The Metric Nobody Questions
Every hospital in America now measures patient satisfaction because CMS requires it. Billions in incentive payments ride on the scores. Press Ganey, the dominant satisfaction survey company, reports more than 26,000 healthcare organizations using its platform worldwide. The logic seems obvious: patients who feel heard get better care, and hospitals that earn high marks deliver better outcomes.
Joshua Fenton, a family physician and researcher at UC Davis, tested that assumption against a dataset of 51,946 people. He found the opposite.
The Numbers
Fenton and colleagues analyzed records from 51,946 adults enrolled in the Medical Expenditure Panel Survey between 2000 and 2007 (DOI: 10.1001/archinternmed.2011.1662), a dataset maintained by the Agency for Healthcare Research and Quality and designed to represent the entire US civilian, non-institutionalized population, with satisfaction measured in each person's first survey year using five items from the Consumer Assessment of Healthcare Providers and Systems, the same instrument CMS uses for its Hospital Consumer Assessment scores. Utilization and spending were tracked the following year. Mortality was followed for 3.9 years on average.
After adjusting for age, sex, race, insurance status, chronic disease burden, health status, and prior-year spending, patients in the highest satisfaction quartile, compared to the lowest, were 12% more likely to be admitted to a hospital (aOR 1.12, 95% CI 1.02–1.23), spent 8.8% more on healthcare overall, spent 9.1% more on prescription drugs, and were 26% more likely to die (aHR 1.26, 95% CI 1.05–1.53).
One finding cut the other direction: the most satisfied patients were slightly less likely to visit the emergency department, but they ended up in hospital beds and on prescription drug regimens at elevated rates instead.
The Replication Made It Worse
Seven years later, Fenton's team returned with a dataset nearly twice as large. Their 2019 follow-up in the Journal of General Internal Medicine (DOI: 10.1007/s11606-019-05058-8) analyzed 92,952 adults from the same MEPS survey spanning 2000 to 2015. The larger sample was explicitly chosen to address two critiques of the original: too few deaths for stable estimates, and potentially inadequate morbidity adjustment.
The effect did not shrink. It grew stronger. After full adjustment for smoking status, disease burden, and healthcare utilization, patients in the highest satisfaction quartile had 57% higher mortality (HR 1.57, 95% CI 1.25–1.98).
The gradient was monotonic: each step up the satisfaction ladder carried higher death risk, and adding extensive illness-severity controls did nothing to attenuate it.
The replication also uncovered a gender asymmetry that the smaller original study lacked the statistical power to detect. The satisfaction-mortality link was significant only in women. In the highest satisfaction quartile, women's typical survival advantage over men vanished entirely.
Why Satisfied Patients Die More
The mechanism Fenton proposed is uncomfortable but simple. Doctors who prioritize satisfaction may be more willing to prescribe unnecessary medications, order discretionary imaging, and admit patients when watchful waiting would be safer.
The evidence supports this. A 2005 randomized trial by Kravitz and colleagues found that physicians frequently accede to patient requests for treatments they consider clinically inappropriate. Compensation tied to satisfaction scores makes it worse: physicians under that pressure are more likely to provide low-value services like advanced imaging for uncomplicated back pain.
Unnecessary hospitalization carries real mortality risk: healthcare-associated infections kill an estimated 75,000 Americans annually, medication errors and falls in unfamiliar corridors compound the danger, and the Dartmouth Atlas has documented that roughly 30% of US healthcare spending goes to discretionary services with no measurable benefit. Fenton's data raises the possibility that the institutional drive to satisfy patients inflates that waste.
An original calculation puts the stakes in perspective. CMS penalizes or rewards approximately 3,000 hospitals annually based on HCAHPS satisfaction scores, and US hospital expenditures totaled $1.4 trillion in 2022. If the behavioral incentive to maximize satisfaction contributes even fractionally to the 8.8% spending increase Fenton observed, the implied excess among the highest-satisfaction cohort exceeds $120 billion annually.
The Strongest Case Against This Finding
The most rigorous critique comes from Xu, Buta, and colleagues at Yale, who reanalyzed the same MEPS dataset in 2015 (DOI: 10.1111/1475-6773.12264). They split deaths into those amenable to medical care and those that were not. The satisfaction-mortality association was significant for non-amenable deaths (cancers with no effective treatment, accidents, suicides) but not for amenable deaths where medical intervention could plausibly have changed the outcome.
If satisfied patients are dying from causes their doctors cannot prevent, the link may have nothing to do with overtreatment and everything to do with selection: characteristics that make certain people rate their care highly (compliance, agreeableness, reluctance to challenge authority) might correlate with unmeasured mortality risk factors that no observational adjustment can fully untangle.
Xu and colleagues also found that satisfaction was remarkably unstable: more than half of respondents changed satisfaction quartiles within a single year, suggesting the measure may be a poor proxy for anything durable about a patient's care.
What We Didn't Prove
Both studies are observational, and neither can prove causation. They establish a robust, replicated, dose-response association between patient satisfaction and mortality, but residual confounding from unmeasured variables remains possible despite extensive adjustment across multiple model specifications.
The MEPS sample excludes institutionalized populations, and the satisfaction instrument captures only five dimensions of patient experience, raising the question of whether these findings would generalize to other survey instruments, healthcare systems, or countries with less fee-for-service incentive pressure.
The gender asymmetry is unexplained. The authors offer no mechanistic account for why the effect appears only in women, and it could reflect unmeasured confounding rather than a genuine biological or behavioral difference.
One final caveat. The 2012 study's confidence interval for the mortality hazard ratio (1.05–1.53) barely excludes 1.0 at its lower bound, and while the 2019 replication tightened the estimate considerably, the original finding alone would warrant caution rather than alarm.
What You Can Do
Separate satisfaction from quality. When choosing a hospital or physician, prioritize outcome metrics published on CMS's Care Compare tool: mortality rates, complication rates, and readmission rates. These measure what happened to patients, not how they felt about the waiting room.
Question unnecessary care. If a doctor orders a test or prescribes a medication without explaining why it is specifically indicated for your condition, ask: "What changes in my treatment if this comes back normal?" The question redirects toward evidence-based decision-making and costs nothing.
Recognize the incentive your doctor faces. Under current CMS rules, hospitals lose revenue when satisfaction scores drop. The physician who says no to a requested scan may be practicing better medicine than the one who says yes with a reassuring smile. A doctor who pushes back on your preferences is not necessarily a bad doctor.
The Bottom Line
Patient satisfaction is not a proxy for healthcare quality. Two nationally representative studies covering nearly 145,000 Americans found that the most satisfied patients consumed more healthcare, spent more money, and died sooner. The billions CMS ties to satisfaction metrics may be rewarding the wrong thing.