Documentation

National DHI Scores

To create the National DHI score for each tract, we first calculated the percentile rank for each of the 21 variables. The percentile represents where a tract falls relative to all other tracts for a given variable. For example, if a tract had a percentile rank of .25 for Percent of the Population on Medicare, it means that 75% of the tracts in the US have a higher value for Percent of the Population on Medicare. For our scoring, higher percentile ranks indicate more vulnerability or a higher likelihood of adverse outcomes. For a few variables, a higher percentile score represented less vulnerability (for example Average Upload/Download Speed). In these cases, the scores were reverse coded such that scores of 0 became 1, .25 became .75, etc.

Other notable exceptions for the percentile calculations were for some of the Access variables. For the percentage of the population that does not have access to an Emergency Department or an Urgent Care facility, since the distribution of the coverage across the US didn’t have a normal distribution, we instead treated the percentage not covered as a percentile. For example, if 75% of the population of a tract was not covered by a 30-minute service area around an emergency department, we treated that coverage as a percentile rank of .75. The variable for Medically Underserved Area/Medically Underserved Population was also treated in a different way because of the way that the source data was coded and the geographic units used for the source data. HRSA data were provided in 2010 Census boundaries. Data were converted to proportionate coverage of 2020 Census boundaries. Most values were 1 or 0 (in or not in an underserved area, respectively) already and the remaining partial intersections were close to those extremes. For example, if a tract in 2020 was in an underserved area, and split into two tracts in 2020, both of the new tracts were designated as in an underserved area. As a result, any tracts that were greater than 50% within an underserved area were assigned a 1, and all other tracts were assigned a 0.

Once the percentile ranks were calculated for each variable, the sum was calculated across all variables for each tract. The result was an aggregate score that could fall between 0 and 21 (0 if a tract was in the very bottom percentile across all variables, or 21 if a variable was in the top percentile for all variables). In reality, the aggregate scores ranged from ~5 to ~18. To make the scores easier to interpret, they were rescaled to fall between 0 and 1, with scores closer to 0 corresponding to areas with less vulnerability and scores closer to 1 corresponding to areas with greater vulnerability. The rescaling formula is as follows:

(aggregate percentile score – minimum percentile score for all tracts) / (maximum percentile score for all tracts – minimum percentile score for all tracts)

Aggregate and rescaled scores were calculated for all variables to create an Overall DHI score, as well as for each of the three themes to create aggregate and rescaled Social, Access, and Digital scores.

State DHI Scores

To better understand how tracts within a state compare to other tracts within that state, state-specific Overall, Social, Access, and Digital DHI scores were calculated following the process outlined above. To create the state scores, only the tracts for a given state were used when calculating the percentile ranks for each variable. In addition, after aggregating the percentile ranks for each variable to create the aggregate percentile scores, the minimum and maximum score for a given state were used to calculate the 0 to 1 rescaled scores. For example, if North Carolina had a minimum aggregate score of 6 and maximum aggregate score of 15, those would be the values used in the rescaling formula. Given the differences between minimum and maximum scores for different states, the state DHI scores should not be used to compare different states. Following the North Carolina example above, if the minimum and maximum scores for Mississippi were 8 and 18, the resulting rescaled scores would have the same range (0 to 1), but would obscure the fact that Mississippi has higher risk when looking at the raw values for the index variables.

Aggregate and rescaled scores were calculated for all variables to create Overall State DHI scores, as well as for each of the three themes to create aggregate and rescaled Social, Access, and Digital state scores.

County DHI Scores

To provide DHI information in geographic units that are more familiar to many users and to help visualize and understand broader trends and patterns, County DHI scores were calculated. Since tracts nest within counties, the tract-level DHI scores outlined above were used to create the county scores. For each county, the mean National and State Overall, Social, Access, and Digital scores were calculated from the tracts that fall within that county. Given that tracts have a target population of 4000 people and are therefore nominally the same size, a simple arithmetic mean was calculated as opposed to a mean calculation that incorporated the size of the tracts in determining the average for a county. The resulting County DHI scores can be used to compare counties from across the US or to compare counties that fall within the same state.