February 2010

posted Sep 7, 2010, 12:38 PM by Angela Jeffers   [ updated Sep 7, 2010, 12:40 PM by PC/SM Chief Resident ]

[click here for a Word document of this review]

 Question: Among adults in the United States, does lack of health insurance predict increased mortality?

 Article: Wilper AP, Woolhandler S, Lasser KE, McCormick D, Bor DH, Himmelstein DU.  Health Insurance and Mortality in U.S. Adults.  AJPH 2009; 99: 2289-2295.

 Reviewer: Aaron Fox, MD

 Methods: This retrospective cohort study utilizes NHANES III data to compare overall mortality of those with non-government health insurance to those uninsured at the time of the intake interview.  The authors excluded study participants with incomplete interview or physical exam data and those with health insurance through government programs such as Medicare, Medicaid, and the Department of Veterans Affairs.  In NHANES III, interviews and physical exams occurred between 1988 and 1994, while mortality data was derived from the National Death Index at the end of 2000.  Using Χ2 analysis, the authors analyzed the relationship between insurance status, demographics, baseline health measures, and mortality.  Next, using Cox proportional hazards survival analysis, the authors evaluated the impact of insurance status on mortality after adjustment for multiple demographic features and health measures, including: income, education, employment, alcohol or tobacco use, leisure exercise, BMI, self related health status, and physician related health.     


Results: 16.2% of the sample was uninsured at baseline.  Overall, 3.1% of the cohort had died by the end of follow-up in 2000.   Lack of health insurance was significantly associated with mortality when adjusted for age and gender (hazard ratio = 1.80; 95% CI = 1.44, 2.26) or when adjusted more extensively (hazard ratio = 1.40; 95% CI = 1.06, 1.84). 


Validity:  The NHANES III cohort is nationally representative and oversamples several minority groups to ensure inclusion.  Sampling did not seem to contain other bias.  The follow-up interval of 6-12 years was appropriate for the outcome of interest, overall mortality.  Within this study, the uninsured cohort includes those lacking health insurance at the time of the intake and there is no additional information on insurance status throughout the follow-up period.  Therefore, there could be overlap between the two groups based on health insurance status, if people gained or lost health insurance coverage during the study.  Despite this limitation, the authors adjust for an impressive number of potential confounding factors, which suggests that insurance status is an independent risk factor for mortality.  The authors perform additional analyses including Medicaid recipients or adjusting for geographical location, marital status, and other factors to demonstrate minimal impact on their findings. 

Generalizability: These findings are generalizable to the adult US population less than 65.  The study examines health insurance status at one point in time making it difficult to determine the association between intermittent health insurance coverage or loss of health insurance and mortality, so the findings should not be extrapolated to the underinsured.     


Advocacy Implication: Some have claimed that the uninsured population consists mostly of young healthy individuals who have little need for health insurance and choose to go without it.  This study contradicts this notion suggesting that the uninsured are at 40% greater risk of death than those with non-government insurance.  By utilizing US Census Data and the hazard ratio determined in this study, the authors estimate that 44,789 deaths were associated with lack of health insurance in 2005.  While, this data cannot be used to predict the impact of extending coverage to those lacking health insurance, it does demonstrate the potential benefit that could come from universal health insurance – almost 45,000 deaths averted.  

PC/SM Chief Resident,
Sep 7, 2010, 12:40 PM
PC/SM Chief Resident,
Sep 7, 2010, 12:40 PM