Table 4.
Adjusted regression models for social contact and support and cognitive health.
| Variable (unit of change) | Cognitive impairment * |
Standardized cognition score (SD) ** |
||
|---|---|---|---|---|
| Poisson model | Linear model | |||
| Named contacts (count) | 0.75 | [0.68, 0.83] | 0.03 | [0.01, 0.04] |
| Communication events per month (SD) | 0.71 | [0.60, 0.83] | 0.05 | [0.02, 0.07] |
| Contact kinship | ||||
| Named kin (count) | 0.84 | [0.77, 0.91] | 0.01 | [0.00, 0.02] |
| Any named non-kin (binary) | 0.55 | [0.42, 0.73] | 0.07 | [0.03, 0.11] |
| Percent named contacts who are kin (10 %age points)† | 1.06 | [1.02, 1.09] | −0.01 | [-0.02, 0.00] |
| Contact distance‡ | ||||
| Same household (count) | 0.85 | [0.75, 0.97] | −0.02 | [-0.05, 0.01] |
| Same village (count) | 0.70 | [0.61, 0.80] | 0.03 | [0.01, 0.05] |
| Agincourt area (count) | 0.75 | [0.63, 0.90] | 0.06 | [0.03, 0.09] |
| Elsewhere South Africa (count) | 0.74 | [0.65, 0.86] | 0.03 | [0.01, 0.05] |
| Egonet effective size | ||||
| Effective size (range 1–5.7)† | 0.87 | [0.63, 1.21] | 0.04 | [0.00, 0.09] |
| Support types | ||||
| Informational, per month (SD) | 0.73 | [0.64, 0.82] | 0.05 | [0.02, 0.07] |
| Emotional, per month (SD) | 0.72 | [0.63, 0.82] | 0.07 | [0.05, 0.10] |
| Financial, per month (SD) | 0.87 | [0.79, 0.96] | 0.02 | [0.00, 0.05] |
| Physical, per month (SD) | 0.78 | [0.69, 0.86] | 0.03 | [0.00, 0.05] |
Note. Each regression coefficient represents results from a different model, with the exception of ‘contact distance’ models (‡), where all four variables were included in a single regression. All models are hierarchical (individuals nested in interviewers) using either Poisson with robust error variance or linear models and are adjusted for age, gender, employment status, household size, household wealth, educational attainment, literacy, marital status, father's occupation, childhood health status, and interview month. †Models for percent of contacts who are kin and egonet effective size are also adjusted for the number of contacts named. *N for ‘cognitive impairment’ models is 4973 and values are prevalence rate ratios (and 95% confidence intervals). **N for ‘cognition score’ models is 4888 and values are the difference in standardized cognition score (in SDs) associated with a one-unit higher value of the variables as shown in Table 2. IQR: Interquartile range; SD: Standard deviation.