Table 3.

Results of negative binomial regression model (no. of observations = 588)

Covariates	Estimate	Std. error	p value	Risk ratio (95% CI)
Urban areaa (AG, MZ, LO, CT)	2.443	0.373	< 0.001	11.50 (5.54–23.89)
Rural areaa (LE, CU)	2.233	0.378	< 0.001	9.33 (4.44–19.58)
Incidence rate (%) in previous month	2.796	0.152	< 0.001	16.39 (12.17–22.07)
Rainfall amount (100 mm) in previous month	0.225	0.052	< 0.001	1.25 (1.13–1.39)
Indoor EIRb (%)	− 0.217	0.194	0.263	0.80 (0.55–1.18)
Outdoor EIRc (%)	0.072	0.016	< 0.001	1.07 (1.04–1.11)
Kdr mutation frequencyd (10%)	0.060	0.019	0.0017	1.06 (1.02–1.10)

The response variable is monthly incidence rate in 7 districts from 2010 to 2016

^aReference area is the offshore island-Príncipe

^bIndoor EIR (%) is calculated by monthly mosquito density of indoor HLCs multiplied sporozoite rate (0.5%) which is referred from Pinto et al. [4]

^cOutdoor EIR (%) is calculated by monthly mosquito density of outdoor HLCs multiplied sporozoite rate (0.5%) which is referred from Pinto et al. [4]

^dKdr mutation frequency is calculated per season using Genepop ver 4.2. Kdr mutation frequency is assumed the same per month in the same season. The risk ratio is modelled by an increase of 10% in kdr mutation frequency