Date of Award
Open Access Dissertation
Computer Science and Engineering
Elizabeth A. Regan
In the age of artificial intelligence and large datasets, information retrieval by querying large databases is an impossible task for the common user due to the information overload. Recommender Systems (RS) for commercial applications like YouTube, Amazon and Netflix were designed to support users by finding items of interest based on their user profiles and various filtering techniques. Health Recommender Systems (HRS) is a category of RSs that provides immense opportunities for application across several healthcare domains and contexts including treatment decision support. Unlike RS applications that focus on analyzing consumer choices, a key differentiator for HRS applications is the decision process that must support a shared decision-making model between patients and healthcare providers. This differentiator poses both challenges and opportunities for extending features for an HRS across different healthcare domains. For instance, orthopedic treatment decisions need dynamic, almost real-time data for treatment effectiveness outcomes comparisons to take place during clinical visits between patients and physicians. The goal is to improve treatment option selection and address Heterogeneity of Treatment Effects (HTE) as defined in the musculoskeletal literature. Moreover, clinical user-centered design, validation, and evaluation of digital health solutions for patient-centered treatment support is a complex and challenging task. A HRS design framework for digital health solutions provides the opportunity to generate data evidence and fill an important research gap. In this dissertation, a digital health design solution for improving treatment decisions in orthopedics is proposed. To achieve this objective, solution options were identified using a design science research framework and HRS design approach was adopted to 1) Research and review HRS design frameworks for health decision support, 2) Collect and front load patients’ treatment outcome preferences into an EHR, 3) Design and validate a HRS prototype for presenting personalized comparative effectiveness data for evidence-based medicine and improved treatments for a leading orthopedic condition, 4) Evaluate the HRS design usability in a clinical visit context, and lastly, 5) Evaluate the HRS design utility for enhancing shared decision-making dynamics in an orthopedic visit. For this purpose, a robust mixed method research methodology was designed and executed. Several design contributions are presented, including the addition of new datasets to digital data collections of health systems; a novel digital tool and mHealth app to collect patients’ treatment outcome preferences using a direct weighting technique; a user evaluated interface design of a health recommender system to select similar patient cohorts and visualize outcomes data for comparative effectiveness; and process contributions such as enhancement of patient provider communication and patient-centered individualized data evidence for improved treatment decisions in a shared decision-making context for an orthopedic clinical visit. The digital health solution artifacts developed through this research offer several benefits over the present treatment decision process in orthopedics and their impact is presented in this manuscript.
Singh, A.(2023). Digital Health Design for Improving Treatment Decisions. (Doctoral dissertation). Retrieved from https://scholarcommons.sc.edu/etd/7467