Malaria knowledge and bed net use in three transmission settings in southern Africa

The study was based on a sample of 7535 participants representing 2066 randomly selected households from Mutasa District in eastern Zimbabwe, Choma District in southern Zambia and Nchelenge District in northern Zambia. The data used were acquired under the auspices of the Southern and Central Africa International Centers of Excellence for Malaria Research (ICEMR) project. The three study sites were specifically chosen by the Southern and Central Africa ICEMR to highlight variability in the epidemiology and transmission of malaria across southern Africa (Fig. 1, Table 1). In Choma District, malaria transmission is seasonal and the prevalence of malaria is low. By contrast, in Nchelenge District, which lies along Lake Mweru and borders the Democratic Republic of Congo, malaria transmission is intense with little or no seasonal fluctuations. Malaria transmission in Mutasa District is highly seasonal, with malaria-related morbidity and mortality peaking during the rainy season.

Several malaria control activities including the distribution of ITNs and application of indoor residual spraying (IRS) occurred during the study period (Fig. 2). In Choma District, mass distribution of free, long-lasting insecticidal nets (LLINs) occurred in 2007, 2012 and 2014 [7, 20]. In Nchelenge District, mass LLIN distribution campaigns took place in 2007, 2011 and 2014. Annual rounds of IRS began in 2006, first with pyrethroids, then carbamates [21]. Since 2014, the organophosphate pirimiphos-methyl has been used for targeted IRS in the study area. Similar to Nchelenge District, there was a programmatic switch from pyrethroid-based to organophosphate-based IRS in Mutasa District in 2014. Following the introduction of pirimiphos-methyl, Mutasa District has experienced moderate reductions in malaria incidence [22]. Universal ITN distribution to the general population was conducted in 2013, with distributions to vulnerable populations (e.g. school age children) in 2014 and 2015.

The study design and procedures have been described elsewhere [23‐25]. Briefly, high resolution satellite images of the study areas were used to establish sampling frames and study households were randomly selected for enrolment in the cross-sectional study arm or the prospective longitudinal cohort. Households enrolled in the longitudinal cohort were visited every other month, while households in the cross-sectional arm were surveyed only once during the study period. Study enrollment began in March 2012 in Choma District, April 2012 in Nchelenge District and October 2012 in Mutasa District (Fig. 2). Across all sites, cross-sectional surveys and longitudinal surveys were conducted during alternating months during the study period. Enrollment of new participants from randomly selected households in Choma District ended in December 2014; thereafter, participants have been recruited under the reactive test and treat programme [26]. Study enrollment is still ongoing in the other sites. For the purposes of this analysis, December 31, 2015 was set as the cutoff date for Choma District and March 31, 2017 for Nchelenge and Mutasa Districts. The analysis was restricted to data from all visits in the cross-sectional arm and the initial baseline visits for the longitudinal cohort.

Data were collected using standardized data collection instruments that were field tested to ensure reliability and validity. In all selected households, the head or another responsible adult member of the household was interviewed after informed consent was given. A complete listing of all residents of selected households was obtained and all residents were eligible to participate regardless of age, gender or pregnancy status. The availability of household assets and characteristics of the house structure were recorded. An interviewer-administered questionnaire was used to gather data on malaria-specific knowledge for adult participants and presenting symptoms, health-seeking behaviours and malaria prevention practices for all participants. At the end of each interview, a blood sample was collected for a malaria rapid diagnostic test (RDT). Participants with positive test results were treated as per the country-specific malaria treatment guidelines.

Several measures reflecting population coverage of bed nets were calculated as recommended by the Roll Back Malaria (RBM) partnership [5, 27], using data on direct observation of bed net ownership and self-reported bed net use. The term ‘bed net’ was used to encompass insecticide-treated and untreated bed nets, although long-lasting insecticidal nets (LLINs) are the standard in all study sites. The primary outcome measure was individual-level bed net use among those with living in a household with any bed net. The primary exposure of interest was knowledge of the benefit of sleeping under an ITN derived from responses to an unprompted open-ended question on the ways to prevent malaria. Demographic characteristics included gender and age of the respondent (< 5, 5–14, 15–34, and ≥ 35 years). Three variables, namely, educational level of the head of the household (none or primary, secondary education and tertiary education or higher), permanent employment status of the head of the household (employed or unemployed) and a wealth index, were used as proxies for socioeconomic status. The wealth index was calculated using principal components analysis based on ownership of assets (radio, television, refrigerator, bicycle, motorcycle, and car or truck) and house characteristics (source of drinking water, source of energy for cooking and floor material) [28]. The index was divided into three tertiles—‘poorest’, ‘less poor’ and ‘least poor’. Other variables controlled for in the analysis were based on prior literature and included number of children under 5 years, number of bed nets in household, and household size.

Frequency distributions were used to describe the sample population, quantify knowledge about the cause, symptoms and prevention measures for malaria, and describe bed net ownership and use. Pearson’s Chi squared test was used to compare sociodemographic and household characteristics across study sites. To determine if knowledge of ITNs as a preventive measure significantly increased the odds of bed net use, logistic regression analyses were performed separately for each study site. To account for within-household correlation, univariate and multivariate logistic regression analyses were conducted using generalized estimated equations (GEE) [29]. Age and gender were included in all multivariate models to control for potential confounding. Calendar year and season (rainy/dry) were also included to account for secular trends and seasonality in the outcomes. Adjusted odds ratios (aOR) with 95% confidence intervals (CI) were computed from the final GEE models. A p value of less than 0.05 was considered statistically significant. Analyses were performed using STATA 14.2 (College Station, Texas).

The analysis included 7535 participants, with 1761 from Choma District, 3405 from Nchelenge District and 2369 from Mutasa District (Table 2). One in five participants was under 5 years of age (19.3%), and slightly more than half of participants were female (55.3%). There were some differences in sociodemographic characteristics by study site. Compared to the other study sites, participants in Choma District tended to reside in larger households, with 46.7% of participants residing in a household with 7 or more members (p < 0.001). Compared to Choma and Nchelenge Districts, a higher proportion of participants in Mutasa District belonged to households headed by individuals who had completed secondary or tertiary education (55.6%, p < 0.001) and were in permanent employment (46.5%, p < 0.001). Participants from Mutasa District were also relatively better off than their counterparts as a higher proportion of participants reported the use of electricity for cooking (7.5%, p < 0.001), piped water for drinking (28.0%, p < 0.001) and a finished floor in the home (88.4%, p < 0.001). A higher proportion of participants in Mutasa District lived in households that had been covered by IRS in the previous 6 months than the two Zambian sites (40.5%, p < 0.001).

Of the 3843 participants aged 16 years or older and eligible to respond to questions related to malaria knowledge, 3836 (99.9%) responded to the malaria knowledge questionnaire (Table 3). The majority (85.0%) of respondents linked malaria to a mosquito bite, with the highest proportion (89.4%) in Choma District, the setting with the lowest malaria burden. A few respondents associated malaria with dirty surroundings (3.9%), drinking bad water (3.6%), and other causes including eating bad food, fresh fruit, maize or sugar cane (4.1%). Among those who correctly linked malaria to a mosquito bite, 2.7% also cited one or more incorrect causes. The most frequent symptoms listed as presumptive for malaria varied by site. In Choma District, respondents most commonly associated malaria with headache (68.3%), chills (62.1%) and fever (47.4%). In Nchelenge District, chills (56.3%), fever (35.8%) and body ache or pain (33.0%) were the most commonly reported symptoms of malaria. By contrast, headaches (70.0%), weakness or fatigue (60.2%) and chills (50.5%) were the most commonly reported symptoms in Mutasa District. Overall, almost all respondents (95.5%) mentioned at least one common symptom of malaria (fever, chills, headache, weakness or fatigue, and body ache or pain), and 29.0% could mention three or more of the common symptoms of malaria. Sleeping under a mosquito net was the most commonly reported measure thought to prevent malaria (73.1%), with the highest level of knowledge of the benefits of net use in Choma District (87.3%) and the lowest in Mutasa District (67.2%). Seeking early treatment (11.5%), keeping surroundings clean (10.9%), burying mosquito breeding sites (8.34%) and indoor residual spraying (6.7%) were other preventive measures reported. A minority of respondents linked eating clean food to the prevention of malaria (3.7%). Information about malaria was commonly received from health workers in health facilities (54.4%), schools (15.7%), and the community (8.6%). Less frequently mentioned sources of information about malaria were radios, newspapers, posters, friends, relatives, non-governmental organizations and the study team.

Bed net ownership, access and use varied by study site, with Mutasa District reporting the lowest levels. At the household level, ownership of any bed net was 69.9%, while ownership of sufficient bed nets (i.e. at least one bed net for every two members) was 39.7% (Table 4). At the population level, access to a bed net within the household was 39.2%, while bed net use was 31.8%. The proportion of the population using bed nets was fairly similar to the proportion of the population with access to a bed net, indicating an average of two users per net. Unavailability of bed nets (50%) and the perceived lack of mosquitoes (26.5%) were the main reasons reported by households for not owning a net, while the perceived lack of mosquitoes (17.4%) and heat (10.1%) were the main reasons for not sleeping under a bed net. By contrast, in the low transmission setting of Choma District, 78.2% of household owned any bed net and 70.8% of the population reported sleeping under a bed net. Indicators of bed net ownership, access and use for Nchelenge District did not vary appreciably from Choma District despite the higher malaria transmission intensity. In both Zambian sites, cost and lack of knowledge of where to obtain a bed net were the main barriers to bed net ownership reported (Choma: 32.1 and 22.5% respectively; Nchelenge: 26.7 and 21.6% respectively). However, the perceived lack of mosquitoes (5.2%) was the most cited reason for non-use of available bed nets in Choma District, while the most common reason in Nchelenge District was the state of the available net (old, dirty or in need of retreatment; 4.8%).

In Choma District, multivariate analyses restricted to individuals residing in households with any bed nets demonstrated marginal evidence of a higher odds of bed net use among respondents with knowledge of ITNs as a preventive measure (aOR 1.40, 95% CI 0.97–2.03) (Table 5). Compared to individuals aged less than 5 years, the odds of bed net use were greater in the ≥ 35 years age group (aOR 2.38; 95% CI 1.55–3.67) and lesser in the 5–14 years age group (aOR 0.57; 95% CI 0.41–0.79). The odds of bed net use decreased with large household size (3–6 members: aOR 0.29; 95% CI 0.14–0.58; 7+ members: aOR 0.32; 95% CI 0.16–0.67 relative to one to two members). Also, residing in a household with three or more bed nets or with at least one child under 5 years increased the odds of bed net use (aOR 2.52; 95% CI 1.75–3.62; aOR 1.42; 95% CI 1.05–1.96, respectively).

Consistent with patterns observed among residents of Choma District, in Nchelenge District awareness of ITNs as a preventive measure was associated with statistically significant increased odds of bed net use (aOR 1.35; 95% CI 1.11–1.64). Associations with bed net use of similar magnitude and significance were observed for age, household size, the presence of at least one child under 5 years and household ownership of three or more bed nets. However, the odds of bed net use were significantly higher among females (aOR 1.34; 95% CI 1.11–1.61) and individuals from households of higher socio-economic status (least poor aOR 1.56; 95% CI 1.19–2.06).

Knowledge of ITNs was predictive of bed net use in Mutasa District (aOR 1.27; 95% CI 1.02–1.58). Age was associated with bed net use, with the odds of bed net use significantly higher among respondents 35 years or older (aOR 1.81; 95% CI 1.18–2.79). The odds of bed net use were reduced by 26% among individuals residing in the least poor households compared to the poorest households (aOR 0.74; 95% CI 0.57–0.96). The presence of at least three bed nets in the household increased the odds of bed net use by 93% (aOR 1.93; 95% CI 1.37–2.72).