Caregivers’ health literacy and management of presumptive malaria in children under five: A cross-sectional study from Ghana


 Background: Low health literacy is associated with poor health status, poor self-management, and poor use of healthcare. This study assessed the associations of caregivers’ health literacy, incidence of malaria in children, use of healthcare and associated household costs in the management of malaria in children under five years.Method: This is a cross-sectional study with data (N=1270) collected in November - December 2017. We used hierarchical cluster analysis to generate health literacy profiles of caregivers based on responses to the Health Literacy Questionnaire (HLQ). We run logistic regression models and generalized linear models with incidence of malaria, desirable use of healthcare and household costs as dependent variables, and health literacy profiles and other socio-demographic and access variables as covariates. Results: We generated 7 caregiver health literacy profiles with Profile 1 characterised by overall high scores, Profile 4 by overall moderate scores and Profile 7 by overall low scores on nine health literacy dimensions. With Profile 4 as reference, children of caregivers in Profile 7 had 69% increased odds of an incident of malaria and no difference in odds for those of Profile 1. Profiles 1 and 7 both had reduced odds of desirable use of healthcare, 26% and 58% respectively. Caregivers in Profile 1 incurred higher spending, while caregivers in Profile 7 incurred lower spending on management of malaria in children as compared to Profile 4. Conclusions: Our findings suggest that general health literacy of caregivers as measured by the HLQ may not be influential in incidence of malaria, desirable use of healthcare and household costs in the management of malaria in children under five years in Ghana where malaria is highly prevalent. The use of hierarchical cluster analysis was feasible in the analysis of comprehensive health literacy and facilitated analysis on use of healthcare and associated costs.

(WHO) recommends testing and con rmation of all suspected malaria cases before treatment with anti-malaria drugs for appropriate treatment and to avoid spread of drug resistance (11) yet presumptive treatment of malaria is still common especially in rural underserved areas.
Not only is malaria a public health problem, it also poses a signi cant economic burden at individual, household, and societal levels especially in developing countries (12)(13)(14)(15)(16). It is estimated that malaria is accountable for a 'growth penalty' of 1.3% per annum in some African countries and, when compounded, this penalty results in considerable difference in the GDP between countries with or without malaria (17). In the long run, this may severely retard the economic growth of the entire region (17). A study on economic costs of malaria in children under ve years in three Sub-Saharan African countries reported the costs for an episode of malaria to range from US$7.99 to US$229.24 in Ghana, from US$5.2 to US$137.74 in Tanzania and from US$11.24 to US$287.81 in Kenya (18).
In Ghana, malaria spurs on as one of the leading causes of morbidity and mortality and in 2018, the country recorded the highest absolute annual increase in malaria cases in the Sub-Saharan African region (10). In 2018, malaria in Ghana accounted for 4% of the global disease burden of malaria and 7% of malaria disease burden in West Africa (19). Despite progress, malaria remains a problem; in 2018 13% of deaths in children under ve years was due to malaria (20).
The economic burden of malaria may also result in catastrophic costs to households. For instance, a study in Ghana showed that about 5% of respondents spent more than 5% of their annual income on the treatment of malaria, which was described as catastrophic cost (15). Tawiah et. al reported average cost to households in Ghana to be US$14.61 per episode of malaria (21). Although the National Health Insurance Scheme (NHIS) was introduced in 2003 to alleviate the economic burden of healthcare including malaria treatments especially for poorer households (22) other factors such as low levels of active subscription and lack of proximity to health facilities reduces the functionality of this insurance coverage. For example, Fenny et al reported that about 15% of insured patients chose to use informal health care over formal health care as a result of lack of proximity to health facilities (14). In addition, it is reported that unfamiliarity with insurance validity and renewable times also contribute to low active enrolment of the Ghana NHIS (23), and thus low protection against nancial costs of malaria.
Health literacy may impact on health behaviour, thus in uences preventive measures against malaria as well as appropriate treatment seeking choices, which could consequently affect malaria incidence, care seeking pathways and associated treatment costs. However, not much research has been conducted on how health literacy levels in uence incidence of disease, health behaviours and associated costs in disease management in Sub-Saharan Africa. Speci cally, no study has been done on health literacy, management of malaria and associated costs. We set to explore the relationship between health literacy levels of caregivers with children under ve years, the incidence of presumptive malaria in their children, their use of health care and total household cost of presumptive malaria in their children under-ves in peri-urban Kumasi, Ghana

Method Study Area
This study was conducted in two districts: Ejisu-Juaben and Kwabre-East, in the Ashanti Region, Ghana. Both districts are farming communities with most of the women engaged in trading of farm products.

Study Design and Sampling strategy
Cross sectional survey data was collected in November -December 2017 for 1270 caregivers. However, we lost responses for 36 caregivers while synchronizing data from the data collection software. The study adopted multi-stage random sampling (sub-districts, communities, households) to select caregivers and details of the sampling strategy is explained in another paper on the cultural adaptation of the Health Literacy Questionnaire in Asante-Twi (24). The sampling process resulted in the selection of 18 communities, 9 from each of the selected districts. One of the communities was excluded from Kwabre-East district for logistic reasons. Data Material (or Data Collection) We used two questionnaires, the Health Literacy Questionnaire (HLQ) and a malaria speci c questionnaire, both in the local language (Asante-Twi) (24). The HLQ covers nine conceptually distinct areas (scales) of health literacy including: The scales are based on 44 question items. The question items of the rst part (scales 1-5) are scored on a 4-point Likert scale (Strongly disagree, Disagree, Agree, Strongly agree), while a 5-point Likert scale is used for the second part (scales 6-9), which rates the ability to perform various tasks (cannot do, very di cult, di cult, easy and very easy).
The malaria questionnaire included questions on recent episode of malaria in under ve-year olds within 6 months, choice of treatment pathway and associated household costs. Other questions covered socio-demographic characteristics of the respondents.

Variables of Interest
Three dependent variables were of interest in this study; incidence of presumptive malaria in children under ve, caregivers' desirable use of health care and total household costs of malaria in children under ve years. Associations of each of the dependent variable was measured in relation to the independent variable of primary interest; health literacy of caregivers (see more details in the analysis section) and other covariates.
Dependent variables Incidents of presumptive malaria were self-reported by the caregivers and here, we asked respondents to report on an episode of presumptive malaria within the past 6 months from the data collection date.
We categorised use of health care as resorting to either a desirable or an undesirable health seeking pathway. In the malaria questionnaire, caregivers were asked their sequence of choice of health care on an incident of presumptive malaria. Thus, on an incident of malaria, "what was your rst treatment source", then, "what next did you do if your child did not get better". This line of question ended on the fourth treatment choice, which mapped out a 4-sequence health seeking pathway as shown in the supplementary Fig. 1. We de ned a desirable pathway as resorting to the use of formal health care as at least second treatment choice, and otherwise undesirable. Formal health care in this study refers to both public and private health facilities. The de nition for desirable pathway was based on the Ministry of Health guideline for treating malaria which recommends using formal health care as the rst source of healthcare (25). We assumed that most respondents used some form of rst aid as rst treatment choice when a child had fever which is a common symptom of malaria, thus our de nition of a desirable approach rested on the rst two treatment choices. We further divided undesirable care seeking into two groups: Undesirable 1 where caregivers resorted to late use (third or fourth choice) of formal healthcare in the treatment seeking pathway and Undesirable 2 where caregivers resorted only to informal treatment including over the counter medication. We did this to assess the difference in costs for the two types of undesirable care pathways.
Total household cost was the accumulation of costs across the 4-sequence health seeking pathway and it included both direct and indirect cost per episode for managing malaria in children under ve years. The direct costs included consultation and medication costs. The indirect costs consisted of the value of days lost to care. Reported days lost to care was valued using the respondents' self-reported income.
We collected cost data in Ghana Cedis and converted to US Dollars based on 2017 currency conversion rate (GH 1,00 = USD 0,22).

Independent Variable of interest
Our primary variable of interest was health literacy. First we calculated the mean score for each of the nine health literacy dimensions, details are provided in another paper (24). To have a combined measure of health literacy, we used hierarchical cluster analysis, to create health literacy pro les of respondents, see details in the data analysis section below. The health literacy pro les were presented as dichotomous variables indicative of whether a caregiver is within a particular pro le or not. We used these as independent variables.

Covariates
Paasche Orlow and Wolf's causal model on health literacy(26) guided the selection of covariates which were grouped into socio-demographic and access variables. The socio-demographic variables included age of the child, sex of respondent, years of schooling, respondents' reported income, and employment status (both indicative of ability to pay). Access variables included reported National Health Insurance Scheme (NHIS) membership and distance to health facility.

Data Analysis
Data was analysed using STATA version 15. Using hierarchical cluster analysis, with Ward's linkage as the clustering method and the Squared Euclidean distance as distance measure, health literacy pro les were created based on the nine health literacy scales. The cluster analysis was performed on the standardized health literacy scores for each scale using zscores to account for the differences in score ranges among scales. The analysis yielded seven distinct pro les where each respondent could only be identi ed in one pro le. These pro les were described based on their average scores for each health literacy scale and later, a description of their socio-demographic characteristics. Representing the most dominant pro le among respondents, and with moderate scores across the nine scales, Pro le 4 was selected as the reference group throughout the analysis.
Using logistic regression, we analysed the associations between caregivers' health literacy pro les and the incidence of malaria in their children under ve, while adjusting for the socio-demographic covariates. This analysis included the total study sample (n = 1234).
The subsequent analysis focused on caregivers of children with an incident of malaria (n = 588), their use of healthcare and total household costs for malaria given their health literacy pro les. Use of health care and mean total household costs by respondent characteristics are presented using descriptive statistics. For binary variables, we tested for equal means using the t-test statistic and for ordinal or categorical variables, we used the ANOVA test. We used Spearman's correlation coe cient to assess the correlation between the health literacy scales and desirable use of healthcare. To minimize the effect of outliers on the estimated relationships, we trimmed the cost data at the 5 percent tails.
We used a logistic regression model to assess the association between desirable use of health care and health literacy pro les, while adjusting for socio-demographic covariates as well as access covariates.
We applied generalized linear model procedures with a log link function and Gamma distribution to measure the association between total household cost and health literacy pro les, as total cost (dependent variable) is not normally distributed. We adjusted for socio-demographic covariates as well as reported NHIS membership and whether the chosen treatment strategy was desirable.
P-values less than 0.05 was used as threshold for statistical signi cance.

Health literacy pro les
In Table 1, we described the average mean scores of the 9 health literacy scales across the 7 health literacy pro les. Health literacy scale scores are colour coded from highest (blue) to lowest (red) within each part of the HLQ.
The hierarchical cluster analysis produced seven distinct clusters and after discussions on the analysis between two of the authors, all the clusters were kept because each covered a good number of respondents for the analysis. Caregivers within pro le 1 (3%) scored high across all nine scales and Pro le 7 (2%) scored relatively low across all scales. In between these pro les, are pro les with moderate to high scores or moderate to low scores. Most respondents (32%) fell within pro le 4 with moderate scores across scales but better at actively managing their health. Pro le 5 had moderate to low scores with least average score in health provider support. Pro les 4, 5, 6, and 7 all had major challenges with health provider support. A detailed description of pro les and their socio-demographic characteristics are presented in the supplementary table 1 in the appendix.
Association between caregivers' health literacy pro les and incidence of malaria in children Table 2 presents the adjusted and unadjusted results of the logistic regression analysis on the associations between the health literacy pro les and incidence of malaria in children under ve years. After adjusting for covariates, caregivers in pro le 7 had increased odds of 69%, thus there is 69% higher likelihood for a child of a caregiver within this pro le, to have an incident of malaria compared to Pro le 4. Similarly, caregivers in Pro le 6 had 44% increased odds for their children to have an incident of malaria as compared to those in Pro le 4. Pro le 3 and 5 showed decreased odds of 12% and 10% respectively as compared to Pro le 4. However, none of the associations were statistically signi cant. Supplementary Table  2 in the appendix is a descriptive table on caregivers' health literacy pro les and the incidence of malaria in children. Associations between health literacy pro les and desirable use of health care Table 3 shows the use of healthcare by characteristics of caregivers with children who had an incident of malaria. Overall, 58% of caregivers resorted to a desirable treatment seeking pathway. In summary, female caregivers from Ejisu-Juaben, who spoke English at home, within the ages of 45-69 years, employed, with no long term illness, with a child between 37 to 48 months and reported having a valid NHIS subscription had a higher chance of choosing desirable pathway to seeking treatment. Desirable use of healthcare increased with increase in years of schooling and was most frequent for health literacy Pro les 3 (66%) and 1 (65%) and least frequent for Pro le 7 (31%). Prior to assessing the associations between the seven pro les and desirable use of healthcare, we investigated the association between the nine health literacy scales and desirable use of healthcare. Spearman's correlation showed a positive association between the nine health literacy scales and desirable use of health care, as desirable use of healthcare increased with increasing levels of health literacy. However, the correlation coe cients were small and only statistically signi cant for scale 9, Understanding health information (ρ= 0.08; p-value=0.04).  Association between health literacy and total household cost of treatment On average total household cost associated with malaria treatment was USD 20.29 with an average direct cost of USD 9.54 and indirect cost of USD 11.10. Table 5 shows the mean total household cost for malaria in children under ve years by the socio-demographic characteristics and health literacy pro les of their caregivers.
Total household costs were higher among older caregivers, caregivers with more than 12 years of schooling, unemployed caregivers, caregivers without long term illness, caregivers who spoke English at home, caregivers with children within 25-36 months age group, caregivers who chose desirable treatment pathway and caregivers without National health insurance. With regard to the health literacy pro les, total household cost for caregivers in Pro le 1 (USD 24.77) was higher than all other pro les with least cost in Pro le 6 (USD 17.93). The mean difference in the total household costs for all the groups were not statistically signi cant except for use of health care (desirable and undesirable). *shows statistically significant difference between total cost incurred by users of desirable healthcare and undesirable healthcare users. Table 6 shows the association between health literacy pro les and total household cost using the generalized linear model. After adjusting for all covariates, caregivers from all pro les incurred less household cost on malaria as compared to Pro le 4 except for Pro le 1 ($0.24). Caregivers in Pro les 6 incurred least cost ($-0.30), then Pro le 7 ($-0.14) as compared to Pro le 4. Pro le 3 incurred less ($-0.11) total household cost on malaria in children than Pro le 4.   In summary, the ndings suggest that the identi ed health literacy pro les 6 and 7 had considerable higher odds of experiencing a child with malaria. Pro le 7 individuals were less likely to use desirable healthcare and as well incurred less on total household costs as compared to caregivers who had moderate scores (Pro le 4) across the nine scales. As compared to Pro le 4, caregivers in Pro le 1 were also less likely to use desirable healthcare, yet, incurred higher total household costs. Caregivers in Pro le 3, however, had reduced odds of an incident of malaria in children, more frequent use of desirable healthcare and in addition, incurred less total household costs on malaria in children as compared to Pro le 4.
None of the associations were, however, statistically signi cant. Although desirable use of healthcare was associated with higher costs, table 7 showed that a caregiver would pay much more for delayed use of formal healthcare.

Discussion
In this section, we discuss the analytical approach, de nitions and measurements of outcome variables and relate the study ndings to existing studies on the same or similar associations. Therefore, our discussions centre on the use of health literacy pro les, the meaning, and implications of desirable use of healthcare, then, the association between health literacy and use of healthcare. We follow up with discussions on our cost estimates and the associations between health literacy and costs. Finally, we present strengths and limitations of this study.
We generated seven health literacy pro les which mapped out the combined strengths and weaknesses of caregivers in terms of the nine health literacy dimensions (27)(28)(29). The pro le descriptions coupled with respondents' socio-demographic characteristics may provide useful information for policies and intervention design (30) . For example, almost all health literacy pro les in our study had challenges with health provider support, so this dimension could generally be addressed. On the other hand, some pro les scored low on almost all dimensions. While it is di cult to target interventions based on intangible characteristics such as health literacy, the identi cation of socio-demographic characteristics of the least scoring pro le, improves the ability to reach the group with the greatest need (30). Thus, the use of this multi-dimensional health literacy questionnaire (HLQ) is essential in identi cation of health literacy needs, still, the use of the pro ling approach could contribute to equitable health literacy interventions (30).
In this study, use of health care was categorized into desirable and undesirable healthcare. Use of health care is usually de ned as either use of formal healthcare (public or private health facilities) or informal healthcare which includes drug stores, herbal and traditional medicine or other therapies acquired outside the hospital for the treatment of a particular disease (14,(31)(32)(33). We found that 42% of caregivers used an undesirable treatment pathway, either not seeking formal care or doing so only as 3 rd or 4 th choice. Other studies in Ghana also suggest a high use of informal care for malaria treatment (15,21,34), with a recent study reporting 28% choosing informal over formal care among caregivers of children under 5 years in Upper West Region (15). This is in line with our nding of a high percentage of undesirable practices as we include late use of formal care. Based on the recommendations by WHO and Ghana's Ministry of Health's malaria treatment guidelines(10, 25) which encourages seeking early treatment for malaria from health facilities, we de ned desirable healthcare as timely use (at least second treatment option) of health facilities in treatment of malaria in children under ve years.
However, in malaria prone areas, presumptive malaria is mostly de ned by the presence of fever (35), and in children, several illnesses could present with fever(36) which may or may not necessarily require a visit to the hospital. Thus, in this study, some caregivers may have used informal treatment as the rst or second option and did not seek further treatments because the child recovered. Such caregivers were categorized as using an undesirable approach, which could be questionable if the child recovered from the illness. Although, our de nition is in line with appropriate and accurate timely diagnosis and treatment of malaria(10), the desirable use as de ned may also result in an in ux of patients at health facilities which yields unnecessary pressure on the limited resources in the Ghanaian health system. In other words, there may be scope for management of some of these presumptive malaria cases at lower level than the health facilities; indeed, this was the basis for programmes like the community case management of malaria in Ghana (35). For countries with limited resources, the proliferated use of informal sources of healthcare in the treatment of malaria could be channelled into an advantage by training and equipping such sources to promote test prior to treatment (37,38). This could serve as a form of screening of mild and severe cases for treatment, thereby making it a desirable treatment option which would likely imply less cost to the individual as well as the health system.
Although health literacy is associated with proper management of health conditions and engagement in healthy behaviours (1,3,4,39), the associations between health literacy levels of caregivers and the use of health care for their children has been mixed (31,40,41). For example, a study conducted in Wisconsin USA, by May et al reports that, certain capabilities including good assessment of severity of illness, navigating the health system, relationship between caregiver and health providers as well as perception of care, could positively in uence the behaviour of caregivers in managing their child's health in mild or acute illnesses with possibilities of waiting for longer appointments (42). Thus, such caregivers are less frequent users of formal health facilities. In a similar study setting in the United States, Sanders et al reported that there was no statistically signi cant association in health care use and costs for children of caregivers with low health literacy or otherwise (31).
In the present study, we found that caregivers in both Pro le 1 (high scores) and Pro le 7 (low scores) were associated with reduced odds of using desirable healthcare. This could be driven by different factors. Given the high health literacy scores and high level of education, caregivers in Pro le 1 may assume to have better judgement of health conditions and only use formal healthcare when necessary as reported in the study by May et al (42). Again, with over 80% of caregivers in Pro le 1 employed, the high indirect costs which in this study comprises value of time spent in care, could possibly also in uence their decisions. However, characterized with low health literacy scores coupled with low education and socio-economic status, cost could be a more in uential determinant in the decisions of caregivers in Pro le 7 in their choice of informal treatments. Other studies also show that caregivers with low socio-economic status use more informal healthcare due to the low cost of care (43,44). However, none of the associations in our study were statistically signi cant as Sanders et al(31) also reported. As we nd high total costs associated with the desirable use of health care, much effort will be needed to stimulate the practice of desirable healthcare by such groups.
With cost being a barrier as mentioned above, we could expect National Health Insurance Scheme (NHIS) membership to be a mediating factor. The mediating effect was evident as we observed a reduction in the differences in the odds ratios after we adjusted for NHIS covariate. Caregivers in Pro le 3 subscribed (94%) to the NHIS more often than all other pro les including Pro le 1 (82%). Thus, it seems that high subscription to NHIS emphasizes the strength of caregivers in Pro le 3 in the active management of health and in their high (57%) use of public health facilities as compared to the 47% use by caregivers in Pro le 1. Agbanyo et al. reports that among expectant mothers' subscribers of the NHIS have 29.4% higher likelihood to use public health facilities and clinics with 2.3% higher likelihood of using private facilities (45). It is also reported that the positive effect of the NHIS is predominant in lower-educated than otherwise (46). Several other factors could also account for the differences in the use of healthcare and cost among the Pro les. These could include, but not be limited to, trust in healthcare providers, the perceptions on various curative measures, birth parity, and the perception of the disease of interest (43,47).
In the present study, household costs per an episode of malaria on the average amounted to $20.29 of which $9.54 in direct treatment costs and $11.10 due to lost income. This appears to be slightly above estimates in other studies that also measured costs for malaria in children under ve. Using a model-based analysis Sicuri et. al, report an expected average direct treatment cost per case of US$7.85 in 2013 (18) and two studies report direct and indirect costs in 2011 around US$14.00 (US$ 13.90, US$14.6) (21,34). In comparison with costs reported from other studies in Ghana, (18,21), the differences in costs estimates could be attributable to the differences in study population, increasing costs over the years, inclusion of expenditures for informal care in our study, valuation of indirect costs and differences in recall period. Out of the identi ed studies on costs estimates on malaria, some focus on estimates for malaria in children (15,18) while others report on cost estimates for the general population (21,34,48). In terms of differences in years of cost estimates, most studies were based on almost 10 years old data (18,34).
Comparison of cost estimates across studies is also challenged by different cost components included in the studies, for example one study reports on only direct costs (15), others report on both direct and indirect costs (18,21,34); some studies report on both informal and formal treatment costs (15,21,34), and others report on only formal treatment costs (18). However, where possible, we compared relevant sub-results from the studies. Generally, compared to the other cost studies with 2 weeks (21) and 1 month recall period (15), our study had a longer recall period of 6 months and this could have in uenced the cost reported by respondents. The cost level apart, our study reports the indirect cost as a greater share of the total costs which is in line with other studies which reports 71% (48), 80% (21) and 74% (34) of costs as indirect expenditure in management of Malaria.
With respect to associations between health literacy and cost of healthcare, the direction did not follow previously reported associations between health literacy and cost of health care (8,9). We found that costs tended to be higher among caregivers within pro les with relatively good health literacy compared to those with relatively poor health literacy. This could re ect that caregivers with high health literacy value health and health care and thus are willing to pay more for appropriate care, but it could of course also re ect a higher ability to pay. Another possible reason could be a difference in the type of undesirable healthcare used between Pro les 1 and 7. Although both high and low scoring caregivers were associated with reduced odds of using desirable healthcare, the use of undesirable healthcare differs between the two extreme pro les. Caregivers in Pro le 1 were dominant in the late use of formal healthcare, which was the most costly use of healthcare, while caregivers in Pro le 7 dominated in sole use of informal healthcare, the least costly use of healthcare.
We found that resorting to an undesirable treatment pathway, on average, costs less than seeking early treatment from a formal source of healthcare. This is because caregivers who chose an undesirable pathway (highly informal), were likely to incur lower direct cost due to the absence of consultation costs and other possible informal payments, including motivations for health staff at health facilities (15). However, the lower cost should be assessed against health outcomes forgone. According to Sumba et al(33), on an episode of malaria in Kenya, children who are sent to health facilities as rst source of treatment are signi cantly more likely to recover and would not need further treatment as compared to those who use informal treatments (33). Thus, resorting to informal treatments could likely lead to failed treatments, hence require further treatment possibly from formal sources of health care, which could in sum result in higher costs of treatment.

Strengths and Limitations
This study used a standardised health literacy questionnaire (24) in measuring health literacy levels. The translation and cultural adaptation of the Health Literacy Questionnaire reports of an acceptable t for the overall nine factor model with less model t for the ninth scale on understanding of health information (24). However, the less model t had no in uence on the pro le groups. This is the rst study to use this tool validated for use in a local language in Ghana and thus assesses the usefulness of the questionnaire.
The cross-sectional design is a limitation as it does not permit close monitoring on costs from the onset of malaria till the child recovered. It also does not allow rm conclusions on causal inference. Our study values work time loss based on selfreported income and a detailed measure of income could have improved this estimate. Due to long time recall period of six months, it is likely that some caregivers did not report the accurate costs they incurred when their child had malaria. This study used six-months recall period because, it has been reported that children are likely to have an episode of malaria within a six months period (33). However, the recall bias may have in uenced the results in this study. We, however, have no reason to believe that this inaccuracy would differ by health literacy level. In terms of the overall relationship between health literacy, incidence of malaria, use of healthcare, and costs, malaria may not have been the best health condition for the assessment considering the familiarity of the disease in this context. A more complex health issue like a non-communicable disease may have shown different associations.

Conclusion
We sought to assess the association between general health literacy levels of caregivers, incidence of malaria in children, use of healthcare and related total household cost of malaria in children under ve years. Our nding showed that caregivers with low health literacy levels more frequently experienced incidence of malaria in their child, less frequently used desirable healthcare and likely incurred less cost on management of malaria in a child. On the other hand, caregivers with higher health literacy levels less frequently experienced an incident of malaria in a child, less frequently used desirable healthcare and likely incurred more cost on management of malaria in a child. However, none of these associations were statistically signi cant. Thus, general health literacy levels of caregivers may not be a signi cant factor in an incident, the use of healthcare and associated costs on malaria in children under ve years. Further studies on health literacy and costs could focus on more complex health conditions like non-communicable diseases.
In this study, the use of hierarchical cluster analysis helped to describe our study population based on their comprehensive health literacy scores from the HLQ, their demographic characteristics, and again facilitated the analysis on use of health care and costs. Such analytical methods appear feasible in the identi cation of vulnerable groups towards equity in healthcare. We received ethical approval from the Committee for Human Research and Publication Ethics, an institutional review board of the Kwame Nkrumah University of Science and Technology (CHRPE/AP/506/17). All methods were carried out in accordance with relevant guidelines and regulations. We obtained written informed consent from all respondents and parents or legal guardian of respondents under 18 as approved by the ethics committee.

Consent to publish: Not applicable
Availability of data and material: The datasets used and analysed during the current study are available from the corresponding author upon reasonable request.