Safety and acceptability of a natural-language AI assistant to deliver clinical follow-up to cataract surgery patients: Proposal for a pragmatic evaluation

Abstract

Background: Due to an ageing population, the demand for many services is exceeding the capacity of the clinical workforce. As a result, staff are facing a crisis of burnout from being pressured to deliver high-volume workloads, driving increasing costs for providers. Artificial intelligence, in the form of conversational agents, presents a possible opportunity to enable efficiencies in the delivery of care.\nAims and Objectives: This study aims to evaluate the effectiveness, usability, and acceptability of Dora - an AI-enabled autonomous telemedicine call - for detection of post-operative cataract surgery patients who require further assessment. The study’s objectives are to: 1) establish Dora’s efficacy in comparison to an expert clinician, 2) determine baseline sensitivity and specificity for detection of true complications, 3) evaluate patient acceptability, 4) collect evidence for cost-effectiveness, and 5) capture data to support further development and evaluation.\nMethods: Based on implementation science, the interdisciplinary study will be a mixed-methods phase one pilot establishing inter-observer reliability of the system, usability, and acceptability. This will be done using using the following scales and frameworks: the system usability scale; assessment of Health Information Technology Interventions in Evidence-Based Medicine Evaluation Framework; the telehealth usability questionnaire (TUQ); the Non-Adoption, Abandonment and Challenges to the Scale-up, Spread and Suitability (NASSS) framework.\nResults: The results will be included in the final evaluation paper, which we aim to publish in 2022. The study will last eighteen months: seven months of evaluation and intervention refinement, nine months of implementation and follow-up, and two months of post-evaluation analysis and write-up.\nConclusions: The project’s key contributions will be evidence on artificial intelligence voice conversational agent effectiveness, and associated usability and acceptability.

Ähnliche Arbeiten