Table 2.
Multivariable linear regression: estimates for log(time delay to testing) for 517 patients in HIV care services in rural Zambézia Province, Mozambique, including persons who did and did not seek initial care from a Traditional Healer
| Estimatea (95% CI) | P-value | |
|---|---|---|
| Female | 1.31 (1.05, 1.63) | 0.019 |
| Age | 0.018 | |
| 20 years | 0.70 (0.55, 0.90) | |
| 30 years (ref) | 1 | |
| 40 years | 1.03 (0.92, 1.15) | |
| 50 years | 0.90 (0.66, 1.21) | |
| Education (per 2 years) | 1.13 (1.00, 1.28) | 0.058 |
| Travel time to clinic (per 30 minutes) | 1.01 (0.94, 1.08) | 0.79 |
| Transportation | 0.66 | |
| Walk (ref) | 1 | |
| Bike | 1.14 (0.84, 1.54) | |
| Car/motorcycle | 1.09 (0.81, 1.45) | |
| Traditional healer (TH) use | <0.001 | |
| Did not consult TH (ref) | 1 | |
| Consulted 1 TH | 1.47 (1.12, 1.93) | |
| Consulted 2 TH | 2.45 (1.73, 3.47) | |
| Consulted 3 TH | 2.12 (1.42, 3.16) | |
| Consulted 4 TH | 2.90 (1.78, 4.73) | |
| Consulted 5 or more TH | 5.69 (4.16, 7.78) | |
| Consulted religious leaderb | 1.13 (0.83, 1.54) | 0.45 |
| Community | <0.001 | |
| Namacurra (ref) | 1 | |
| Inhassunge | 0.52 (0.40, 0.66) |
a
Because duration from symptom to testing is log-transformed, the model effect is the ratio of the geometric mean for the comparison and reference groups. Thus, we are able to summarize delay time in terms of percent change.
b
There is little evidence of a hypothesized interaction between use of traditional and/or religious healers (p=0.78); the interaction term was not included in this final model.