143 - Human-Centered Design and Evaluation of an EHR-Enabled Stewardship Clinical Decision Support Tool for Pediatric Lower Respiratory Tract Infections

Publication Number: 143.5355

Sabrina E. Carro, Monroe Carell Jr. Children's Hospital at Vanderbilt, Nashville, TN, United States; Carrie Reale, Monroe Carell Jr. Children's Hospital at Vanderbilt, Franklin, TN, United States; Jason M.. Slagle, Vanderbilt University Medical Center, Nashville, TN, United States; Matthew B. Weinger, Vanderbilt University School of Medicine, Nashville, TN, United States; Russell Beebe, Vanderbilt University Medical Center, Nashville, TN, United States; Justine Stassun, Monroe Carell Jr. Children's Hospital at Vanderbilt, Nashville, TN, United States; Alison R. Carroll, Monroe Carell Jr. Children's Hospital at Vanderbilt, Nashville, TN, United States; Jassmine Reill VanSise, University of California, San Francisco, School of Medicine, Bend, OR, United States; Sunitha V. Kaiser, University of California, San Francisco, School of Medicine, San Francisco, CA, United States; Stephen Gradwohl, Vanderbilt University, Nashville, TN, United States; Derek Williams, VUMC, nashville, TN, United States

Background: Diagnostic uncertainty is common in lower respiratory tract infections (LRTI), such as pneumonia, bronchiolitis, and infectious asthma exacerbations, creating a barrier to evidence-based care in children’s hospitals. Electronic health record (EHR)-enabled clinical decision support (CDS) is a potential vehicle for reducing uncertainty and improving care.

Objective: To design an EHR-enabled CDS tool incorporating key determinants influencing antibiotic decision-making for LRTI and evaluate usability with end-users via case-based simulation.

Design/Methods: We employed a mixed-methods, human-centered design process at three children’s hospitals. We conducted direct observations in the emergency, acute, and critical care areas and individual, semi-structured interviews with faculty and trainee clinicians (n=18) to define determinants influencing uncertainty and decision-making in LRTI. We used contextual inquiry to examine needs, workflow, and decision-making processes. Next, iterative design cycles with clinician end-users were used to develop production-ready prototypes. Usability testing (n=16) with adapted Single Ease Questions was completed using deidentified vignettes derived from actual cases to evaluate the final prototype.

Results: User research identified 50 determinants influencing LRTI decision-making. Native EHR workflow barriers included lack of data curation and poor trend visualization. The final prototypes incorporated 34/50 determinants, including 1) a unified dashboard display with timeline-oriented views summarizing clinical course trends (Figure 1); and 2) a Stewardship Advisor to address uncertainty in LRTI decision-making and organize important considerations for antibiotic use (Figure 2). Determinants deemed to have low importance or unavailable due to EHR functionality constraints were excluded. Usability testing identified a mean ease-of-use score of 4.3/5 (median 4.0) and a mean satisfaction score of 4.1/5 (median 4.0). After viewing all aspects of the LRTI CDS tool, participants provided open-ended feedback and optimization recommendations.

Conclusion(s): A human-centered design process facilitated identification of important determinants in antibiotic decision-making for children with LRTI. Usability test findings promoted refinements and successfully incorporated many identified determinants into a single, usable EHR tool to meet the needs and preferences of clinicians. Next steps include clinical impact evaluation in a multi-site, pragmatic clinical trial.

Figure 1: Integrated Lower Respiratory Tract Infection Dashboard
The dashboard displays pertinent determinants identified during user research including timeline-oriented views to show important trends. Abbreviations: EHR: Electronic Health Record. LRTI: Lower Respiratory Tract Infection. WBC: White Blood Cell. ANC: Absolute Neutrophil Count. CRP: C-Reactive Protein.

Figure 2: Stewardship Advisor
This interactive clinical practice guideline cascades open as information is filed by the clinician. Hovering over a button will provide more decision support as shown above for the “Procalcitonin Level < 0.25 ng/mL” selection. These inputs then guide antibiotic treatment recommendations at the bottom that filter in real time and curate an order set with the recommended antibiotics based on this advisor.