# Technical Notes for the Healthy People 2020 Overview of Health Disparities

The *Healthy People 2020 *(HP2020) *Final Review* includes an evaluation of how health disparities have changed over time during the HP2020 tracking period for 611 objectives. The analysis included data on changes in disparities for the following six population characteristics: sex, race and ethnicity, educational attainment, family income, disability status, and geographic location. This document provides technical information on the criteria used for population characteristics and data years in the analyses, and the measures used to compute disparities and changes over time.

## Criteria for Including Objectives and Determining Baseline and Final Years for the Disparities Analysis

The inclusion of objectives and the baseline and final years of data used for the disparities analysis was based on the following stepwise criteria.

- The objectives had at least two nonoverlapping time points by sex, race and ethnicity, educational attainment, family income, disability status, or geographic location.
- For specific population characteristics, the following criteria were additionally applied:
- Sex, disability status, and geographic location: An objective must include at least two identical groups for at least two nonoverlapping time points to be included in the analysis.
- Race and Ethnicity: An objective must include the following “minimum set” of three groups for at least two nonoverlapping time points to be included in the analysis: White, not Hispanic or Latino, Black, not Hispanic or Latino, and Hispanic or Latino.
- Educational attainment and family income: An objective must include at least three identical groups for at least two nonoverlapping time points to be included in the analysis. Additionally, the objective was required to have data for the lowest and highest family income or educational attainment groups to be included, because these extremes were considered to likely have a large influence on driving disparities for many objectives (for example, less than 100% and 600% or greater than the poverty threshold for family income, and less than high school and advanced degree for educational attainment). Given the variability in how these population characteristics were measured across HP2020 data sources, the specific groups that represent the lowest and highest categories differ across objectives in the analysis.

The earliest and latest years during the HP2020 tracking period that met the minimum set criteria listed in step 2 were used as the baseline and final years of the disparities analysis. For race and ethnicity, educational attainment, family income, and geographic location, additional groups beyond the minimum number were included in the disparities analysis if they had data available at both the baseline and final years. When possible (see next section), population groups included in disparities analyses were defined using mutually exclusive categories.

The baseline and final years used in the disparities analysis for different population characteristics within the same objective, and for the same population characteristic across objectives (even among those that have the same data source), may differ depending on population group reporting and data availability during the years in the HP2020 tracking period. Additionally, after applying the criteria described, the baseline and final years for each objective in the disparities analysis do not necessarily correspond to the HP2020 baseline and final years for the total population that were used in the HP2020 Progress Table or the HP2020 Progress by Population Group analysis to evaluate target attainment.

The following table presents the number of objectives for each of the six population characteristics that were included in the disparities analysis.

Population characteristic | Number of objectives included in the disparities analysis^{1} |
---|---|

Sex | 529 |

Race and ethnicity | 506 |

Educational attainment | 223 |

Family income | 292 |

Disability status | 156 |

Geographic location | 324 |

^{1} A total of 611 objectives with data that met the criteria for at least one of the six population characteristics were included in the disparities analysis. The number of objectives with available data that met the inclusion criteria for each population characteristic individually is shown in this Table. Data for selected informational objectives and measurable objectives that provided data as counts (as opposed to proportions or rates) were excluded from the disparities analysis.

## Population Group Considerations for Race and Ethnicity in the Overview of Health Disparities Analysis

For several HP2020 objectives, data were reported for racial and ethnic groups during the same year that were not mutually exclusive from each other. The Overview of Health Disparities analyses were conducted using independent groups for race and ethnicity, using the following criteria for scenarios where this overlap in groups occurred. Specifically, data for the separate groups Asian and Native Hawaiian or Other Pacific Islander (with or without specification of Hispanic or Latino ethnicity, see below) were included when available, and were prioritized over data for the combined group Asian or Pacific Islander (API) when reported in the same year. However, if an objective reported data for the combined API category and not Asian only and Native Hawaiian or Other Pacific Islander at the baseline and final year, the API data were then used for the disparities analysis instead.

The groups American Indian or Alaska Native, not Hispanic or Latino; Asian or Pacific Islander, not Hispanic or Latino; Asian, not Hispanic or Latino; Two or more races, not Hispanic or Latino; or Native Hawaiian or Other Pacific Islander, not Hispanic or Latino were used instead of American Indian or Alaska Native, Asian or Pacific Islander, Asian, Two or more races, or Native Hawaiian or Other Pacific Islander, respectively, when data by Hispanic origin were reported for those race groups at the baseline and final time points. When data by Hispanic origin for those race groups were not available at the baseline and final time points, then data for the groups American Indian or Alaska Native, Asian or Pacific Islander, Asian, Two or more races, or Native Hawaiian or Other Pacific Islander were used instead if available.

Some objectives only included data for the groups White or Black, with no specification of Hispanic ethnicity. These objectives were not included in the assessment of disparities by race and ethnicity, given the potential for a high degree of overlap between the groups White or Black and Hispanic or Latino. However, data for these groups are still provided in the Disparities Tool on the HP2020 website.

## Statistical Measures and Formulas Used in the Overview of Health Disparities Analysis

HP2020 has defined and applied a suite of measures to evaluate health disparities during the HP2020 tracking period. The theoretical background and calculations pertaining to this suite of measures are discussed in Healthy People Statistical Note 27 (1). The focus of the Overview of Health Disparities analysis was to quantify the change in disparities and assess the progress in eliminating health disparities over the HP2020 tracking period. This section focuses on the measures and corresponding formulas that were used specifically for the change in disparities analysis. Disparities data for individual HP2020 data years are available using the Disparities Tool on the HP2020 website, and in the Overview of Health Disparities downloadable data file for the baseline and final data year(s) included in this analysis.

### Disparities calculations for baseline and final HP2020 data year(s)

For population characteristics with three or more groups (for example, race and ethnicity, educational attainment, and family income), the summary rate ratio (*RR*_{ave}) measure was used for the disparities analysis. *RR*_{ave} was computed separately for the HP2020 baseline and final data years by taking the ratio of the “best” (most favorable or least adverse) group rate and the average rate of all other groups for that population characteristic. Note that “rate” in this case may refer to a statistical rate expressed per unit population or a proportion, depending on how the HP2020 objective was defined. The procedure for computing *RR*_{ave} is as follows [1]:

[1] |
${R}_{A}=\mathrm{}\frac{{R}_{1}+{R}_{2}+...+\mathrm{}{R}_{(K-1)}}{K-1}$ ${RR}_{\mathrm{a}\mathrm{v}\mathrm{e}}=\mathrm{m}\mathrm{a}\mathrm{x}\left\{\frac{{R}_{B}}{{R}_{A},}\frac{{R}_{A}}{{R}_{B}}\right\}$ |

where *K* represents the total number of population groups included for analysis within a given population characteristic (for example, sex, race and ethnicity), *R _{A}* represents the average of the

*K*-1 population group rates other than the best rate, and

*R*represents the group with the best rate.

_{B}

The value of *RR*_{ave} will be greater than or equal to 1, regardless of whether the objective is expressed in terms of a favorable outcome to be increased, in which case *RR*_{ave} *= $\frac{{R}_{B}}{{R}_{A}}$*, or an adverse outcome to be decreased, in which case *RR*_{ave} = $\frac{{R}_{A}}{{R}_{B}}$. Assuming the population group rates are independent and *K* is greater than or equal to 3, the standard errors (SE) of *R _{A}* [2] and

*RR*

_{ave}[3] are evaluated using the following formulas, respectively:

[2] |
$S{E}_{A}=\sqrt{\frac{S{E}_{1}^{2}+S{E}_{2}^{2}+S{E}_{K-1}^{2}}{K-2}}$ |

[3] |
$S{E}_{R{R}_{\mathrm{a}\mathrm{\nu}\mathrm{e}}}=R{R}_{\mathrm{a}\mathrm{v}\mathrm{e}}\mathrm{}\times \mathrm{}\sqrt{{[\left(\frac{S{E}_{A}}{{R}_{A}}\right)}^{2}+{\left(\frac{S{E}_{B}}{{R}_{B}}\right)}^{2}]}$ |

For population characteristics with only two groups, disparities for the baseline and final data year(s) were calculated instead using the maximal rate ratio (*RR*_{max}). *RR*_{max} is computed by taking the ratio of the best group rate and the “worst” group rate (the group with the least favorable or most adverse rate) for that population characteristic. The formulas [4] for *RR*_{max} and the SE [5] are as follows, where *R _{B}* represents the best group rate, and

*R*represents the worst group rate:

_{W}[4] |
${RR}_{\mathrm{m}\mathrm{a}\mathrm{x}}=\mathrm{m}\mathrm{a}\mathrm{x}\left\{\frac{{R}_{B}}{{R}_{W},}\frac{{R}_{W}}{{R}_{B}}\right\}$ |

[5] |
$S{E}_{R{R}_{\mathrm{m}\mathrm{a}\mathrm{x}}}=R{R}_{\mathrm{m}\mathrm{a}\mathrm{x}}\mathrm{}\times \mathrm{}\sqrt{{[\left(\frac{S{E}_{W}}{{R}_{W}}\right)}^{2}+{\left(\frac{S{E}_{B}}{{R}_{B}}\right)}^{2}]}$ |

Because the best group rate for all population characteristics was computed separately for both the baseline and final years in the disparities analysis, the best group is not necessarily the same at both data points for a given objective. In this scenario, the computation of *RR*_{ave }or *RR*_{max} follows the same procedure as described previously, but differences between the baseline and final years regarding the population group with the best rate should be considered when interpreting the HP2020 objective-specific results in the data file.

### Handling “ties” for population groups with the same rate

For the* RR*_{ave} calculation, which was used for population characteristics with three or more groups, the population group with the best rate was compared with the average rate of all other population groups for a given population characteristic. For *RR*_{max}, which was used for population characteristics with only two groups, the population group with the best rate was compared with the rate for the other population group within that characteristic. Data used for the disparities analysis were unrounded when available, although some objectives only had rounded data in the HP2020 database. If two population groups had the same rates that resulted in a tie when identifying the best group, then the group with the lower SE for that rate was identified as the best group. If no SE was reported, then the population group with the larger denominator from the survey question or the group with the larger population size in the United States during the data-collection year was used as second- and third-tier criteria, respectively, to identify the best group. Note that the other tied population group(s) that was (were) not identified as having the best group rate was still included in the calculation of *RR*_{ave} or *RR*_{max}, so the value of these measures would be equivalent regardless of which of the tied group rates was reported as the best.

### Changes in disparities during the HP2020 tracking period

In the Overview of Health Disparities analysis, changes in disparities were quantified by taking the absolute difference in the computed rate ratios (*RR*_{ave} for population characteristics with three or more groups and *RR*_{max} for population characteristics with two groups) between the baseline and final HP2020 data years in the disparities analysis [6], where *RR* represents either *RR*_{ave} or *RR*_{max}. When measures of variability were available for individual population group estimates in the data system for an objective, and so, the SEs of *RR*_{ave} and *RR*_{max }could be calculated, *z* scores [7] were computed to evaluate the statistical significance of the change in disparities. The change in the rate ratio was considered statistically significant at the 0.05 level using a two-sided test if |*z*| > 1.96.

[6] |
Disparities change (DC) = RR (final data year(s)) – RR (baseline data year(s)) |

[7] |
$z=\frac{R{R}_{\left(\mathrm{f}\mathrm{i}\mathrm{n}\mathrm{a}\mathrm{l}\right)}-R{R}_{\left(\mathrm{b}\mathrm{a}\mathrm{s}\mathrm{e}\mathrm{l}\mathrm{i}\mathrm{n}\mathrm{e}\right)}}{\sqrt{S{E}_{RR\left(\mathrm{f}\mathrm{i}\mathrm{n}\mathrm{a}\mathrm{l}\right)}^{2}+S{E}_{RR\left(\mathrm{b}\mathrm{a}\mathrm{s}\mathrm{e}\mathrm{l}\mathrm{i}\mathrm{n}\mathrm{e}\right)}^{2}}}$ |

### Change in disparities status categories

The changes in disparities for the HP2020 Overview of Health Disparities analysis were summarized using three categories: decrease, little or no detectable change, and increase. The criteria used to define these categories were as follows:

- Decrease: Difference in the summary rate ratio or the maximal rate ratio between the final and baseline data year was negative and statistically significant when this could be assessed, or negative and |
*DC*| ≥ 0.10 when statistical significance could not be assessed. - Little or no detectable change: Difference in the summary rate ratio or the maximal rate ratio between the final and baseline data year was not statistically significant when this could be assessed, or |
*DC*| < 0.10 when statistical significance could not be assessed. - Increase: Difference in the summary rate ratio or the maximal rate ratio between the final and baseline data year was positive and statistically significant when this could be assessed, or positive and
*DC*≥ 0.10 when statistical significance could not be assessed.

## Reference

- Talih M, Huang DT. Measuring progress toward target attainment and the elimination of health disparities in Healthy People 2020. Healthy People Statistical Notes, no 27. Hyattsville, MD: National Center for Health Statistics. 2016. Available from: https://www.cdc.gov/nchs/data/statnt/statnt27.pdf.