Do malaria vector control measures impact disease-related behaviour and knowledge? Evidence from a large-scale larviciding intervention in Tanzania

Abstract

Background: Recent efforts of accelerated malaria control towards the long-term goal of elimination had significant impacts in reducing malaria transmission. While these efforts need to be sustained over time, a scenario of low transmission could bring about changes in individual disease risk perception, hindering adherence to protective measures, and affecting disease-related knowledge. The goal of this study was to investigate the potential impact of a successful malaria vector control intervention on bed net usage and malaria-related knowledge.

Methods: Dar es Salaam's Urban Malaria Control Program was launched in 2004 with the aim of developing a sustainable larviciding intervention. Larviciding was scaled-up using a stepped-wedge design. Cross-sectional and longitudinal data were collected using a randomized cluster sampling design (2004-2008). Prevalence ratios (PR) for the effect of the larviciding intervention on bed net usage (N = 64,537) and household heads' knowledge of malaria symptoms and transmission (N = 11,254) were obtained from random effects regression models.

Results: The probability that individuals targeted by larviciding had used a bed net was reduced by 5% as compared to those in non-intervention areas (PR = 0.95; 95% credible intervals (CrI): 0.94-0.97) and the magnitude of this effect increased with time. Larviciding also led to a decline in household heads' knowledge of malaria symptoms (PR = 0.88; 95% CrI: 0.83-0.92) but no evidence of effect on knowledge of malaria transmission was found.

Conclusion: Successful control interventions could bring about further challenges to sustaining gains in reducing malaria transmission if not accompanied by strategies to avoid changes in individual knowledge and behaviour. This study points to two major research gaps. First, there is an urgent need to gather more evidence on the extent to which countries that have achieved significant decline in malaria transmission are also observing changes in individual behaviour and knowledge. Second, multidisciplinary assessments that combine quantitative and qualitative data, utilizing theories of health behaviour and theories of knowledge, are needed to optimize efforts of national malaria control programmes, and ultimately contribute to sustained reduction in malaria transmission.

Figure 1

Prevalence of bed net usage stratified by survey round and larviciding status. Confidence intervals are based on 9,999 bootstrap replicates at the TCU levels. (The time frame of larviciding phases and survey rounds do not overlap perfectly. Thus, due to small sample size and the geographically limited extent of data collection (only one ward), results for 697 data points in the larviciding area in survey round 3, and 744 data points in control area in survey round 6 are not shown).

Figure 2

Proportion of household heads knowing at least five symptoms of malaria, stratified by survey round and larviciding status. Confidence intervals are based on 9,999 bootstrap replicates at the TCU levels. (Prevalence estimates based on small sample size and geographically limited extent of data collection are not represented).

Figure 3

Proportion of household heads that know that mosquitoes transmit malaria, stratified by survey round and larviciding status. Confidence intervals are based on 9,999 bootstrap replicates at the TCU levels. (Prevalence estimates based on small sample size and geographically limited extent of data collection are not represented).

Figure 4

Effect modification of the larviciding intervention by age, gender, and socio-economic status on bed net usage, knowledge of malaria symptoms, and knowledge of malaria transmission. Statistically significant results are bolded. To account for the fact that the coefficients of the ward fixed effects exhibited slow convergence for the ‘Bed net usage’ models, the number of iterations used for inference was doubled to 120,000. † Models for the bed net usage outcome are adjusted for: age, gender, dummy for being a follow-up observation, use of insect repellent, use of sprays, use of coil, living in a house with window screens, socio-economic status, and weekly rainfall lagged by two weeks (with quadratic term). Models also include: a semiparametric time trend, random effects at household and TCU levels, and fixed effects at the ward level (as in Model 2). ‡ Models for the knowledge of malaria symptoms and malaria transmission outcomes are adjusted for: age, gender, dummy for being a follow-up observation, and socio-economic status. Models also include: a semiparametric time trend, random effects at TCU level, and fixed effects at the ward level (as in Model 2).