Not Every Hospital Program Should Make Money: But every loss should be deliberate, funded, and measured.
In my last essay, I used FIFA and the World Cup to argue that mission requires margin. The harder question is what leaders should do with it.
Not every hospital program should make money. Some should. Some never will. Some should close.
That last sentence is the one healthcare leaders avoid. We are comfortable defending a program because its purpose is worthy. We are less comfortable asking whether the program still works, whether another model would better serve patients, or whether history is its only strategy.
I have spent much of my career building clinical programs, research, education, international care, and the systems that support them. I have also sat in the rooms where those programs are reduced to a line on a financial report. Once the slide appears, the language shifts. Green services become good services. Red services become problems. The color feels conclusive. I have never trusted that slide.
A margin report is essential but incomplete. It can show where money is earned and lost, but it cannot tell us whether the loss reflects poor design, a deliberate mission subsidy, or an investment whose value will appear later. Leaders must make that distinction. The spreadsheet cannot make it for them.
Here is the controversial lesson: a noble mission does not rescue a flawed operating model. Mission should make us more disciplined, not less. Sometimes the most responsible act is to fund a program that will never support itself. Sometimes it is to close one that everyone loves.
The Red Column
Healthcare organizations often divide service lines into two columns. The green programs generate margin, while the red programs consume it. Then the budget discussion begins as if the colors have already answered the strategic question. They have not.
A negative margin tells us that a program costs more than it directly earns, but not why. Billing may be weak. The schedule may waste expensive space. Leadership may be absent. Staff may spend hours repairing a process that should have been designed correctly from the start. Cases may be canceled because the patient entered through the wrong door, the authorization arrived late, or the equipment was never ready. That is an operating failure.
Another program may lose money because the payment system undervalues the care. Pediatrics is the clearest example. In the United States, Medicaid covers nearly four in ten children, while its fee-for-service physician rates average about two-thirds of Medicare rates. A pediatric service can remain busy, deliver excellent care, and still lose money.
But Medicaid is only the American version of a wider problem. In public, insurance-based, and mixed systems, children’s services often require specialist teams, family support, dedicated equipment, and constant readiness. Much of their value appears years later and outside the hospital budget. The same problem affects psychiatric emergency care, community diabetes programs, trauma readiness, and language access.
That is not failure. It is a deliberate subsidy for work the institution, or the health system, has decided must exist.
A third program may be in the red because it is building capacity the institution will need next: a registry, a fellowship, a new surgical service, a better clinical front door, an international partnership, or the shared data required to use artificial intelligence safely.
That is an investment, but only if leaders define what it should produce, who owns it, how long it should take, and when they will judge the results. Without those conditions, “investment” becomes a convenient label for an expense no one wants to challenge.
Failure, subsidy, and investment can all appear in red. They require different decisions. Failure needs repair or closure. Subsidy needs an explicit source of support. Investment needs a time frame, milestones, and someone with enough authority to protect it through the first difficult quarter.
Calling everything mission protects poor management. Calling everything a loss destroys useful work.
Strategy begins by telling them apart.
This Is a Global Problem, Even When the Payment Rules Differ
Funding models vary from country to country. My career has spanned Mexico, the United Kingdom, and the United States. Public budgets, national insurance, commercial insurance, philanthropy, employer funding, and direct payment create different pressures. The financial language changes. The allocation problem does not.
In the United States, for example, Medicare covers only part of the direct cost of graduate medical education. Its separate adjustment for teaching hospitals also recognizes duties such as treating more complex patients and maintaining standby capacity for trauma and burn care, according to the Association of American Medical Colleges.
Other countries fund teaching and readiness through different channels. None has found a way to make staff, training, equipment, data, and time appear without cost.
Every system therefore makes the same choice, whether it admits it or not: which work will be supported even when the revenue attached to the patient does not cover it?
A system can make that choice openly. It can also bury it in an annual budget and pretend that no choice was made.
The second approach is common. It is also why supposedly essential programs disappear when the margin tightens.
Two Systems, One Lesson
At NYU Langone Health, I learned that a program can succeed clinically and still carry limited financial weight within a large academic system.
Our pediatric orthopedic division grew. More patients came. Referring physicians trusted us. Clinical care, teaching, and research improved. Yet pediatric orthopedics remained small relative to cancer, transplant, cardiovascular care, adult reconstruction, imaging, neuroscience, and the many other demands on capital.
That was not a criticism of the program. It was a lesson in scale. Clinical importance does not automatically produce financial influence. Physician leaders must understand both.
At Shriners Children’s, I learned the opposite risk.
Shriners has sustained a clear charitable mission for more than a century. Philanthropy and endowment income protect care from some of the pressures created by fee-for-service medicine. But donated money is not free money. It carries an obligation. Someone gave it because they believed the institution would use it well.
That makes discipline more important, not less. A philanthropic model does not excuse weak data, outdated processes, idle capacity, or programs that persist without clear results. Waste does not become acceptable simply because the money came from a donor rather than an insurer.
The two systems taught me the same lesson from opposite directions. At NYU Langone, clinical importance could not replace a sound financial case. At Shriners Children’s, available resources could not replace discipline.
Mission needs a financial model. The financial model needs a reason.
The Next Mission Investment May Be Invisible
The next essential investment may not be another clinic, operating room, or building. It may be the shared data and technology that help every existing service work better.
That investment is difficult to defend because its first appearance in the budget is often unattractive. Clean data cost money. System integration costs money. Redesigning intake, scheduling, prior authorization, follow-up, supply management, and the revenue cycle takes time from people who already have full workloads.
The return does not arrive because someone purchased software. It arrives only when the work changes.
Health systems have repeated the same error for years. They buy a tool, run a pilot, count the pilot as progress, and leave the operating model untouched. The tool sits alongside the old process. Staff now manage both. The institution has added technology and created more work.
Artificial intelligence gives us the chance to stop doing this. It also allows us to repeat the same mistake faster and across the entire institution.
A March 2026 McKinsey report on health-system operating models surveyed 100 U.S. health-system leaders. Seventy percent ranked operating-model redesign among their five highest priorities, while 81 percent said their current model was neither effective nor efficient. The report also identified inconsistent use of technology and AI as part of the problem, not a cure sitting outside it.
That finding matches what I have seen. Most struggling systems do not lack effort or tools. They lack a clear account of how work gets done, who owns each decision, where patients get stuck, and which measures show whether the process works.
Adding AI before answering those questions does not modernize the system. It automates its confusion.
Technology Is Not a Strategy
A hospital can now buy ambient documentation, automated scheduling, denial management, patient messaging, supply forecasting, image analysis, and clinical decision support from multiple vendors.
Each may solve a real problem. Together, they can also create another set of disconnected systems, contracts, dashboards, and handoffs.
The wrong question is: Where can we use AI?
That question produces a list of tools.
The useful question is: Which part of the operating model must change, and what role should technology play in that change?
McKinsey makes this point directly in The Health System CEO Imperative. Health systems have focused on isolated tools. Greater value comes from choosing one or two important areas and redesigning the work from beginning to end.
A second report, The Real Future of Work in Healthcare, states the risk plainly: AI placed on top of a broken system can increase the inefficiency already present.
Consider the clinical front door. A weak front door sends patients to the wrong clinic, hides urgent cases among routine referrals, loses outside records, repeats imaging, and forces specialists to spend time sorting demand that should have been organized before the visit. Installing a chatbot does not fix that.
A serious AI-supported front door would begin with a redesigned pathway. What information is required? Which warning signs change urgency? Who owns the triage rule? Which cases require a specialist? Which can be handled elsewhere? What happens when the system is uncertain? How does the clinician view the recommendation, correct it, and improve the system? Who confirms that the patient reached the next step?
Only then does the technology have a role. It can organize records, identify missing information, flag risk, route demand, prepare the clinician, and track whether the patient completed the pathway. It should reduce delays before the visit and protect time for judgment during it.
The same logic applies to operating rooms, supply chains, the revenue cycle, and follow-up. The goal is not to speed up a single task. It is to remove the delays, duplication, and uncertainty that keep the entire pathway from working.
AI also needs limits. The National Academy of Medicine’s AI Code of Conduct calls for accountable, safe, reliable, and human-centered use. In practical terms, that means a named owner, visible performance metrics, a way to detect changes in performance, a clear path for human review, and the authority to stop the system when it behaves badly. A vendor’s assurance is not governance.
Cross-Subsidy Is a Choice
A service with a strong contribution margin generates more revenue than the direct cost of delivering care. That surplus helps pay for shared hospital functions, reserves, capital, and services that the institution considers essential, even though they cannot cover their own costs.
There is nothing shameful about that. If the institution believes both services are needed, reallocating resources from one to the other is not a strategic distortion; it is strategy.
The danger lies in keeping the subsidy invisible. When no one defines the subsidy, the supported program may stop tracking its costs. The service generating the surplus can be weakened without anyone acknowledging what else depends on it. A new leader can remove a red line and later discover that the line contained access, training, coordination, or clinical readiness that had never been measured.
Technology investment faces the same problem. A dozen small pilots may each seem affordable, while their combined licensing, integration, security, training, and support costs remain hidden. A shared data system may seem expensive even though it can replace several disconnected products and support the entire institution.
Poor accounting can make fragmentation look cheap and shared infrastructure look wasteful.
Leaders must name the trade-off. What are we supporting? Why? For how long? What is funding it? What result are we buying?
If the institution cannot answer these questions, it does not have a strategy. It has inherited a collection of old decisions and new subscriptions.
The Mission Investment Statement
Every program receiving institutional support should have a one-page Mission Investment Statement. So should every major technology program.
This should not become another committee document. It should force a decision stated in plain language.
1. What are we buying?
State the purpose precisely. “Supporting pediatrics” is too vague. Are we buying access for children with complex disease, trauma readiness, training capacity, care coordination, a research registry, or a referral pathway that keeps urgent patients from waiting behind routine cases? For technology, specify the work that will change. Do not describe the software. Describe the work.
2. Is this failure, subsidy, or investment?
A structural payment shortfall is different from poor scheduling. A five-year effort to build new capacity is different from a program that has missed its targets for a decade. The category determines the response.
3. What does it truly cost?
Include staff, space, integration, maintenance, security, training, support, and the time clinicians and operators must spend changing the process. Do not pretend that donated money is free or that software costs end with the license fee.
4. Where will the support come from?
Name the source: another service line, unrestricted philanthropy, a restricted gift, public funding, endowment income, an industry partnership with proper safeguards, or operating funds. “The hospital” is not a funding plan. It is a refusal to identify the trade-off.
5. What work must change?
This is the question most technology proposals avoid. Which steps will disappear? Which roles will change? Who owns the exceptions? What old tool or process will stop? If nothing stops, the institution has likely added cost rather than capacity.
6. What result should the investment produce?
Use measures that align with the purpose. Access may require waiting time, referral completion, and proper routing. A navigator may reduce cancellations and patients lost to follow-up. AI-enabled scheduling may improve use of appointment slots, reduce manual steps, and shorten the time between referral and care.
Financial results belong on the page, but so do quality, access, staff time, and whether patients complete their care.
7. What are the clinical and ethical limits?
State where human review is required, how errors will be detected, which data may be used, who can override the system, and what would trigger suspension. Clinical accountability cannot be outsourced through a software contract.
8. Who owns the decision, and when will it be reviewed?
Name one accountable leader. Set the time frame. State what evidence would justify expansion, redesign, partnership, or closure. Programs survive with vague ownership because no one has the authority to stop them. Good investments die for the same reason because no one has the authority to protect them.
Some Programs Should Close
There is a risk in this argument. Clinicians may use it to justify every program they value, and technology leaders may do the same for every platform they have already purchased, which misses the point.
Some programs should close. Some pilots should end. A service may have low demand, poor outcomes, excessive costs, weak leadership, or no clear place in the institution. Another hospital may provide the service more effectively. A partnership may make more sense than ownership.
The patient need may be real while the current program remains the wrong answer.
Likewise, an AI tool may work technically and still fail strategically. It may save minutes in one department while adding work elsewhere. It may produce a convincing demonstration but never connect to the medical record, the scheduling system, the people doing the work, or the measures leaders use.
The correct decision is not another pilot; it is to stop. Closing a weak program is not always a betrayal of mission. It can free staff, space, attention, and capital for work that delivers greater value. Ending a failed technology project is not resistance to innovation. It is evidence that someone is paying attention.
The test is simple: knowing what we know now, would we build and fund this program in this form today?
If the answer is yes, support it deliberately.
If the answer is no, history and sunk cost should not become strategy.
Physicians Have to Enter the Room
Physicians often arrive at budget meetings with clinical stories, while finance arrives with numbers. Both are incomplete. Only one usually controls the decision.
Physician leaders do not need to become accountants. They do need to understand contribution margin, payer mix, cost per case, budget variance, restricted and unrestricted funds, and the source of the subsidy supporting their work.
They also need to understand enough about data, workflow, and AI to challenge a technology proposal that lacks a clinical owner and an operating plan.
Financial ignorance does not protect patients from finance. Technical ignorance will not protect them from technology. Both simply remove physicians from the decisions that shape care.
Finance leaders have a matching duty. A CFO should distinguish between expenses, subsidies, and investments rather than treating every dollar without immediate revenue as waste. Technology leaders should show how the work will change, not just what the product can do. The board should protect long-term capacity even when the quarterly report cannot yet reflect its full value. The chief executive should make the trade-offs explicit. The physician leader should keep the investment tied to clinical reality.
That is governance. It is also the difference between praising a mission and funding one.
Margin Creates the Choice
A hospital must generate enough surplus to maintain staff, equipment, reserves, data, training, and the capacity to respond when conditions change. Without that room, every difficult quarter threatens the work that does not bill well but matters greatly.
The harder question is what the institution does with the margin once it has it.
It can chase profitable volume without asking whether the care is needed. It can preserve cash while research, teaching, access, and coordination weaken. It can buy technology without changing work.
Or it can use financial strength to fund services the payment system undervalues and build capacity the institution will need before the crisis arrives.
That choice should not be buried in an overhead allocation, a philanthropic fund, or a technology budget. It should be named, funded, measured, and governed.
Not every hospital program should make money.
But every loss should be a decision.
Margin creates the ability to choose. Mission is the discipline to choose well.
That is the central argument of my book, Mission Requires Margin: A Physician's Guide to Building Healthcare Organizations That Work. Margin is not the mission, nor is it proof of virtue. It is the financial room needed to protect useful work, stop failed work, and build what patients will need next.
Selected Sources
1. KFF. Medicaid and Children’s Health: 5 Issues to Watch Amid Recent Federal Changes.
2. Medicaid and CHIP Payment and Access Commission. Provider Payment and Delivery Systems.
3. Association of American Medical Colleges. Graduate Medical Education: Payments to Teaching Hospitals.
4. McKinsey & Company. What It Takes to Build a High-Performing Health System Operating Model. March 2026.
5. McKinsey & Company. The Health System CEO Imperative: Turning AI’s Promise into Performance. June 15, 2026.
6. McKinsey & Company. The Real Future of Work in Healthcare. July 2, 2026.
7. National Academy of Medicine. Health Care Artificial Intelligence Code of Conduct. 2025.