Connecting Medical Academy, Organizations, and Industry

No single institution controls the physician as a profession; dozens do, partially, and none is accountable for the whole.

The length of the list is itself the diagnosis. The medical academy, meaning schools, teaching hospitals, and their faculties. The major professional societies. Accreditation bodies for undergraduate and graduate medical education. State licensing boards. Specialty certification and recertification boards. Health system leaders and networks. Health insurers. Governmental regulatory and funding bodies. Medical and philanthropic foundations. And the pharmaceutical, biotechnology, device, and healthcare delivery companies whose products and platforms increasingly define the daily texture of practice.

Each acts rationally within its own remit. The aggregate is a system in which no actor can authorize the change this report describes, and every actor can veto it. What is needed is coordinated evolution, beginning with exploratory pilots on the way to a rethinking of physicianship. A new supervening authority will not be created and would not be trusted.

There are precedents: engineering reorganized its accreditation around demonstrated outcomes instead of prescribed course content. Aviation built a shared safety culture across fiercely competitive carriers, manufacturers, and regulators. Architecture restructured education, supervised practice, and licensure into a modular sequence with multiple entry points. Each achieved substantive change without a controlling authority, over roughly a decade, by the same mechanism: a small coalition of credible institutions demonstrated an alternative, measured it publicly, and let the accrediting and certifying bodies follow evidence.

On that model, this report calls for two reforms. A new approach to accreditation, shifting from process compliance toward demonstrated competency outcomes, with room for programs organized around identities and roles in place of a standard course inventory. A new approach to certification, moving from point-in-time examination of static knowledge toward continuous, portfolio-based demonstration of capability, including competencies this report has argued are now core: supervision of AI systems; communication of probabilistic risk; cross-disciplinary collaboration; and discovery literacy.

Both must be phased, from parallel-track pilots under existing accreditation, through formal experimental categories with reciprocity for graduates, to standards rewritten around what the pilots have shown. The precondition is that the profession decide what it wants before others decide for it.

A recommendation addressed to accreditors and certifiers needs someone to move first, and the precedents indicate who. In each case, the sequence began with a small number of institutions credible enough to be copied, acting jointly rather than singly, running the alternative openly and publishing what it produced. Five or six schools with that standing could bootstrap the alternative, opening parallel-track programs under existing accreditation, agree in advance on common outcome measures so the results are comparable, and commit to publishing them whether or not they flatter the experiment. Nothing in that sequence requires permission from a certifying board, which is what makes it doable.

The diagnosis is not idiosyncratic, and three bodies of work reach it independently and are illustrative. The National Academy of Medicine’s 2024 special publication The State of the U.S. Biomedical and Health Research Enterprise finds an enterprise constrained less by scientific limits than by structural ones, with funding streams and research agendas set in isolation, where siloed funding by its nature prohibits the collaboration across agencies and areas of expertise that complex health problems demand. Its remedies are coordination rather than money: a national advisory body; streamlined federal funding; and an explicit expansion of convergence science to break down silos.⁸⁰

That word carries an institutional literature. The National Research Council's Convergence: Facilitating Transdisciplinary Integration of Life Sciences, Physical Sciences, Engineering, and Beyond, together with MIT's The Third Revolution, argue that convergence is a structural merging of fields, and so demands institutional redesign rather than goodwill: shared facilities, joint appointments, funding that does not force a project into a single disciplinary home, and promotion criteria that do not punish the investigator working between departments.⁸¹

The third critique comes from outside biomedicine. Einer Elhauge's The Fragmentation of U.S. Health Care: Causes and Solutions holds that fragmentation is the predictable product of how decision rights and payment are divided, not an accident to be deplored. No actor internalizes the cost of poor coordination, so none invests in repairing it. Exhortation is not a strategy, and remedies must alter who decides and who bears consequences.⁸²

The capabilities described in Part I are almost all integrative ones, among them theranostics, digital twins, learning health systems, exposomics, physical AI, and N-of-1 inference, and the value of each scales with the breadth of what it may connect. A digital twin built on one institution's data models one institution's patients. A learning health system that cannot see past its own walls learns a local artefact and calls it evidence. Fragmentation is thus the limiting factor on the technologies this report projects.

The Physician's Part in This

Consortia are described in the language of charters and data-use agreements, which obscures what determines whether one works: who populates it. The future physician is the natural connective tissue, and the reason lies in where physicians already stand. The physician is the one actor routinely standing inside several of these institutions at once, as clinician, faculty member, investigator, sometimes advisor to a company or foundation, and always the person the patient believes is theirs.

That position widens what will be asked of physicians. Physicians will increasingly be asked to work not only as clinicians and scientists but as institutional reformers, as designers of the systems through which care is delivered, and as advocates for organizational models that do not yet exist. None of those is a departure from clinical work, and each is a condition of it, since the arrangements that decide what a physician may do are designed objects with authors.

The four identities of Part II map onto four bridging functions. The technoscientific physician carries the clinical problem into the engineering room and a realistic account of the engineering back out. The consilient physician performs the translation without which an ensemble of expert opinions is but parallel monologues. The transcendent physician carries the patient's interest into rooms the patient is not in. The global physician supplies the equity and validation discipline.

The traffic runs both ways. A connected system supplies what no single institution can: frontier tools and compute beyond any department's budget; cohorts large enough that a rare phenotype becomes a study; validation against populations elsewhere; and career paths that let a physician work at the frontier without leaving the academy. The physician creative of Chapter VII, otherwise dependent on the luck of employer, can place an observation into a federated cohort and make it count.

Six Precedents

Six standing arrangements illustrate the model, two from Japan, two from the United States, and two from Europe, and within each pair one with commercial partners at the table and one without.

Japan, with industry: T-CiRA. The Takeda-CiRA Joint Program for iPS Cell Applications paired Kyoto University's Center for iPS Cell Research and Application with Takeda across a decade from fiscal 2016, on twenty billion yen and facilities at Takeda's Shonan Research Center, where roughly a hundred researchers from both institutions worked side by side under Shinya Yamanaka's leadership. Aimed at a recognized national difficulty in moving university science into development at speed, it answered by co-location rather than contract.⁸³

Japan, without industry: the National Center Biobank Network. Established in 2011, the NCBN federates the biobanks of Japan's six National Centers for advanced and specialized medicine. Each retains its own collection and specialty. What is shared is an open-access catalog through which investigators search across all six by disease, sample type, patient characteristics, and genomic data, amounting by March 2022 to over four hundred thousand bioresources from some hundred and twenty thousand patients. Cooperation without merger; interoperability without central custody.⁸⁴

The United States, with industry: the Accelerating Medicines Partnership. Launched in 2014 through the Foundation for the National Institutes of Health, AMP joins the NIH and FDA with ten or more biopharmaceutical companies and non-profit partners to identify and validate biological targets, first in Alzheimer's disease, type 2 diabetes, rheumatoid arthritis, and lupus, later in Parkinson's disease, schizophrenia, and ALS. Two choices do the work: the collaboration is precompetitive, and contributed data become public by prior agreement.⁸⁵

The United States, without industry: the CTSA Consortium. Begun in 2006 and housed since 2011 at the National Center for Advancing Translational Sciences, the Clinical and Translational Science Awards fund more than sixty hubs mandated to re-engineer the national translational enterprise. Its transferable lesson is methodological. From 2016, sixty-four hubs implemented common metrics, among them review-board duration, trainee career outcomes, and publications and follow-on funding from pilot awards, the first shared performance measurement across an NIH consortium.⁸⁶

Europe, with industry: IMI and the Innovative Health Initiative. The Innovative Medicines Initiative launched in 2008 between the European Commission and the European pharmaceutical industry association, with two billion euros and a second phase of some three and a quarter billion across roughly two hundred projects. Its organizers describe the central achievement as collaboration among parties who do not normally collaborate, including companies large and small, regulators, patient organizations, and public research institutions, made possible by a neutral platform on which all partners carry the same rights and responsibilities.⁸⁷

Europe, without industry: the European Reference Networks. Created in 2017 under the cross-border healthcare directive, the twenty-four ERNs link more than nine hundred specialized units in over three hundred hospitals, each covering a family of rare or complex conditions on the principle that the knowledge should travel rather than the patient. A first independent evaluation of the networks and their eight hundred and thirty-six providers found the ecosystem working as intended, while flagging uneven integration with national systems and thin institutional recognition of network activity.⁸⁸

Four properties are common to all six and amount to a design specification. Each is a federation rather than a merger. Each has an explicit neutral convener no participant controls. Each defines in advance what is shared and what is not. And each publishes, which is what allows the arrangement to be judged.

The specification earns its keep by being applied, so apply it to two proposals this report has already made. A shared curriculum commons would need three things. The first is a neutral convener that no school controls, which points to an association of medical colleges or a consortium constituted for the purpose, not to the strongest institution in the group. The second is a definition agreed in advance of what is shared, which is modules, assessments, and outcome data, not admissions or reputation. The third is published results, without which no participant can tell whether the arrangement is worth its overhead.

The certifying function for health data proposed in Chapter IV fails the specification as usually imagined, since a body convened by industry or housed inside a single regulator has no neutral convener at all. Constituted as a federation of societies, systems, and boards with regulators present, defined scope, and published determinations, it satisfies all four. What the test mostly reveals is which proposals have not yet decided who convenes them.

A seventh arrangement solves a problem the other six do not address: the latency of a shared standard. cIMPACT-NOW, the Consortium to Inform Molecular and Practical Approaches to CNS Tumor Taxonomy, was founded by David Louis of Massachusetts General Hospital and Harvard Medical School. Louis was also primary editor of successive editions of the World Health Organization's classification of central nervous system tumors.

The difficulty it was built to solve is one of timing. A WHO classification is the diagnostic vocabulary of a whole field, revised on a cycle of roughly a decade, while the molecular findings that change how a tumor should be classified arrive continuously. Between editions, a diagnosis can be known to be wrong and still be the standard. The consortium answers with a standing international panel that issues numbered updates in the interval, each vetted and published, which the next formal edition absorbs. Its product is a common currency, neither data nor funding, and the design generalizes well beyond neuro-oncology. Wherever a field depends on a shared classification, someone has to own the gap between revisions, or practice will lag the science by exactly the length of the publishing cycle.

Two organizations warrant a closer look, marking the ends of the connective spectrum. The Future of Health is a standing international community of chief executives and senior leaders of health systems, co-founded by Charles N. Kahn III of the Federation of American Hospitals and Eyal Zimlichman of Sheba Medical Center, with members across some thirteen countries, Sheba, Karolinska, Charité, Mount Sinai, Cedars-Sinai, Duke, Stanford Health Care, SingHealth, and Albert Einstein in São Paulo among them.

It exists to create a shared vision for care delivery for 2035 and beyond, developing annual research topics through multi-sector workgroups and publishing them with an academic partner, first the Duke-Margolis Institute and more recently through a four-year partnership with Stanford Health Policy. What distinguishes it is the layer it reaches: the people who operate delivery systems, below ministries and specialties.⁸⁹

FasterCures, a center of the Milken Institute, occupies the other end. Founded in 2003 as the Center for Accelerating Medical Solutions, it grew from the recognition that the bottlenecks between science and therapy were systemic and cross-disease, not specific to any one condition, and its niche was convening. A 2004 retreat brought together, for many of them the first time, disease-specific patient organizations that had operated in isolation; from it came TRAIN, the Research Acceleration and Innovation Network, founded in 2005 to help disease foundations take strategic approaches to funding research and to build networks among themselves. Its distinctive move is to treat the patient organization as an institutional actor rather than as an audience.⁹⁰

Disease-targeting foundations need separate treatment, as the most under-described actor in the list this chapter opened with. A foundation funded by patients and their families works on a lifetime horizon rather than a grant or product cycle, because its constituents live with the disease and their children may inherit it. Being disease-specific, it funds the unglamorous connective infrastructure that neither a promotion case nor a business case rewards: registries; natural history studies; biorepositories; standardized and patient-reported outcome measures; and the trial networks that make a scattered rare-disease population addressable. It carries moral standing that lets it convene competitors, and it holds the one asset every other party requires and none controls: the patients themselves.

The record shows the consequence: the Cystic Fibrosis Foundation's venture-philanthropy model, pairing a national registry and a therapeutics development network with direct investment in industry programs; the Michael J. Fox Foundation's progression-markers initiative; and the Multiple Myeloma Research Foundation's genomic study, which made a multi-site resource common property. In each the foundation acted less as grant-maker than as consortium builder.

The research enterprise makes the same case more sharply, because there the stakes include the survival of the academic contribution itself. The ecosystem has become atomized and corporatized, with entangled power vectors that place academic research at a structural disadvantage. Frontier capability, meaning the largest compute clusters, the most complete proprietary datasets, and the deepest engineering teams, increasingly sits inside firms rather than universities, and the terms of academic access are set by parties with different objectives.

This is not an argument against industry, whose capabilities academia cannot replicate. It is an argument about asymmetry, and about what is lost if the only well-resourced research is research with a commercial sponsor: the unpatentable intervention, the negative result, the study whose finding is inconvenient.

Coordination is therefore needed at two levels. Nationally: shared research infrastructure no institution can afford alone; federated data architectures permitting joint learning without pooling raw records; common standards for provenance, consent, and model validation; pooled negotiating position with commercial platform providers; and funding vehicles for the translational middle identified in Chapter II. Globally: interoperable data standards across jurisdictions; coordinated validation of models across populations; equitable partnership structures locating leadership where the work is done; and shared access to frontier tools, which Exhibit 7.1 identified as the single most consequential democratizing move. Consortium is the only posture from which the academy can set terms.

Barriers and Risks

It would be dishonest to present consortium-building as straightforwardly good. Transaction cost and time: charters, intellectual-property schedules, and multi-institutional scheduling consume the scarcest input in academic medicine. Every precedent above is measured in decades, and that is no accident.

Free-riding: federations struggle where contribution is voluntary and benefit common, and the mitigant is public measurement. Capture: where one party is much larger, partnership quietly becomes procurement, and the answers visible in the precedents are structural, namely a neutral platform binding all partners equally and strict precompetitive scoping. Good intentions are not among them. Data governance: federation reduces but does not remove the problems of provenance, secondary use, and re-identification.

Agenda-setting: industry participation is legitimate and irreplaceable, but the tractable question with a product at the end is likelier to be posed. Published agendas and a governance table including patient organizations are the correction. Extraction: international consortia can reproduce the pattern Chapter X warns against, with data flowing one way and authorship the other. Homogenization: common models and agreed measures make joint work possible and also narrow what can be noticed, and such systems succeed when they amplify diversity and fail when they homogenize it. Fragility: consortia rest on a few individuals and time-limited funding, so succession, institutional credit, and a durable funding model belong in the design.

The relevant comparison is never a consortium against an ideal. It is a consortium against the counterfactual: a dozen institutions each optimizing locally, which is the present arrangement, and which this report has identified as the single greatest obstacle to everything it proposes.