Signaling Sepsis: Improving Sepsis Diagnosis through User-Centered Clinical Decision Support
Project Lead: Kristen Miller, DrPH, MSPH, CPPS
Project AIM(s): The aim of the project is to launch and evaluate a novel user-centered sepsis-specific clinical decision support tool with an enhanced visual display that supports diagnosis through display of important clinical features (referred to as the Sepsis Patient-Level Mpage). The SMART (Specific, Measurable, Achievable, Relevant, and Time-Bound) goal is to reduce errors associated with sepsis diagnosis and optimal, timely clinical management at MedStar Health as evidenced by a 10% increased adherence to the Centers for Medicare & Medicaid Services bundles by September 2021 and by an average clinician-rated System Usability Scale (SUS) score of greater than 68 after two months of use.
Narrative Description
We designed, developed, and deployed a novel CDS tool (Sepsis Patient-Level Mpage) for a subset of nurses responsible for identifying and managing patients with sepsis. Working with the informatics team, this included a robust design (more robust than a simple qSOFA alert) to include important clinical elements, trends in vital signs, and potential actions for clinical management. In addition to the Sepsis Patient-Level Mpage, updates were made to a sepsis dashboard which provides clinicians with one location to access and collectively interpret clinical findings and treatment in real-time. The tool reduces the burden of “hunting and gathering” in the EHR for important and relevant information. Sepsis advanced practice providers and rapid response nurses can identify potentially septic patients and prioritize clinical response. The landing screen serves as a quick reference tool and allows the care team to prioritize patients by adherence to the Centers for Medicare and Medicaid Services (CMS) sepsis bundle, length of stay, or sepsis alert type.
We measured the outcomes of this project in three distinct but complementary ways: preference (e.g., end-user feedback through pre and post surveys), process (e.g., audit data regarding usage of the tool), and clinical outcomes (e.g., sepsis specific bundle compliance and mortality). Use of the Sepsis Patient-Level Mpage varied but the majority of respondents are currently using is sometimes (n=4) or always (n=2). Participants indicated that the new Sepsis Patient-Level Mpage made it easy to visualize patient data and sped up chart review time, allowing the clinician to triage and respond to patients more effectively. Participants were asked if they trusted the Sepsis Patient-Level Mpage and all responded neutrally (n=4) or positively (n=5). Overall, feedback about the Sepsis Patient-Level Mpage was very positive but challenges remain for communicating with the bedside team. One participant stated, “Still frequently talk with nurses who don't know what to do when their patient alerts for sepsis - they don't know what qSOFA is, reach out to primary team instead of getting lactic acid”, while another stated: “My biggest challenge remains promptness/sense of urgency from bedside RNs. The MPage is only as good as the accuracy and timeliness of data entered and I often find challenges with both”.
During a three-month pilot period, the Sepsis Patient-Level Mpage was accessed 300 times by 51 distinct users. Access ranged from one instance (primarily during orientation of the Mpage) to 33 instances with an average of 5.7 uses per user. Daily use of the tool ranged from 32 (first of two orientation days) to 0, with an average launch of 3 times a day. Across the 51 users, 2 physicians used the Mpage 31 times, 15 nurse practitioners used the Mpage 103 times, 17 physician assistants used the Mpage 64 times, and 16 registered nurses used the app 93 times.
The sepsis mortality index, which compares the observed to expected mortality rate had been trending downwards and hit its lowest measure at 1.25 down from 1.93 6 months prior. In evaluating patients who received antibiotics within 3 hours of alert (CMS bundle requirement), we compared the baseline in 2021 where 76% of patients received antibiotics within 3 hours with the new process (Feb 15, 2022 – March 31, 2022) where 85% of patients received antibiotics within 3 hours. It is important to note that there are constant improvements in sepsis care so these improvements can not be completely attributed to the Sepsis Patient-Level Mpage but given the evaluation as a whole, we are confident it was a contributing factor.
Our research methodology including focus groups, data analysis, and problem mapping which was critically important to understand the actual challenges and contributing factors to delays in identification and care of sepsis patients leading to poor patient outcomes. By engaging with clinical end-users, we elicited their needs which ranged from improved visualization of data to strategies when engaging with the bedside care team. We identified that different clinical stakeholders have different needs and it can be challenging to develop a single solution to address their needs. A key finding is the need for deeper investigation into personalized and tailored informatics solutions that fit seamlessly into clinical workflow. In addition to clinical end-users, our partnership with leaders of the Sepsis Advisory Committee provided stakeholder buy-in, helped to expedite technical and recruitment requests, and will likely contribute to sustained and scaled interventions of our project.
Diagnostic quality problem type, failure, or category (symptoms, observed problems, gaps in performance) addressed by the intervention
- Failure in information gathering
- Failure in information integration
Root causes/causative factors addressed by the intervention
- Health information sharing and accessibility via health IT
Setting of the diagnostic quality improvement intervention
- Acute care hospital inpatient area
- Inpatient non-intensive care area (e.g., adult or pediatric medical/surgical, maternity, psychiatric unit)
- Special care area (e.g., ICU, CCU, NICU, step-down unit)