Clinician Preimplementation Perspectives of a Decision-Support Tool for the Prediction of Cardiac Arrhythmia Based on Machine Learning: Near-Live Feasibility and Qualitative Study

Abstract

Background Artificial intelligence (AI), such as machine learning (ML), shows great promise for improving clinical decision-making in cardiac diseases by outperforming statistical-based models. However, few AI-based tools have been implemented in cardiology clinics because of the sociotechnical challenges during transitioning from algorithm development to real-world implementation. Objective This study explored how an ML-based tool for predicting ventricular tachycardia and ventricular fibrillation (VT/VF) could support clinical decision-making in the remote monitoring of patients with an implantable cardioverter defibrillator (ICD). Methods Seven experienced electrophysiologists participated in a near-live feasibility and qualitative study, which included walkthroughs of 5 blinded retrospective patient cases, use of the prediction tool, and questionnaires and interview questions. All sessions were video recorded, and sessions evaluating the prediction tool were transcribed verbatim. Data were analyzed through an inductive qualitative approach based on grounded theory. Results The prediction tool was found to have potential for supporting decision-making in ICD remote monitoring by providing reassurance, increasing confidence, acting as a second opinion, reducing information search time, and enabling delegation of decisions to nurses and technicians. However, the prediction tool did not lead to changes in clinical action and was found less useful in cases where the quality of data was poor or when VT/VF predictions were found to be irrelevant for evaluating the patient. Conclusions When transitioning from AI development to testing its feasibility for clinical implementation, we need to consider the following: expectations must be aligned with the intended use of AI; trust in the prediction tool is likely to emerge from real-world use; and AI accuracy is relational and dependent on available information and local workflows. Addressing the sociotechnical gap between the development and implementation of clinical decision-support tools based on ML in cardiac care is essential for succeeding with adoption. It is suggested to include clinical end-users, clinical contexts, and workflows throughout the overall iterative approach to design, development, and implementation.

User-Centered Design of a Machine Learning Dashboard for Prediction of Postoperative Complications

“…In settings ranging from primary care clinics to hospital wards, explanations for ML predictions have been necessary to aid physician understanding and build trust. [22][23][24] Anesthesiology clinicians using a postoperative hypotension prediction system felt that different decision thresholds were needed for different patient groups, 8 similar to our finding that different thresholds should be used in different surgical populations. However, some user needs depend on the specific use case.…”

Intensive Care Unit Physicians’ Perspectives on Artificial Intelligence–Based Clinical Decision Support Tools: Preimplementation Survey Study

“…We recommend focusing on the important local and sociotechnical context of each preimplementation site to meet the challenge of the "last mile" of implementation [11,21,22,30,31]. The positive attitudes and willingness to use AI-CDS tools we observed are positive indicators of the acceptance of this new technology [32,33], but they also underpin the idea that expectations should be aligned with the intended use of the AI-CDS tool to be adopted [17]. Moreover, it will be of value to repeat our questionnaire after the implementation of the AI-CDS tool for discharging ICU patients, as it has previously been observed that physicians showed As illustrated by the differences in familiarity and enthusiasm toward AI-CDS at the development and nondevelopment sites, sufficient attention should be paid to training and informing physicians on the use of the AI-CDS tool in their daily practice [10].…”

“…A recent scoping review of guidelines for the development of AI-CDS tools concluded that more focus on implementation strategy is needed for effective integration in the clinical setting [29]. Human-factors research, in the form of qualitative interviews and questionnaires, may enhance the uptake of AI-CDS tools, as this approach may improve the system's design, training process, and implementation strategies [12,17]. We recommend focusing on the important local and sociotechnical context of each preimplementation site to meet the challenge of the "last mile" of implementation [11,21,22,30,31].…”

“…There is a need to study human factors for the safe and effective implementation of AI-CDS tools, as high predictive performance does not ensure acceptance of these technologies [12,13]. Physicians' perspectives on clinical AI have been investigated in survey studies in the fields of psychiatry [14], gastroenterology [15], diagnostic pathology [16], and cardiology [17], as well as across specialties [18][19][20]. In general, strong interest and favorable attitudes toward the use of AI-CDS were reported, but no study has focused solely on the application of AI-CDS tools in the ICU in terms of willingness to use such a tool in clinical practice and how it would fit into clinical workflows.…”

Multimodal Healthcare AI: Identifying and Designing Clinically Relevant Vision-Language Applications for Radiology

“…Developing AI systems for healthcare is a complex space with many, wide-ranging sociotechnical challenges [3,9,46,60,150], spanning: (i) concerns about patient autonomy and ability to explicitly consent or withdraw from healthcare data uses, and its privacy protection in AI development or use [123,134]; (ii) investigations into AI workflow integration [9,21,27] and how best to configure clinician-AI relationships to effectively empower care providers [50,54,125,141,147]; as well as (iii) challenges around acceptance, trust and adoption of AI insights into clinical practice [52,60,86,114,139]. This is mostly addressed in the field of eXplainable AI (XAI) through research into AI transparency via explanations and other mechanisms to help clinicians contest [53] or learn about AI outputs [24] to be able to develop an appropriate mental model of AI capabilities and their limitations.…”