffs on the chart. Dashboard infrastructure was built in Tableau, due to Hopkins limitations with PHI.

Above: The anatomy of each outcome in LOOP followed a similar structure, prioritizing common time junctures, allowing filtering by provider level and other relevant information needed to jog a physician's memory about a patient.

From this development design process, we found that there are three major areas physicians wanted to know:

• the population perspective: What are the high-level department trends?

• the specific application: How do I stack up to those trends?

• and the patient specific: Who are the people behind the trends?

Similarly, we found common independent variables which better supported a physician's navigation: time (7 days, 30 days and 60 days), physician time (primary and secondary caretaker), location of outcome (in ED, post-ED).

User testing

I worked alongside a resident and a core team researcher to create interview guides and initiate user testing for the dashboard in February 2021. To read more about our user testing process and results, please see our JMIR paper which dives into the details. (Link coming soon, article in pre-publishing)

Launch + Implementation

The dashboard launch in April 2021 with the core outcomes of in-hospital mortality, elevations in care within 24- and 48-hours, and patients who return within 72 hours and those who are admitted upon return.

As part of the implementation plan, I created weekly emails for providers, which helped nudge and drive track to the dashboard when new patients qualified for outcomes in LOOP.

Post-launch, I created an internal User Dashboard using the logs from the server to better understand user engagement. The User Dashboard gave insight into what kind of physicians viewed their information, how long they spent using LOOP, whether they engaged with email nudges, and any correlation with increase in patient load and LOOP use.

The User Dashboard also scoped my outreach to clinicians as I conducted Usability Check-ins. These conversations were informal, but helped us better understand barriers in the usability and helped scope scaling efforts.

Currently, LOOP has added two outcomes — Sepsis 3 and Antibiotic Prescribing behaviors — and has also added a new user set to support internal QI teams at Hopkins. The dashboard is also being used as a tool for resident education and humanities-based curriculum development.

----

More about the team:

The Connected Emergency Care (CEC) Patient Safety Learning Lab formed to reduce health and financial harms caused by a fragmented and disconnected emergency care system. Our work brings together an interdisciplinary team of engineers, designers, digital architects and medical practitioners to create human-centered solutions to systemic and structural challenges.