The Assumption Nobody Questions
When you're wheeled into a hospital, you want the gray-haired attending. The one who's seen everything. Medical dramas built an entire genre around the premise that decades of clinical experience produce a near-omniscient diagnostic wisdom, and word of mouth from friends, family, and coworkers reinforces the same assumption from the opposite direction: the doctor with more years under their belt has encountered your condition before, knows what to watch for, will anticipate complications, and will ultimately make the call that keeps you alive.
The evidence says otherwise, and the dataset behind that claim is enormous. In 2017, Yusuke Tsugawa and colleagues at Harvard analyzed 736,537 Medicare hospitalizations across the United States from 2011 to 2014 and found a clear, monotonic relationship between a physician's age and the likelihood that their patients would die within 30 days (DOI: 10.1136/bmj.j1797).
The Numbers
The study tracked 18,854 hospitalist physicians treating elderly Medicare patients admitted with medical conditions. By comparing outcomes within the same hospital, the researchers controlled for differences in hospital quality, patient population, and regional variation. Patients were effectively quasi-randomized to hospitalists based on work schedules, not physician preference.
The adjusted 30-day mortality rates tell the story in a single, uncomfortable gradient: 10.8% for physicians under 40, rising to 11.1% for those aged 40 to 49, then 11.3% for those 50 to 59, and finally 12.1% for physicians 60 and older. Every additional decade in a physician's career added roughly 0.4 percentage points to their patients' probability of dying, and for doctors over 60, the adjusted odds ratio reached 1.17 compared to those under 40.
These are not small differences measured in a handful of patients. The full Medicare hospitalist population during this period included an estimated 154,000 admissions managed by physicians 60 and older, and applying the 1.3-percentage-point mortality gap between the youngest and oldest physician groups yields approximately 2,000 excess deaths over those four years that would not have occurred at the younger physicians' mortality rate, a number that grows substantially when you account for Medicare Advantage enrollees and non-Medicare hospital admissions excluded from the sample.
A Pattern Found Twice
The Tsugawa findings didn't emerge from nowhere. Twelve years earlier, Niteesh Choudhry and colleagues at Harvard Medical School published a systematic review of 62 studies examining physician experience and care quality in the Annals of Internal Medicine (DOI: 10.7326/0003-4819-142-4-200502150-00008). Their results: 32 of 62 evaluations (52%) found declining performance across all measured outcomes as years of practice increased. Another 13 (21%) found decline for at least some outcomes. Only a single study out of 62 reported that more experience consistently improved performance.
The Choudhry review measured process quality: guideline adherence, diagnostic accuracy, knowledge test scores, while Tsugawa measured the hardest outcome imaginable: whether the patient survived. Both pointed in the same direction, and taken together they represent two decades of evidence converging on the same uncomfortable conclusion.
Why Clinical Skills Decay
Two mechanisms are routinely proposed to explain the pattern, and both concern the distance between training and practice. Medical knowledge becomes obsolete at a startling rate: one estimate pegs the doubling time of medical knowledge at just 73 days, meaning a physician who finished residency in 1985 trained before evidence-based medicine became standard methodology, before hospitalist medicine existed as a specialty, and before many current treatment protocols were developed. Keeping current requires deliberate, sustained effort, and the sheer volume of new literature makes that effort increasingly overwhelming even for physicians who try.
The second mechanism is subtler but potentially more dangerous: clinical intuition calcifies into routine. Physicians develop pattern-recognition shortcuts early in their careers that initially serve them well, but over decades those shortcuts persist even when they're no longer optimal. The attending who still reaches for a familiar drug class "because it always worked" may be relying on evidence that has been superseded by three subsequent trials, never having encountered the newer data in a setting that forced a rethink.
The Volume Exception
The mortality gap between young and old physicians vanished entirely among high-volume doctors treating more than 200 patients per year.
For low-volume physicians (fewer than 90 cases annually), the effect was stark: patients of doctors over 60 had 17.0% mortality versus 12.7% for doctors under 40, an adjusted odds ratio of 1.59. For high-volume physicians, the relationship flattened to statistical nonsignificance (OR 1.01, 95% CI: 0.99–1.03).
The implication is specific. Regular, sustained clinical practice acts as a form of continuing education. Physicians who see patients frequently enough appear to keep their skills current through repeated exposure to evolving standards of care. The physicians at risk are those whose clinical volume has declined, perhaps from shifting toward administrative roles, part-time practice, or outpatient work with occasional inpatient coverage.
The Strongest Case Against These Findings
The most serious challenge is the age-versus-cohort problem, and it is not a trivial objection. Older physicians didn't just age; they trained in a different era, when medical education, residency structure, and evidence standards were fundamentally different from what exists today. The study cannot distinguish between "skills atrophy over time in individual physicians" and "older cohorts were trained under educational standards that produced systematically weaker clinicians," and if medical education improved substantially between 1975 and 2005, then cohort effects alone could explain the entire gap without any individual decline occurring at all.
The patients also weren't truly randomized. Hospitalist assignments were quasi-random based on scheduling, but senior physicians might self-select into certain hospital environments or roles. Unmeasured confounders — including end-of-life care preferences and do-not-resuscitate orders — could bias results despite extensive statistical adjustment.
Letters to the Annals of Internal Medicine responding to the Choudhry review raised a methodological concern: the review included studies with varying quality standards and did not apply a priori criteria to exclude weak designs, potentially inflating the apparent relationship between experience and declining performance.
What We Didn't Prove
This is an observational study. It cannot establish that physician age causes higher mortality, only that the two are associated after extensive statistical adjustment. Residual confounding from unmeasured patient severity is possible despite hospital fixed effects and comorbidity controls.
The study focused exclusively on hospitalized Medicare patients aged 65 and older treated by hospitalists. Whether the same pattern holds for outpatient care, surgical specialties, pediatrics, or younger patient populations remains untested.
The 1.3-percentage-point absolute mortality difference between the youngest and oldest physician groups means that the vast majority of patients survived regardless of physician age. At the individual patient level, the increased risk is small even though the relative increase is meaningful at population scale.
Readmission rates showed no association with physician age. Whatever mechanism drives the mortality difference does not appear to affect whether patients return to the hospital after discharge.
What You Can Do
Ask about case volume. A physician who treats hundreds of patients per year in your condition has better predicted outcomes than a less-active colleague of any age, and asking "How many patients with my condition do you see per year?" is a reasonable question for any hospitalized patient or their family.
Check board certification status. Physicians who maintain active certification must pass periodic knowledge assessments; the American Board of Medical Specialties maintains a searchable verification tool at certificationmatters.org.
Look at hospital-level data. CMS publishes facility-level mortality and readmission data through its Care Compare tool, and the hospital you're admitted to may matter more than the individual doctor treating you.
Recognize the nuance before acting on it. This research describes a statistical pattern across hundreds of thousands of cases, not a verdict on any individual physician. Many older doctors maintain excellent outcomes, particularly those with high patient volumes and active engagement with continuing medical education. The risk factor this study identifies is not age alone but age combined with low clinical volume, and those are two very different things.
The Bottom Line
The instinct to seek the most seasoned physician is intuitive but unsupported by the largest available data. In a sample of three-quarters of a million hospitalizations, patients treated by younger doctors were less likely to die, with one critical exception: older physicians who maintained high clinical volumes had outcomes indistinguishable from their younger peers. The most dangerous combination isn't raw inexperience or the passage of time — it is experience that has gone stale.