Reason
· collected 2026-10-07 · by Jeffrey A. Singer
I recently saw a patient in my clinic who told me she made an appointment with me because I was a highly rated surgeon—at least according to her Garner app.
uncertain
I → see → app
I had never heard of Garner.
asserted
I → hear → Garner
So I looked into it.
asserted
I → look → it
I discovered that a private company had been analyzing my performance without my knowledge, comparing me with other surgeons and steering patients my way.
asserted
company → discover → patients
Naturally, I was happy to learn that Garner considered me a good surgeon.
asserted
me → learn → ?
But I was curious about how it decided which doctors were good.
asserted
doctors → decide → ?
Garner Health is part of a growing industry that helps patients navigate the healthcare system.
asserted
patients → grow → system
Companies such as Included Health, Quantum Health, and Transcarent help patients navigate benefits, coordinate care, obtain second opinions, and find providers.
asserted
patients → help → providers
Others, such as Carrum Health, direct patients to designated centers of excellence for major procedures.
asserted
Others → direct → procedures
Consumer-facing platforms such as Healthgrades and Zocdoc also help patients choose doctors by using factors such as patient reviews, credentials, and convenience.
asserted
patients → face → reviews
Self-insured employers purchase Garner's service and offer it to their workers as a benefit.
asserted
employers → insure → benefit
But rather than relying primarily on patient reviews, reputation, or hospital prestige, Garner analyzes vast amounts of insurance claims data to identify individual doctors who achieve better outcomes at lower overall costs.
uncertain
who → rely → costs
Garner says its database includes more than 60 billion de-identified claims involving roughly 320 million patients and that it evaluates doctors using more than 550 metrics across over 80 specialties.
uncertain
it → say → specialties
It then designates high-scoring doctors as "Top Providers."
asserted
It → designate → Providers
Garner also uses AI to review medical research and help keep the clinical measures it uses to evaluate doctors up to date.
asserted
it → use → date
Employers give their workers a reason to pay attention to those recommendations.
asserted
Employers → give → recommendations
Depending on the employer's plan, patients who choose a Garner-recommended doctor may get reimbursed for some or all of their copays, coinsurance, or deductible expenses.
uncertain
who → depend → copays
Patients receive information about physician performance and can save money by acting on it.
asserted
Patients → receive → it
Employers hope that steering workers toward doctors who achieve better outcomes will reduce their healthcare costs.
asserted
who → hope → costs
Doctors whom Garner rates highly may attract more patients.
uncertain
Garner → rat → patients
Garner appears to have found a market for that proposition.
asserted
Garner → appear → proposition
The company raised $100 million in May at a $2.74 billion valuation and said it now serves more than 2.5 million people.
uncertain
it → raise → people
But measuring an individual doctor's quality is notoriously difficult, and as a surgeon, I can readily see some of the problems.
asserted
I → measure → problems
Insurance claims were designed primarily for billing, not for evaluating clinical performance.
uncertain
claims → design → performance
They can tell an analyst a great deal about what happened to a patient, but they don't necessarily capture disease severity, anatomy, functional status, operative complexity, or other clinical details that influence outcomes.
asserted
that → tell → outcomes
Risk adjustment creates another problem.
asserted
adjustment → create → problem
Imagine that out of two surgeons, one routinely accepts frail 85-year-olds, difficult reoperations, and complicated referrals that other surgeons decline.
asserted
surgeons → imagine → that
The other operates mainly on relatively healthy patients with straightforward problems.
asserted
other → operate → problems
Even an elaborate statistical model might struggle to determine whether differences in their outcomes reflect differences in surgical skill or in the patients they were willing to treat.
uncertain
they → struggle → patients
That matters because physician report cards can create perverse incentives.
asserted
cards → matter → incentives
If doctors know that poor outcomes will hurt their rankings, some may become reluctant to treat the very patients most likely to experience them.
uncertain
some → know → them
Garner says it addresses this problem by adjusting outcomes for patients' comorbidities and demographics and by excluding unusually complex cases when it cannot reliably adjust for their risk.
asserted
it → say → risk
Whether any algorithm can fully account for the patients whom doctors actually choose to treat is another question.
asserted
doctors → account → whom
To a physician, it's slightly unsettling to discover that an algorithm you didn't know existed has been grading you.
asserted
you → discover → you
Physicians can't simply log on to Garner to see their individual scores or learn exactly how the company arrived at them.
asserted
company → log → them
Still, compare Garner's admittedly imperfect approach with the traditional ways patients choose doctors: word of mouth, a friend's recommendation, hospital reputation, an online review from someone angry about spending 45 minutes in the waiting room, or a photograph on a billboard.
asserted
patients → compare → billboard
None of those methods includes reliable risk adjustment either.
asserted
None → include → adjustment
That's what makes Garner interesting.
asserted
Garner → make → ?
It aims to reduce one of healthcare's most persistent information asymmetries: Patients usually know far less than providers about the quality of the care they're buying.
asserted
they → aim → care
Garner doesn't have to devise a perfect method for identifying good doctors to provide useful information.
asserted
Garner → have → information
…and 14 more, not listed.
The Aporia
