Comprehensive or Wasteful? Clinical AI and the Up-front Workup

AbstractClinical AI is being adopted faster than it is being evaluated, on the promise that it will improve health and reduce waste. Aran and colleagues show that large language models recommend more up-front testing, referrals, and treatment than the physicians who actually saw the patients and at more than three times cost, at least in the short-term. But practicing differently is not the same as practicing worse: whether this reflects wasteful excess or appropriate thoroughness is unknown, as are the consequences for long-run health and spending. On one hand, more up-front care could reduce diagnostic errors, catch disease earlier, deliver guideline-concordant treatment that clinicians overlook, and avert costly late-stage care. On the other hand, it could surface incidental findings that spark anxiety and follow-up cascades, add treatments of marginal benefit, and raise spending through added visits and more intensive billing and diagnosis coding. Neither vignettes nor single-visit studies can settle this. Clinical AI deserves the standard we would demand of any new therapy: trials that randomize practices to these tools and measure long-run effects on patients' health and spending.