Applying Artificial Intelligence to create risk stratification visualization for underserved patients to improve population health in a community health setting

Abstract

<title>Abstract</title> Current healthcare visualizations utilize unstructured Electronic Health Record (EHR) data that lack user requirement analysis for intuitive visualizations. Since clinicians are the end-users, it is crucial to consider end users’ input to improve clinical workflow efficiency. In this paper, we developed a user-centered design through cognitive task analysis (CTA) to visualize unstructured EHR data using artificial intelligence (AI) and natural language processing (NLP). The research team conducted CTA with 8 clinicians. The interviews were transcribed, and a content analysis was performed. Themes were coded, and user requirements helped us to understand the clinical workflow. The CTA resulted in 5 different themes: 1) Gathering patient information, 2) Filtering and searching for necessary information, 3) Subjective, objective assessment, plan of the patient, 4) Visualization of unstructured EHR data, and 5) Progression of trends and comparisons in patients. The intuitive visualization dashboard utilizing unstructured EHR population data was successfully developed. Design elements included an interactive dashboard with a snapshot and basic information for all patients, filter and search keywords, and visualizations of patient trends and lab results through graphs. Finally, a system usability scale (SUS) survey was completed to assess the usability of the dashboard by 20 participants. The completed 20 SUS surveys resulted in an average score of 80.9, which concluded that the platform had a high usability. The health analytics dashboard demonstrated unstructured data containing diverse information that can support identifying underserved populations, enhance workflow efficiency and create intuitive design interface.

Ähnliche Arbeiten