Clinical AI tools are entering health care, and Axios reporters Caitlin Owens and Adriel Bettelheim ask whose interests their recommendations will serve when cost and patient benefit pull in different directions. Their account draws on an argument by Bob Wachter, chair of the Department of Medicine at the University of California, San Francisco: There is rarely one right decision in medicine.
As Axios recounts it, Wachter warns that setting AI to optimize care for each patient could put society on a path toward bankruptcy. Make cost-effectiveness the default instead, he argues, and talk of “AI death panels” could return. These are warnings about possible outcomes, not reports that either has occurred. Wachter’s proposed response is to show clinicians multiple recommendations and the tradeoffs of each.
Axios also describes lawsuits challenging insurer predictive tools. Medicare Advantage patients and families are suing UnitedHealth over allegations involving nH Predict and decisions about when post-acute rehabilitation or skilled nursing care should end. UnitedHealthcare denies using nH Predict in coverage decisions. Axios describes a case against Humana that alleges a predictive tool applied a standard recovery time to patients whose recovery took longer. Those claims are allegations, distinct from Wachter’s proposal for how clinical tools should present options.
