Ambient AI Documentation in Clinical Genetics: Perspectives on Implementation and Impact on Burnout
Narain, A.; Misurac, J.; Van Tiem, J.; LaSpisa, C.; Campbell, C. A.
Show abstract
Objectives: To assess genetic counselors perspectives on ambient AI adoption and its impact on counselor burnout. Materials and Methods: We utilized a mixed methods approach, surveying burnout using the validated Stanford Professional Fulfilment Index (PFI) before and after ambient AI adoption and exploring adoption perspectives through semi-structured interviews. Results: 64% of participants (16/25) completed the pre-survey, with eleven completing post-surveys (69% response rate for completion of all three surveys). 14/25 participants completed interviews. Ambient AI use was associated with reduction in burnout after 90 days; respondents who reported using ambient AI (vs. non-use) had burnout scores 1.05 points lower, on average (p=0.008). Benefits of adoption included effective use with interpreters, memory aid, summarization of non-templated note sections (e.g. family/social history), and improved patient engagement. Challenges included template customization, variable accuracy, oversimplified medical language, and rapport disruption during consent. Ethical and regulatory considerations included data privacy, bias, awareness of training resources, and concerns about job displacement. Discussion: Ambient AI documentation can reduce documentation burden and burnout among genetic counselors. By evaluating both outcomes and real world implementation considerations, our study provides evidence to guide scalable integration of AI enabled documentation tools in clinical genomic medicine. Conclusion: Ambient AI can help support the sustainability of the clinical genetics workforce as genomic medicine initiatives are scaled across health systems. Addressing genetics-specific documentation needs while prioritizing patient trust, transparency, and provider oversight is essential for responsible ambient AI implementation.
Matching journals
The top 7 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Using Coincidence Analysis to Identify Causal Chains of Factors Associated with Implementation and Optimization of Lynch Syndrome Tumor Screening Across Multiple Health care Systems 94%
- Development and evaluation of a novel educational program for providers on the use of polygenic risk scores 94%
- GUIA: a digital platform to facilitate result disclosure in genetic counseling 92%
Similar papers in this journal
- “It reminds me and motivates me” : Human-centered design and implementation of an interactive, SMS-based digital intervention to improve early retention on antiretroviral therapy: usability and acceptability among new initiates in a high-volume, public clinic in Malawi 93%
- Clinical code sets and the problem of redundancy in code set repositories 93%
- Sources of Successful Participant Engagement in a Public Health Research Study: a Focus on a Latino Community 93%
Similar papers in this journal
- If you build it, will they use it? Use of a Digital Assistant for Self-Reporting of COVID-19 Rapid Antigen Test Results during Large Nationwide Community Testing Initiative 92%
- Low adherence to existing model reporting guidelines by commonly used clinical prediction models 91%
- Low prevalence of COVID-19 Exposure is Coincident with Self-reported Compliance with Public Health Guidelines among Essential Employees at an Institute of Higher Education 90%
Similar papers in this journal
- A Qualitative Study Exploring the Consumer Experience of Receiving Self-Initiated Polygenic Risk Scores from a Third-Party Website 94%
- Benefits and barriers to implementing precision preventive care: results of a national physician survey 93%
- Defining the Critical Educational Components of Informed Consent for Genetic Testing: Views of US-Based Genetic Counselors and Medical Geneticists 93%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.