Skip to main content

Table 2 Adjusted effects of mosquito-disseminated pyriproxyfen on adult-mosquito catches (Aedes aegypti + Culex quinquefasciatus): top-ranking (smallest-BIC) generalized linear mixed model

From: Measuring mosquito control: adult-mosquito catches vs egg-trap data as endpoints of a cluster-randomized controlled trial of mosquito-disseminated pyriproxyfen

Term Estimate SE 95% CI
Lower Upper
Fixed effects
Intercept (CC, BP)a 0.776 0.352 0.087 1.465
Intervention period (IP)b 0.028 0.341 − 0.641 0.697
Intervention cluster (IC) − 0.436 0.317 − 1.056 0.185
IP × ICc − 1.086 0.227 − 1.532 − 0.641
Temperatured 0.721 0.139 0.448 0.994
Random effects SD
Dwelling ID 0.956 0.773 1.183
Month 0.456 0.303 0.686
  1. aThe intercept estimates the (log-scale) expected mean number of mosquitoes caught per 10 min aspiration in the CC, in the typical dwelling and at typical temperatures, during the BP; the other fixed-effect slope coefficients estimate changes in this expectation associated with period, cluster, intervention, and temperature effects
  2. bNote that the model estimates a near-zero change in (log) mean mosquito-catch as the CC entered the IP (but received no intervention); the estimated incidence rate ratio is e0.028 = 1.028, or a 2.8% increase in mean mosquito-catch, with the 95% CI spanning zero
  3. cThe ‘IP × IC’ interaction coefficient estimates the (log) change in expected mean mosquito-catch that can be attributed to the intervention – deployment of 150 pyriproxyfen dissemination stations over 13 months (the intervention period ‘IP’) in the intervention cluster ‘IC’. Here, the model estimates an e− 1.086 = 0.337 incidence rate ratio, indicating that the intervention resulted in a 100 −  33.7 = 66.3% reduction (95% CI: 47.3–78.4%) of the expected mean mosquito-catch
  4. dSpecified as the (standardized) mean of minimum daily temperatures in the month before each sampling occasion (‘tmin_m’); the original variable had mean = 17.39°C and SD = 1.73°C. Given our focus on estimating adjusted intervention effects, we considered weather covariates as confounders; ‘tmin_m’ yielded better-performing models, as measured by BIC scores, than other measures of temperature and rainfall
  5. Abbreviations: BIC, Bayesian information criterion; SE, standard error; 95% CI, 95% confidence interval (lower/upper limits); CC, control cluster; BP, baseline period; IP, intervention period; IC, intervention cluster; SD, standard deviation; ID, identity of each sampling dwelling