Infant and child mortality in relation to malaria transmission in KEMRI/CDC HDSS, Western Kenya: validation of verbal autopsy

• Published in: Malaria Journal
• Main Authors: Nyaguara O. Amek, Annemieke Van Eijk, Kim A. Lindblade, Mary Hamel, Nabie Bayoh, John Gimnig, Kayla F. Laserson, Laurence Slutsker, Thomas Smith, Penelope Vounatsou
• Format: Article in Journal/Newspaper
• Language: English
• Published: BMC 2018
• Subjects: Childhood mortality; Bayesian inference; Malaria entomology data; Verbal autopsy; Health and demographic surveillance system; Arctic medicine. Tropical medicine; RC955-962; Infectious and parasitic diseases; RC109-216; Arctic
• Online Access: https://doi.org/10.1186/s12936-018-2184-x; https://doaj.org/article/3da2c79f19444ac08fc80d130fa93669

Abstract Background Malaria transmission reduction is a goal of many malaria control programmes. Little is known of how much mortality can be reduced by specific reductions in transmission. Verbal autopsy (VA) is widely used for estimating malaria specific mortality rates, but does not reliably distinguish malaria from other febrile illnesses. Overall malaria attributable mortality includes both direct and indirect deaths. It is unclear what proportion of the deaths averted by reducing malaria transmission are classified as malaria in VA. Methods Both all-cause, and cause-specific mortality reported by VA for children under 5 years of age, were assembled from the KEMRI/CDC health and demographic surveillance system in Siaya county, rural Western Kenya for the years 2002–2004. These were linked to household-specific estimates of the Plasmodium falciparum entomological inoculation rate (EIR) based on high resolution spatio-temporal geostatistical modelling of entomological data. All-cause and malaria specific mortality (by VA), were analysed in relation to EIR, insecticide-treated net use (ITN), socioeconomic status (SES) and parameters describing space–time correlation. Time at risk for each child was analysed using Bayesian geostatistical Cox proportional hazard models, with time-dependent covariates. The outputs were used to estimate the diagnostic performance of VA in measuring mortality that can be attributed to malaria exposure. Results The overall under-five mortality rate was 80 per 1000 person-years during the study period. Eighty-one percent of the total deaths were assigned causes of death by VA, with malaria assigned as the main cause of death except in the neonatal period. Although no trend was observed in malaria-specific mortality assessed by VA, ITN use was associated with reduced all-cause mortality in infants (hazard ratio 0.15, 95% CI 0.02, 0.63) and the EIR was strongly associated with both all-cause and malaria-specific mortality. 48.2% of the deaths could be attributed to malaria by analysing ...