Table 2.
Longitudinal regression analysis of the pre-post control group design
| LLIN | IRS | EVM | ||
| Model | Parameter | Coefficient (P-value) | P-value | P-value |
| Simple (crude estimates) | Intervention effect† | -0.43 (0.042) | -0.95 (<0.001) | -0.49 (0.024) |
| Reduction in counts (95% CI)‡ | -4.34 (-8.57,-0.10) 43.7% |
-8.92 (-13.20,-4.64) 72.4% |
-5.55 (-10.57,-0.53) 42.0% |
|
| Full | Intervention effect | -0.42 (0.044) | -0.94 (0.001) | -0.49 (0.025) |
| Type of wall | -0.02 (0.881) | -0.17 (0.260) | 0.01 (0.925) | |
| Type of dwelling | 0.00 (0.996) | 0.27 (0.032) | 0.26 (0.020) | |
| Type of dwelling | Intervention effect | -0.43 (0.042) | -0.95 (<0.001) | -0.49 (0.024) |
| Type of dwelling | -0.01 (0.970) | 0.25 (0.043) | 0.26 (0.021) | |
| Type of wall | Intervention effect | -0.43 (0.044) | -0.94 (<0.001) | -0.49 (0.025) |
| Type of wall | -0.02 (0.876) | -0.16 (0.289) | 0.02 (0.875) | |
The intervention effect and covariates are tested in four different longitudinal models; simple not controlling for any covariates, full model controlling for type of wall and type of dwelling and the two semi controlled models. Intervention effect in terms of sandfly counts and percentage reduction is given for the simple model. P-values for the regression parameters are presented. Take into account the multiplicative structure of the regression model. (n = 840)
* (B-A)- (D-C), with A and B being baseline and 5 months sandfly count in the intervention group and C and D being baseline and 5 months sandfly count in the control group. Percent reduction = level of reduction (e.g. 4.34) divided by baseline intervention group count (e.g. 9.92) times 100 (43.7% in this example).
Note: The pooled analysis of EVM has to be interpreted with caution as two different EVM methods were applied: lime plastering in India and Nepal; mud plastering in Bangladesh (see site specific differences in the text)
†The intervention effect and covariates are tested in four different longitudinal models; simple not controlling for any covariates, full model controlling for type of wall and type of dwelling and the two semi controlled models. Intervention effect in terms of sand fly counts and percentage reduction is given for the simple model. P-values for the regression parameters are presented. Please observe the multiplicative structure of the regression model. (n = 840)
‡ Percentage reduction
LLIN, long-lasting insecticide treated nets; IRS, indoor residual spraying; EVM, environmental modification; CI, confidence interval