Table 4. Selection of variables for inclusion in multi-variable mixed logit regression models designed to test the association between Management and Uncontrolled Hypertension.
| Potential predictors of Uncontrolled Hypertension | Model contributions LR test χ2 (p-value) | d.f. (Degree of freedom) |
|---|---|---|
| Key variable of interest | ||
| • Management | 44.7 (p<0.001) | 1 |
| Potential confounders | ||
| • District | 227.0 (p<0.001) | 33 |
| • Income | 26.1 (p<0.001) | 5 |
| • Compliance | 81.2 (p<0.001) | 2 |
| • Exercise | 12.9 (p = 0.012) | 4 |
| • Education | 29.5 (p<0.001) | 7 |
| • Wellbeing | 40.1 (p<0.001) | 1 |
| • Age | 8.4 (p = 0.004) | 1 |
| • Smoking | 10.6 (p = 0.031) | 4 |
| Random effects | ||
| • Street / Township clusters (ICC a ± s.e.) | 9.0e-35 ± 7.4e-19 | 1 |
| • Neighbourhood committee / Village (ICC a ± s.e.) | 0.19 ± .04 | 1 |
| • LR test [χ2 (p-value)] | 46.6 (p<0.001) | 1 |
| Full model | ||
| • Combined predictors | 1,235.0 (p<0.001) | 59 |
| • Sample size | 7,867 | --- |
| Variables excluded from the final model | ||
| • Alcohol | 2.9 (p = 0.089) | 1 |
| • Gender | 1.0 (p = 0.327) | 1 |
| • Insurance | 3.7 (p = 0.593) | 5 |
| • Hukou | 1.4 (p = 0.245) | 1 |
| • Urban / Rural residence | <0.1 (p = 0.899) | 1 |
a Intra-class correlation.