Reconsidering indoor residual spraying coverage targets: A retrospective analysis of high-resolution programmatic malaria control data

Abstract

Indoor residual spraying (IRS) is one of the core vector control interventions available to malaria control programs. Normative and scientific guidance has long held that very high IRS coverage (at least 80 to 85% houses sprayed) is necessary to provide community protection, but there is little evidence backing these recommendations, in large part due to the operational and ethical concerns that conducting appropriate trials of differing IRS coverage levels would raise. The present study leverages data from four years of targeted IRS implementation on Bioko Island, Equatorial Guinea, to estimate a dose-response curve of IRS coverage. Due to the observational nature of the data, a double robust causal inference technique was utilized. The results suggest that at all spatial scales examined a threshold providing community protection was reached at much lower coverage levels than previously assumed (30 to 50%). Sensitivity analysis corroborated this result across multiple methods, but there was less agreement on whether extremely high coverage ([Formula: see text]85%) has additional benefit. A secondary analysis of the impact of changing operational coverage targets found that significantly reducing coverage targets (to 30 to 60%) could provide nearly the same protection as maintaining the existing 80% target in Bioko. While these findings are limited in strength by the observational nature of the data and may be specific to the context of Bioko Island, they raise important questions for further research on how IRS coverage impacts epidemiological outcomes and on how malaria control programs should set programmatic IRS coverage targets.

Keywords: indoor residual spraying; malaria; vector control.

Fig. 1.

Spray coverage achieved 2017–2020: The IRS coverage achieved in the greater Malabo area based on the 200 m buffer zone from 2017 to 2020. All inhabited sectors are shown, regardless of whether they were sampled in the MIS.

Fig. 2.

Overall dose–response curves: mean odds ratio as a function of coverage, defined on the sector of interest, and using buffers of 200 m, 400 m, and 600 m around this sector. Estimates obtained with the double-robust method using all data, and the 95% CI (shaded area) was estimated by bootstrap. The black dotted line corresponds to no effect (OR $=$ 1), and the purple dashed line corresponds to the Cochrane systematic review estimate transformed to the OR scale.

Fig. 3.

Dose–response curves in Malabo and periphery: mean odds ratio as a function of coverage, defined on the sector of interest, and using buffers of 200 m, 400 m, and 600 m around this sector in urban Malabo (orange) and the rural periphery (light blue). Estimates obtained with the double-robust method using all data, and the 95% CI (shaded area) was estimated by bootstrap. The black dotted line corresponds to no effect (OR $=$ 1), and the purple dashed line corresponds to the Cochrane systematic review estimate transformed to the OR scale.

Fig. 4.

Estimated impact of changing coverage target: Expected OR in sectors targeted during at least one year of the study as a function of coverage target, defined on the sector of interest, and using buffers of 200 m, 400 m, and 600 m around this sector overall (black), in urban Malabo (orange) and the rural periphery (light blue). Coverage targets examined were 20% to 90% in increments of 10, with 80% corresponding to the data as observed and used as the reference for computing odds ratios. Bars show 95% CIs of OR, based on asymptotic normality of the estimator, and the black dashed line corresponds to no effect (OR $=$ 1).