Uncertainty and Risk at the Point of Care: Implications of Patient-Generated ECGs and Algorithmic Interpretations for Clinical Decision Making

Paper Title

Uncertainty and Risk at the Point of Care: Implications of Patient-Generated ECGs and Algorithmic Interpretations for Clinical Decision Making

Publication Info

• Topic area: Clinical decision-making with patient-generated ECGs and algorithmic interpretations.
• Keywords: Patient-generated data, ECG, algorithmic interpretation, clinical decision-making, uncertainty, risk, wearable technology, primary care, emergency medicine, atrial fibrillation.

Background and Problem

• Problem / challenge: The increasing use of patient-generated ECGs and algorithmic interpretations from consumer wearables introduces diagnostic uncertainty for clinicians, particularly non-specialists, due to concerns about data legitimacy, interpretation challenges, and lack of integration into clinical workflows.
• Significance: Understanding how clinicians perceive and use patient-generated ECGs is critical as these technologies become more prevalent, potentially impacting patient safety, healthcare resource allocation, and clinical workflows.
• Motivation and related work: Prior research has focused on specialist cardiologists’ use of patient-generated data (PGD), but little is known about how general practitioners (GPs) and emergency clinicians, who are often the first point of contact, navigate these data. Existing barriers include concerns about accuracy, reliability, and integration into clinical workflows.

Solution

• Proposed approach: A vignette-based study exploring how 33 primary care and emergency clinicians perceive and anticipate using patient-generated ECGs and algorithmic interpretations in clinical decision-making.
• Novelty:
  1. Empirical insights into how non-specialist clinicians perceive and interpret patient-generated ECGs and algorithmic outputs.
  2. Identification of four key factors influencing decision-making: legitimacy concerns, interpretation challenges, diagnostic confidence, and the duality of patient and professional risk.
  3. Introduction of the concepts of clinical uncertainty and duality of risk as frameworks for understanding and designing for PGD in clinical settings.
• Procedure and key techniques: Semi-structured interviews using five case vignettes, each presenting different scenarios involving patient-generated ECGs and algorithmic interpretations, to elicit clinician perceptions, decision-making processes, and strategies for managing uncertainty and risk.

Results

• Concrete findings:
  • Clinicians often viewed patient-generated ECGs as familiar and clinically compelling but questioned their legitimacy due to consumer origins and lack of guideline endorsement.
  • Algorithmic interpretations were approached with skepticism, shaped by prior experiences with medical-grade ECGs.
  • Trust in patient-generated data was conditional on corroborating evidence, such as symptoms or confirmatory clinical tests.
  • Decision-making was heavily influenced by balancing patient risk (e.g., stroke) against professional risk (e.g., legal liability).
• Advantage over baselines: This study uniquely highlights the duality of risk and the propagation of uncertainty in decision-making with PGD, offering a nuanced understanding of non-specialist clinicians’ perspectives.
• Experiments / evaluation: Conducted 33 interviews with UK-based general practitioners and emergency clinicians, analyzing 165 case scenarios using reflexive thematic analysis. Key themes included legitimacy concerns, interpretation challenges, building diagnostic confidence, and managing risk.
• Limitations and future work:
  • Findings may not generalize across healthcare systems with different access to specialists.
  • Participant sample was skewed toward primary care clinicians.
  • Vignettes lacked real-world contextual noise, and future work should explore clinician competencies and the role of illness narratives alongside PGD.

Summary

This study investigates how primary and emergency care clinicians perceive and use patient-generated ECGs and algorithmic interpretations in decision-making. It identifies key factors shaping their responses, including legitimacy concerns, interpretation challenges, and the duality of patient and professional risk. The findings highlight how PGD introduces diagnostic uncertainty and complicates decision-making, with risk often serving as the final arbiter. The study proposes the taxonomy of clinical uncertainty and duality of risk as frameworks for designing technologies to better support clinicians. These insights are critical as wearable technologies and AI-enabled diagnostics become more prevalent in healthcare.