Table 4.
Effect of intervention on female Phlebotomus argentipes densities adjusted for covariates by longitudinal regression analysis
| Time/Model | Parameter | IRR [95% CI] (P-value) | ||
|---|---|---|---|---|
| NKTA | FC + DWL | FC + ITN | ||
| 1-Month follow-up | ||||
| Simple model | Crude intervention effect* | 0.48 [0.27, 0.85] (0.012) | 0.27 [0.11, 0.67] (0.005) | 0.19 [0.04, 0.96] 0.044) |
| Full model | Adjusted intervention effect | 0.47 [0.26, 0.84] (0.012)† | 0.24 [0.09, 0.62] (0.003)‡ | 0.20 [0.04, 0.97] (< 0.046)§ |
| 3-Month follow-up | ||||
| Simple model | Crude intervention effect* | 0.46 [0.26, 0.81] (0.007) | 0.40 [0.18, 0.91] (0.028) | 0.19 [0.04, 0.92] (0.003) |
| Full model | Adjusted intervention effect | 0.49 [0.27, 0.87] (0.016)† | 0.31 [0.12, 0.83] (0.020)‡ | 0.19 [0.04, 0.93] (0.040)§ |
| 6-Month follow-up | ||||
| Simple model | Crude intervention effect* | 0.46 [0.25, 0.87] (0.017) | 0.44 [0.19, 0.98] (0.0.47) | 0.41 [0.09, 1.93] (0.260) |
| Full model | Adjusted intervention effect | 0.40 [0.22, 0.72 (0.002)† | 0.29 [0.11, 0.81] (0.018)‡ | 0.39 [0.08, 1.87] (0.242)§ |
| 9-Month follow-up | ||||
| Simple model | Crude intervention effect* | 0.59 [0.32, 1.11] (0.103) | 0.50 [0.21, 1.14] (0.100) | 0.32 [0.07, 1.51] (0.152) |
| Full model | Adjusted intervention effect | 0.56 [0.31, 1.05] (0.073)† | 0.53 [0.23, 1.22] (0.139)‡ | 0.32 [0.07, 1.51] (0.151)§ |
| 12-Month follow-up | ||||
| Simple model | Crude intervention effect* | 0.38 [0.19, 0.77] (0.008) | 0.31 [0.11, 0.94] (0.038) | 0.34 [0.07, 1.76] (0.201) |
| Full model | Adjusted intervention effect | 0.26 [0.16, 0.42] (< 0.0001)† | 0.21 [0.07, 0.69] (0.010)‡ | 0.34 [0.07, 1.74] (0.792)§ |
CI = confidence interval; FC + DWL = fever camp and installation of durable wall lining; FC + ITN = fever camp and insecticide-treated net; NKTA = no kala-azar transmission activity; IRR = incidence rate ratio.
* The intervention effect and covariates are tested in two types of longitudinal regression models (generalized estimating equation with negative binomial model) at five different follow-up times: simple not controlling for any covariates and full model controlling covariates. The variables that varied significantly between intervention and control areas are considered as covariates for full model. Only incidence rate ratio with 95% CI and P-values for the regression parameter of intervention effect are presented. Regression analysis was performed by considering the clustering effect due to the index-based approach.
† Full model adjusted by the covariates: humidity in the bedroom, mud wall.
‡ Full model adjusted by the covariates: humidity in the bedroom, household head occupation, mud wall.
§ Full model adjusted by the covariates: crack in the wall.