Does Health Insurance Affect Health? Evidence of Medicare’s Impact on Cancer Outcomes Srikanth Kadiyala, Ph.D. RAND Erin Strumpf, Ph.D. McGill University.

Slides:



Advertisements
Similar presentations
The Role of Health Coverage in Eliminating Disparities in Care Marsha Lillie-Blanton, DrPH Associate Research Professor GWU School of Public Health and.
Advertisements

K A I S E R C O M M I S S I O N O N Medicaid and the Uninsured Figure 0 Medicaid: The Essentials Diane Rowland, Sc.D. Executive Vice President, Henry J.
Multinational Comparisons of Health Systems Data, 2007 Bianca K. Frogner, Meghan Bishop, and Gerard F. Anderson, Ph.D. Johns Hopkins University November.
Figure 0 The Role of Public Programs in Health Reform Diane Rowland and Robin Rudowitz Henry J. Kaiser Family Foundation for Congressional Health Care.
Health Care Spending Growth
Access and Affordability: An Update on Health Reform in Massachusetts as of Fall 2008 Sharon K. Long Urban Institute August 10, 2009 Alliance for Health.
K A I S E R C O M M I S S I O N O N Medicaid and the Uninsured Figure 0 Health Reform Primer: Who are the Uninsured? Diane Rowland, Sc.D. Executive Vice.
THE COMMONWEALTH FUND Adults Ages and the Affordable Care Act Sara R. Collins, Ph.D. Vice President, Affordable Health Insurance AARP and Alliance.
Medicare: The Essentials Juliette Cubanski, Ph.D. Principal Policy Analyst Kaiser Family Foundation for Alliance for Health Reform Washington, D.C. March.
K A I S E R C O M M I S S I O N O N Medicaid and the Uninsured Figure 0 Childrens Coverage: The Role of Medicaid & SCHIP Diane Rowland, Sc.D. Executive.
Figure 0 K A I S E R C O M M I S S I O N O N Medicaid and the Uninsured Medicaid: A Primer Robin Rudowitz Associate Director Kaiser Commission on Medicaid.
The Decline in Employer- Sponsored Health Insurance for Retirees and Its Impact on Older Americans Erin Strumpf Harvard University
The Pent-up Demand for Health Care of the Uninsured Near Elderly Approaching Age 65 Li-Wu Chen, Ph.D. University of Nebraska Medical Center AcademyHealth.
Understanding Changes in Local Public Health Spending Glen Mays, PhD, MPH Department of Health Policy and Management University of Arkansas for Medical.
Health Insurance Coverage of Young Adults John Holahan The Urban Institute February 4, 2008 THE URBAN INSTITUTE.
The Impact of Drug Benefit Caps Geoffrey Joyce, PhD.
THE COMMONWEALTH FUND Enhancing Value in Medicare Stuart Guterman Senior Program Director Program on Medicares Future The Commonwealth Fund Bipartisan.
THE COMMONWEALTH FUND Figure Million Uninsured in 2008; Increase of 7.9 Million Since 2000 Number of uninsured, in millions Source: U.S. Census Bureau,
THE COMMONWEALTH FUND Figure 1. Health Insurance Coverage and Uninsured Trends Data: Analysis of the U.S. Census Bureau, Current Population Survey Annual.
Chartpack National Scorecard on U.S. Health System Performance, 2011
THE COMMONWEALTH FUND National Scorecard on U.S. Health System Performance: Complete Chartpack Cathy Schoen, Senior Vice President Sabrina K. H. How, Research.
THE COMMONWEALTH FUND The Future of Employer-Sponsored Health Insurance The Commonwealth Fund and The Century Foundation Business and National Health Care.
National Scorecard, 2008: Chartpack This Chartpack presents data for all indicators scored in the National Scorecard on U.S. Health System Performance,
Time for Change: The Hidden Cost of a Fragmented Health Insurance System Karen Davis President, The Commonwealth Fund Testimony to the Senate Aging Committee.
THE COMMONWEALTH FUND Why Not the Best? Results from a National Scorecard on U.S. Health System Performance September 20, 2006 Cathy Schoen Senior Vice.
Figure ES-1. How Well Do Different Strategies Meet Principles for Health Insurance Reform? Principles for Reform Tax Incentives and Individual Insurance.
Using Cancer Registry Data for Comprehensive Cancer Control Christie Eheman, PhD, National Program of Cancer Registries, Division of Cancer Prevention.
SOURCE: Kaiser Family Foundation analysis of Urban Institute tabulations of March 2012 Current Population Survey, Bureau of the Census. Medicare and Medicaid.
Diabetes in Idaho BRFSS 2009 Data collected from Behavioral Risk Factor Surveillance System Idaho Department of Health and Welfare, Division of.
Young and Uninsured: Insurance Patterns of Adolescent and Young Adult (AYA) Cancer Survivors 2014 Applied Demography Conference Susanne Schmidt, Helen.
THE COMMONWEALTH FUND Multinational Comparisons of Health Systems Data, 2013 David Squires The Commonwealth Fund November 2013.
The Impact of Diabetes Mellitus in the United States
THE COMMONWEALTH FUND 1 Benefit Design for Public Health Insurance Plan Offered in Insurance Exchange Current Medicare benefits* New Public Health Insurance.
The Almanac of Chronic Disease 2008 Edition. 2 Table of Contents I.The Human Cost Today II.The Economic Cost Today III.The Cost Tomorrow IV.Opportunity.
Uninsured now 15% 24 million 17% 30 million 18% 32 million 20% 37 million 19% 36 million 16% 29 million Insured now 85% 138 million.
Exhibit 1. Percent of Adults Under Age 65 with a Usual Source of Care, by State, 2012 Percent Data source: 2012 Behavioral Risk Factor Surveillance System.
Eligible Women and Participation in the Women’s Health Network Ellen M. Kramer ScD RD April 7, 2005.
The Impact of Insurance Status on Hospital Treatment and Outcomes David Card, Carlos Dobkin and Nicole Maestas.
Asthma Prevalence in the United States National Center for Environmental Health Division of Environmental Hazards and Health Effects June 2014.
Millions of U.S. women ages 19–64 Source: Analysis of the March 2001–2010 Current Population Surveys by N. Tilipman and B. Sampat of Columbia University.
David Card, Carlos Dobkin, Nicole Maestas
What is the Impact of the Internet on Medical Care Use and Cost? Implications of Value Based Benefit Design from a Consumer Driven Health Plan Stephen.
Genomics Alexandra Hayes. Genomics is the study of all the genes in a person, as well as the interactions of those genes with each other and a person’s.
Disparities in Cancer September 22, Introduction Despite notable advances in cancer prevention, screening, and treatment, a disproportionate number.
Patient Protection and Affordable Care Act and Health Outcomes Samuel A. Kleiner Cornell University and NBER Human Capital Research Collaborative Fall.
The National Program of Cancer Registries: Enhancing Cancer Incidence Data … Hannah K. Weir, PhD Division of Cancer Prevention and Control Centers for.
Cancer Healthy Kansans 2010 Steering Committee Meeting May 12, 2005.
Why is Cultural Competency Important in the Practice of Medicine? Karen E. Schetzina, MD, MPH.
“The African American Prostate Cancer Crisis in Numbers”
Spillover Effects of State Mandated-Benefit Laws The Case of Outpatient Breast Cancer Surgery June 5, 2007 John Bian, Ph.D., Atlanta VAMC, American Cancer.
Employer-Sponsored Health Insurance for Early Retirees: Impacts on Retirement, Health and Health Care Erin Strumpf, Ph.D. McGill University AcademyHealth.
Causal Effect of Managed Care on Health Care Quality: Evidence from Cancer Screening Guideline Discontinuities Srikanth Kadiyala* Grant Miller** Harvard.
Figure 1 K A I S E R C O M M I S S I O N O N Medicaid and the Uninsured Dual Eligibles: The Basics Barbara Lyons, Ph.D. Director, Kaiser Commission on.
Incorporating Multiple Evidence Sources for the Assessment of Breast Cancer Policies and Practices J. Jackson-Thompson, Gentry White, Missouri Cancer Registry,
1 An Overview of Colorectal Cancer in Delaware Delaware Health Care Commission November 3, 2011.
State Trends in Premiums and Deductibles, : Eroding Protection and Rising Costs Underscore Need for Action Cathy Schoen Senior Vice President.
Rural Health Research Center S outh C arolina Greater Rurality Increases Barriers to Primary Health Care: Evidence of a Gradient in Access or Quality Janice.
Exhibit Million Uninsured Adults Ages 50–64 in 2009, Up by 1.1 Million in Last Year Millions uninsured, adults ages 50–64 Source: Analysis of the.
The Impact of Insurance Status on Hospital Treatment and Outcomes David Card, Carlos Dobkin and Nicole Maestas.
Trends in Colorectal Cancer Incidence Rates by Race, Age and Indices of Access to Medical Care in the U.S., Yongping Hao, PhD 1 Ahmedin Jemal,
Definitions: Definitions: Obesity: Body Mass Index (BMI) of 30 or higher. Obesity: Body Mass Index (BMI) of 30 or higher. Body Mass Index (BMI): A measure.
Arnold School of Public Health Health Services Policy and Management 1 Women’s Cancer Screening Services Utilization Versus Their Insurance Source Presenter:
State-Specific Prevalence of Asthma Among Adults, by Industry and Occupation — Behavioral Risk Factor Surveillance System, 21 States, 2013 Katelynn E.
Blacks account for 13% of the population in the United States.
WA OR ID MT ND WY NV 23% CA UT AZ NM 28% KS NE MN MO WI TX 31% IA IL
Colorectal Cancer Screening, Medicare and Disability
Health and Health Care for Blacks in the United States
Private Sector Participation in Medicare: Exceeding Expectations
Dual Eligibles Across the States
Colorectal cancer survival disparities in California
Presentation transcript:

Does Health Insurance Affect Health? Evidence of Medicare’s Impact on Cancer Outcomes Srikanth Kadiyala, Ph.D. RAND Erin Strumpf, Ph.D. McGill University McGill Institute for Health and Social Policy March 27, Preliminary results: please do not cite Draft available at:

Health Insurance and Health Does health insurance affect health outcomes? –Lower price paid for medical care –Increased access to care –Higher quantity (quality?) of care consumed –If marginal health benefits from increased care are positive, then we expect an effect on health

Measuring the Impacts Empirically, identifying effects of health insurance on health is complicated –Many contexts have little to no variation in health insurance status (national HI systems) –In the U.S., insurance status is endogenous Are the risk-averse and healthy more likely to be insured? Or are the sick more likely? –If the benefits from marginal health care are small, detection of any health benefit is difficult

How do we study this question? –Experimental designs are rare RAND HIE (1970s), Oregon HIE (ongoing) –Descriptive studies lacking identification are plentiful –Better evidence from quasi-experimental studies* Find that insurance reduces mortality and improves health status Most focus on acute conditions: heart attack, stroke, car accidents, pregnancies –Suggests effects result from better treatment conditional on diagnosis Some examine impact of insurance on differences by race, education, or previous insurance status * Currie and Gruber 1996, Decker and Rappaport 2002, Decker 2005, Doyle 2005, Polsky et al 2006, Card et al 2009, McWilliams et al 2009

Our Research Question Does Medicare affect cancer health outcomes? –Approximately 12% of the U.S. population is uninsured at ages (~4 million)* –At age 65 nearly everyone becomes eligible for Medicare What is the effect of Medicare coverage on cancer detection? –Cancer is the 2 nd leading cause of death in the U.S. –Medicare accounts for 45% of all spending on cancer treatment** –Cancer care is ~10% of Medicare spending *Kaiser Family Foundation, Health Insurance Coverage for Older Adults, May 2009 Birnbaum and Patchias 2008 estimate 5% of the age 65+ population is ineligible for Medicare **Cancer Action Network, ACS, Cancer and Medicare Chartbook 2009

Mechanisms: Detection vs. Treatment Cancer is not necessarily a symptomatic condition –Does any effect of health insurance on health work via disease detection? –The fact that it’s not acute or symptomatic might make it more possible to identify differential effects for the uninsured Medicare reduces the price of both screening tests and physician visits –Screening rates and physician visits increase at age 65* *Lichtenberg 2002, McWilliams et al 2003, Ward et al 2007, Card et al 2008

Cancer Detection Data U.S. Surveillance Epidemiology and End Results (SEER) database, –Cancer detection from 25% of the U.S. population (12 states) CA, CT, GA, HI, IA, KY, LA, MI, NJ, NM, UT, WA –Detailed information on staging, size of the tumor and other measures of cancer severity Behavioral Risk Factor Surveillance System (BRFSS), –Provides covariate data at year*age level

Methodology Regression Discontinuity Design –Uses the discontinuity in insurance status at age 65 and compares cancer detection rates on either side of this age threshold Assumes: –Smoothness in other determinants of cancer detection across the cutoff Education, marital status, employment, etc –In the absence of treatment (insurance), smoothness in the outcomes (detection) Cancer risk or unobserved true incidence is smooth

Analysis Graphical Evidence Estimate the magnitude of the discontinuity using regression –Cancer detection = α + β 1 (Mcare) + β 2 (Mcare*a-65) + β 3 (1-Mcare*a-65) + … + ε –Adjust for age, age 2 and age 3 ; sex, race, education, income, marital status, employment at year*year-of- age level from BRFSS Examine cancers with screening tests separately from those that do not –Breast, colorectal, prostate, cervical (BCPC) Assess heterogeneous impacts by pre- Medicare insurance status

Impact of Medicare on All Cancer Detection OLS Poisson Medicare Cutoff101.5***108.4***92.9*** Age Trend Above75.8*** Age Trend Below87.4***85.86***179.0 % Increase relative to Detection rate at age %6.8%5.9% Covariates Controls Included NoYes N77 N=77, * p<=.05, ** p<=.01, *** p<=.001; Data: U.S. SEER (12 states); Controls: sex, race, income quartiles, education (4), marital status (5), employment. Cutoff coefficient is robust to adding age 2 and age 3. Poisson model includes age 2 and age 3.

Effect of Medicare on BCPC Detection OLS Poisson Medicare Cutoff65.7***70.3***63.2*** Age Trend Above21.4***22.8***192.3 Age Trend Below40.2***41.9***206.6 % Increase relative to detection rate at %9.6%8.7% Covariates Controls Included NoYes N77 * p<=.05, ** p<=.01, *** p<=.001; Data: U.S. SEER (12 states); Controls: sex, race, income quartiles, education (4), marital status (5), employment. Cutoff coefficient is robust to adding age 2 and age 3. Poisson model includes age 2 and age 3.

Effect of Medicare on Non-BCPC Cancer Detection OLS Poisson Medicare Cutoff35.8***38.0***29.7** Age Trend Above54.4***53.0***-31.8 Age Trend Below47.2***44.0***-17.8 % Increase relative to detection rate at age %4.5%3.5% Covariates Controls Included NoYes N77 * p<=.05, ** p<=.01, *** p<=.001; Data: U.S. SEER (12 states); Controls: sex, race, income quartiles, education (4), marital status (5), employment. OLS cutoff coefficient is robust to adding age 2 and age 3. Poisson model includes age 2 and age 3.

Stage at Detection for BCP Cancers 80% of newly detected breast cancer cases are at local or regional stages 65% of newly detected CRC cases are at regional or distant stages 81% of newly detected prostate cancer cases are at the local stage About 45% of newly detected cases for these three cancers are at “treatable” stages* where we expect diagnosis to lead to significant health benefits –Conservative estimate based on concerns about over-diagnosis and treatment * Local and regional for breast; in-situ, local and regional for colorectal; regional for prostate

Initial Conclusions Medicare increases the cancer detection rate by a substantial amount Medicare’s effects on cancer detection are larger for cancers with recommended screening tests –Breast cancer detection rate: 6% increase –Colorectal and prostate: 9% increases –No impact on cervical cancer detection Impacts for non-screening cancers as well An important share of newly detected cases are treatable and health improvements are likely

Effects by Pre-Medicare Insurance Status We expect differential detection effects for those uninsured pre-Medicare –Large change in insurance status –But quality of insurance may change for those previously insured (ie, screening mandates) SEER does not include insurance status

Insured vs. Uninsured Analysis State Insurance Rates –Variation in insurance rates at age 64 by state, % in Louisiana, 92% Michigan –Correlate RD estimates of the increase in insurance coverage from age 64 to 65 with RD estimates of the change in cancer detection –Control for state and time fixed effects

Correlation of RD Estimates

Detection Linked to Changes in Insurance Status State-level variation in health insurance discontinuities due to Medicare can explain 37-71% of the state-level variation in BCPC detection due to Medicare We did not identify a similar relationship with respect to Non-BCPC detection Suggests that the extensive margin is important, but it’s possible that quality of insurance coverage also plays a role

Conclusions Health insurance plays a role in improved health in the context of chronic, latent disease Medicare increases the cancer detection rate by 6.4%, about 100 cancers per 100,000 individuals Larger detection effects for screening cancers, but also effects for non-screening The increase in insurance rates at age 65 can account for a significant share of the increase in screening cancer detection