Correlates of social behavior change communication on care-seeking behaviors for children with fever: an analysis of malaria household survey data from Liberia

In partnership with the National Malaria Control Programme and the National Health Promotion Division of the MOH, the RBHS project developed early case management materials and audio messages as part of the “Healthy Baby, Happy Mother” campaign [5]. Campaign materials included five radio spots promoting four key case management messages that encouraged early detection of child fever, home management of child fever, prompt care-seeking at a health facility, and full treatment compliance according to the directions of a health provider [9]. Case management radio messages were aired on two national radio stations and on eight community-based radio stations [9] in the five project counties (Marietta Yekee, personal communication January 23, 2017). Between October 2010 and October 2014, the radio spots were aired 7776 times—a total of 129.6 playing hours [9]. In addition, posters and brochures were distributed in two of the study counties: Bong and Grand Cape Mount. Approximately 30,000 posters and 50,000 brochures were distributed to health facilities, gCHVs, and school children [5] in five total counties, including Bong and Grand Cape Mount. The brochure specified that ACT was a ‘new’ medicine in Liberia that treated malaria. The poster focused on home management and instructed caregivers to give a febrile child paracetamol (acetaminophen) and a cold sponge bath and promptly seek care at a health facility.

The data analysed in this manuscript came from a 2014 cross-sectional household survey implemented to inform malaria prevention and treatment programmes in Liberia. For this survey, all 15 Liberian counties were separated into a low- or high-malaria transmission group based on whether the prevalence of parasitaemia among children 6–59 months of age was below or above the median county prevalence (52%) [10]. Two counties from each group were then selected at random. The selected counties from the lower prevalence group were Bong (41%) and River Cess (50%), and the counties from the higher prevalence group were Cape Mount (59%) and Grand Kru (70%). Within each county, 15 clusters or villages were selected based on their population size, as reported in the 2008 National Population and Housing Census data. Female caregivers from 600 households in each transmission group were interviewed in order to detect differences between malaria prevention and treatment behaviors across levels of malaria transmission of 10% points or greater. In each cluster, 20 households that were home to children under five years old were selected to participate in the survey. One female caregiver from each household was interviewed.

In 2014, Subah-Belleh Associates, a local research firm, collected data from 1200 households (300 from each county) and interviewed 1200 female caregivers between March 31 and April 22, 2014. Only households that were home to a child under five who was under the care of a female caregiver were eligible for the interview. In each eligible house, after obtaining informed consent, interviewers approached the head of the household and asked him or her list all of the people who slept in the household the night before the interview. Data collectors also used the household questionnaire to record sociodemographic characteristics of the household and information about the household’s physical dwelling. From the household listing, data collectors randomly selected a child between the ages of 0 and 4 years old and interviewed the child’s primary female caregiver using an individual questionnaire. Depending on the composition of a household, the head of the household was sometimes also the caregiver who responded to the individual questionnaire; in other words, a single individual could have completed both questionnaires.

The individual questionnaire was used to gather information about the female caregiver and if her child had recently had fever. If her child had been ill with fever, she was asked if and where she sought care, whether or not the child received a blood or finger stick for malaria diagnosis, and what medications the child was given for fever, if any. Interviewers asked caregivers to indicate if a child in their care had been given sulfadoxine–pyrimethamine, chloroquine, quinine, ASAQ, aspirin, paracetamol, or any other drugs for their fever. If a caregiver’s child had suffered from fever recently and had received ASAQ, interviewers asked her if the child received ASAQ the same or next day as the fever’s onset, to determine whether an ACT had been received promptly. Interviewers did not ask caregivers if they had sought care the same or next day as fever onset. Interviewers also asked caregivers about their exposure to any malaria prevention or treatment message as well as to the “Healthy Baby, Happy Mother” campaign, and what campaign messages they recalled. Data from the hard copy questionnaires were entered into a Census and Survey Processing System data shell.

Analysis was limited to 607 female caregiver and child pairs in which the child had experienced fever within the two weeks prior to data collection. The authors used Stata (version 14.0) to analyse all data. First, all caregivers were nested within their sampling cluster or community and ran bivariate analysis to examine how each potential correlate was associated with the two dependent outcomes. Using the variables that were significant from this step, multivariate multilevel models were run to detect association with the two dependent variables. For each dependent variable, an empty model was run with no covariates to assess the variance in the outcome at the cluster level. Based on significance of independent variables in bivariate analysis, a model was constructed for each dependent variable by adding the same set of independent variables to a full model.

While the campaign encouraged caregivers to seek care promptly at a health facility, the questionnaire design prohibited an investigation of the relationship between a caregiver’s exposure to the campaign and prompt care-seeking at a health facility for her child’s fever. The reason for this was that the individual questionnaire asked caregivers if they had taken their child with fever in the last two weeks to a health facility, but not if they had taken the child immediately after the child had become ill. In contrast, the questionnaire did ask the caregivers if the child had received an ACT and then asked them if this had occurred the same or next day as fever onset. In order to examine prompt care-seeking behavior among female caregivers, this analysis examined correlates of a caregiver taking her child with fever during the last two weeks to a health facility and correlates of that child receiving ACT the same or next day as fever onset. Two binary dependent variables of interest were created to determine if a woman sought care at a health facility at any time during a child’s fever and whether or not a caregiver’s child receiving ASAQ or an ACT the same or next day as fever onset. Caregivers who sought care at a hospital, community health center, health post, mobile clinic, or pharmacy were considered as having sought treatment at a health facility.

Independent variables included a woman’s educational level, religion, age, marital status, household size, household’s wealth quintile, county of residence, exposure to radio, recall of campaign messages, and recall of any malaria prevention or treatment message. Wealth quintiles were constructed based on household characteristics and ownership of assets. Each head of household described their household in terms of various physical characteristics, such as the type of roof, floor, and water infrastructure. Principal component analysis was used to categorize each household as belonging to one of five wealth quintiles, with each quintile reflecting the relative wealth of each household compared to all households sampled. To account for the possible influence of other anti-malarial communication interventions, interviewers first asked caregivers if they had heard or seen any malaria prevention or treatment messages in the past year. To measure exposure to the campaign, interviewers also asked caregivers if they had heard or seen the phrase “Healthy Baby, Happy Mother” in the last 6 months. Interviewers then asked caregivers how they had been exposed to this campaign. Possible responses included radio, posters, television, community events, a friend/family/neighbor, and a health-care worker. The last response potentially included not only facility-based health providers, but also gCHVs. Next, interviewers asked exposed caregivers to recall what health messages they associated with the phrase. A categorical variable was created to assign all female caregivers to one of four groups according to their ability to recall anti-malarial and/or campaign messages. The first group was comprised of female caregivers who were not able to recall any campaign messages or any messages relating to malaria prevention and treatment. The second group included caregivers who could not recall any campaign messages, but recalled hearing or seeing a message relating to malaria prevention and treatment. The third group included caregivers who could recall one or more campaign messages, but said they had not heard or seen any anti-malarial prevention and treatment messages. The fourth group included caregivers who recalled one or more campaign messages and reported hearing or seeing an anti-malarial prevention or treatment messages.

Table 1 presents characteristics of interviewed female caregivers whose children had been sick with fever during the past two weeks and caregivers whose children had not. The proportion of caregivers between the ages of 18 and 24 years old was smaller among caregivers with sick children (29.0%) than those without (38.4%). Over half of caregivers with sick children (55.5%) and without (54.5%) had either no education or had not completed primary school. Roughly three-quarters (76.1%) of the caregivers with sick children were Christian and almost one-fifth (19.0%) were Muslim. Among caregivers without sick children the proportions reporting different religions varied. Most of the female caregivers (86.8%) with sick children and without sick children (83.3%) were married. About three-fifths (60.5%) of the caregivers with sick children listened to the radio at least once a week, while a smaller proportion of caregivers (45.9%) without sick children listened to the radio at least once a week. The distribution of caregivers with and without children varied across the four study counties. The size of caregivers households varied between those with caregivers with and without sick children in the past two weeks.

Table 2 describes caregiver exposure to anti-malarial messages as well as the “Healthy Baby, Happy Mother” campaign. Among all the female caregivers, about a third (35.4%) were either not exposed to or could not recall any health messages associated with the “Healthy Baby, Happy Mother” campaign (results not shown in Table 2). About one-tenth (11.2%) of caregivers did not report hearing or seeing a malaria prevention or treatment message in the past year or hearing or seeing the “Healthy Baby, Happy Mother,” campaign in the past 6 months. Slightly more (12.2%) had heard or seen a malaria prevention or treatment message in the past year, but had not heard or seen the phrase “Healthy Baby, Happy Mother.” A small proportion (5.1%) of caregivers had heard or seen the campaign phrase, but reported not hearing or seeing malaria prevention or treatment message. Most commonly (71.5%), caregivers reported hearing or seeing the phrase “Healthy Baby, Happy Mother” in the 6 months before their interview and having seen a malaria prevention or treatment message. The most common channels through which women were exposed to the campaign were the radio (59.4%) and health-care workers (57.8%)—a category that included both facility-based health workers and gCHVs. Caregiver exposure to the campaign through other communication channels, such as posters (26.2%); discussions with friends, neighbors, or family members (18.9%); and community events (14.2%), was less common. In terms of how exposure to the campaign channels varied across counties, the proportion of exposed caregivers who heard the phrase, “Healthy Baby, Happy Mother” on the radio varied across River Cess (69.6%), Bong (62.2%), Grand Cape Mount (50.0%), and Grand Kru (50.0%) counties. The proportion of exposed caregivers who saw the campaign phrase on a poster was greater in Bong (36.7%) and Grand Cape Mount (55.9%) counties than in Grand Kru (7.7%) and River Cess (13.0%), which were counties where the poster was not widely distributed (results not shown in Table 2). Exposed caregivers recalled an average of 1.4 of the four campaign case management messages (results not shown in Table 2). Among exposed women, the most commonly recalled message was taking a full dose of treatment for malaria (60.0%). A third of exposed caregivers recalled the message that instructed mothers to seek care promptly for a child’s fever (33.1%), give the child a sponge bath before seeking care (28.0%), and detect fever early (15.3%).

Ideally, children taken to a health a facility would be tested for malaria and prescribed an ACT if the child was diagnosed with malaria. Results indicated the vast majority (98.5%) of female caregivers sought advice or treatment for children who had suffered from fever within the two weeks before the survey interview. The two most common places caregivers sought out treatment were at a government hospital (52.1%) or community health center (20.7%). Caregivers less frequently sought treatment for child fever from a travelling medicine seller (8.5%), pharmacy (7.7%), private hospital/clinic (7.5%), or traditional health practitioner (6.0%). Table 3 displays the prevalence of various care-seeking, diagnosis, and treatment behaviors among interviewed caregivers. Four-fifths (80.6%) of sick children were taken to a health facility. Nearly three-quarters (72.6%) of children received a blood test for malaria—likely using either an RDT or microscopic examination of a blood smear. Roughly three-fifths (60.4%) of children received a positive malaria diagnosis.

Almost all children (95.8%) for whom treatment or advice was sought were given medicine for their illness. Two-thirds (66.7%) of children received an ACT, however, only 38.0% were given the medication on the same day or the day following the fever’s onset. The most common places caregivers obtained ACT were at a government hospital (55.4%) or at a community health center (18.0%), a category that included community-based centers and clinics. Roughly fourth-fifths (86.2%) of caregivers whose children received an ACT also reported obtaining the medication from a government hospital or community health center. Caregivers less frequently sought out treatment for child fever from a family or friend (7.8%), traveling medicine seller (6.5%), different public health-care source (6.3%), or private hospital/clinic (6.0%) (not shown). Results indicated that a greater number of children received medication for fever than were taken to a health facility or diagnosed for malaria. Over two-fifths (43.1%) of children who received an ACT did not take it promptly.

Table 4 displays the results of multilevel analysis for the two behavioral outcomes. The interclass correlation coefficient (ICC) of 0.42 in the empty or null model (Table 4, Model 1) indicated that over two-fifths (42%) of the variance in a caregiver taking a child with fever to a health facility was due to unmeasured correlates. The full model (Table 4, Model 2) with all potential correlates indicated that a caregiver’s residence in Grand Kru County (OR 0.21, 95% CI 0.073, 0.632). The ICC for this model indicates that 19% of the variation in odds of seeking care at a health facility was dependent on correlates operating at the cluster or community level, such as distance from a health facility.

The second empty multilevel model (Table 4, Model 3), predicting whether a febrile child received ACT promptly, indicated that 17% of variance in this outcome was due to unmeasured differences at the community level. The full model (Table 4, Model 4) with all potential correlates indicated that listening to the radio at least once a week (OR 1.99, 95% CI 1.359, 2.929) was associated with increased odds of prompt ACT treatment. It also showed that recalling one or more campaign messages was significantly and positively correlated with increased odds of a caregiver’s child receiving ACT the same or next day as fever alone (OR 3.62, 95% CI 1.398, 9.372) or in combination (OR 2.11, 95% CI 1.090, 4.131) with hearing or seeing any malaria prevention or treatment message in the past year. Results indicated that children of caregivers who listened to the radio at least once a week were about twice as likely to receive an ACT promptly than those of caregivers who did not listen to the radio at least once a week. The ICC for this Model indicates that variation in the odds of prompt ACT at the community or cluster level was negligible.

This study has several limitations. All data is from a cross-sectional survey, which means associations did not signify causal attribution. The first Ebola cases were reported in Liberia during data collection and may have influenced the care-seeking behaviors for children with fever. However, relatively few suspected Ebola cases or confirmed deaths from Ebola were reported in April 2014 [22] when data collection took place. In addition, very few Ebola cases occurred in children under five (6%) [23]. The accuracy of the authors conclusions is subject to caregivers ability to recall campaign and the specific medicine that their child received. The study’s ability to describe variance in the two behavioral outcomes is limited by the omission of supply-side factors from the study questionnaires. As stated earlier, the study was not able to measure the campaign’s association with prompt care-seeking at a health facility—one of the campaign’s key target behaviors. Instead, this analysis examined similar case management behaviors: seeking care at a health facility and promptly receiving ACT.