Table 2. Results of Mixed-Effects Regression Models Associating Black PCP Representation and County-Level Covariates With Study Outcomes and Moderation Analysisa.
| Variable | Survival outcome for Black individuals | |||
|---|---|---|---|---|
| Model 1: life expectancy, y (95% CI) | Model 2: all-cause mortality rate (95% CI) | Model 3: log(mortality rate disparity between Black and White individuals) (95% CI) | Model 4: statistical moderation, life expectancy, y (95% CI) | |
| Log(Black PCP workforce ratio) | ||||
| Between counties | 0.88 (0.55 to 1.22) | −133.37 (−154.93 to −111.82) | −12.19 (−13.43 to −10.95) | 0.59 (0.39 to 0.79) |
| Within counties | 0.04 (−0.21 to 0.30) | −35.34 (−58.86 to −11.81) | −2.44 (−3.65 to −1.23) | 0.06 (−0.19 to 0.32) |
| Interaction of between-counties log(Black PCP workforce ratio) × Poverty rate | NAb | NA | NA | 0.04 (0.01 to 0.07) |
| Poverty rate | ||||
| Between counties | −0.10 (−0.15 to −0.06) | 1.30 (−3.04 to 5.63) | −0.14 (−0.41 to 0.12) | −0.09 (−0.13 to −0.04) |
| Within counties | 0.00 (−0.07 to 0.06) | 2.25 (−5.10 to 9.59) | 0.16 (−0.20 to 0.52) | −0.00 (−0.06 to 0.06) |
| Uninsured rate | ||||
| Between counties | −0.06 (−0.11 to −0.01) | 1.55 (−3.36 to 6.46) | 0.14 (−0.16 to 0.44) | −0.06 (−0.11 to −0.01) |
| Within counties | −0.14 (−0.23 to −0.06) | 0.35 (−9.45 to 10.15) | 0.05 (−0.42 to 0.52) | −0.14 (−0.22 to −0.05) |
| Timec | −0.23 (−0.52 to 0.06) | −17.57 (−51.43 to 16.29) | −0.03 (−1.65 to 1.59) | −0.22 (−0.51 to 0.07) |
| Ratio of men per 100 women, % | 0.04 (0.02 to 0.06) | −3.88 (−5.65 to −2.10) | −0.21 (−0.32 to −0.11) | 0.04 (0.02 to 0.06) |
| Rural status | 1.06 (0.63 to 1.50) | −24.10 (−65.56 to 17.35) | −1.83 (−4.36 to 0.70) | 1.06 (0.63 to 1.49) |
| Home value, $, median (95% CI) | 0.02 (−0.00 to 0.04) | 0.56 (−1.58 to 2.70) | 0.20 (0.07 to 0.34) | 0.02 (−0.00 to 0.04) |
| Hispanic population, % | 0.04 (0.02 to 0.06) | −1.12 (−3.26 to 1.02) | −0.02 (−0.15 to 0.11) | 0.03 (0.01 to 0.06) |
| Unemployed rate, % | 0.11 (−0.05 to 0.26) | 2.01 (−12.67 to 16.68) | 0.73 (−0.17 to 1.62) | 0.10 (−0.05 to 0.25) |
| Less than high school education, % | 0.03 (−0.03 to 0.09) | 0.02 (−5.86 to 5.90) | −0.09 (−0.45 to 0.27) | 0.02 (−0.04 to 0.09) |
| Adult obesity, % | −0.08 (−0.12 to −0.04) | −0.80 (−4.54 to 2.94) | −0.20 (−0.42 to 0.03) | −0.07 (−0.11 to −0.03) |
| Adult smoking, % | −0.11 (−0.20 to −0.03) | 21.45 (12.95 to 29.95) | 0.57 (0.05 to 1.09) | −0.13 (−0.22 to −0.04) |
| Medicare enrollment, % | −0.02 (−0.06 to 0.02) | 0.00 (−4.01 to 4.01) | −0.08 (−0.32 to 0.17) | −0.02 (−0.06 to 0.02) |
| Air pollution | −0.13 (−0.24 to −0.01) | 20.74 (9.31 to 32.16) | 1.27 (0.57 to 1.98) | −0.13 (−0.24 to −0.01) |
| Log(No. of hospital beds) | −2.65 (−0.95 to 4.35) | −95.62 (−268.90 to 77.66) | −1.20 (−12.10 to 9.71) | −2.46 (−0.75 to 4.16) |
Abbreviations: NA, not applicable; PCP, primary care physician.
Median age was not included as a final study covariate due to issues regarding collinearity and a zero-order Pearson correlation between median age and Medicare enrollment percentage that exceeded 0.85. Unstandardized fixed effects with corresponding 95% CIs for between-county influences reported in Table 2 will neither match previously reported results of a 30.61-day increase in life expectancy, nor match between-county influence results for specified reductions in all-cause mortality and mortality rate disparities between Black and White individuals associated with a 10% increase in the (log-transformed) Black PCP workforce ratio.
Cells with NA denote cases where mixed-effects regression models did not include tests for statistical moderation, such that the interaction between the Black PCP workforce representativeness ratio and county-level poverty rates was not examined.
Time refers to a constructed variable generated such that 2009 was coded as 0, 2014 was coded as 1, and 2019 was coded as 2. This variable was used to assess whether each survival outcome exhibited statistically significant increases or decreases over time.