AI Care Plans Cost Three Times More Than Doctors’ Decisions, Study Finds

Artificial intelligence systems recommended tests, referrals and medicines costing more than three times as much as the care documented by physicians, according to research published in NEJM AI.

Researchers asked 24 AI systems to produce care plans from 200 anonymised US outpatient and emergency department notes. Among the 20 general-purpose systems, the estimated cost of tests and specialist referrals averaged $314 per visit, compared with $97 for physicians, a 3.2-fold difference.

Much of the gap arose because the systems suggested investigations in cases where doctors had ordered none. Physicians recorded no diagnostic testing in around two-thirds of the visits, yet every AI system proposed tests in at least some of those cases. The models also recommended more specialist referrals and almost twice as many new medicines.

Asking the systems to consider cost or take a more restrained approach reduced the difference but did not eliminate it. However, the findings do not establish that every additional recommendation was unnecessary: some AI plans identified tests that appeared consistent with clinical guidance but had been omitted by the treating doctor.

The analysis was based on written notes rather than full consultations and measured estimated US healthcare costs, not patient outcomes. Researchers said AI-generated suggestions should therefore remain subject to clinical judgement, warning that introducing such systems without proper oversight could increase unnecessary investigations and pressure on healthcare budgets.

Stay informed with the stories shaping health and social care.

Register with The Daily Round for the latest sector news, inspections & reports, wellbeing content, workforce insights, expert features and practical updates—delivered free to your inbox.