Table 4.
Growth Curve Models Predicting Frailty Index for Race/Ethnic Subgroups With Statistically Significant Interaction Effects
| White (non-Hispanic; N = 54,178 obs.) | Blacks (non-Hispanic; N = 9,484 obs.) | |
|---|---|---|
| Fixed effects | ||
| Intercept | 28.297*** (0.660) | 29.219*** (1.417) |
| Household residence | ||
| Multigenerational household | 0.166 (1.113) | 0.232 (2.349) |
| Skipped generation household (Ref. cat. = no G’Child in household) | 9.469*** (1.398) | 11.843*** (3.259) |
| Socioeconomic status | ||
| Education (years) | −0.807*** (0.038) | −0.748*** (0.088) |
| Income (Ln) | −0.124** (0.044) | −0.202* (0.103) |
| Net wealth (/100,000) | −0.005 (0.005) | −0.258 (0.136) |
| Social support | ||
| Frequency of interaction/week | −0.049*** (0.009) | 0.021 (0.014) |
| Interaction effects | ||
| Multigen.HH × Education (years) | 0.020 (0.088) | −0.354** (0.127) |
| Skip.Gen.HH × Education (years) | −0.721*** (0.114) | −0.014 (0.163) |
| Multigen.HH × Income (Ln) | 0.399 (0.221) | |
| Skip.Gen.HH × Income (Ln) | −0.912** (0.312) | |
| Multigen.HH × Net Wealth | 0.162 (0.125) | |
| Skip.Gen.HH × Net Wealth | −0.342** (0.124) | |
| Multigen.HH × Freq. of Interaction | 0.047*** (0.011) | 0.016 (0.024) |
| Skip.Gen.HH × Freq. of Interaction | −0.196** (0.072) | −0.234** (0.088) |
| Random effects | ||
| Level 1 | ||
| Within-person | 75.787*** | 108.070*** |
| Level 2 | ||
| In intercept | 0.272** | 0.919** |
| In linear growth rate | 0.356*** | 0.439*** |
| Goodness of fit | ||
| AIC | 331,949.000 | 58,619.800 |
| BIC | 331,979.000 | 58,643.200 |
Notes. AIC = Akaike information criterion; BIC = Bayesian information criterion; Multigen.HH = multigenerational household; Skip.Gen.HH = skipped generation household. All models include individual-level covariates (age, age squared, female, foreign born, caregiving hours, household residence, education, income, net wealth, has long-term care insurance, currently receiving pension, currently working for pay, married/partnered, relative living nearby, friend living nearby, frequency of interaction, deceased, loss to follow-up). Results adjusted for propensity score weighting, respectively, for each race/ethnic subgroup. Results for Hispanic sample are not presented because no significant interaction effect is found.
*p < .05. **p < .01. ***p < .001.