Why internal medicine isn’t producing
primary care physicians

The pipeline is thinning

Because internal medicine has no fellowship gate on general practice, entry is not a question of competitiveness — it is the largest entry specialty in US graduate medical education, and a graduating resident who wants outpatient primary care can have it. The problem is upstream of the job market: fewer and fewer residents choose the field at all.

The best current measurement comes from Paralkar and colleagues, who analyzed American College of Physicians In-Training Examination surveys from 61,991 internal medicine residents in 2019–2021¹. Read the columns carefully, because they describe different populations.

* The study analyzed only the two 3-year IM program types — categorical (58,028; 93.6%) and primary-care track (3,963; 6.4%); preliminary interns (PGY-1 by definition) and other types were excluded. Of all residents, 19,624 (31.6%) were PGY-3. The article did not specify the categorical-PGY-3 count these percentages are based on; applying the 93.6% categorical share to those PGY-3 residents estimates it at approximately 18,369. The two columns describe different populations; do not read one against the other.

Under one in ten internal medicine residents planned a general-internal-medicine career, whether you look across all residents (9.4%) or at graduating categorical PGY-3s (9.5%). The figures do not sum to 100 because a remainder were undecided or headed elsewhere. One movement deserves emphasis, and this analysis returns to it below: hospital-medicine intent rises over the course of residency. The 15.1%-versus-21.2% gap in the table cannot establish that on its own — the two columns differ in training year and program mix — but a within-person study can, and does, as the next sections show.

The halving is the part that matters most. A decade earlier, West and Dupras found 19.9% of categorical final-year residents planning general internal medicine — 21.5% across all 16,781 final-year residents they surveyed in 2009–2011² — against roughly 9.5% of categorical PGY-3s now. Matched denominator to denominator, the categorical share fell by about half. The generalist decline is older still: Garibaldi and colleagues recorded PGY-3 generalist plans falling from 54% in 1998 to 27% in 2003, with 2003’s incoming interns already down at 19%³ — the erosion was underway well before today’s residents arrived.

You can see the same shift from a different vantage point. An American Board of Internal Medicine study by Gray and colleagues followed 67,902 general internists through Medicare claims from 2008 to 2018 and found the once-standard mixed inpatient-and-outpatient practice collapse from 52% of general internists to 23%, as hospital medicine (25% to 40%) and outpatient-only practice (23% to 38%) pulled apart⁴. Among general internists newly certified by 2018, 71% were practicing as hospitalists and only 8% were outpatient-only. A 2026 Annals commentary argues, as this analysis does, that the decline is structural rather than a failure of values in the residents themselves⁵. The two structural forces are the pay gap, taken up at the end, and what residency actually trains — which the program requirements let us examine directly. These two do not act alone: they interact with the culture and role models of an inpatient-dominated training environment, the preferences residents arrive with, and the six-figure debt that makes choosing the lower-paying path harder. But those are the slower levers to move; the accreditation requirements and the payment system are the ones policy can change, and they are the ones this analysis pursues.

What residency trains — and what it does not

A resident who wants to become an outpatient generalist has to be trained toward it, and here the specialty’s own accreditation requirements are revealing. Compare internal medicine with family medicine, the other primary-care residency. The contrast is not subtle.

Family medicine pins its residents to a measured, age-balanced, longitudinal primary-care practice. Internal medicine requires the functions and duration of a continuity practice — residents must serve as the primary physician for a panel across chronic, acute, and preventive care for the length of training — but it prescribes no quantitative minimum: no minimum hours, no continuity percentage, no panel composition, no panel size. This absence is not an inference from silence — targeted searches of the 2026 internal medicine requirements for a required number of continuity hours, a continuity percentage, or a panel composition return nothing; the specialty simply does not set them.

Two further provisions widen the gap. First, internal medicine requires at least ten months of outpatient experience — but a “month” is defined as 40 half-day sessions, so continuity clinic layered across inpatient blocks counts toward it, and the Review Committee states plainly that it “has not established specific guidelines detailing what counts toward the outpatient requirement.” Subspecialty clinics, urgent care, even the longitudinal continuity panel itself — the requirements explicitly allow that panel to be held in a subspecialty clinic such as an HIV clinic — can satisfy the outpatient floor⁷. A resident can therefore meet much of the outpatient-time requirement through subspecialty and episodic clinics while holding only the minimal functional version of a longitudinal generalist practice. Second, supervision is capped in the clinic and not on the wards: a preceptor may teach no more than four learners at once, and one supervising more than two “must not have other patient care responsibilities” — so that faculty member cannot run a separate independent clinic at the same time. (The encounters they precept can still be billed — Medicare’s primary-care exception allows it — so the cost is opportunity, not lost revenue.) Inpatient supervision caps patient census, not learners: one attending can run a large resident team. The asymmetry is written into the rules, not merely the culture.

Set against this backdrop, three real-world pressures push programs to maximize inpatient time, and it is worth stating them as the reasoning they are rather than as proven claims. First, a resident ward team supplies substantial supervised service capacity at low marginal cost. Second, a large ward team is easier to supervise than a one-to-four clinic. Third, inpatient decisions are more discrete and observable, arguably easier to teach and to test than the longitudinal assessment, prevention, communication, and management of uncertainty that outpatient care demands — competencies that are harder to measure, not necessarily less teachable. The first two have direct support in the supervision provisions above; the third is offered as the author’s reasoning from experience, not as a measured finding.

Why hospital-medicine intent rises over training

The clearest fact here is one of timing. In a within-person study of 9,501 residents, 66.4% of eventual hospitalists first chose that career in their PGY-3 year, and fewer than one in ten held the hospitalist plan across all three years⁸. Hospital medicine is not, for most, a starting intention that persists; it is a destination residents arrive at late, after three years spent overwhelmingly on the wards. Several plausible reasons compound — and they are plausible readings, not tested mechanisms: the ward is the environment residents know best and are most comfortable transitioning into; hospital medicine is a natural fallback when a desired subspecialty does not materialize; and a seven-on, seven-off schedule offers the clear work–life separation this generation of trainees increasingly prioritizes.

There is also a sharper, and more speculative, way to name what makes outpatient work wear on residents — one this author observed repeatedly in late-PGY-3 residents as a program director, and which the literature circles without quite testing. On the wards, the resident makes a decision and enters it as an order for a professional team obligated to carry it out; the implementer is the system. In the clinic, the implementer is the patient: outpatient care is co-produced over time, and a plan works only if the patient can understand it, afford it, and sustain it against cost, transportation, work and caregiving demands, side effects, and competing priorities. The encounters residents find most frustrating are often the ones where that follow-through does not happen — sometimes because a patient questions the evidence, more often because the conditions to act on the plan are not there. The nearest published sentence is an aside in a residency-curriculum paper: residents “frequently feel more comfortable intensifying medical management than they do collaborating with patients to promote healthy behavior”⁹. That is an experienced observation, not a proven mechanism; no study has isolated who implements the plan as the variable, and this analysis presents it as a theory worth testing, not a finding.

Why the clinic never competes

It would be easy to conclude that residency actively converts trainees away from primary care. The evidence supports something more specific, and more interactive: many residents arrive with preferences that already point away from the clinic, and residency does little to recruit toward it — even as career sorting continues through training, with hospital medicine chosen late (as the Ratelle timing showed). Selection and late sorting are not rival explanations so much as two stages of the same failure: the clinic never competes. Residents rate the continuity clinic lowest of their major training experiences — 4.1 on a five-point value scale, against 4.7 for the medical wards, 4.6 for the ICU, and 4.3 for the coronary care unit — and, tellingly, planning a general-internal-medicine career was not associated with valuing clinic more highly¹⁰. When residents were asked directly, 28% said their continuity-clinic experience made them less likely to enter general internal medicine, against 11% who said more likely¹¹. And the structural reason the clinic seldom sells itself is that it is often not experienced as a team: residents at medical-home-certified sites were no more likely than others to work with social workers, pharmacists, or diabetes educators¹². Certification is not the same as team experience; the team tends to exist around the clinic, not around the resident.

The contrast with the wards runs deeper than who sits on the team — it is that learning there is a shared, social act. The inpatient services, the ICU, and many specialty consult rotations operate as resident teams: residents ask each other questions, teach one another, and are challenged by one another, while the attending — typically present at the bedside for the examination, the procedure, and the family conversation — reasons aloud with the whole group at once. The continuity clinic inverts it — one resident with one patient while the preceptor is divided among several — so there is little shared learning and none of the team camaraderie. In the author’s experience the inpatient structure builds more robust growth, faster skill acquisition, and greater confidence in management, and a resident gravitates toward the setting where they feel most capable; the clinic’s deeper failure is not that it repels so much as that it never competes. Read together, the picture is less a machine that turns generalists into hospitalists than a training environment that never reliably shows residents the primary care it is nominally preparing them for.

What actually works

The diagnosis so far describes what categorical training does not do. But the same evidence carries its own counter-case, and it is the most hopeful part of the picture: internal medicine is not intrinsically incapable of producing generalists — standard categorical training is simply far less effective at it than a purpose-built primary-care track.

In the very Paralkar survey above, categorical PGY-3s planned general internal medicine at 9.5% — but primary-care-track PGY-3s planned it at 31.8%, more than three times the rate, an association that survived adjustment (odds ratio 4.22; 95% CI 3.71–4.76)¹. That is intent; the practice outcomes track it. A 2025 multi-institutional study of graduates of seven internal-medicine primary-care programs (2014–2019, 256 of 314 responding) found 66% still practicing primary care or focused primary care — and the graduates named what sustained them: continuity relationships, breadth of practice, mentorship, a primary-care peer community, and lifestyle, while administrative burden, workload, and burnout pushed the other way¹³.

Read against the accreditation contrast, the lesson is not subtle. When continuity, patient ownership, mentorship, a generalist peer group, and a coherent professional identity are deliberately built into the training — the things a purpose-built track supplies and a default rotation schedule does not — the pipeline responds. The categorical program’s low yield is a feature of its design, not a fixed property of the specialty.

The value the system underpays

The pay gap is the other structural force, and its irony is worth stating plainly: the care the evidence most rewards is the care the payment system funds least. In a fee-for-service system, reimbursement rewards procedures and volume, not cognitive complexity or continuity, so a primary care physician billing office visits will earn on the order of half what a proceduralist earns — a gap covered in dollars on the career profile. Recent adjustments have narrowed that gap without closing it: the 2021 Medicare revaluation raised the office-visit codes primary care depends on, but budget-neutrality rules offset much of the gain with a conversion-factor cut, and value-based and team-based models remain promising but partial — the core fee-for-service misalignment, procedures and volume over cognition and continuity, persists¹⁴. What that gap obscures is the value on the other side of it.

The evidence that primary care improves population health is among the most consistent in health-services research. In a US analysis, every 10 additional primary care physicians per 100,000 population was associated with a 51.5-day increase in life expectancy, against 19.2 days for the same increase in specialists — even as primary care physician supply per capita fell over the study decade¹⁵. Areas with more primary care physicians have long shown lower total health care costs and a more equitable distribution of health¹⁶. The National Academies put it most plainly: primary care is “the only health care component where an increased supply is associated with better population health and more equitable outcomes,” and yet it accounts for about 35% of health care visits while receiving only about 5% of health care expenditures¹⁷. The contrast is sharper still against total spending: the United States spends more on health than any other wealthy country — 17.2% of its GDP in 2024, far above the next-highest¹⁸ — yet directs only that small share to primary care itself. Its health dollars flow to high prices, specialist and hospital-based ambulatory care, and administration; OECD’s own analysis finds US primary-care quality lags peer nations even as overall ambulatory spending runs well above them¹⁹. The field the data most reward is the one the payment system least does — the structural fact beneath the whole career-choice story.

The levers that would move it

A diagnosis this specific implies its own remedies, and they follow the two structural forces rather than the residents. On the training side, family medicine is the template: internal medicine could require what family medicine already does — a minimum dose of continuity practice, a continuity percentage, a panel of defined size and age mix — so that a longitudinal generalist experience is mandated rather than merely permitted. Short of rewriting the requirements, the cheaper lever is the one the evidence already validates: expand and fund purpose-built primary-care tracks, which produce generalists at several times the categorical rate, and build the clinic into a genuine team so that residents meet primary care as it is actually practiced rather than as a solo afterthought. On the payment side, the task is to value cognitive and continuity care closer to what procedures earn — through further evaluation-and-management revaluation, and through the value-based and team-based models that fund the panel management fee-for-service ignores — so that choosing the field does not mean choosing against a doubled income. None of these is speculative; each targets a mechanism named above. What they share is that they are choices, not fixed features — which is the more hopeful reading of a decline that can look, from the trend alone, like an inevitability.

What it means for a resident

None of this is a verdict on the work. It is a verdict on how the work is paid and how residency is structured to train it. The thinning pipeline is what happens when a payment system underpays cognitive, continuity, whole-person care, and a training system — by its own requirements — never obliges itself to point residents toward the clinic. A resident who dismisses primary care as a lesser or fallback choice has usually absorbed the signal of that structure, not weighed the career on its merits.

The practical counsel that follows — what the career actually holds, what it pays, how to enter it, and why it is the most visa-navigable path in American medicine for an international graduate — is on the companion general internal medicine and primary care career profile. If the reasons in this analysis ring true and the work still calls to you, that is worth taking seriously: the country needs this field more than almost any other, and it is among the most open doors in medicine.

References

1. Paralkar, N., LaVine, N., Ryan, S., Conigliaro, R., Ehrlich, J., Khan, A., & Block, L. (2023). Career plans of internal medicine residents from 2019 to 2021. JAMA Internal Medicine, 183(10), 1166–1167. https://pmc.ncbi.nlm.nih.gov/articles/PMC10463168/ ACP In-Training Examination surveys, N = 61,991. The subgroup denominators for categorical PGY-3 residents are not printed in the letter; only the marginal percentages are. Primary-care-track PGY-3 residents planned general internal medicine at 31.8% vs 9.5% for categorical (adjusted OR 4.22; 95% CI 3.71–4.76). The survey’s career-choice options did not include hospice and palliative medicine, a point raised in a published comment (Shalev D, Robbins-Welty G, Sinclair CT. JAMA Intern Med. 2024;184[3]:332–333). ↩₁ ↩₂
2. West, C. P., & Dupras, D. M. (2012). General medicine vs subspecialty career plans among internal medicine residents. JAMA, 308(21), 2241–2247. https://pubmed.ncbi.nlm.nih.gov/23212502/ IM-ITE surveys, October 2009–2011. The 19.9% categorical figure is the like-for-like anchor for the decade comparison; match the denominator before comparing. ↩
3. Garibaldi, R. A., Popkave, C., & Bylsma, W. (2005). Career plans for trainees in internal medicine residency programs. Academic Medicine, 80(5), 507–512. https://pubmed.ncbi.nlm.nih.gov/15851467/ ↩
4. Gray, B. M., Vandergrift, J. L., Stevens, J. P., & Landon, B. E. (2022). Evolving practice choices by newly certified and more senior general internists: A cross-sectional and panel comparison. Annals of Internal Medicine, 175(7), 1022–1027. https://www.acpjournals.org/doi/10.7326/M21-4636 N = 67,902 general internists, Medicare claims 2008–2018 (subscription access may be required). ↩
5. Graham, K. L., Fazio, S. B., & Laine, C. (2026). Revitalizing primary care internal medicine: Addressing the hidden curriculum. Annals of Internal Medicine. Advance online publication. https://www.acpjournals.org/doi/10.7326/ANNALS-26-01780 A commentary arguing that the decline in generalist careers is structural (subscription access may be required). ↩
6. Accreditation Council for Graduate Medical Education. (2026). ACGME program requirements for graduate medical education in family medicine (including FAQs). https://www.acgme.org/globalassets/pfassets/programrequirements/2026-prs/120_familymedicine_2026.pdf ↩
7. Accreditation Council for Graduate Medical Education. (2026). ACGME program requirements for graduate medical education in internal medicine (including FAQs; effective July 1, 2026). https://www.acgme.org/globalassets/pfassets/programrequirements/2026-prs/140_internalmedicine_2026.pdf ↩₁ ↩₂
8. Ratelle, J. T., Dupras, D. M., Alguire, P., Masters, P., Weissman, A., & West, C. P. (2014). Hospitalist career decisions among internal medicine residents. Journal of General Internal Medicine, 29(7), 1026–1030. https://pmc.ncbi.nlm.nih.gov/articles/PMC4061354/ ↩
9. Abramowitz, S. A., Flattery, D., Franses, K., & Berry, L. (2010). Linking a motivational interviewing curriculum to the chronic care model. Journal of General Internal Medicine, 25(Suppl 4), S620–S626. https://pmc.ncbi.nlm.nih.gov/articles/PMC2940440/ The quoted sentence appears in the paper’s introduction as the authors’ framing, not as a study finding. ↩
10. Sisson, S. D., Boonyasai, R., Baker-Genaw, K., & Silverstein, J. (2007). Continuity clinic satisfaction and valuation in residency training. Journal of General Internal Medicine, 22(12), 1704–1710. https://pmc.ncbi.nlm.nih.gov/articles/PMC2219834/ ↩
11. Peccoralo, L. A., Tackett, S., Ward, L., Federman, A., Helenius, I., Christmas, C., & Thomas, D. C. (2013). Resident satisfaction with continuity clinic and career choice in general internal medicine. Journal of General Internal Medicine, 28(8), 1020–1027. https://pmc.ncbi.nlm.nih.gov/articles/PMC3710375/ ↩
12. Block, L., LaVine, N., Verbsky, J., Sagar, A., Smith, M. A., Lane, S., Conigliaro, J., & Chaudhry, S. A. (2017). Do medical residents perform patient-centered medical home tasks? A mixed-methods study. Medical Education Online, 22(1), 1352434. https://pmc.ncbi.nlm.nih.gov/articles/PMC5653944/ ↩
13. Factors influencing primary care career choice: A multi-institutional cross-sectional survey of internal medicine primary care residency graduates. (2025). Journal of General Internal Medicine. https://link.springer.com/article/10.1007/s11606-024-08846-z Graduates of seven internal-medicine primary-care programs (2014–2019); 256 of 314 responded (82%). 66% were practicing primary care or focused primary care. Positive influences named: continuity relationships, breadth of practice, mentorship, a primary-care peer community, and lifestyle; discouraging: administrative burden, workload, and burnout. ↩
14. Centers for Medicare & Medicaid Services. (2021). Physician Fee Schedule (PFS) payment for office/outpatient evaluation and management (E/M) visits fact sheet. https://www.cms.gov/files/document/physician-fee-schedule-pfs-payment-officeoutpatient-evaluation-and-management-em-visits-fact-sheet.pdf Effective January 1, 2021, CMS revalued the office/outpatient E/M codes (99202–99215), raising the work RVUs for the established-patient visits that dominate primary care; because the change was budget-neutral, a conversion-factor reduction offset much of the gain, and Congress partially softened that cut for 2021. The net effect on primary-care payment was a modest, partial adjustment, not a correction of the procedure-versus-cognition gap. ↩
15. Basu, S., Berkowitz, S. A., Phillips, R. L., Bitton, A., Landon, B. E., & Phillips, R. S. (2019). Association of primary care physician supply with population mortality in the United States, 2005-2015. JAMA Internal Medicine, 179(4), 506–514. https://pmc.ncbi.nlm.nih.gov/articles/PMC6450307/ Observational; the paper reports associations, not causation. ↩
16. Starfield, B., Shi, L., & Macinko, J. (2005). Contribution of primary care to health systems and health. The Milbank Quarterly, 83(3), 457–502. https://pmc.ncbi.nlm.nih.gov/articles/PMC2690145/ A synthesis of observational research. The cost findings are in the paper’s “Costs of Care” section, not the abstract. ↩
17. National Academies of Sciences, Engineering, and Medicine. (2021). Implementing high-quality primary care: Rebuilding the foundation of health care. National Academies Press. https://www.ncbi.nlm.nih.gov/books/NBK571800/ ↩
18. Organisation for Economic Co-operation and Development. (2025). Health at a Glance 2025: OECD indicators. OECD Publishing. https://www.oecd.org/en/publications/2025/11/health-at-a-glance-2025_a894f72e/full-report/health-expenditure-per-capita_affe6b0a.html The United States spent 17.2% of GDP and USD 14,885 per person on health in 2024 — the highest of any OECD country (the next-highest, Germany, spent 12.3% of GDP). ↩
19. Organisation for Economic Co-operation and Development. (2022). Understanding differences in health expenditure between the United States and OECD countries. OECD Publishing. https://www.oecd.org/content/dam/oecd/en/publications/reports/2022/09/understanding-differences-in-health-expenditure-between-the-united-states-and-oecd-countries_cafc404c/6f24c128-en.pdf US ambulatory care spending is relatively high — about a third of total health spending, and in per-capita terms roughly 2.5× the average of the other G7 countries — while the paper finds the quality of US primary care “fares less well” than peers. OECD does not publish a primary-care spending share for the United States. ↩

Figures here are survey- and model-based and shift over time. This analysis critiques incentives and requirements, not any individual program or institution, and is educational rather than career, financial, or policy advice.