Table 2.
Total Effect and Controlled Direct Effect of Home Owners’ Loan Corporation Ranking on Women’s Breast and Cervical Cancer Stage at Diagnosis and Residual Disparity, Using Census Tract Index of Concentration at the Extremes for Racialized Economic Segregation as a Mediator, Massachusetts, 2001–2015a
| Breast Cancer | Cervical Cancer | |||
|---|---|---|---|---|
| Comparison | RR | 95% CI | RR | 95% CI |
| Total Effect b | ||||
| HOLC ranking | ||||
| Green + blue (best; referent) | 1.00 | Referent | 1.00 | Referent |
| Yellow | 1.07 | 0.99, 1.15 | 0.90 | 0.69, 1.19 |
| Red | 1.07 | 0.98, 1.17 | 0.98 | 0.73, 1.33 |
| Mixed HOLC grades | 1.02 | 0.93, 1.11 | 0.76 | 0.54, 1.09 |
| No grade assigned | 1.03 | 0.96, 1.11 | 0.88 | 0.66, 1.17 |
| Controlled Direct Effect c | ||||
| HOLC ranking by CT ICE tercile | ||||
| ICE tercile 1 | ||||
| Green + blue | 1.00 | Referent | 1.00 | Referent |
| Yellow | 1.09 | 0.97, 1.22 | 0.76 | 0.44, 1.33 |
| Red | 1.15 | 0.95, 1.39 | 0.57 | 0.23, 1.40 |
| Mixed | 0.98 | 0.87, 1.10 | 0.89 | 0.48, 1.64 |
| No grade | 1.04 | 0.94, 1.14 | 0.75 | 0.48, 1.17 |
| ICE terciles 2 and 3 | ||||
| Green + blued | 1.23 | 1.08, 1.40 | 1.22 | 0.73, 2.01 |
| Green + blue | 1.00 | Referent | 1.00 | Referent |
| Yellowe | 0.94 | 0.84, 1.05 | 0.83 | 0.57, 1.21 |
| Rede | 0.94 | 0.83, 1.06 | 0.92 | 0.62, 1.37 |
| Mixede | 1.02 | 0.88, 1.17 | 0.67 | 0.42, 1.07 |
| No gradee | 0.94 | 0.84, 1.06 | 0.87 | 0.59, 1.29 |
| Residual Disparity f | ||||
| HOLC ranking | ||||
| Green + blue (best; referent) | 1.00 | Referent | 1.00 | Referent |
| Yellow | 1.03 | 0.95, 1.12 | 0.80 | 0.54, 1.11 |
| Red | 1.07 | 0.93, 1.21 | 0.72 | 0.00, 1.13 |
| Mixed HOLC grades | 0.99 | 0.90, 1.09 | 0.78 | 0.46, 1.15 |
| No grade assigned | 1.00 | 0.93, 1.09 | 0.80 | 0.62, 1.10 |
Abbreviations: CI, confidence interval; CT, census tract; HOLC, Home Owners’ Loan Corporation; ICE, index of concentration at the extremes; RR, risk ratio.
aMassachusetts Cancer Registry data (2001–2015) for the 28 municipalities with HOLC rankings (1937–1938).
bTotal effect of HOLC ranking estimated using quasi-Poisson models with indirect age standardization and adjustment for sex/gender and race/ethnicity. All models used HOLC categories weighted by land area. All tests for statistical significance were 2-sided.
cControlled direct effect estimated using quasi-Poisson models with indirect age standardization and adjustment for sex/gender and race/ethnicity. Results represent the CT ICE stratum for racialized economic-segregation–specific HOLC effects, sex/gender- and race/ethnicity-adjusted. All models used HOLC categories weighted by land area. Tercile cutpoints based on the total distribution of CT ICE for race/ethnicity + income for the Massachusetts population (2001–2015) were 0.17 and 0.34.
dRelative to green + blue in CT ICE category T1 (tercile 1). This reminds us that the baseline risk of late-stage diagnosis is substantially elevated in green + blue areas in CT ICE category T2 + T3 (terciles 2 and 3).
eRelative to green + blue in CT ICE category T2 + T3. These are controlled direct effects.
fResidual disparity estimated using logistic regression for ICE for race/ethnicity + income (mediator) and quasi-Poisson models for the outcome. All models adjusted for race/ethnicity and sex/gender and used indirect age standardization. This residual disparity was estimated using the nonparametric simulation approach described by Imai et al. (41) and can be interpreted as the health disparity between persons in CTs with HOLC grades of yellow, red, or mixed or no grade and persons in CTs with HOLC grades of green + blue, if the ICE distribution of the persons in HOLC yellow, red, mixed, or no-grade CTs were set to a random draw from the distribution among HOLC green + blue CTs.