The open door: why U.S. graduates pass
on the subspecialties the country needs most

One match, two markets

Internal medicine trains one class of residents and then sorts them, through a single fellowship Match each year, into more than a dozen subspecialties. It is tempting to read that Match as one competition. It is closer to two. Some fields turn away applicants; others cannot fill the seats they offer. The 2026 appointment-year results make the split unmistakable¹.

The National Resident Matching Program said the quiet part in its own release: geriatric medicine’s fill rate “declined to 38.9 percent”; adult infectious disease “filled at 60.9 percent, a decrease of 9.3 percentage points” in a single year; and addiction medicine’s rate slipped “from 66.7 to 63.8 percent” even as the field added positions and program tracks². These are not marginal shortfalls. A field that fills two of every five seats is not choosing to be small; it is failing to recruit at the level its patient population demands. The question this piece answers is not whether the gap is real — the numbers settle that — but what U.S. graduates are responding to when they walk past these doors, because they are responding to something real.

What the empty seats have in common

Start with the temptation to explain the gap by intellectual content — that graduates avoid the “cognitive” fields and flock to the procedural ones. That is half true, and the half that is false is instructive. It is true that the fully-filled end of the table is dominated by fields with a procedure or a well-reimbursed technical stream: cardiology’s catheterizations, gastroenterology’s endoscopy, oncology’s infusion and drug margin. And it is true that the empty end is dominated by fields whose core work is cognitive, longitudinal, and delivered largely to public-payer populations — the old, the chronically infected, the patient in kidney failure, the patient in addiction. But endocrinology sits at 97.7% — a cognitive, non-procedural, largely outpatient field that fills nearly every seat. So the dividing line is not simply procedure versus thought.

The sharper pattern is this: the empty seats are where low pay meets a second deterrent. Endocrinology pays modestly but asks little extra beyond the training — predictable clinic hours, limited inpatient acuity, no unusual stigma or on-call burden. The underfilled fields each pair low compensation with something more: nephrology’s dialysis-tethered, high-acuity, weekend-heavy practice; infectious disease’s outbreak-facing complexity and a pay structure that, as its own society concedes, has fallen behind; addiction medicine’s stigma and its newness as a recognized field; geriatrics’ documented paradox in which the fellowship can lower lifetime earnings. Where the compensation penalty stands alone, graduates absorb it. Where it compounds with acuity, geography, stigma, or a training investment the market does not reward, the seats sit empty. The rest of this analysis walks those deterrents in turn, because each one is measurable.

What the payment system rewards — and what it does not

The clearest of the deterrents is the one the fields themselves are loudest about. American physician compensation runs on the work relative value unit — the wRVU — a currency that pays for discrete, billable, largely procedural encounters and pays comparatively little for the cognitive labor of managing complexity over time. Infectious disease has made the arithmetic explicit. In a 2023 Clinical Infectious Diseases analysis, ID physicians attending on the general-medicine wards earn $30,000–70,000 less than hospitalists who work fewer hours; an ID physician practicing HIV primary care can earn $30,000–60,000 less than a generalist doing primary care; and AAMC data place ID mean salaries $40,000–80,000 below other divisions in departments of medicine — because, in the authors’ words, the wRVU model “poses challenges for nonprocedural specialties like ID”³. That is a specialty describing, in dollars, why the extra two years of training do not pay for themselves.

Geriatric medicine is the more startling case, because there the training carries a negative return. Geriatrics is, in the words of a 2023 JAMA analysis, “one of very few specialties for which fellowship training and board certification result in a lower salary than if neither had been pursued”: the 2022 AAMC faculty salary survey put the median geriatrician’s salary 9% below that of general internists and 14% below that of hospitalists⁴. A resident who completes a geriatrics fellowship can expect to earn less than the classmate who stopped at internal medicine. The subspecialty survey data point the same way: geriatrics averaged $291,968, infectious disease $320,730, and endocrinology $290,606, all at or below the general-internal-medicine figure of $326,116 — the internal-medicine subspecialties for which more training buys a lower average paycheck⁵. The procedural fields at the full end of the table pay far more; but endocrinology is the tell — it carries the same pay penalty as geriatrics and infectious disease, yet fills nearly every seat, the first sign that low pay by itself is not what empties a field. Still, the gradient is real: with endocrinology the exception, the fields that fill fastest tend to be the ones that pay most, and the empty fields are the ones where the training does not raise — or actively lowers — the paycheck. The Match table is the rational sum of thousands of those calculations.

The arithmetic of another year

The pay gap does not act on a blank slate; it acts on a balance sheet already carrying six figures of debt. 71% of the medical-school class of 2024 graduated with education debt, a median of $205,000 among those who borrowed⁶. Fellowship does not pause that debt so much as deepen its opportunity cost. A fellow earns a trainee stipend — a median of $65,100 in the first post-MD year⁶ — for one to three additional years, forgoing an attending salary several times larger. For a field that pays well afterward, that deferral is an investment with a return. For a field that pays at or below what internal medicine already offered, the deferral is a cost with no offsetting gain: the graduate spends years at a fellow’s wage to arrive at a salary they could have earned two years earlier without the fellowship at all.

What sharpens this is that in several of the underfilled fields, the credential the extra year confers is one the job market does not require. Geriatrics is the clearest example: board certification, the JAMA analysis notes, “does not meaningfully influence job opportunities or practice activities” — a general internist can care for older adults, and most of the country’s older adults are cared for by physicians who never did the fellowship⁴. Addiction medicine now sits in a similar position for its signature treatment. Since Section 1262 of the Consolidated Appropriations Act, 2023 eliminated the federal “X-waiver,” any clinician with a standard DEA registration may prescribe buprenorphine for opioid use disorder, with no waiver, training gate, or patient cap⁷. That change was a public-health victory — it put the most effective outpatient treatment for opioid use disorder into every practice — but it also means the subspecialty fellowship is no longer the ticket of entry to the work it was built around — the fellowship still confers real expertise in complex and co-occurring substance use disorders, withdrawal management, consultation, and program leadership, but the common outpatient treatment no longer requires it. When the market does not price a credential for the work most physicians will do, a debt-carrying graduate is unlikely to spend a year and a stipend to earn it.

Prestige, visibility, and a young field

Money and debt do most of the explaining, but not all of it. Two softer forces round out the picture, and they are worth naming precisely rather than gesturing at.

The first is prestige, which in medicine tracks pay and procedure closely and is transmitted, powerfully, during training. A resident forms a specialty identity by watching the specialists around them — and the fields that fill are the fields whose consultants are most visible, most highly compensated, and most often described, in the informal hierarchy of the wards, as the “competitive” ones. The underfilled fields suffer a visibility problem on top of a pay problem: geriatrics and addiction medicine in particular are underrepresented on the inpatient services where residents spend most of their time, so a trainee may finish residency having rarely seen the career practiced at its best.

The second is novelty. Addiction medicine was recognized as an ABMS subspecialty only in 2016 — the board voted in October 2015 and announced it in March 2016, making it one of the youngest recognized fields in American medicine, and a multidisciplinary one open to physicians from any of the 24 member boards⁸. A field that young has not had the decades the established subspecialties have had to build fellowship infrastructure, role models, a research identity, and a settled place in the training hierarchy. Some of its underfilling is not rejection so much as a pipeline still under construction — which is a more hopeful reading, and a reason the numbers there may move faster than the others if compensation and visibility catch up.

The geography of need

There is one more structural mismatch, and it is spatial. The need these fields serve is not distributed the way applicants are. Kidney disease, chronic infection, addiction, and the care of the very old fall heaviest on rural and lower-income communities — precisely the places with the fewest physicians and the least appeal to graduates choosing where to build a life and service a mortgage. HRSA projects a national shortage of 187,130 full-time-equivalent physicians by 2037, with the shortfall most severe in nonmetropolitan areas and in primary care⁹; the AAMC’s own supply-and-demand modeling projects a shortfall of 13,500 to 86,000 physicians by 2036¹⁰. So the fields with the emptiest fellowship seats are, disproportionately, the fields whose work is needed in the places graduates are least willing to go. Underfilling and maldistribution are not two problems here; they are one problem seen from two angles.

Who walks through the open door

An open door is not the same as an empty one. The underfilled fellowships do not sit vacant — they are filled, disproportionately, by international medical graduates, and the pattern is the tell. In the 2026 Match, “non-U.S. IMGs were the second most frequently matched applicant group in all Internal Medicine specialties, representing 28.4 percent”² — and the concentration is sharpest exactly where U.S. graduates thin out. Nephrology, the third-least-filled field on the table, had the highest share of positions filled by non-U.S.-citizen international graduates of any specialty in the entire Match¹.

It would be a serious misreading to take that as a statement about the caliber of the physicians who fill these seats. The direction of causation runs through the same structure this whole piece describes: these are the fields with the most unfilled positions, and an applicant pool facing a steeper path into the most competitive specialties will rationally concentrate where the seats are open and the need is greatest. International graduates are not filling these fields because they are lesser physicians willing to take lesser jobs; they are filling them because the jobs are open, the work is essential, and many are willing to serve the underserved communities where the need is concentrated — a willingness the visa pathways into these fields both reward and, at times, require. The fields themselves say so. In the same statement lamenting its Match results, the Infectious Diseases Society of America asked policymakers for “immigration and training pathways that allow qualified international graduates to join the workforce,” alongside “loan repayment for ID physicians” — because “reduced funding, lower reimbursement and increasing clinical complexity make practicing ID especially challenging,” and because a strong ID workforce “is not optional; it is fundamental to patient care, outbreak response and the nation’s health security”¹¹. When a specialty society names international graduates as part of the answer to its own shortage, it is confirming the arithmetic: the door is open, the country needs it filled, and the graduates most willing to walk through it are increasingly the ones who trained abroad.

What this is not

Two guardrails, because a piece like this is easy to misread. First, none of this is a verdict on the medicine or on the physicians who choose it. Geriatrics, infectious disease, nephrology, and addiction medicine are among the most intellectually demanding and most needed fields in internal medicine; the graduates who enter them — U.S. and international alike — are choosing harder, lower-paid, more essential work, and the point of naming the structure that deters others is to honor that choice, not to diminish it. The empty seats are an indictment of the incentives, not of the field or of anyone in it.

Second, the structure is not a law of nature. Every deterrent named here is a policy choice with a lever attached: the wRVU weightings that undervalue cognitive care, the debt load a graduate carries into the decision, the loan-repayment and visa pathways that could tilt it the other way, the training exposure that determines which careers a resident ever sees practiced well. The fill rates are the sum of those choices, and they would move if the choices did — which is the more useful way to read the table than as a fixed ranking of which fields graduates “want.” They do not lack for wanting. They lack for a payment and training system that makes wanting them affordable. Where you land on which lever to pull is a policy judgment; that the levers exist, and that they — not the merit of the medicine — are what the Match is measuring, is what the data show.

References

1. National Resident Matching Program. (2026). Results and data: Specialties Matching Service, 2026 appointment year. https://www.nrmp.org/match-data/2026/02/results-and-data-specialties-matching-service-2026-appointment-year/ The 2026 fellowship Match. Overall, 15,358 positions were offered and 12,963 (84.4%) filled. Internal-medicine subspecialty fill rates: cardiovascular disease 100.0%, gastroenterology 99.5%, hematology and oncology 99.5%, rheumatology 99.0%, pulmonary disease and critical care 98.8%, endocrinology 97.7%, nephrology 66.5%, infectious disease 60.9%, geriatric medicine 38.9%. Nephrology had the highest share of positions filled by non-U.S.-citizen international medical graduates of any specialty in the Match. ↩₁ ↩₂
2. National Resident Matching Program. (2025). NRMP celebrates results for the 2025 Medicine and Pediatric Specialties Match [Press release]. https://www.nrmp.org/about/news/2025/12/nrmp-celebrates-results-for-the-2025-medicine-and-pediatric-specialties-match/ Geriatric medicine’s “fill rate declined to 38.9 percent”; adult infectious disease “filled at 60.9 percent, a decrease of 9.3 percentage points”; addiction medicine’s fill rate “declined by nearly three percentage points, from 66.7 to 63.8 percent” even as it added positions; and “Non-U.S. IMGs were the second most frequently matched applicant group in all Internal Medicine specialties, representing 28.4 percent.” ↩₁ ↩₂
3. Swartz, T. H., & Aberg, J. A. (2023). Preserving the future of infectious diseases: Why we must address the decline in compensation for clinicians and researchers. Clinical Infectious Diseases, 77(10), 1387–1394. https://academic.oup.com/cid/article/77/10/1387/7223213 ID physicians attending on general-medicine wards are “compensated $30 000–70 000 less than hospitalists who work fewer hours annually”; an ID physician practicing HIV primary care “can earn $30 000–60 000 less” than a generalist doing primary care; and AAMC data place ID mean salaries “$40 000–80 000 below those in other divisions” in departments of medicine. The work-RVU compensation model “poses challenges for nonprocedural specialties like ID.” ↩
4. Gurwitz, J. H. (2023). The paradoxical decline of geriatric medicine as a profession. JAMA, 330(8), 693–694. https://jamanetwork.com/journals/jama/fullarticle/2808221 Board-certified geriatricians fell from 10,270 in 2000 to 7,413 in 2022. In the 2022 fellowship match, “only 177 positions were filled of 411 positions offered (43%), the lowest percentage of all fellowships across 71 specialties of medicine.” Per the 2022 AAMC faculty salary survey, the median geriatrician salary was “9% lower than that of general internists and 14% lower than that of hospitalists,” and board certification “does not meaningfully influence job opportunities or practice activities.” (Subscription access may be required.) ↩₁ ↩₂
5. Doximity. (2025). Physician compensation report 2025 (2024 data). https://www.doximity.com/reports/physician-compensation-report/2025 Self-reported/modeled survey (over 37,000 responses, 2024 data). Infectious disease averaged $320,730, geriatrics $291,968, and endocrinology $290,606 — all at or below general internal medicine ($326,116) — even though each requires one to two years of fellowship beyond an internal-medicine residency. Endocrinology fills nearly every fellowship seat despite that same pay penalty, underscoring that compensation alone does not drive the fill gap. ↩
6. Association of American Medical Colleges. (2024). Debt, costs, and loan repayment fact card, class of 2024. https://students-residents.aamc.org/media/12846/download 71% of the class of 2024 graduated with education debt; among those with debt, the median was $205,000 and the mean $212,341. The 2024 first-post-MD-year (first-year resident or fellow) median stipend was $65,100. ↩₁ ↩₂
7. Substance Abuse and Mental Health Services Administration. (2023). Waiver elimination (MAT Act). https://www.samhsa.gov/substance-use/treatment/resources/mat-act Section 1262 of the Consolidated Appropriations Act, 2023 eliminated the federal “X-waiver” and its patient caps. Any clinician holding a current DEA registration that includes Schedule III authority may now prescribe buprenorphine for opioid use disorder, subject to state law. ↩
8. American Board of Medical Specialties. (2016). ABMS officially recognizes addiction medicine as a subspecialty [Press release]. https://www.abms.org/newsroom/abms-officially-recognizes-addiction-medicine-as-a-subspecialty/ The ABMS board voted to recognize addiction medicine at its October 2015 meeting; the recognition was announced March 14, 2016. Sponsored by the American Board of Preventive Medicine, it is multidisciplinary — open to physicians certified by any of the 24 ABMS member boards — and among the youngest recognized subspecialties in American medicine. ↩
9. Health Resources and Services Administration, Bureau of Health Workforce. (2024). Physician workforce: Projections, 2023–2038. https://bhw.hrsa.gov/sites/default/files/bureau-health-workforce/data-research/physicians-projections-factsheet.pdf HRSA projects a national shortage of 187,130 full-time-equivalent physicians by 2037, with shortages most severe in nonmetropolitan areas and in primary care disciplines. ↩
10. Association of American Medical Colleges. (2024). The complexities of physician supply and demand: Projections from 2021 to 2036. https://www.aamc.org/media/75236/download Projects a shortfall of 13,500 to 86,000 physicians by 2036. ↩
11. Infectious Diseases Society of America & Pediatric Infectious Diseases Society. (2025). IDSA and PIDS statement on the 2025 Infectious Diseases fellowship Match results. https://www.idsociety.org/news--publications-new/articles/2025/idsa-and-pids-statement-on-2025-infectious-diseases-fellowship-match-results/ “Reduced funding, lower reimbursement and increasing clinical complexity make practicing ID especially challenging.” The societies call for “loan repayment for ID physicians” and for “immigration and training pathways that allow qualified international graduates to join the workforce,” adding that a strong ID workforce “is not optional; it is fundamental to patient care, outbreak response and the nation’s health security.” ↩

Match figures are for the 2026 appointment-year fellowship Match and change every year; compensation figures are survey- and model-based and shift over time. This analysis examines incentives and policy, not any individual, program, institution, or the patients these fields serve; it is educational and is not career, financial, or policy advice.