Background
Malaria remains a global health problem with an annual incidence of 214 million cases and 438,000 deaths; more than 88% of the total morbidity is in the WHO Africa region [
1‐
4]. Bednets have been demonstrated to reduce malaria-related morbidity and mortality in sub-Saharan Africa [
3,
5‐
8]. Models suggest that, the proportion of the population who had access to bednets increased from 4 to 67% between 2004 and 2015 in sub-Saharan Africa [
9]. WHO has recommended universal coverage of bednets (defined as 1 bednet per two persons) because of the effectiveness of bednets in malaria prevention [
10‐
12]. Centralized mass distribution campaigns have been the cornerstone of the effort to achieve universal bednet coverage [
10].
In Kilifi, Kenya, bednet coverage was estimated to be less than 6% in the early 1990s [
13]. A randomized controlled trial of bednets was conducted in part of Kilifi in 1993, where bednets were distributed to residents in a defined geographical area representing approximately 30% of the population. Bednet use in other areas remained low. The trial found that bednet use was associated with a 33% reduction in childhood mortality and a 44% reduction in severe malaria morbidity [
14]. In 2005, subsidized nets were made available to children and pregnant women through maternal and child health clinics, and in September 2006, a free bednet distribution campaign by the then Kilifi District Health Management Team increased coverage across Kilifi county from about 0.25 to 0.5 insecticide-treated nets per person [
15]. After the campaigns, bednets continued to be provided to the community through maternal and child health clinics at a cost of 50 Kenyan Shillings (~ $0.50). This was the case until January 2009, when a new policy was implemented to offer bednets at government health clinics to pregnant women and children aged < 5 years free of charge and to the rest of the population for 50 Kenyan Shillings. Bednets were again distributed through mass campaigns in July 2012 and in October 2015 as part of the efforts of achieving universal coverage for people at risk of malaria.
Household possession of bednets is an indicator of the extent to which distribution channels are enhancing coverage. However, use of bednets, rather than possession, is what affords protection and is therefore a more valid predictor of epidemiological impact [
16,
17]. The objective of this analysis was to evaluate trends in ownership and usage of bednets in the context of mass distribution campaigns, and to assess the effect of bednet ownership and usage on the incidence of hospitalized malaria under routine, non-trial conditions.
Discussion
Malaria remains a major cause of morbidity and mortality, particularly in sub-Saharan Africa. Many countries, including Kenya, have attempted to reduce the burden of malaria through policies and practices that promote bednet ownership and use. We report that bednet ownership and use remained relatively stable over the 8-year period and temporarily increased corresponding to mass distribution campaigns. We found that bednet use was associated with a 33% reduction in the incidence of malaria hospitalization among children < 5 years.
Pooled data from controlled trials of insecticide treated nets have consistently demonstrated reductions in mortality, morbidity and malaria parasitemia of 17, 50 and 13%, respectively [
6,
25]. Although a number of studies have postulated a causal relationship between bednet use and reductions in childhood morbidity and mortality, there is little data on the protection against malaria hospitalization afforded by bednet use under routine conditions. Our estimate of the protection against malaria hospitalization associated with bednet use is in agreement with a randomized controlled study conducted in the study area that documented a 42% (95% CI; 21, 57) protective efficacy against severe malaria [
14]. It has been estimated that substantial protective community effects of net ownership are observed with coverage ≥ ~ 50% [
7,
26,
27]; in the study area overall bednet ownership/usage was observed to be > 50% over the 8-year period. Bednet use was also associated with a reduction in deaths related to malaria hospitalization although this was not statistically significant. The number of malaria-related deaths was low; this may reflect local care-seeking and case-management practices rather than bednet use ownership and use.
The protective association of bednet use has been shown to vary by setting, making a multi-faceted approach to malaria control essential [
28]. In our setting, bednet usage was highest among children < 5 years and women of reproductive age [
29]. This pattern of bednet use has been observed in other areas in Africa [
30‐
32] and likely reflects population sub-groups targeted by net distribution campaigns. Bednet usage among women of reproductive age was 60% in the 2008/2009 KHDSS survey – similar to the 58% usage reported in a 2007 survey in western Kenya [
17] - and increased to 89% in 2015 following a distribution campaign.
Observing a mounted bednet is considered a more reliable indicator of bednet use than relying solely on self-reported use; however, the time required to do a thorough assessment of the household, along with local customs that typically prohibit strangers from entering the household, can preclude conducting such surveys at scale. In a contemporaneous survey of a randomly selected subset of children under 16 years in the KHDSS, we observed a bednet in 86% of children for whom net ownership was reported as “yes”. This suggests that self-reported net-use the night prior to the survey may overestimate actual net use to a small degree in our setting but is a reasonable proxy. The median time lapse between the routine KHDSS enumeration rounds and sero-study interviews was 1 month (IQR: 0.4, 2.5). We considered the time lapse to be reasonable to make a comparison between observed and reported bednet ownership and usage; however, the gap in time may have contributed to the lack of perfect agreement in the data.
Our study has several other limitations. Because we did not require each resident to be present to respond to the questions, data on bednet ownership and usage may have been incorrectly reported. We attempted to minimize this potential misclassification by instructing fieldworkers to interview only residents of the same homestead regarding bednet ownership and usage. Although non-significant, there was a trend toward a reduced protective effect as the period of observation was lengthened 1, 2, and 3 months before and after the KHDSS enumeration period. This suggests that limiting the analysis to the KHDSS enumeration periods was an appropriate choice to minimize misclassification of exposure status that could arise due to variation in bednet use or ownership over time. In this study, we asked about individual bednet ownership – defined as availability of a bednet for an individual’s usual sleeping space – and bednet use the night prior to enumeration. This limits comparability to studies that report ownership at a household level. While we found good agreement between bednet ownership reported during the KHDSS routine enumerations and bednet observation during the sero-survey, this does not directly validate reported bednet usage estimates. The results on the incidence of malaria hospitalization may be biased and confounded by other unmeasured factors (e.g., variation in housing, urbanization, socio-economic status, mother’s education, care-seeking behaviors). Therefore, the estimates obtained could be an overestimation or underestimation of the true incidence. Because we included all bednets, regardless of quality or insecticidal efficacy, we likely underestimated the protection afforded by usage of a high-quality, long-lasting insecticide treated net.
The strengths of the study include use of demographic surveillance data of a sizeable population, the longitudinal nature of the study, which allowed us to evaluate trends in bednet ownership and use over time in association with changes in government policy on provision of bednets and mass distribution campaigns, and the validation of the demographic surveillance data using the sero-survey data. In addition, the diagnostic and laboratory evaluation of children admitted to KCH was the same throughout the study period.
Acknowledgements
We are grateful to the KHDSS field workers and the KCH clinical and laboratory staff who conducted the bednet surveys and collected admission information. We are grateful to Donald Akech and the PCVIS fieldworkers who conducted the observational surveys. We also thank Kevin Marsh, Robert Snow and Ally Olotu for comments on earlier versions of this manuscript. This paper is published with the permission of the director of KEMRI.