Why Is U.S. Healthcare So Expensive—and Can AI Help?
Noah Vandal and Dr. Joseph Yoon examine U.S. healthcare costs, hospital spending, rural access, prescription drugs, prevention, and AI.
Why U.S. Healthcare Costs So Much - and Where the Money Goes | Ep. 6.
This episode asks a bigger question than what is on one medical bill
Episode 1 of the AI and Healthcare Podcast looked at why an individual healthcare bill can be so hard to understand. Episode 6 steps back and asks a broader question: why does the United States spend so much on healthcare in the first place? In this conversation, recorded May 25, 2026, Noah Vandal and Dr. Joseph Yoon examine what the country receives for that spending and why the answer is complicated. Modern hospitals can deliver specialized cardiac care, intensive monitoring, advanced imaging, and treatments that were not widely available several decades ago. Those capabilities can save lives, but they require trained people, expensive facilities, sophisticated equipment, and continuous readiness. The episode also looks at the parts of the system that do not translate as clearly into better health: administrative growth, uneven competition, pressure to use costly equipment, fragmented access, and a model that often responds more aggressively after people become seriously ill than before. The important distinction is between cost and value. A service can be expensive and still be clinically valuable. A larger budget can also be used inefficiently. Understanding U.S. healthcare spending requires asking both what the money buys and whether it reliably improves outcomes.
U.S. spending is high even before debating whether it is wasteful
The Centers for Medicare & Medicaid Services reports that national health expenditures reached $5.3 trillion in 2024, or $15,474 per person. Hospital expenditures accounted for about $1.63 trillion, while prescription-drug spending reached $467 billion. Those figures help ground the episode's opening estimate of roughly $15,000 per person. They also show why one simple explanation cannot account for the full total. Healthcare spending is distributed across hospitals, physicians and clinical services, prescriptions, insurers, public programs, long-term care, and other categories. International comparisons add another layer. OECD data shows that the United States spends substantially more per person than other member countries while producing a mixed performance across access, quality, and health outcomes. That does not mean U.S. medicine fails at everything, or that every additional dollar is waste. It means the relationship between spending and population health is not consistently strong enough to assume that higher cost equals higher value. The useful question for healthcare leaders is therefore not merely, “How do we cut spending?” It is, “Which spending improves health, which spending creates avoidable friction, and what changes can preserve care while reducing preventable harm or duplication?”
Modern hospital capability has a real price
Dr. Yoon describes how hospital medicine has changed during his career. A patient with a heart attack may now receive rapid imaging, catheter-based treatment, specialized nursing, intensive monitoring, and coordinated follow-up. Maintaining those services requires staff with different expertise, equipment that must be purchased and maintained, and facilities that remain ready even when demand varies. That fixed-cost structure matters. An MRI scanner, intensive-care unit, emergency department, or specialty team cannot be evaluated only by the marginal cost of the next patient. The organization must also finance the capacity to have the service available when the patient arrives. At the same time, an expensive capability can create pressure to keep it in use. The episode raises the possibility that organizations may perform more tests after making large capital investments. That should not be turned into a blanket claim that advanced imaging is unnecessary. The better question is whether each use is clinically appropriate, whether results change management, and whether a lower-cost option could provide the information needed. AI enters this part of the discussion as a possible support tool. It may help clinicians interpret certain images or prioritize findings, but performance depends on the device, clinical task, patient population, workflow, and evidence. FDA authorization for an AI-enabled medical device is an important regulatory signal; it is not proof that every deployment will reduce total healthcare spending.
Administration and consolidation create tradeoffs, not one easy villain
Independent practices, imaging centers, and hospitals have increasingly become part of larger systems. Integration may improve coordination, resource sharing, purchasing, access to capital, and the ability to support specialized teams. It can also add management layers or weaken competition in a local market. The Federal Trade Commission has summarized evidence that some forms of hospital consolidation and state-protected merger arrangements can raise prices and reduce competition. The effect of a particular transaction still depends on the market, the organizations involved, and how care is structured afterward. That nuance matters because “administration” includes both avoidable bureaucracy and work that a complex healthcare system genuinely needs: billing, scheduling, compliance, cybersecurity, quality reporting, care coordination, staffing, and safety processes. Automation should target measurable friction rather than treating every administrative role as waste. For AI projects, that means starting with a specific workflow. A tool might reduce repetitive documentation, improve access to approved information, or make follow-up more reliable. If it instead creates more alerts, more review work, or another disconnected system, it can add cost even when the model itself is inexpensive.
Rural care exposes the economics of readiness
Rural providers face a difficult version of the same fixed-cost problem. A smaller community may need emergency care, imaging, transportation, and trained staff but have fewer patients across whom to spread those costs. When local capability is unavailable, a transfer to a distant specialty center can add time, risk, and substantial expense. The answer is not to assume every rural hospital should duplicate every urban specialty service. Regional partnerships, telemedicine, transportation planning, and carefully selected technology may help local clinicians access expertise without recreating an entire tertiary-care center. AI could contribute by supporting specific imaging tasks, helping organize patient information for a remote specialist, or reinforcing follow-up instructions after a patient returns home. Those uses still depend on reliable infrastructure, clinical oversight, and clear responsibility. Technology cannot replace the people, broadband, transportation, and facilities that rural care requires. Current federal investment in rural-health transformation reflects the scale of the challenge. The meaningful outcome is not whether a community buys more technology; it is whether patients gain timely access to safe care and avoid preventable deterioration or transfers.
Prescription-drug policy has changed since part of the history discussed
The episode connects prescription-drug spending to the growth of chronic disease, the increasing number of available treatments, and U.S. pricing structures. It also discusses the historical restriction that prevented Medicare from directly negotiating prices for covered drugs. That policy context needs an update. Medicare negotiated prices for an initial group of selected high-expenditure Part D drugs, and those prices took effect January 1, 2026. The change does not settle the broader debate over drug development, public research investment, access, and affordability, but it means the historical restriction should not be described as the current policy without qualification. For healthcare organizations, the operational challenge remains medication access and understanding. A treatment cannot produce its intended benefit if a patient cannot obtain it, does not understand it, or cannot follow the plan. Cost, side effects, transportation, complexity, health literacy, and competing priorities can all affect adherence. This is where the episode connects directly to the earlier discussion of AI and medication adherence. A clinician-supervised system may reinforce approved instructions, answer routine questions in plain language, and identify barriers that need a pharmacist, nurse, physician, or care manager.
Prevention is the most credible path from AI to better value
Dr. Yoon's strongest argument is that U.S. healthcare remains too centered on expensive acute care. Hospitals are essential when people are seriously ill, but a system that waits for deterioration will continue to spend heavily on crises that might sometimes have been prevented or managed earlier. AI could support a more preventive model by making education available between visits, encouraging medication adherence, helping patients recognize when to seek help, and identifying patterns for clinician review. In a well-designed workflow, the AI does not replace medical judgment. It extends the reach of an approved care plan and makes human follow-up easier. Those benefits must be measured rather than assumed. A responsible evaluation should consider clinical outcomes, avoidable utilization, total cost, patient experience, staff workload, privacy, safety, bias, and escalation performance. A system that improves one metric while creating hidden work or unsafe reassurance is not delivering better value. The opportunity is practical: use technology to help patients understand and follow care before a preventable problem becomes an emergency. That is a narrower claim than saying AI will fix U.S. healthcare costs, but it is also a more testable and useful place to begin.
What healthcare organizations should take from Episode 6
The episode does not identify one villain or one technological cure. High U.S. healthcare spending reflects real medical capability, difficult fixed costs, policy choices, market structure, chronic disease, administration, and uneven access. Healthcare organizations considering AI should begin with a defined problem and a clinical owner. Ask what action the system supports, who reviews uncertainty, how patients reach a person, what data is retained, and which outcome will demonstrate value. Cost reduction should never be separated from safety and quality. The durable lesson from Episode 6 is that prevention and communication deserve more attention. AI may help healthcare teams explain care, reinforce plans, and notice problems earlier. Its best role is clinician-supervised support that makes existing expertise more available—not an autonomous replacement for the professionals responsible for care.