Associations between ideational variables and bed net use in Madagascar, Mali, and Nigeria

Study setting

This study is based on analysis of household survey data collected in Madagascar, Mali, and Nigeria between 2014 and 2015 by the Johns Hopkins Center for Communication Programs (CCP), under the United States Agency for International Development (USAID)-funded HC3 project and with the support of the President’s Malaria Initiative (PMI). The surveys were designed specifically for the purpose of comparing ideational influences on the use of bed nets for malaria prevention as part of CCP’s mandate under this grant. These countries were chosen for the study because they are PMI priority countries.

CCP designed the questionnaires, but field work was conducted by trained interviewers through a contract with a private research firm in each country, selected through competitive procurement. Questionnaires were translated from the original English into local languages. Depending on respondents’ preferences, the interviews were conducted in English, Pidgin English, Igbo or Hausa in Nigeria; French or Bambara in Mali; and Malagasy in Madagascar. Data were collected during the dry season/early rainy season in Madagascar (between September and November 2014), the rainy season in Mali (between July and September 2015), and the rainy season in Nigeria (between July and September 2015).

The three countries have differing levels of malaria parasitemia in children 6–59 months—measured through microscopy—varying from 9% in Madagascar [49] to 27% in Nigeria [50] and 36% in Mali [51]. While malaria transmission in Nigeria is stable and perennial, northern Sahalian areas experience increases in transmission during 3 months or less of seasonal rains [52]. Mali’s Sahalian belt experiences stable, seasonal transmission while in the northern desert region transmission is low, unstable and epidemic-prone [53]. In Madagascar, the sub-desert and highland zones experience shorter seasons of increased transmission, the latter with pockets of high elevation where autochthonous transmission is rare; the tropical and equatorial zones experience stable malaria transmission most of the year [54]. Cyclones occur between December and April and often result in increased risk of malaria infection. The majority of malaria cases are caused by Plasmodium falciparum parasites in all three countries.

Mass bed net distributions took place in each of the three countries prior to survey data collection. While sensitization campaigns accompany most mass distributions, details about SBCC coverage, message exposure, and message recall are only available for a program in Nigeria, namely the Support to the National Malaria Programme (SuNMaP) funded by UKaid, and the NetWorks project funded by the U.S. President’s Malaria Initiative [55]. In that instance, over half of respondents in ten different states were exposed to multi-channel SBCC and positive attitudes towards nets were positively associated with the number of messages recalled.

Sampling design and participant characteristics

The sampling design was similar across the three surveys. Study participants were selected through a multi-stage random sampling process that involved first selecting—with probability proportional to size—districts, local government areas, or communes depending on the country, then clusters or enumeration areas, and finally households with at least one child under 5 years old. In each selected household, a child under 5 years old was randomly selected and the mother of that child was invited to participate in the caregiver survey. In our experience, household heads and caregivers are able to describe not only their own behavior, but that of other members of the family with a fairly high level of reliability. This is especially true for observable behaviors like sleeping under a bed net. We anticipated little additional informational value from multiple interviews in the same household, so opted to interview the caregiver of only one under-five child in each household and distribute the sample across a wider variety of households. In addition, in one third of the selected households, the head of household (if he/she was not the same as the caregiver) was also interviewed to obtain the perspective of the key decision-maker in the household. The multistage sampling design was taken into consideration in calculating the sample size. We applied a design effect of 2.0. Sample weights were not available; so, they were not used in the analyses. The sampling strategy yielded 2390 households in Madagascar, 3202 in Mali, and 3616 in Nigeria. Using a structured household survey questionnaire, information on ownership and use of bed nets was collected on all household members: 12,834 in Madagascar, 19,345 in Mali, and 16,832 in Nigeria. In this manuscript, we merged household-level data with data from the caregiver survey to assess the correlates of household members’ net use. Specifically, from the household data, we derived bed net use information and sociodemographic characteristics of individual household members, while data on female caregivers’ sociodemographic and ideational variables were derived from the individual caregiver questionnaire.

Variables

A cumulative ideation score was then calculated as the sum of the number of ideational variables that are positively associated with bed net use in each country (among the items with * above), producing an ideation score.

In addition to the main models that included the individual ideational variables, other models that substituted a composite core—the sum of the number of positive individual ideational variables for each respondent—were tested to assess the Sp: cumulative predictive power of ideation. In addition, we calculated an interaction term between age and sex in the estimated models for all three countries.

Analysis

We used both bivariate and multivariate analytic methods. The bivariate method compares net use across sociodemographic groups and reports the significance of the differences. Logistic regression was the main multivariable analytic method used and was limited to individuals from households with at least one net and at least one child under 5 years old, representing 73% of households in Madagascar, 97% in Mali, and 79% in Nigeria. We report both the odds ratios and the fully-standardized beta coefficients for the regression models.

Madagascar

Multivariate analysis

Results of the multivariate logistic regression indicate that the significant demographic predictors of net use among household members are the individual’s age, zone of residence, household wealth, household size, and number of nets in the household. Also significant were the female caregiver’s television viewing and her level of exposure to malaria-related communication messages. There were no significant interactions between age and gender (Table 1).

Predictor	Madagascar		Mali		Nigeria
	Odds Ratio	Betaa	Odds Ratio	Betaa	Odds Ratio	Betaa
Age group
Under-5 (RC)	1.00	–	1.00	–	1.00	–
5–17	0.411***	− 0.192	0.332***	− 0.265	0.502***	−0.162
Adult	0.934	−0.016	0.374***	−0.242	0.448***	−0.202
Gender
Male (RC)	1.00	–	1.00	–	1.00	–
Female	1.155	0.034	0.839‡	−0.044	1.044	0.011
Age group/Gender Interactions
Under-5 X Male (RC)	1.00	–	1.00	–	1.00	–
5–17 X Female	0.978	− 0.004	1.459***	0.073	1.084	0.015
Adult X Female	1.282‡	0.049	3.296***	0.247	1.667***	0.107
Household wealth quintile
Lowest (RC)	1.00	–	1.00	–	1.00	–
Second	0.971	−0.006	0.913	−0.019	0.960	−0.008
Middle	0.849‡	−0.029	0.852*	−0.031	0.962	−0.008
Fourth	0.801**	−0.041	0.711***	−0.065	0.950	−0.011
Highest	0.989	−0.002	0.646***	−0.088	0.879‡	−0.026
Household sizeb			0.831***	−0.316
Household sizeb
2–4 (RC)	1.000	–
5–6	0.452***	−0.173
7–8	0.246***	−0.274
9 +	0.147***	−0.338
Household sizeb
2–4 (RC)					1.00	1.00
5–6					0.529***	−0.150
7–9					0.365***	−0.234
10 +					0.207***	−0.313
Number of nets in householdc
1–2 (RC)			1.00	–
3–4			2.722***	0.240
5 or more			5.931***	0.415
Number of nets in householdc
1–2 (RC)	1.00	–
3 or more	2.793***	0.218
Number of nets in householdc
0–1 (RC)					1.00	–
2					0.893***	−0.028
3 or more					0.646***	−0.100
Zone of residence
Highlands (RC)	1.00	–
Sub-desert	3.595***	0.278
Tropical	2.928***	0.231
Equatorial	5.076***	0.352
Region of residence
Koulikoro			1.00	–
Sikasso			1.787***	0.130
Mopti			1.231***	0.045
Bamako			0.952	−0.010
State of residence
Akwa Ibom (RC)					1.00	–
Kebbi					1.487***	0.098
Nasarawa					1.099‡	0.022
Caregiver’s education
Primary or less (RC)	1.00	–	1.00	–	1.00	–
Post-primary	0.925	−0.011	0.849**	−0.032	0.957	−0.010
Caregiver’s religion
Non-Christian (RC)	1.00	–			1.00	–
Christian	0.888‡	−0.026			0.865*	−0.037
Caregiver’s radio listening habits
Fewer than once a week (RC)	1.00	–	1.00	–	1.00	–
At least once a week	0.911	−0.022	0.977	−0.006	1.024	0.006
Caregiver’s TV watching habits
Fewer than once a week (RC)	1.00	–	1.00	–	1.00	–
At least once a week	1.348**	0.041	1.227***	0.051	1.034	0.008
Caregiver’s exposure to messages on malaria
No exposure (RC)	1.00	–	1.00	–
Low	0.975	−0.006	0.962	−0.009
High	1.447***	0.073	0.899	−0.018
Heard messages on malaria					0.960	−0.010
Caregiver’s perceived severity of malaria
Lower (RC)	1.00	–	1.00	–	1.00	–
Higher	0.971	−0.006	0.973	−0.007	0.885**	−0.031
Caregiver’s perceived susceptibility to malaria		–				–
Lower (RC)	1.00		1.00	–	1.00
Higher	0.998	−0.000	0.977	− 0.006	0.927‡	0.019
Caregiver’s perceived self-efficacy to prevent malaria
Lower (RC)	1.00	–	1.00	–	1.00	–
Higher	1.567***	0.061	1.068	0.016	1.034	0.008
Caregiver’s perceived self-efficacy to detect a serious case of malaria
Lower (RC)	1.00	–	1.00	–	1.00	–
Higher	1.162**	0.033	0.940	−0.014	0.986	−0.003
Caregiver’s perceived response-efficacy of nets
Lower (RC)	1.00	–	1.00	–	1.00	–
Higher	1.137	0.017	1.060	0.012	0.919*	−0.021
Caregiver’s awareness that fever is a symptom of malaria
Not aware (RC)	1.00	–	1.00	–
Aware	0.985	−0.003	1.155***	0.035
Caregiver’s awareness that mosquito is the cause of malaria
Not aware (RC)	1.00	1.00	1.00	–	1.00	–
Aware	0.982	−0.003	1.365***	0.048	0.933	−0.010
Caregiver’s perceived self-efficacy to purchase enough nets
Lower (RC)	1.00	–	1.00	–
Higher	1.344***	0.067	1.200***	0.044
Discussed malaria with friends/relations	1.057	0.013			0.996	−0.001
Participates in decisions about net allocation	1.343***	0.061			0.998	−0.001
Attitudes towards nets
Negative (RC)	1.00	–			1.00	–
Positive	1.020	0.005			0.968	−0.008
Perceived net use as the norm among in community	1.391***	0.069			1.024	0.006
Knows where to buy nets	1.284***	0.058			1.023	0.005
Willing to pay for nets	1.015	0.003			1.045	0.011
Pseudo-R2	17.1%		9.9%		8.6%
Hosmer-Lemeshow GOFd (X2/ p)	15.3/0.08		15.5/0.08		14.9/0.09
Number of Observations	9260		19,345		15,463

Notes: ‡ p < 0.1; * p < 0.05; ** p < 0.01; *** p < 0.001

^aFully (XY) standardized beta coefficients

^bHousehold size was not normally distributed in Madagascar and Nigeria. The variable was differently categorized in the two countries in a way that helps to ensure model fit

^cNumber of nets was not normally distributed in any of the study countries. The variable was differently categorized in the three countries in a way that helps to ensure model fit

^dGOF Goodness of Fit

Among males, household members aged 5–17 years were 59% less likely than children under 5 years old to sleep under a net. There was no significant difference in net use between adult males and male children under 5 years old. Among children under 5 years old, gender did not make a difference in the odds of net use. Additionally, being female did not significantly moderate the relationship between age and net use. The relationship of net use and household wealth was such that the only significant difference was found between those in the lowest quintile and their peers in the fourth quintile. The relationship between net use and household size was negative. For example, people from households with nine or more members were 85% less likely than those from households with fewer than five members. Furthermore, having three or more nets in the household increased the odds of net use almost three-fold compare to having one or two nets.

Whereas female caregiver’s regular radio listenership (at least once a week) made no difference in individual member’s net use by 15%, regular television viewership (at least once a week) increased the odds by 35%. The female caregiver’s low exposure to malaria-related communication messages (one or two messages recalled) did not make a difference compared to no exposure; however, a high level of exposure (three or more messages recalled) was associated with a 45% percent higher likelihood of household member’s net use.

Significant ideational variables for female caregivers included perceived self-efficacy to prevent malaria, perceived self-efficacy to detect malaria, perceived self-efficacy to obtain enough nets for members of her household, perceived response efficacy of nets, descriptive norm about nets, awareness about where to procure nets, and level of participation in decisions regarding net allocation within the household. Residents of households where the female caregiver demonstrated perceived self-efficacy to prevent malaria were 57% more likely to sleep under a net than those in households where the female caregiver did not demonstrate such belief. Similarly, the female caregiver’s self-efficacy to detect a severe case of malaria was associated with a 16% increase in the odds of household member use of a net while the caregiver’s perceived self-efficacy to obtain enough nets for her household was associated with a 34% increase in the odds. Female caregiver’s belief that net use was the norm in her community increased the odds of household member’s net use by 39% while her awareness of where to procure a bed net increased the odds by 28%. Finally, members of a household in which the female caregiver participates in net allocation decisions were 34% more likely to sleep under a net than those in households where the woman did not participate in such decisions.

A look at the magnitude of the fully standardized beta coefficients reveals that the most important predictors of household members’ net use were the individual zone of residence, number of nets in the household, and household size and age, in order of importance. Among the female caregiver’s ideational variables, the most important were descriptive norms (perceptions about how other people behave), perceived self-efficacy to procure enough nets for the household, participation in household net allocation decisions, perceived self-efficacy to prevent malaria, and awareness of where to procure nets.

Figure 2 shows a graduated, dose-response relationship of net use with ideation score, controlling for all other non-ideational variables. The percent who slept under a net on the night before the survey increased from 54.5% in households where the female caregiver had zero positive ideational variables (out of six) to 83.2% in households where the female caregiver had all six positive ideational variables. When the composite ideation score (mean = 4.13, SD = 1.18) was substituted for the individual ideational variables in a logistic regression, the results indicate a strong positive relationship, controlling for all other non-ideational variables. Specifically, a one-point increase in the caregiver’s ideation score increases the household member’s odds of sleeping under a net by about 37%.

Mali

Nigeria

The role of independent demographic variables

In terms of demographic predictors of bed net use, there were more similarities than differences across the three countries analyzed in this study, even though Madagascar and Mali had much higher rates of bed net use (73% and 82%, respectively) compared to Nigeria (34%). In all three countries, children under 5 years old were more likely to sleep under a bed net compared to older children aged 5–17 years (and adults, except in Madagascar). Also in all three countries, the odds of bed net use were lower on average in larger households. The documented relationships with age and household size are consistent with what other studies have found in Africa and elsewhere [14, 57–60]. Also in Madagascar and Mali, bed net use was higher in the poorest households compared to wealthier households. Only in Mali were families whose caregivers had lower levels of education significantly more likely to use bed nets than families whose caregivers had more education. In Nigeria, Christians were less likely to use bed nets than Muslims. Findings of this nature can help programmers to target subgroups that are at higher risk.

In Madagascar and Mali, households with caregivers who regularly watched television were more likely to report bed net use. Consistent with findings from other studies, in Madagascar, caregivers’ higher malaria message recall was associated with higher rates of bed net use [4, 14, 16, 61]. The reasons for the lack of significant association between caregiver’s exposure to malaria messages and bed net use in Mali and Nigeria were not clear; we know little about the intensity or content of malaria messages except in Nigeria, so it is hard to attribute the findings to the reach or quality of the information provided. The overall lower use of bed nets in Nigeria may result in lower salience of malaria messaging, which would reduce motivations to pay attention to and process malaria information, even though communication campaigns were underway.

Effects of independent ideational variables

In terms of ideational predictors of bed net use, multiple factors were positively correlated with net use in Mali and Madagascar. Counterintuitively, in Nigeria, the higher the caregiver’s perceived severity of a malaria infection or her perceived response efficacy of nets (i.e., belief that bed net use can prevent infection), the lower the reported bed net use. Some conceptual frameworks that focus on threat and efficacy as predictors of behavior see high perceived risk as potentially debilitating—a frightened person may not take action to counter a disease threat if there is no accompanying belief in one’s ability to manage the threat [27, 62–64]. Sometimes self-efficacy can result in lower rates of behavior if it takes the form of overconfidence to control threats and achieve health outcomes [65–68]. This may be the case in Nigeria, where there was no evidence that self-efficacy perceptions—self-efficacy to detect or prevent malaria or self-efficacy to obtain bed nets—were associated with bed net use. People may feel so confident in their ability to detect and treat a severe case of malaria that they feel less threatened by the disease, and are therefore less motivated to sleep under a net.

In Mali, consistent with findings from many previous studies, knowledge about malaria played a role in motivating behavior: people who knew that fever is a symptom of malaria and who knew that mosquitoes spread malaria were more likely to report bed net use [57, 69, 70]. This finding was in contrast to what we found for the other two study countries. Without more information about existing malaria communication in Mali, it is difficult to know why these differences exist. Some malaria programs in Sub-Saharan Africa emphasis rapid careseeking in response to fever as part of large scale “test and treat” campaigns. Such programs would draw attention to information about malarial fever caused by mosquitos versus non-malarial fever and the importance of being tested to know the difference. In such a case, a stronger association between mosquitos and fever could result in greater interest in bed net use to prevent both. Future analysis of how malaria campaigns with different emphases may reinforce (or interfere with) each other would be useful.

In Madagascar, self-efficacy to prevent malaria and self-efficacy to obtain bed nets were associated with higher reported net use, echoing findings from other studies, including a similar study conducted in Liberia [14]. Unlike Mali, in Madagascar self-efficacy to detect a severe case of malaria was positively associated with net use. Also in Madagascar three additional ideational variables were positive predictors of bed net use: the belief that bed net use was the community norm, knowing where to obtain a bed net, and participation at the household level in decision-making about who would sleep under a bed net. Why more ideational variables were significant in Madagascar compared to Mali or Nigeria is not clear. Greater familiarity or experience with an issue is often associated with greater cognitive elaboration; if malaria communication (whatever its content) is more intensive or far-reaching in one place compared to another, that could explain more nuanced perceptions about disease threat and prevention. On the other hand, where exposure to information has been low, the introduction of an information campaign can have novelty effects—people learn a lot very quickly—whereas familiarity with an issue sometimes results in more counterarguing and rejection of new information. More research on the history and content of malaria communication in these countries would be needed to determine which explanation is more likely.

Madagascar and Mali share only one of the ideational predictors: self-efficacy to obtain enough bed nets for your household. In Nigeria, perceived severity of malaria and perceived response efficacy of nets were the only ideational variables associated (negatively) with bed net use. This finding suggests that structural and demographic barriers as well as psychosocial influences may need to be addressed as determinants of bed net use in that country.

It is important to note that another study in Nigeria that included measures of female caretakers’ malaria-related ideation [71] found it to be significantly associated with the caretaker’s own use of a bed net, rather than bed net use by members of her household. Other factors may also explain why household bed net use is more weakly correlated with caretaker’s ideation in Nigeria. For example, it may be that female caregivers in Nigeria have less agency to influence bed net use by other members of the household compared to female caregivers in Mali and Madagascar, thus weakening the causal link between caregiver ideation and household member behavior.

Implications for effective malaria social and behavior change strategies

It is essential to note that age, region, household size, and number of nets in the household were the strongest predictors of bed net use, demonstrating that families are prioritizing younger children, consistent with recent studies [72], and that the primary barriers to bed net use are still largely due to access. Controlling for these structural barriers permits SBCC programs to then isolate pertinent ideational variables. Of all the ideational variables examined in these three countries, those that were the most consistently significant predictors of bed net use were related to efficacy perceptions: self-efficacy to obtain bed nets for your household and self-efficacy to detect malaria. This suggests that efficacy-based communication strategies are likely to produce desirable results. Programs should emphasize increasing and reinforcing the head of household’s and caregiver’s confidence in responding to the threat of malaria and their correct and consistent net use as a preventive measure. The widely validated social learning approach of modeling effective behavior [21] could be used to show or describe locally acceptable and identifiable examples of effective net use and other preventive behaviors in an effort to build local self-confidence in malaria prevention. Showing how nets can be obtained and used may be particularly important in places like Nigeria where net use and the access to use ratio are low to begin with, and might complement expanded distribution efforts and accelerate uptake. Finally, while we cannot confirm the relationship between female agency and decision making using the current datasets, it may be important in settings where caregivers have less household authority to influence decision makers or other female caregivers in the same household. In such settings, modeling efficacious communication in the household about bed net use could be beneficial.

This study also provides support for the hypothesis of a cumulative impact of ideational variables. In Madagascar and Mali, where multiple ideational variables were significant predictors of bed net use (controlling for the effect of household and sociodemographic variables), bed net use was higher when more ideational variables were positive. This suggests that it may be effective to bundle more than one ideational variable into program messaging, for example, by addressing multiple aspects of self-efficacy and malaria knowledge in an integrated communication strategy.

Limitations of this study

The use of bed nets by household members was determined through an enumeration of household residents, of the nets available in the household, and through subsequent questions about who slept under each of those nets on the night prior to the survey. This produced an outcome variable representing whether an individual slept under a net the night before the survey. Yet, data on ideational variables were collected only from the female caregivers of a selected child under 5 years old in each household. Our analysis assumes that what a particular female caregiver—who may be only one of several caregivers in the household—knows, thinks, and feels about malaria and bed net use will have an impact on the bed net use patterns of the entire household. Ultimately, bed net use is a volitional behavior, particularly for older children and adults who may have the power to make their own choices and are therefore influenced by their own individual attitudes, beliefs, and preferences as opposed to those of the female caregiver who was interviewed. While our data indicate that there is a positive relationship between female caregiver ideation and household net use, controlling for numerous other household and demographic variables, the actual causal pathways are undoubtedly more complicated. Future studies may wish to collect ideational data from more than one individual in a household or, perhaps, include more questions about the content of interpersonal discussion of malaria prevention and net use among household members, in order to better understand the broader household context of net use decisions.