Original Research · Healthcare
What Physicians on X Actually Post About
The assumption: doctors online mostly vent about pay. We read 10,169 posts from 100 active US physicians to check.
Bottom line
Pay is not the main thing. Only 25% of active physicians posted about pay in a year. Far more posted about training, AI, misinformation and burnout. And only about 1 in 5 physician posts is about the job at all.
The ranking
| Topic | Physicians who posted on it | Share of work posts | |
|---|---|---|---|
| Training & workforce * | 64% | 18.9% | |
| AI & tech | 54% | 16.7% | |
| Misinformation & public trust * | 43% | 25.0% | |
| Burnout & well-being | 40% | 10.2% | |
| Bosses, ownership & careers | 39% | 14.8% | |
| Lawsuits & regulation * | 31% | 6.8% | |
| Pay & reimbursement | 25% | 13.7% | |
| Paperwork & EHR | 20% | 6.0% | |
| Insurance & prior auth | 17% | 5.9% | |
| NP/PA scope | 15% | 2.1% |
* Less reliable tag (model agreement below 0.6). A post can carry more than one topic.
Findings
Pay is not what physicians on X talk about most
Only 25% of active physicians posted about pay or reimbursement even once in twelve months. That ranks pay #7 of 10 work topics. Counted by posts instead of people, pay is 13.7% of work-related posts, still mid-pack.
Pay tagging agreement κ 0.81 · strongAdversarial check. Maybe pay talk lives in replies, which we excluded. Possible, but the same rule applied to every topic, so it does not explain why pay trails training, AI, misinformation and burnout.
Training, AI, misinformation and burnout come first
The topics the most physicians raised at least once: training & workforce (64%), ai & tech (54%), misinformation & public trust (43%), burnout & well-being (40%). Insurance and prior authorization, a frequent headline grievance, reached only 17%.
Moderate · training and misinformation tags are less reliable (κ < 0.6)Adversarial check. Training is broad: mentoring and "proud of my residents" posts count, not just complaints. Read #1 as "most talked about," not "most complained about."
Most physician posts are not about the job at all
Only 19.5% of all posts touched a work-life topic. The rest: clinical medicine and science (34.0%), personal and other (36.9%), and politics (14.2%). A post can carry more than one topic.
High · large categories, κ 0.73–0.78Adversarial check. Physicians who share cases and studies may simply be the ones who stay on X. That is part of the finding: the platform's physician voice is mostly clinical.
The loud pay posts come from a few accounts
Just three accounts wrote 74% of all pay posts in the sample. For misinformation it was 56%. A post count rewards a few prolific voices; counting physicians is the steadier measure.
High · direct countAdversarial check. Capping each account at 200 posts was meant to limit this. It helped, but did not remove it, which is why we lead with the share of physicians.
When physicians do talk money, half of it is Medicare and insurer rates
Pay posts split almost evenly: 51% about their own pay (salary, RVUs, contracts, debt) and 47% about what Medicare and insurers pay for care.
Moderate · sub-split set by the taggerAdversarial check. Counting payer rates as "pay" made pay more likely to rank high. It still did not.
Most physician accounts on X have gone quiet
Of 2,246 verified physician accounts checked, 1,223 (54%) posted nothing original in twelve months, and 75% posted fewer than ten. The active physician voice on X is a small group.
High · direct countAdversarial check. Some reply instead of posting, so they are present but not originating. Either way, a small minority sets the tone.
Method
| Step | What we did |
|---|---|
| Find | Searched X profiles by profession words only (“physician,” “MD,” 40 specialties), never topic words, then grew the list through who those physicians follow. Rules plus an AI screen removed clinics, programs, students, NPs/PAs and non-US profiles. |
| Sample | 100 randomly ordered active physicians (20 from profile search, 80 from the follow network). Active = 10+ original posts (no reposts, no replies), Oct 2025–Oct 2026. Up to 200 posts each: 10,169 posts. |
| Verify | A random 100 checked against the federal NPI registry: 78 matched a physician, 14 did not, 8 used no real name (84.8% of named accounts). |
| Tag | Claude Haiku tagged each post against a fixed 13-topic codebook, tested on a pilot, revised once, then frozen. |
| Check | Claude Opus re-tagged 200 posts independently. Agreement on “is this about the job”: κ 0.72. Topics marked * scored below 0.6. This is model-to-model agreement, not human review. |
Limits: 100 physicians is a small sample; adjacent ranks are effectively ties. Physicians on X skew academic and opinionated, and most of the sample came through the follow network. Topics only, not tone. The US filter relied on profile location and is imperfect.
Open question
Does the ranking hold at 1,000 physicians, and in replies?
The next run: the full 1,000-account sample, plus replies, where much physician argument on X happens. That would tighten the intervals and show whether pay talk hides in conversation rather than in original posts.
Sources
- Original data: 10,169 public X posts from 100 US physician accounts, Oct 2025–Oct 2026, collected via twitterapi.io. Individual accounts are not named.
- CMS NPPES NPI Registry — physician verification.