Health Coverage Correlations
Health insurance is a major factor in determining whether or not people have access to lifesaving health care. According to the CDC's Healthy People 2020 initiative, uninsured people are more likely to have poor health status, less likely to receive medical care, more likely to be diagnosed later, and more likely to die prematurely.
LiveStories examined data on the uninsured rate across the 3,142 counties in the United States. We correlated the percent of uninsured residents with a social and economic indicators from the American Community Survey. We also correlated the uninsured rate with several mortality measures from the Centers for Disease Control and Prevention. We’ve calculated Pearson’s correlation coefficient—also known as r—for each measure.
Note that correlation is not causation. Any relationships between these measures and the uninsured population may point to some deeper connection—or they may be entirely incidental.
Visually, you can think of the correlation coefficient as a scale that tells you: “how much do these data points form a diagonal line?” A perfect upward-sloping diagonal line has a coefficient of 1—for example, the scatterplot you see when the “Uninsured Rate” measure is correlated with itself. A diagonal line sloping downward, on the other hand, would have a –1 coefficient: a perfect negative correlation.
Explore the data below. Select a measure, and the scatterplot chart will change to show how well it correlates to the percent of uninsured residents (x-axis) for the 1,000 largest counties.
Data is for the 5-year period, 2013-2017. Note that only 1,000 counties are shown in the scatterplot. (See "About the Data" below.)
About the Data
Mortality rate data was queried from CDC Wonder. The rates are the reported age-adjusted rates per 100,000 population, covering the 5-year period from 2013 to 2017. "Deaths of Despair" consists of alcohol, drug, and suicide deaths.
All other data is from the American Community Survey (ACS), 5-year estimates (2013-2017), published by the U.S. Census Bureau. We used the following ACS tables: Median Household Income (B19013); Bachelor's Degree or Higher and Less than High School Graduate (B16010); Median Home Value (B25077); Disability (DP02); Unemployment (B23025); Can't Speak English Well (B16004); and Poverty (B17001). For the English-speaking measure, we summed all "Not well" and "Not at all" columns and calculated the value as a percentage.
Although the interactive scatterplot chart only shows data points for 1,000 counties, the correlation coefficients were calculated based on data for all available counties for each pair of measures. While ACS data is available for all counties, the CDC does not report death rates with numerators less than 20. Counties without data for a given cause of death—which tend to be counties with small populations—have been excluded from the correlation coefficient calculations.