The Health Care industy: high spending, poor outcomes

[chinaexpert1]

Investigate High U.S. Healthcare Spending and Poor Health Outcomes

Overview

The United States spends more on healthcare than any comparable country but receives unusually poor population-health returns. In 2024, U.S. healthcare spending reached approximately 18% of gross domestic product—nearly twice the OECD-country average—while the United States continued to experience relatively low life expectancy, high avoidable mortality, limited primary-care capacity, and major geographic and socioeconomic disparities.

This contradiction is sometimes described as a healthcare failure, but the explanation may involve two related systems:

1. A high-cost medical system that pays unusually high prices, uses complex administration, and frequently rewards service volume rather than health improvement.
2. A poor population-health system in which chronic disease, inequality, environmental conditions, weak prevention, and uneven access generate more illness before patients reach medical care.

Recent health-policy analysis argues that poor population health increases the need for care, but that high U.S. spending is driven primarily by high prices and administrative complexity rather than unusually high service volume.

The central research question is:

Why does the United States convert exceptionally high healthcare spending into comparatively weak gains in life expectancy, healthy life expectancy, avoidable mortality, and functional health?

The investigation should determine:

• Which categories of spending have the weakest measurable relationship with outcomes;
• Whether price, utilization, administration, chronic disease, or unequal access explains the largest share of the spending-outcome gap;
• Which U.S. regions achieve unusually good outcomes for their level of spending;
• Which interventions are associated with lower costs and better health;
• Whether healthspan provides a more useful policy target than life expectancy alone.

The recommended primary unit of analysis is the county-year or hospital-referral-region-year. State- and country-level comparisons should provide broader context.

---

Action Items

Initial Evidence

Established evidence supporting the investigation includes:

• U.S. healthcare spending per person and as a share of GDP substantially exceeds spending in peer countries.
• The United States has among the lowest life expectancy and highest avoidable-mortality rates among comparable wealthy nations.
• Approximately 27 million people remained uninsured in the latest international comparison, and many insured Americans still delayed or skipped care because of cost.
• In 2026, fewer than half of surveyed U.S. adults were classified as consistently able to afford healthcare, the lowest level in five years.
• The United States has comparatively few primary-care physicians and medical graduates per capita.
• U.S. mortality performance varies sharply by state and county. A 2025 JAMA viewpoint reported that age-adjusted mortality in the worst-performing state was approximately twice that of the best-performing state in 2021.
• From approximately 2010 through 2019, U.S. income continued rising while life expectancy stagnated or declined, suggesting a weakening relationship between economic resources and longevity.
• Federal experts convened by the Government Accountability Office identified primary care, workforce capacity, pricing, and payment reform as major opportunities to increase healthcare value.

These findings establish the spending-outcome contradiction but do not prove any single explanation.

---

Hypotheses

H1 — High prices explain more excess spending than high utilization

The United States may not consume dramatically more care than peer countries but may pay more for hospital services, physician services, pharmaceuticals, administration, and insurance.

Evidence to seek:

• Price-adjusted utilization is closer to peer-country levels than spending;
• Higher hospital-market concentration predicts higher commercial prices;
• High-spending regions do not consistently use proportionally more services;
• Standardizing payment rates substantially reduces regional spending differences.

H2 — Administrative complexity consumes resources without improving health

Multiple insurers, billing systems, prior-authorization requirements, coding rules, and network arrangements may create a large nonclinical cost burden.

Evidence to seek:

• Administrative staffing grows faster than clinical staffing;
• Administrative intensity predicts higher spending without better outcomes;
• Providers operating under simpler payment structures have lower overhead;
• Prior-authorization burden predicts delayed treatment or clinician attrition.

H3 — Weak primary care increases expensive downstream treatment

Insufficient access to continuous primary and preventive care may increase emergency-department use, avoidable hospitalization, and late diagnosis.

Evidence to seek:

• Primary-care supply predicts fewer avoidable admissions;
• Regions with better continuity of care have lower per-beneficiary spending;
• Preventive-care access predicts improved diabetes and cardiovascular outcomes;
• Primary-care shortages precede higher emergency utilization.

H4 — Medical spending is concentrated after preventable disease has developed

The system may devote more resources to treating advanced disease than to preventing obesity, diabetes, cardiovascular disease, substance-use disorders, and other chronic conditions.

Evidence to seek:

• High chronic-disease prevalence explains a large portion of regional spending;
• Preventive investment is weak relative to acute-care spending;
• Counties improving risk factors later experience slower spending growth;
• High spending on advanced disease produces limited healthy-life gains.

H5 — Market concentration raises prices without reliably improving quality

Hospital consolidation and physician-practice acquisition may increase negotiating power and prices without proportional improvements in survival, readmission, safety, or patient access.

Evidence to seek:

• Higher concentration predicts higher risk-adjusted prices;
• Acquisitions produce price increases without measurable quality improvement;
• Independent and consolidated markets differ after matching on population and case mix;
• Competition effects differ between rural and urban markets.

H6 — Insurance coverage alone is insufficient when deductibles and networks limit access

Nominally insured patients may still delay care because of deductibles, copayments, uncovered services, narrow networks, or provider shortages.

Evidence to seek:

• Underinsurance predicts delayed care independently of insurance status;
• High-deductible enrollment reduces both low-value and high-value care;
• Cost-related nonadherence predicts preventable complications;
• Coverage expansions have larger effects where provider capacity is adequate.

H7 — Social conditions explain more geographic health variation than medical spending

Income, education, housing, environmental exposure, transportation, food access, and community conditions may predict health outcomes more strongly than local medical spending.

Evidence to seek:

• Social variables substantially improve prediction of life expectancy;
• High-spending areas with poor social conditions retain poor outcomes;
• Medical spending coefficients weaken after social determinants are introduced;
• Community-level interventions produce measurable health improvements.

H8 — The United States underinvests in prevention and public health

High clinical spending may coexist with insufficient public-health infrastructure, vaccination, disease surveillance, health education, and environmental prevention.

Evidence to seek:

• Public-health funding predicts lower preventable mortality;
• Local public-health staffing predicts better outbreak or vaccination performance;
• Prevention spending has delayed but measurable health returns;
• Funding instability predicts deteriorating surveillance or response capacity.

H9 — Measuring lifespan alone conceals weak healthspan

Medical treatment may extend life without producing equivalent gains in disability-free, disease-free, or functionally independent years.

Evidence to seek:

• Spending growth exceeds growth in health-adjusted life expectancy;
• Disability prevalence rises despite stable or improving lifespan;
• Certain disease categories have high lifetime spending per healthy year gained;
• Regions with similar life expectancy differ sharply in functional health.

A 2026 health-value study specifically recommends examining lifetime spending relative to health-adjusted life expectancy and identifying diseases with high spending but limited healthy-life gains.

H10 — The major causes differ across places and populations

There may be no single national explanation. Rural areas may suffer primarily from access shortages, while wealthy metropolitan areas may experience high prices and administrative waste.

Evidence to seek:

• Distinct regional clusters emerge;
• Spending drivers differ by rurality, age, income, and insurance market;
• The same spending level produces different outcomes across demographic groups;
• National averages obscure high-performing local systems.

---

Research Plan

1. Define measurable outcomes

Use several outcomes rather than a single ranking:

• Life expectancy;
• Health-adjusted life expectancy;
• Disability-free life expectancy;
• Avoidable mortality;
• Preventable mortality;
• Treatable mortality;
• Infant and maternal mortality;
• Chronic-disease prevalence;
• Self-reported physical and mental health;
• Avoidable hospital admissions;
• Emergency-department use;
• Functional limitations;
• Workforce absence or disability.

Create a principal value measure:

Because the numerator may be expressed in different units, the final index should use standardized values and should always be accompanied by the original metrics.

2. Build a geographic healthcare panel

Recommended units:

• County-year for population health;
• Hospital referral region-year for Medicare use and spending;
• State-year for insurance and policy comparisons;
• Country-year for OECD benchmarking.

Suggested period:

• 2000 through the latest available year;
• A more detailed post-2010 module where data definitions are consistent.

Core variables should include:

• Per-capita spending;
• Hospital spending;
• Physician spending;
• Prescription-drug spending;
• Service utilization;
• Insurance coverage;
• Out-of-pocket burden;
• Primary-care supply;
• Hospital capacity;
• Market concentration;
• Demographics;
• Chronic disease;
• income and education;
• Housing and environmental conditions;
• Health outcomes.

3. Separate price from utilization

Decompose spending:

Where possible, compare:

• Number of visits;
• Procedures;
• Admissions;
• Days of care;
• Prescription volume;
• Payment per service;
• Service intensity;
• Site of care.

CMS geographic-variation datasets provide standardized spending that removes geographic payment-rate differences, supporting comparisons of underlying resource use.

4. Estimate regional healthcare value

For each county or hospital referral region:

1. Adjust spending for age, sex, health risk, and local payment rates;

2. Adjust outcomes for demographic and baseline-health differences;

3. Calculate residual spending after expected spending;

4. Calculate residual outcomes after expected outcomes;

5. Classify regions into:

   • Lower spending, better outcomes;
   • Higher spending, better outcomes;
   • Lower spending, worse outcomes;
   • Higher spending, worse outcomes.

The most useful comparison group consists of regions achieving better outcomes with similar populations and fewer resources.

5. Measure primary-care access

Construct a primary-care capacity index using:

• Primary-care physicians per resident;
• Federally Qualified Health Center capacity;
• Appointment availability;
• Travel time;
• Continuity of care;
• Preventive visits;
• Insurance acceptance;
• Rural hospital and clinic closures.

Test whether primary-care capacity predicts later changes in:

• Avoidable hospitalization;
• emergency use;
• chronic-disease control;
• mortality;
• total spending.

6. Analyze market concentration

Calculate hospital and insurer concentration using the Herfindahl–Hirschman Index:

Where (s_j) is the market share of provider or insurer (j).

Test:

Potential outcomes:

• Commercial hospital prices;
• Medicare spending;
• premiums;
• staffing;
• closures;
• quality;
• patient travel distance.

Acquisition events may support difference-in-differences analysis if credible comparison markets can be identified.

7. Model population health separately from healthcare performance

Estimate two linked models:

This prevents poor health before treatment from being incorrectly attributed entirely to the healthcare system.

8. Estimate healthspan

Combine mortality with disability or self-reported health states using multistate life tables or Sullivan’s method:

Compare healthcare spending with:

• Total life-years;
• Healthy life-years;
• Disability-free years;
• Functional independence.

9. Compare the United States with peer countries

Use OECD data to compare:

• Spending per capita;
• Spending as a percentage of GDP;
• Hospital and physician prices;
• Utilization;
• Primary-care capacity;
• insurance coverage;
• avoidable mortality;
• life expectancy;
• healthy life expectancy.

Avoid comparing the U.S. only with the OECD average. Construct a peer group using income, demographics, and economic structure.

OECD health datasets are accessible through an SDMX-based API that supports JSON and CSV outputs.

10. Test policy changes

Potential quasi-experiments include:

• Medicaid expansion;
• Hospital mergers;
• rural hospital closures;
• primary-care payment reforms;
• prescription-drug caps;
• public-health funding changes;
• state insurance-market reforms;
• accountable-care organization entry;
• all-payer or global-budget programs.

Use:

• Difference-in-differences;
• Synthetic controls;
• Interrupted time series;
• Event studies;
• Instrumental variables only where assumptions are defensible.

11. Conduct sensitivity analyses

Repeat results using:

• Medicare spending only;
• Total personal healthcare spending;
• Price-standardized spending;
• Different age groups;
• Different health outcomes;
• Urban and rural samples;
• Pre- and post-pandemic periods;
• State, county, and hospital-referral-region units;
• Alternative deprivation indexes;
• Alternative definitions of avoidable mortality.

12. Produce the final PowerPoint

Recommended structure:

1. The U.S. spending-outcome contradiction;
2. International comparison;
3. Definitions of healthcare value and healthspan;
4. Geographic spending variation;
5. Geographic outcome variation;
6. Price versus utilization;
7. Primary care and preventable illness;
8. Hospital and insurer concentration;
9. Social determinants versus medical care;
10. High-value and low-value regions;
11. Which hypotheses survived testing;
12. Policy interventions and measurable targets.

The final report must distinguish:

• Association from causation;
• Medical-system performance from population-health conditions;
• Prices from utilization;
• Health insurance from effective access;
• Lifespan from healthy lifespan.

---

Resources

Data, APIs, and Where to Get Data

CMS Open Data APIs

CMS publishes downloadable and API-accessible data on:

• Medicare spending;
• Utilization;
• quality;
• providers;
• hospitals;
• physicians;
• prescription drugs;
• geographic variation.

The Medicare Geographic Variation datasets include demographic, spending, utilization, and quality indicators at state, county, and hospital-referral-region levels.

Use for:

• Risk-adjusted spending;
• Price-standardized spending;
• Service utilization;
• Avoidable admissions;
• Hospital spending per beneficiary;
• Provider-level comparisons.

CMS’s provider datasets expose Open Data API endpoints that support filtering, querying, and aggregation.

CDC WONDER

CDC WONDER provides mortality, birth, population, and disease-related statistics through web queries and an XML-based web service.

Use for:

• Cause-specific mortality;
• Age-adjusted mortality;
• Infant mortality;
• Maternal mortality;
• Avoidable and preventable death measures;
• Geographic and demographic comparisons.

Carefully follow suppression rules and dataset-specific geographic limitations.

CDC PLACES

Provides local estimates for:

• Chronic disease;
• health behaviors;
• preventive services;
• disability;
• mental health;
• health status.

Use for county-, place-, census-tract-, and ZIP Code Tabulation Area-level population-health modeling.

Behavioral Risk Factor Surveillance System

Use for:

• Self-reported health;
• chronic disease;
• disability;
• insurance;
• access;
• health behaviors.

Microdata require survey-weighted analysis.

National Health Interview Survey

Use for:

• Insurance and underinsurance;
• delayed care;
• functional limitations;
• health status;
• medical access;
• socioeconomic characteristics.

Medical Expenditure Panel Survey

Use for:

• Individual and household medical spending;
• utilization;
• insurance;
• out-of-pocket expenses;
• prescription use;
• health conditions.

MEPS is particularly valuable for distinguishing spending concentration and underinsurance.

National Health Expenditure Accounts

Published by CMS.

Use for:

• National spending by payer;
• service category;
• sponsor;
• historical trend;
• projected spending.

Hospital Price Transparency Data

Hospitals are required to publish machine-readable pricing files.

Use for:

• Negotiated prices;
• cash prices;
• insurer-specific variation;
• hospital-market comparisons.

Substantial data cleaning and standardization will be required.

Transparency in Coverage Data

Insurers publish machine-readable negotiated-rate files.

Use for:

• Commercial prices;
• insurer-provider contracts;
• price variation.

These files are extremely large and may require cloud storage, DuckDB, Polars, Spark, or targeted sampling.

Medicare Spending per Beneficiary

CMS publishes hospital-, state-, and national-level measures covering spending around inpatient episodes, with risk adjustment and payment standardization.

Area Health Resources Files

Published by the Health Resources and Services Administration.

Use for:

• Physicians;
• primary-care capacity;
• hospitals;
• health professions;
• demographics;
• county-level healthcare resources.

Health Professional Shortage Area Data

Use for:

• Primary-care shortages;
• mental-health shortages;
• dental shortages;
• underserved populations and regions.

Federal Qualified Health Center Data

Use for:

• Patients served;
• service capacity;
• payer mix;
• clinical outcomes;
• underserved-community access.

American Community Survey API and PUMS

Use for:

• Insurance status;
• income;
• poverty;
• disability;
• employment;
• race and ethnicity;
• housing;
• transportation;
• local demographic controls.

BEA Regional API

Use for:

• Personal income;
• regional price levels;
• GDP;
• employment;
• geographic economic controls.

Bureau of Labor Statistics APIs

Use for:

• Healthcare employment;
• wages;
• staffing trends;
• occupational shortages;
• producer prices;
• consumer medical-care prices.

OECD Health Statistics API

OECD Health Statistics provides the principal international comparison source for health spending, health workforce, health status, utilization, and mortality.

AHRQ Healthcare Cost and Utilization Project

HCUP provides hospital utilization, diagnoses, procedures, charges, and outcomes through query tools and research databases.

Some detailed HCUP files require purchase and a data-use agreement, so the core reproducible analysis should rely on free query outputs or other public sources.

County Health Rankings

Use for consolidated county-level measures involving:

• Health outcomes;
• clinical care;
• social and economic conditions;
• physical environment.

Prefer the original source when a measure is available directly.

Environmental Protection Agency Data

Use for:

• Air pollution;
• water quality;
• toxic exposure;
• environmental-justice indicators.

Social Security and Census Disability Data

Use for:

• Disability prevalence;
• working-age disability;
• labor-force withdrawal;
• geographic changes in functional health.

Foundational Resources

• Commonwealth Fund, U.S. Health Care from a Global Perspective, 2026.
• Government Accountability Office, Reducing Spending and Enhancing Value in the U.S. Health Care System.
• JAMA, Setting Public Health Priorities in the United States.
• JAMA Health Forum, To Improve U.S. Life Expectancy, a New North Star Is Needed.
• Berkowitz, What’s Keeping the U.S. From Better Population Health?
• The U.S. Mortality Crisis as a Preston Curve Reversal.

Central Scientific Caution

A region with high spending and poor outcomes is not necessarily wasting money. It may have an older, sicker, poorer, or more medically complex population. Conversely, a low-spending region may appear efficient because patients cannot obtain care.

The analysis must therefore compare:

1. Health need;
2. Effective access;
3. Prices;
4. Utilization;
5. Quality;
6. Outcomes;
7. Healthy years gained.

The strongest policy finding will likely come from identifying positive deviants: U.S. counties, health systems, or hospital-referral regions that achieve better risk-adjusted outcomes at lower standardized spending than otherwise comparable places.

• If this issue requires access to 311 data, please answer the following questions:
  • Do you need a one-time or ongoing dump of the data?
  • Do you need subset of data (i.e. certain years) or the entire data set (approx. 4 million rows or 11 GB)?
    • If a subset is needed, please define subset characteristics (i.e. date range, etc.)
  • Do you need online access via an API or a download of data?

[orinmukherjee2009]

1. I have been looking at APIs and data sets. I decided I will try to answer the following question: Which US regions get the best health outcomes per spending, and what do they do differently from other regions? In this upcoming week, I will begin coding for this project, and start finding trends within the data.

Unfortunately, two of the datasets I downloaded cause an error when I try to extract the ZIP file, so I will have to debug or find a workaround for that.
3. I will likely work on the project for around 5-7 hours this week.
4. ETA: I also cannot come to this week's weekly meeting since I have soccer practice at that same time. I think I should be able to finish the project around 9/12-9/19 range.
5. Pictures: NA at this stage.