Ideational elements related to internet care behaviour: a multi-country evaluation | Malaria Journal


Study population and setting

This study analysed data from three cross-sectional malaria behaviour surveys in Cameroon, Côte d’Ivoire and Sierra Leone, implemented as community-based household surveys. The surveys were designed to estimate the prevalence of malaria-related prevention, care-seeking and treatment behaviours and their relationship with ideational factors, including attitudes towards caring for and using nets. The three surveys used nearly identical data collection tools and sampling protocols. The tools assessed malaria prevention strategies adopted by each country: ITN use, intermittent preventive treatment of malaria in pregnancy using sulfadoxine-pyrimethamine (IPTp-SP), case management, seasonal malaria chemoprevention (SMC)/intermittent preventive treatment of malaria in infants (IPTi), and indoor residual spraying (IRS) modules were included in the Sierra Leone and Côte d’Ivoire tools, while Cameroon omitted the IRS module. There were minor revisions to wording of questions and response options for contextual relevance.

Decisions about the scope of the survey were made in consultation with the Ministry of Health and National Malaria Control Programme in each country. In Côte d’Ivoire and Sierra Leone, the Government Offices of Statistics helped to select the study enumeration areas in urban and rural areas with probability proportional to size sampling and supplied enumeration area sketch maps. In Cameroon, the National Institute of Cartography provided this service. The survey was national in Côte d’Ivoire but conducted in two districts in Sierra Leone and two regions in Cameroon. In all three countries, a multi-stage, stratified sampling approach was used; this approach started with stratifying the study locations (zones in Côte d’Ivoire, districts in Sierra Leone, regions in Cameroon) into urban and rural areas. Subsequently, a number of enumeration areas were selected within each stratum using probability proportional to size techniques. Field workers visited each selected enumeration area, listed all households in the enumeration area and randomly selected between 21 and 23 eligible households for inclusion in the survey. In each selected household, all women of reproductive age were targeted for interview; in one third of selected households, the spouse of one of the recruited women was targeted for interview. Detailed information on MBS methods is available at

The survey in Côte d’Ivoire was nationally representative while the other two focused on specific locations in Sierra Leone (Bo and Port Loko districts) and Cameroon (North and Far North regions). The study population included the head of each household or their representative, women of reproductive age within each household, and for a portion of participating women, a male partner/husband was also interviewed.

Sample size

The sample size was calculated for the three cross-sectional surveys to provide estimates of certain indicators with a specified degree of certainty and to produce programmatically useful data. Sample sizes for each country were calculated based on the prevalence of three key indicators using the formula below:

$$n = d times frac{{Z_{{1 – frac{alpha }{2}}}^{2} times p(1 – p)}}{{delta^{2} times R_{h} times R_{i} }}$$


  • n is the required sample of individuals (e.g., women)

  • Z is the Z value corresponding to the desired 95% confidence level

  • d is the design effect due to departure from simple random sampling (2.0 in Sierra Leone, 2.5 in Cameroon, and 3.0 in Côte d’Ivoire).

  • p is the estimated (expected) outcome of interest. The same three outcome indicators were used in each country:

    • proportion of women of reproductive age with positive attitudes towards consistent use of mosquito nets

    • the proportion of women of reproductive age that slept under a net on the night before the survey

    • the proportion of children under five years of age that had fever in the last two weeks

      For outcomes that were not available in the publicly available data sets (e.g. Malaria Indicator Survey (MIS) or the Multiple Indicator Cluster Survey (MICS), a proportion of 50% (p = 0.5) was assumed for maximum variability and consequently maximum sample size.

  • δ is the desired margin of error (δ = 6% in Cameroon, 5% in Côte d’Ivoire and 5% in Sierra Leone)

  • Rh is the response rate for households; a response rate of 90% was assumed in all three countries.

  • Ri is the response rate for women in selected households; a response rate of 96% was assumed in all three countries.

The number of households that would need to be approached for each sub-national region: Port Loko and Bo (Sierra Leone), North zone, Central zone, South zone, and Abidjan (Côte d’Ivoire), and Far North and North regions of Cameroon was calculated for all three indicators. In each country, the maximum number of households that would need to be approached was selected from each indicator estimate as a target for data collection teams in that sampling area. The sampling strategy for the MBS is described in further detail at

Data collection

Trained teams implemented the same sampling procedures to collect survey data during the rainy season in each country. Data collection took place between September and October 2019 in Cameroon, between September and November 2018 in Côte d’Ivoire, and between September and October 2019 in Sierra Leone. Teams visited each sampled enumeration area and listed households to facilitate sample selection and data collection. Systematic random sampling was used to select households within each enumeration area. Interviewers used mobile devices to collect responses from household heads or their representatives in households where at least one woman of reproductive age lived. Household heads provided information about the household’s socio-economic status, the number of household members, and the number of ITNs in the household. In each selected household, data collectors interviewed all eligible women of reproductive age. In every third household, one participating woman’s male partner/spouse was also interviewed. More women than men were sampled to better capture malaria prevention behaviours usually performed by women such as accessing antenatal care (ANC) and IPTP-SP, accepting IPTi, and care-seeking for fever in young children. By interviewing men in one-third of the households, the survey also was able to better understand men’s views as key household decision-makers. Interviewers asked each adult about how they cared for the nets in the house as well as their attitudes and other perceptions related to net care, net use, and malaria in general (see Table 1).

Table 1 Description of independent variables examined in analysis

Field work was conducted by local data collectors in each country who were trained on ethical considerations for human subjects’ research by the study principal investigator, including informed consent, the study protocol and data collection tools. Data collectors used mobile devices and open data kit-based software (Survey-To-Go in Cameroon and Côte d’Ivoire; Survey CTO in Sierra Leone) to collect data. The study principal investigator for each country supervised field work and monitored data quality. The data were submitted to a secure online data management system on a daily basis.

The total population of households from which data were collected was 8,566 in Côte d’Ivoire, 3836 in Sierra Leone, and 4514 in Cameroon. As this paper focuses on ideational factors associated with net care, the analysis includes only the sub-set of households that had at least one ITN: 4264 households in Côte d’Ivoire, 1892 households in Cameroon, and 2077 households in Sierra Leone. This corresponded to a final sample of 6164 adults (1378 men and 4786 women) in Côte d’Ivoire, 2,995 adults (640 men and 2355 women) in Cameroon, and 2730 adults (456 men and 2274 women) in Sierra Leone. Refusal rates were 2% or less for household and individual participants in each country.

Human subjects

Each study was approved by the Johns Hopkins School of Public Health Institutional Review Board as well as the appropriate national research ethics authority in each country prior to contact with study participants, namely: the National Ethics and Research Committee in Côte d’Ivoire, the Office of the Sierra Leone Ethics and Scientific Review Committee in Sierra Leone, and the National Human Health Research Ethics Committee in Cameroon. Data collection teams received training on ethical guidelines and the rights of human research participants. As part of the informed consent process, trained data collectors verbally explained the purpose of the survey, the types of questions that would be asked, the potential risks associated with participating in the survey, and the actions the study team will take to protect the privacy of the participants and keep the data confidential. In addition, data collectors informed participants that they did not have to participate in the study, they could decide at any point to discontinue their interview, and they did not need to answer any questions they did not want to. Written consent/assent was obtained in each of the three countries. Recruitment, consent and assent procedures were designed and implemented to ensure voluntary participation.

Data analysis

Dependent variable

Researchers asked respondents what they did, if anything, to prevent nets from tearing or getting holes in them. The binary dependent variable reflects whether or not the respondent reported folding up or tying their net when not in use, in response to this question.

Independent variables

Analysis focused on the association between the dependent variable and household, sociodemographic, and ideational variables. Sociodemographic characteristics of the respondents included their gender, age and exposure to messages related to malaria in the last six months. Household demographic characteristics included household wealth quintile, household size, number of nets in the household, place of residence (i.e., rural or urban), and region or district of the country. Ideation variables included attitudes towards net use, net-care attitudes, perceived severity of malaria, perceived susceptibility to malaria, malaria-related discussion, perceived self- and response-efficacy, and perceptions of net use as a community norm. Lastly, the association between the dependent variable and whether a respondent reportedly slept under a net every night was also examined (see Table 1).

The questions linked with each ideational factor were converted into independent binary variables for each ideational element. This was done in a similar fashion as the procedure described in the 2017 Roll Back Malaria Partnership Malaria SBC Communication Indicator Reference Guide [35]. The questionnaire allowed for the creation of the binary ideational variables (see Additional file 1).

Most of the ideational variables were measured by asking respondents to indicate agreement or disagreement with Likert statements. Each respondent received a score for each question based on their response: (− 1) disagree, (0) don’t know/not sure/missing, and (+ 1) agree. If disagreement with the statement corresponded to a favourable response, the scoring for that particular statement was reversed. Thus, the respondent received a positive score for disagreeing (+ 1) with the statement. Second, an index score was calculated to reflect how each individual responded to the overall set of questions for an ideational factor. Each respondent’s index score was summed across the statements or questions. For example, three Likert scale statements were used to measure the perceived efficacy of nets to prevent malaria, the resulting index score would be integer values ranging from (− 3) to (+ 3). Third, based on their index score, each respondent was classified as having expressed a favourable ideational construct or not. A binary variable was created by classifying respondents with a zero or negative index score as not, for example, believing nets effectively prevent malaria (low response efficacy related to net use). Conversely, those with a positive score (above zero, 1) were considered to have high response efficacy and believe that nets were effective in preventing malaria.

Analytic methods

All data were cleaned and analysed using Stata 16.0. The data were analysed and compared by country rather than in aggregate due to differences in the scale and representativeness of the data between countries. Descriptive, bivariate and multivariate analysis were conducted on the survey data. Geographic and sociodemographic variables that are known to influence behavioural outcomes were included in the final models as well as variables that were significantly associated with the outcome. The specific descriptive and multivariate results shown reflect the key results of our study. Adjusted Wald tests were used to compare the proportions of different groups of respondents who reported they tied or folded up their net when not in use with different background characteristics (e.g., region, wealth quintile, gender). Multilevel logistic regression was used to account for the similarities among respondent responses within each household and the similarities among households within each sampled cluster. Intra-class cluster coefficients (ICC) were calculated to assess the need to incorporate random effects in each model.

Leave A Reply

Your email address will not be published.