Ontario family doctor says new AI notetaking saved her job
AI-powered note‑taking tools are being introduced into primary care to relieve doctors of heavy documentation workloads, exemplified by an Ontario pilot where an “AI scribe” generates SOAP notes from consultations. Commenters welcome the potential time savings and burnout relief but raise deep concerns about data privacy, medical accuracy, liability, and the risk that overburdened clinicians will stop properly checking AI‑generated records. Many argue that the underlying problem is misaligned healthcare incentives and excessive billing-driven paperwork, and warn that AI may paper over systemic issues rather than fix them.
Healthcare incentives & workload
- Several commenters blame Canada’s (and other countries’) payment models: fee‑for‑service and low capitation rates push doctors to short visits and high volume.
- Documentation and admin work are legally required but often unpaid, driving burnout and exits from family medicine.
- Some note recent reforms (e.g., in BC) that move away from pure fee‑for‑service, but say systemic pressure and shortages remain.
Perceived benefits of AI scribes
- Many see AI note‑taking as analogous to human scribes: capturing history, exam, and decision‑making so clinicians can focus on patients.
- Reported benefits include 30–120 minutes saved per day, reduced after‑hours charting, and better capture of secondary details mentioned in visits.
- Supporters argue that even imperfect tools can improve overall care if they reduce delays and cognitive load.
Risks: accuracy, hallucinations & overreliance
- Strong concern about transcription and summarization errors, with examples of dangerous or embarrassing mistakes in other systems.
- Fear that as AI tools get “good enough,” clinicians will stop thoroughly checking notes, similar to autopilot complacency in aviation or cars.
- Some argue reading and lightly editing AI drafts is still less work than writing from scratch; others counter that this mindset itself is risky.
Privacy & data use
- Multiple comments worry about medical conversations being sent to cloud providers (e.g., big US tech), questioning legality and consent.
- HIPAA (and Canadian equivalents) are described as strict but leaky in practice: complex vendor chains, frequent breaches, and low caps on penalties.
- Patients often sign broad data‑sharing forms without understanding them; some report explicit attempts to route data to non‑compliant third parties.
Liability & regulation
- General agreement that AI systems themselves shouldn’t be legally liable; responsibility lies with clinicians, institutions, and vendors who deploy them.
- Some argue that AI tools that alter clinical text should be treated as medical devices, requiring rigorous certification—possibly hard for LLMs.
EMRs, billing, and system design
- Many see EMRs as primarily billing and compliance tools, not patient‑care tools; documentation is optimized for reimbursement codes.
- Critics say the real fix is to pay clinicians for charting and simplify requirements, not add another layer of tech that they must supervise.
Open-source and market dynamics
- Interest in open‑source scribes (e.g., SOAP generators using Whisper/LLMs) and local processing to improve privacy and lower cost.
- Skepticism that such tools will remain open: expectation that larger vendors will buy and bundle them into expensive proprietary systems.