The Bright Spot Analysis

The Creating a Culture of Health in Appalachia initiative defines a Bright Spot as an Appalachian county that has better-than-expected health outcomes given its characteristics and resource levels—that is, the socioeconomics, demographics, behaviors, health care facilities, and other factors that influence health outcomes.

The Bright Spot identification process used residual analysis to identify those Appalachian counties that have “unexpectedly” better health outcomes given their characteristics and resources.

In this analysis, we did not aim to identify healthier counties with high levels of resources and the sorts of characteristics that support positive health outcomes but, rather, counties encompassing a wide range of resource levels and characteristics that all managed to find a way to have better health outcomes than expected.

Identification Methodology

The methodological approach used in this analysis assumes we can

  • broadly measure health in a community;
  • compare actual health outcomes to expected health outcomes; and
  • determine whether a community’s health outcomes exceed expectations.

The methodology employed four steps.

  1. We first identified 19 county-level outcome measures that together capture the overall health of a community.

    Some examples of these measures include the infant mortality rate, the cancer mortality rate, the percentage of adults who are obese, the prevalence of diabetes, and the prevalence of depression among Medicare beneficiaries.

    The outcome measures selected represent both physical and behavioral health, as well as diagnosed and perceived health.

  2. We then identified 29 county-level drivers known to affect individual and community health.

    The drivers were organized into broad categories, including social determinants, health behaviors, and access to health care services.

    Examples of the drivers are median income, the percentage of adults with some college education, the percentage of the population under age 65 who are uninsured, the number of primary care physicians per 100,000 population, and the percentage of adults who smoke.

  3. A multivariate regression analysis then determined the relationship between the 19 health outcome measures and the 29 driver measures.

    The analysis produced one expected outcome value for each of the 19 health outcome measures for each of the Appalachian Region’s 420 counties. The counties were separated into a metropolitan group and a nonmetropolitan group; a separate multivariate regression was run for each of these groups.

    The expected outcomes were then compared with the actual, observed outcomes for each county to identify counties that performed better than expected. In most counties, some of the 19 outcomes were better than expected, and some were worse than expected.

    The difference between the actual, observed value and the expected (or predicted) value is called a residual. We used the residual to determine how each county performed on each measure, relative to what the model predicts.

  4. Each outcome residual was then standardized into a z-score to allow comparison across all health outcome measures.

    We reversed signs on the outcomes so that positive z-scores reflected “good health.” By assessing the degree to which a county’s observed health outcomes exceeded expected values, the Bright Spots model identified counties that either did very well on a few outcomes or exceeded expectations—perhaps only marginally—across many outcomes.

A county whose average of all 19 standardized outcome residuals fell in the top 10 percent of averages in either the metropolitan or the nonmetropolitan county group was classified as a Bright Spot.

Bright Spots are places that exceed expectations, regardless of the values of the drivers. This is a strength of the identification approach, and allowed the research to focus on the positive aspects of communities relative to their own characteristics and resource levels.

The Bright Spot Counties

Counties with an average standardized residual score in the top 10 percent of scores for either metropolitan or nonmetropolitan Appalachian counties were identified as Bright Spots. The analysis identified a total of 42 Appalachian Bright Spot counties: 15 in the metropolitan group and 27 in the nonmetropolitan group.

The Bright Spot counties are located in all five of Appalachia’s subregions (North, North Central, Central, South Central, and Southern) and convey the diversity of communities across the Appalachian Region.

Table 1 below lists the metropolitan Bright Spot counties, their average standardized residual score, and the health outcome with the highest residual, which reflects the county’s greatest over performance for an outcome relative to available resources. Table 2 shows the same information for nonmetropolitan Bright Spot counties.

The higher the residual score for a county, the more the county had outperformed expectations. The standardized residual scores represent standard deviations. For example, outcomes in a county with an average residual score of 0.47 were, on average, 0.47 standard deviations above the expected outcomes.

Table 1
Metropolitan Appalachian Bright Spot Counties, Ranked by Average Outcome Residual

Rank County State Average Standardized Residual Score a Highest Individual Residual b
1 Wirt West Virginia 0.47 Injury mortality 1.58
2 Clay West Virginia 0.40 Heart disease mortality 1.51
3 Henderson North Carolina 0.35 % obese adults 0.98
4 Hale Alabama 0.35 Depression prevalence 1.10
5 Sequatchie Tennessee 0.31 Poisoning mortality 1.22
6 Floyd Virginia 0.30 COPD mortality 1.08
7 Sullivan Tennessee 0.30 Poisoning mortality 1.23
8 Marshall Mississippi 0.30 % opioid Rx claims 1.58
9 Madison North Carolina 0.29 % obese adults 1.26
10 Whitfield Georgia 0.29 Depression prevalence 0.97
11 Tioga New York 0.27 Stroke mortality 0.87
12 Schoharie New York 0.25 Average HCC risk score 0.83
13 Beaver Pennsylvania 0.25 Average HCC risk score 1.00
14 Jefferson Tennessee 0.24 Average HCC risk score 1.06
15 Catoosa Georgia 0.24 Stroke mortality 0.90

Notes:

  1. Average residual score for the regression analysis involving 152 Appalachian metro counties
  2. Highest of the 19 standardized residual outcome scores for each county

Table 2
Nonmetropolitan Appalachian Bright Spot Counties, Ranked by Average Outcome Residual

Rank County State Average Standardized Residual Score a Highest Individual Residual b
1 Wayne Kentucky 0.72 Stroke mortality 1.79
2 Noxubee Mississippi 0.58 COPD mortality 2.19
3 Calhoun West Virginia 0.58 Injury mortality 2.02
4 Grant West Virginia 0.49 Cancer mortality 1.88
5 McCreary Kentucky 0.45 Poisoning mortality 1.94
6 Potter Pennsylvania 0.45 Heart disease mortality 1.44
7 Taylor West Virginia 0.42 Heart disease hospitalizations 1.20
8 Rockbridge Virginia 0.41 Heart disease hospitalizations 1.37
9 Pulaski Kentucky 0.40 Poisoning mortality 1.64
10 Green Kentucky 0.40 YPLL 1.38
11 Lee Virginia 0.40 Poisoning mortality 2.29
12 Russell Kentucky 0.40 Heart disease hospitalizations 1.68
13 Bledsoe Tennessee 0.39 Cancer mortality 1.88
14 Grayson Virginia 0.39 Injury mortality 1.83
15 Hardy West Virginia 0.38 % opioid Rx claims 1.21
16 Johnson Tennessee 0.38 Poisoning mortality 1.52
17 Lincoln Kentucky 0.37 % obese adults 1.37
18 Meigs Tennessee 0.36 % opioid Rx claims 2.17
19 Pendleton West Virginia 0.36 Poisoning mortality 1.48
20 Choctaw Mississippi 0.35 Cancer mortality 1.69
21 Adair Kentucky 0.35 Injury mortality 1.57
22 Lewis Kentucky 0.34 Depression prevalence 1.78
23 Roane West Virginia 0.33 Heart disease hospitalizations 1.35
24 Monroe Tennessee 0.32 COPD mortality 1.18
25 Alleghany North Carolina 0.31 YPLL 1.18
26 Chickasaw Mississippi 0.31 Stroke mortality 1.61
27 Morgan Kentucky 0.28 Injury mortality 0.92

Notes

  1. Average residual score for the regression analysis involving 268 Appalachian nonmetro counties
  2. Highest of the 19 standardized residual outcome scores for each county

Key Findings and Observations

Bright Spot Patterns and Clusters

The Bright Spots are not distributed evenly among the Appalachian states; Kentucky and Mississippi have proportionately more Bright Spot counties than other states. The model did not identify any Bright Spot counties in Ohio, a state with 32 Appalachian counties. The other two states with no identified Bright Spot counties, South Carolina and Maryland, have only a few Appalachian counties: six and three, respectively. The absence of Bright Spots in these two states may be the result of small sample sizes, whereas the Ohio result suggests a pattern of lower-than-expected outcomes.

Several Bright Spot counties appear in geographic clusters, suggesting that factors leading to better-than-expected health outcomes may prevail across broad, multicounty areas. Clustering suggests the presence of some common factor that has improved the health of the cluster. The unit of analysis, the county, may be a proxy for a larger “community.” These communities may be in the service area of a particularly effective program, health care provider, or other resource. Alternatively, other factors, such as environment, local culture, and tradition, may also support a culture of health.

Correlation of Specific Outcomes with Overall Health

Our approach allows us to broadly measure health in a community and determine whether that community exceeds expectations. We developed the average standardized outcome residual because it captures the degree to which a county’s outcomes exceeded expectations.

It is important to keep in mind that the average standardized residual does not represent the entire composition of a county’s health status. For individual outcomes, even among counties identified as Bright Spots, there were still lower-than-expected values. These results suggest that community health cannot be painted with one broad brushstroke; rather, it is more accurately represented as a multidimensional combination of many different aspects of health.

By using the actual value of outcomes, we were able to find certain individual outcome measures that were more highly correlated with overall good health outcomes. Results for 3 of the 19 health outcome measures were consistently better than expected in the Bright Spot counties:

  • Premature mortality;
  • Unintentional injury mortality; and
  • Poisoning mortality.

Premature mortality (described as “years of potential life lost,” or YPLL) had the highest correlation with the average standardized residual, supporting its use as a comprehensive measure of community health. Further, outcomes such as injury mortality and poisoning mortality were highly correlated with average standardized outcome residuals in the top-performing counties. Bright Spot counties—those with average outcome residuals in the top ten percent—tended to have better-than-expected poisoning mortality rates.

Poisoning mortality includes deaths due to drug overdose. Among the ten Appalachian counties with the lowest average standardized residuals in both the metropolitan and nonmetropolitan groups—20 counties altogether—only one county performed better than expected on poisoning mortality; many others had much higher poisoning mortality rates than expected.

This suggests that poisoning mortality—and by extension, substance abuse—may have an important link to overall health for all counties.

Seven High-Impact Drivers

The results of this analysis suggest that the following seven drivers predicted the most variation in the 19 health outcomes. The relative driver level generally associated with better health is shown in parentheses (for example, a higher median income level is associated with better health outcomes).

  • Median income (higher);
  • ARC Economic Index value (lower);
  • Poverty rate (lower);
  • Percentage of adults who smoke (lower);
  • Percentage of adults who are physically inactive (lower);
  • Percentage of the population receiving disability payments (lower); and,
  • Teen birth rates (lower).

These seven drivers were better predictors of health outcomes in the Bright Spot counties than drivers describing the supply of health resources, such as the supply of primary care physicians or the supply of specialty physicians.

These findings suggest that focusing on improvements in these seven drivers may lead to the greatest overall impact on health in a community.

Complete information on the Bright Spot methodology and analysis is available in the report Identifying Bright Spots in Appalachia: Statistical Analysis. See the data files for that report for the analysis results.