This study indicates that the overarching perception among caretakers was that although malaria was seen as an easily curable and uncommon illness, bed-nets were still valued for avoiding mosquito nuisance, especially during the rainy season. Thus, despite the dramatic reduction in malaria burden in the past few years , bed-nets were still desirable and their use was stated to be maintained.
Awareness of malaria and its symptoms was high among the caretakers interviewed. Susceptibility to malaria was perceived as low, especially in the dry seasons. Despite the perception that malaria could develop into a severe illness that could lead to disability and even death, the caretakers were confident that the risk of severity remained low as long as prompt appropriate treatment was sought. The acknowledgement of severe forms of malaria and the importance of treatment illustrate a better understanding of malaria than what was previously shown in coastal Tanzania . In a review of the HBM from the 80s, Janz and Becker showed that perceived susceptibility was especially important in preventive health behavior, whereas perceived severity was the least powerful predictor and was more related to sick-role behavior and, therefore, less relevant for a preventive behavior . Thus, the low perceived susceptibility, more than the low perceived severity might have a negative influence on bed-net usage.
Children, who were viewed as more susceptible to mosquito bites and malaria, were prioritized when it came to bed-net usage. Although prioritizing bed-net usage in small children is contrary to previous findings from a randomized control trial in Kenya where children were less likely to use bed-nets compared to older individuals , our finding does concur with more recent findings where under-five children were more likely to use bed-nets [12, 13]. This is probably due to the delivery systems in recent years that have focused on distributing bed-nets to the vulnerable under-five children and pregnant women, which was also the case in the free distribution campaign in Zanzibar.
Benefits of bed-nets included malaria prevention as well as protection against mosquito nuisance, which has also been reported from Kenya [31, 32]. In this study, due to the strong association made between mosquito bites and malaria, it was difficult to differentiate between the two benefits. Nonetheless, it was common that protection against mosquito bites was described independently of malaria transmission, and thus served as an additional benefit.
Although caretakers were well aware of the fact that mosquitoes transmitted malaria, other causes of malaria were also reported. Beliefs in alternative causes of malaria and their prevention could potentially lower the perceived benefit of bed-nets for malaria prevention [16, 31, 32]. In this study, the women in particular mentioned several alternative causes of malaria (i.e. dirt, poor hygiene, diet, etc.) and described prevention strategies against these causes. While the alternative causes mentioned cannot actually cause malaria, they may lead to other illnesses with similar symptoms, such as gastroenteritis. The fact that the female informants mentioned alternative causes more often than men could indicate that women are more exposed to both seeing different illnesses in children as well as to receiving health education messages provided by health workers, often resulting in a mix of biomedical and local pre-existing ideas and logics [19, 33].
While complementary interventions for malaria and mosquito prevention such as IRS might lower the perceived benefit of bed-nets, the caretakers in this study did not see them as substitutes for the long-lasting physical barrier that a bed-net provides. This could partly be explained by the common belief that the insecticide used in both ITNs and IRS only affects the malaria-carrying mosquitoes and that other mosquitoes were perceived to be resistant; thus bed-nets were seen as a necessary barrier against all other mosquito types and insects. Additionally, the perceived short-lived effect of IRS hinders its eligibility to replace bed-nets; findings which concur with those found in a recent study in Mozambique, where IRS was perceived to have similar limitations .
As was previously documented by Winch et al. (1994) in Tanzania, seasonal usage of nets was attributed to the inconvenience of sleeping under the net in the heat, the perceived low mosquito density and perceived low risk of malaria in the dry seasons . The discomfort of sleeping under the net during the hot season was also previously shown in Kenya . While the heat barrier could be reduced by the use of nets with a large mesh size, a study that was conducted in the Solomon Islands demonstrated that large mesh-size was a cause of concern to the users who believed that mosquitoes might more easily penetrate the net . While this concern, which has also been documented previously in Zanzibar , was raised in our study, the large mesh size was also acknowledged as a positive characteristic of LLINs as a way to increase ventilation and reduce heat. The finding that continuous use of bed-nets was more common in children suggests that the perceived higher risk of malaria in children outweighed the caretakers' perceived discomfort of being hot. According to Janz and Becker, the perceived barrier component was the most powerful single predictor among the HBM elements; and although benefits were important, they more strongly predicted sick-role behavior than preventive health behavior .
The high cost of the bed-nets was identified as a barrier for bed-net ownership. Although all informants received at least one free LLIN, the cost barrier would need to be addressed to ensure sustainable bed-net usage when bed-nets need to be replaced. Since this study took place just over a year after the LLIN distribution, and since the expected lifespan of the LLINs is about 4–5 years, cases in which the LLINs had been discarded or worn out were not encountered. This, however, can be expected to happen with time. The cost barrier is especially relevant since the low malaria transmission may reduce donors’ willingness to continue free distributions of bed-nets in Zanzibar in the future. Therefore, strategies to provide subsidized and affordable nets through the private and public sector should be implemented both during periodical free distributions as a "keep-up" strategy  and as a sustainable delivery mechanism when future funding for free distribution campaigns ebb out.
Women in this study described having high self-efficacy and decision-making power concerning bed-net use for children and they were recognized as a source of knowledge on health issues. The self-efficacy component was added to the HBM in the 80s, and is especially relevant when applied to long-term behavioral changes . Therefore, it can be expected that women’s high self-efficacy and strong decision-making power are likely to facilitate regular bed-net use in small children.
Theoretical frameworks are useful in planning and designing health interventions, with the HBM being one of the most widely used theories to explain health service uptake and behavior . Although the model has been used extensively, it has been criticized for simplifying health-related representational processes and solely focusing on individuals' attitudes and beliefs without taking into account social, economic and even emotional factors that may also influence health-related actions. In addition, there is dispute as to which of the HBM components have more influence on health-behavior and what the relationship between these components are [21, 24]. In this study, the HBM was used as the framework to understand the community attitudes, perceptions and beliefs on malaria and bed-nets and not to describe the association between these beliefs and the actual behavior of caretakers.
Limitations of this study included the fact that the interviews were conducted by foreign researchers using interpreters. The presence of a foreigner may have compromised the comfort of the informant during the interview. Although mitigated by using an interviewer of the same gender as the informant, as well as adhering to appropriate dress-code, this limitation may have increased desirability bias, whereby the informants answered less truthfully in order to please the interviewer. Trustworthiness could also have been affected by the interpreter who may impact the findings . This was mitigated by training the interpreter on the interview guide and by transcribing only the Kiswahili sections of the interview. In this way the text included only what the informant had been asked and what the informant had answered regardless of the original question asked by the researcher.
Also, due to logistical constraints, the first contact with the informants was handled by the Shehas (village leaders). Although the Shehas were asked to introduce the study as a discussion around children's health, the researchers did not have direct control over how the study was presented to the informants. An additional concern is that the interviewees were selected due to their extensive answers during a previous survey. Therefore, there may have been a selection bias where more empowered women were chosen. This could have overestimated the observed high self-efficacy and decision-making power, and this finding is probably not transferable to all Zanzibari women.
Despite the limitations, this study gives an overview of the perceptions of malaria and bed-net use in Zanzibar. It shows that, although malaria is no longer viewed as a major health issue, seasonal usage of bed-nets might be sustained due to the perceived benefit of bed-nets in reducing mosquito bites, especially during the rainy season. This study, however, did not link caretakers' perceptions to actual use, and therefore sustained bed-net usage despite low risk perceptions should be further investigated through quantitative surveys.