Key Takeaways
- AI is already transforming healthcare, often without patients realizing its presence.
- Ambient scribing reduces administrative burdens, freeing physicians for patient interaction.
- AI offers clinical decision support, augmenting human judgment in complex diagnoses.
- Physicians remain liable for AI-assisted outputs, necessitating human oversight.
- AI can strengthen the doctor-patient relationship by reducing EMR time.
- Patient consent, data privacy, and potential AI bias are critical ethical considerations.
- AI's impact on healthcare billing is creating an 'arms race' among stakeholders.
- Cybersecurity for AI tools and generational adoption patterns are key implementation challenges.
Deep Dive
- Dr. Anthony Mazzarelli, co-CEO of Cooper University Health Care, states AI is transforming medicine, often unseen by patients.
- 'Ambient scribing' is the fastest-spreading AI, predicted for 60% physician adoption by year-end.
- This AI generates initial clinical notes from conversations, significantly reducing physician administrative burden.
- AI tools for clinical decision support assist physicians with differential diagnoses at the bedside.
- EPIC's 'Clinical Diagnosis Advisor' uses a 300-million-patient de-identified database for medication suggestions.
- AI functions as 'augmented intelligence,' supporting human judgment rather than replacing it.
- Human verification of AI outputs is crucial due to potential 'hallucinations' or inaccuracies.
- Physicians remain responsible for AI-generated notes they sign, similar to other information sources.
- Concerns about liability increase when AI automates decisions or is fully relied upon, especially with over 340 FDA-approved AI devices.
- Consulting AI for diagnoses may become a standard of care, requiring AI to be updated with new evidence to avoid outdated recommendations.
- Dr. Mazzarelli argues AI can strengthen physician-patient relationships by reducing Electronic Medical Record (EMR) time.
- Ambient scribing increased patient and physician satisfaction, and led to physicians seeing more patients.
- A fear exists that AI's potential accuracy might lead to insufficient human review of generated notes.
- Generative AI creates concise medical notes, filtering extraneous conversation to combat 'note bloat'.
- Concerns include health systems selling de-identified patient data, used for training large language models, to third parties.
- Patient consent for AI use in medical settings is emphasized as both a legal and good practice.
- AI's potential role in mental health diagnosis is explored, particularly for brain health lacking lab values.
- Caution is advised as AI trained on existing data may struggle with diagnoses not yet fully understood.
- An insurance IT department caller reported $500 million in losses this year, $250 million last year, due to increased provider billing.
- AI systems at provider offices identify and bill for multiple issues in a single visit, leading to insurer losses.
- This rise in billing could project a 30-80% price increase for services.
- Dr. Mazzarelli notes AI is creating an 'arms race' among patients, health systems, and insurance companies.
- Concerns are raised about medical device or drug companies biasing AI, and if a non-AI doctor might be considered unbiased.
- Potential for intentional or unintentional bias in AI, due to pre-trained internet data, is a critical implementation consideration.
- The discussion highlights AI's potential to overwhelm human intuition, stressing the importance of the doctor-patient connection.
- Dr. Mazzarelli notes that resistance to change is common in medicine, sometimes for valid patient protection reasons.
- A caller raises concerns about AI and patient privacy regarding personal health information for AI training and data breaches.
- Dr. Mazzarelli states using AI tools does not significantly increase cybersecurity risk, as most valuable data is already in health systems.
- AI scribing saw higher adoption among less tech-savvy, older doctors, contrary to typical early adopter patterns.
- Resistance to AI is present in medicine, sometimes for valid reasons such as specific phrasing preferences.