AI May Drive Health Costs Up, Doc-Economist Says

MedPage Today
MedPage TodayMar 26, 2026

Why It Matters

AI’s dual potential to cut per‑unit costs while inflating volume and reimbursement could reshape health‑care spending, affecting insurers, providers, and patients alike.

Key Takeaways

  • AI could automate clinician tasks, potentially lowering service costs.
  • Efficiency gains may spur higher volume of tests, offsetting savings.
  • AI-driven coding optimization may increase hospital reimbursement rates.
  • Revenue-maximizing AI could raise overall healthcare expenditures significantly.
  • Caution advised: AI might exacerbate cost issues rather than solve them.

Summary

The video features a “doc‑economist” warning that artificial intelligence may not lower health‑care spending as hoped, and could even push costs higher.

He outlines how AI can automate routine clinician work—AI scribes, rapid EKG and radiology interpretation, and AI‑assisted coding—potentially reducing per‑service expenses. Yet he notes that efficiency often leads to higher utilization, with clinicians ordering more tests because they are faster and cheaper to read.

He cites examples such as cutting scan‑reading time from a minute to two seconds and using AI to generate optimal diagnosis codes that maximize reimbursement, illustrating how the same technology can boost revenue while inflating overall expenditures.

The implication is that policymakers and health‑system leaders must treat AI adoption cautiously, balancing productivity gains against the risk of volume‑driven cost inflation and intensified coding intensity, lest AI exacerbate the very cost pressures it promises to alleviate.

Original Description

Will AI help the U.S. control healthcare costs? Maybe not.
In an interview with MedPage Today, Ben Sommers, MD, PhD, a primary care physician and professor of healthcare economics at Harvard T.H. Chan School of Public Health, talks about the AI "arms race" in medicine and how it may end up driving healthcare costs up.

Comments

Want to join the conversation?

Loading comments...