Patterns and predictors of malaria care-seeking, diagnostic testing, and artemisinin-based combination therapy for children under five with fever in Northern Nigeria: a cross-sectional study

Nigeria—the most populous country in Africa with 140 million people [16]—is comprised of six geopolitical zones (GPZs), 36 states, and 774 Local Government Areas (LGAs). Northern Nigeria is historically underserved and suffers the highest infant mortality rate (IMR) and under five mortality rate (U5MR) [11]: the North East and North West GPZs have an IMR of 77 and 89 per 1,000 live births [11] and a U5MR of 160 and 185 per 1,000 live births [11], respectively. This is higher than the national IMR of 69 per 1,000 live births and U5MR of 128 per 1,000 live births [11].

In order to address these regional disparities, the Targeted States High Impact Project (TSHIP)—a five-year reproductive, maternal, newborn and child health and family planning (RMNCH/FP) program—was implemented in Bauchi, a state with a population of 4.6 million in the North East GPZ, and Sokoto, a state with a population of 3.7 million in the North West GPZ [16]. TSHIP—implemented by JSI Research and Training Institute, Inc. from 2009 through 2015—was the first United States Agency for International Development (USAID) full-state approach to providing RMNCH/FP services in Nigeria. This approach addressed the challenges of effective donor collaboration, program efficacy, and health systems strengthening. Malaria interventions, especially for pregnant women and children under five, were implemented in these states through TSHIP. Bauchi and Sokoto experience the highest transmission of malaria from April to October.

In order to contextualize malaria care-seeking, understanding the structure of the Nigerian health care system is necessary, as well. In the public sector there are dispensaries, primary health facilities (e.g., Primary Health Clinics [PHCs]), secondary facilities, and tertiary hospitals [8]. The private sector is made up of patent medicine vendors (PMVs), private clinics and hospitals. The Ministry of Health (MOH) is responsible for providing operational supervision to the private sector to ensure compliance standards. However, enforcing these standards is difficult due to inadequate numbers of qualified personnel to carry out this task [17].

Despite poor regulation and standards of care, the largest source of treatment for children (and adults) with fever is PMVs (45.6%) and the second largest are government health facilities (33.8%) [11]. This is alarming due to the fact that PMV staff lack formal training and supervision [17] and serve a large proportion of patients.

In addition to high mortality rates and poor service delivery, commodity management is also poor in this region. Drug supply management has been an on-going problem, especially at the PHC level [13]. A stock-out analysis of ACTs and RDTs done by TSHIP in Sokoto showed a deficit of 2 million ACT doses and 700 thousand RDTs to meet the current malaria burden [18]. Thus, data to help identify patterns in care-seeking will allow us to inform programmatic and policy activities to strengthen malaria case management and to better serve rural and underserved populations.

In November and December 2012, a relatively low period of malaria transmission, a cross-sectional lot quality assurance sampling (LQAS) survey was administered proportionally throughout each of the 20 LGAs in Bauchi and 23 LGAs in Sokoto (Figure 1[19]). This was the baseline survey of an annual survey for monitoring and evaluation for TSHIP.

LQAS originated in the manufacturing sector to test the quality of a lot of goods. However, several studies and health programmes have used LQAS as a methodology to assess programme coverage and performance indicators in public health [5, 20]. This method of sampling divides a programme area into smaller geographic “lots,” or supervision areas, and then samples households from each lot. The results are then compared with a decision rule, or benchmark value, to help local TSHIP planners identify areas that are meeting or failing the benchmark value. This ability to prioritize the most vulnerable areas is crucial when resources are limited.

Nine separate questionnaires were implemented as part of the LQAS survey, which covered topics on maternal, newborn, and child health; malaria; and family planning to assess TSHIP’s programmatic activity. The data from “Questionnaire 8: Mothers of children 0 to 59 months with fever in the last 2 weeks” as well as “Questionnaire 1: Women of reproductive age with children aged 0 to 59 months” are used in this paper.

In Bauchi and Sokoto, each LGA acted as a lot or a supervision area. Within each lot, 19 settlements (villages) were randomly selected, proportional to population and representative of each LGA, to complete one comprehensive LQAS survey. LQAS methodology is based on a decision rule of choosing 12–30 settlements per lot. Nineteen settlements per lot have been proven to give a good fit for an annual coverage target of 50% or below with a sensitivity of 95% [21]. A household in each settlement was randomly selected and all applicable questionnaires were completed. If a selected household was not eligible to answer all nine questionnaires, the surveyor moved on to the next household to finish the remaining questionnaires. This approach, called “parallel sampling”, is unique to LQAS. If a settlement was too small to complete all nine questionnaires, a neighboring settlement was visited to complete the survey.

In a household where two or more eligible respondents were found, one respondent was randomly selected. Interviewers were trained on the LQAS methodology and use of the pre-tested tools over a period of five days. Questionnaire 1 contained detailed maternal demographic questions. The questionnaire (number 8) on malaria included questions about maternal and child’s age, whether treatment was sought outside of the home, promptness of treatment, location of treatment, whether the child received a diagnostic test (either microscopy or RDT), whether the child took a medication, and what type of medication the child took.

In this analysis, we combine all of the small samples to create an unbiased random sample of 814 households representing the states of Bauchi (n = 379) and Sokoto (n = 435). With this much larger sample size, there is more power to describe care-seeking and treatment patterns, and identify differences between the two states. Using the chi-square test, a 5% significance level, and the given sample sizes, there is 80–90% power to identify a 10% or greater difference between the states across the range of possible response percentages. Moreover, percentages reported for an individual state, or both states combined, have margins of error ranging from 2–5%.

Both the first and malaria questionnaires included women of reproductive age (15–49 years old) with children aged 0–59 months. The malaria questionnaire only included children who had a fever in the last two weeks. Detailed descriptive and demographic information is reported from the first questionnaire, since it was not collected on the malaria questionnaire. Some, but not all, mothers completed both questionnaires. Since both questionnaires focused on the same target population in the same villages the demographics should represent the women who completed the malaria questionnaire. Maternal and child age, and child’s sex, which were ascertained on both questionnaires, were statistically similar between the two survey samples.

Prompt treatment was defined as treatment on either the same or following day (24 hours) of fever onset based on recommendations from the RBM Partnership [6]. In addition, if the mother said the child received treatment, but indicated location of treatment as home and provider as self, the child was re-categorized as not receiving treatment, since treatment was not sought outside of the home.

Treatment location was categorized into formal and informal sectors. Formal treatment locations included health facilities and private doctors. Informal treatment locations included PMVs, community based health volunteers (CBHV), traditional practitioners, and other. Included in the informal “other” category are unauthorized drug sellers, friends/relatives, and other.

To assess the primary aim of defining current care-seeking and treatment patterns for children under five with fever, frequencies of survey question responses were computed. Responses were also cross-classified by state, and the chi-square test was used to assess the statistical significance of geographic differences. To address a secondary aim of comparing these results to the NMCP/WHO standard case management pathway, results were organized according to the standard and the number and percent of respondents meeting each component of the case management pathway was estimated. Epi Info version 7 [22] was used for data entry and data cleaning and Stata/IC 12.1 [23] was uses for data analysis.

To assess another secondary aim—identifying whether maternal age, child age and gender, and state help predict if a child under five with fever is taken to treatment—a binary logistic regression was performed with seeking treatment as the dependent variable. Akaike information criterion and likelihood ratio tests (LRTs) were used to assess the explanatory power of the model and identify factors that contributed significantly to each model.

Ethical approval was granted by the Sokoto and Bauchi State Health Research Ethics Committee and by local leaders in the study area. Approval from the National Health Research and Ethics Committee is not required when a study is done in less than three states. Participants in the survey gave written consent before answering the questionnaire.

Table 1 showcases the descriptive statistics of the study population. Stratification of the data by state shows significant demographic differences between states (p <0.05). Therefore, further analyses were done with all data combined and stratified by state. Of the 818 women and children pairs that answered this survey, four were excluded because they did not meet age criteria. Therefore, 814 woman and child pairs were included in the analysis.

Mean maternal age was 26.1 (15–48) years. The vast majority of mothers had no education (79.8%, 650) and only a small proportion had completed some higher education (1.7%, 14). The proportion of uneducated women was significantly greater in Sokoto (89.0%, 387) than Bauchi (69.4%, 263) (p <0.001). Most mothers were Muslim (96%, 781) and married (98%, 798). Children’s ages ranged from 0–59 months with a mean age of 17.0 months. Of the children, 53.2% (433) were male and 46.8% (381) were female. In addition, 15.8% (129) of mother and child dyads were part of a multifamily household.

Of children under five with fever in the past two weeks, 76.7% (624) were brought to treatment. Of those children who were brought to treatment outside of the home (624), 61.5% (384) followed WHO and NMCP standards to do so promptly, 46.0% (287) received treatment in the formal sector and 54.0% (337) received treatment in the informal sector. These behaviours did not differ by state.

Treatment was most often sought at PMVs (45.5%, 284), followed by health facilities (43.8%, 273). The remainder sought care at CBHVs, private doctors, traditional practitioners, and other sources (3% or less of the time, respectively) (Tables 2 & 3). Treatment location differed by state only for CBHVs, with children accessing them 4.6% (15) of the time in Sokoto and 1.7% (5) of the time (p = 0.04) in Bauchi (Table 2).

Of children who sought treatment, only 9.8% (61) received a blood test (Table 2). Children were most likely to receive a blood test if they received treatment at a health facility (15.0%, 41), next at a private doctor (14.3%, 2), traditional practitioner (6.7%, 1), and, last, a PMV (6.0%, 17) (Table 3).

Of children who received treatment, 39.9% (249) received an anti-malarial, however 29.8% (186) received a non-ACT anti-malarial and only 26.4% (165) received a prompt anti-malarial (Table 2). A child was most likely to receive an anti-malarial if they went to a health facility (46.5%, 127), next at a PMV (38%, 108), private doctor (35.7%, 5), CBHV (25%, 5), and traditional practitioner (13.3%, 2) (Table 3).

An even smaller proportion of children received a WHO and NCMP recommended ACT (10.1%, 63), and only 7.2% (45) received a prompt ACT (Table 2). Children were most likely to receive an ACT (14.3%) if they received treatment at a health facility (39) or private doctor (2), next from a CBHV (10.0%, 2) and then a PMV (7.0%, 20) (Table 3, Figure 2). Children were most likely to receive a prompt ACT if they received treatment at a health facility (10.6%, 29), next a private doctor (7.1%, 1), and last a PMV (5.3%, 15) (Figure 2). Treatment results did not differ by child’s sex (p ≥0.17), but did occasionally differ by state (p <0.02) (Table 2).

Figure 3 illustrates the NMCP/WHO standard treatment pathway. Of children under five who had a fever in the last two weeks (814), 47.2% (384) sought prompt treatment; of those that sought prompt treatment, 4.8% (39) of the total sample received a blood test; of those that received a blood test, 1.0% (8) of the total sample received an ACT.

In addition, Figure 4 illustrates which proportion of children received the NMCP/WHO standard of care stratified by treatment location. Of all locations, health facilities provided the greatest percentage of NMCP/WHO standard care (0.9%, 7). PMVs and traditional practitioners each treated one patient according to the NMCP/WHO standard, accounting for 0.1% of children with fever, respectively.

Table 4 presents the results of the logistic regression created to identify factors that help predict whether a child received treatment outside of the home. The results show that neither state, mother’s age, nor child’s sex was significantly related to treatment seeking. Child’s age was somewhat predictive of treatment seeking, but the pattern was not strictly linear. The clearest indication from the data was that treatment seeking was less likely for children under the age of six months. There was a significantly greater odds of seeking treatment for 7–59 month olds (OR: 1.74, p = 0.003) compared to 0–6 month olds, adjusted for mother’s age, child’s sex, and state.

Rates of treatment, prompt treatment, and formal sector treatment in Northern Nigeria are 77%, 47%, and 35% respectively (N = 814). This rate of treatment seeking is consistent with a 2013 study in Nigeria in which 70% of parents of children under five sought advice or treatment [11]. These rates are also slightly higher than rates discovered in a similar study of malaria treatment in Senegal where 62% of children under five with fever in the last two weeks received treatment, 40% received prompt treatment, and 32% received treatment from the formal sector [15]. However, there is much room for improvement since the malaria treatment seeking rates in Benin, DRC, Madagascar, Uganda, and Zambia are significantly higher (83 – 95%) [22]. Since only 47% of caregivers sought prompt care and 35% sought formal care, future programmatic emphasis may be placed on educating caregivers where and how quickly they need to seek treatment for children under five with fever. Future research in this area may also focus on assessing recognition of malaria symptoms and knowledge of case management strategies among families with children under five years old.

PMVs were found to be the preferred treatment providers, with 46% of all children under five with fever receiving treatment from this location. This is consistent with the national rate of 46% [11] and other areas such as the Kabale district in Uganda (53%) [23]. One reason for this may be that families feel that local private providers are more accessible, are more sensitive to the patients’ needs, and spend more time with them than public sector providers [24, 25]. However, there are key problems with this sector including inappropriate storage [25], knowledge, and dosing of drugs [24, 26].

Another downfall of seeking treatment at a PMV is that they are not authorized to administer diagnostic tests [personal communication, Dr. Mohammed Ibrahim]. This was evident by a recent survey that found microscopy completely unavailable at PMVs and RDTs available at 1% of PMVs [27]. In order to promote all steps in the malaria case management pathway, policy recommendations are to facilitate stocking of and training of staff in the use of RDTs at PMVs. In addition, when RDTs were implemented in the private sector, providers were not knowledgeable in treatment for non-malaria fever, which resulted in treatment with an ACT despite a negative RDT [28]. Therefore, capacity development and quality assurance of this sector is an important piece to improving malaria case management.

Less than half (44%) of the parents sought treatment for their children from a health facility, even though health facilities provided the largest proportion of children with recommended case management (Table 3). This rate is actually higher than the national rate of children receiving care at a public facility (35%) [11].

Despite a national benchmark to promptly test 80% of children under five with fever [8], in this study, children rarely received a blood test (8% of sample, 10% of care-seeking children). This is lower than the national rate of 11% [11], and much less than the average rate across 13 sub-Saharan African countries (17%) [29] and Rwanda (30%) [30], but higher than the rate of testing in Sierra Leone (1%) [31].

Although health facilities tested the largest proportion of children in this study (15%), that proportion is far below the testing rate of 95% in health facilities in Zambia [32]. In order to bridge the current gap between NMCP/WHO standard and current treatment, policies and programmes may focus on making blood tests and ACT more consistently available. A survey done by ACTwatch throughout Nigeria in 2011 showed that microscopic testing was more available than RDTs [27], therefore policies and initiatives to improve diagnostic testing should be consistent with commodity availability and acceptability. More specifically, of locations that stocked anti-malarials in the last three months, only 26% of public health facilities, 37% of private for profit facilities, and 1% of PMVs offered any testing services [27]. These stock-outs are a key weakness is the ability of facilities to provide diagnostic blood testing consistent with WHO/NMCP standards for malaria case management.

Of anti-malarials given, 75% were non-ACT, much higher than the national average of 26% [11] This is of great concern because malaria is resistant to many anti-malarials in Nigeria [8] and ACT has been the standard of care since 2005 [6]. The rate of standard treatment using prompt ACT is 7%; higher than the national rate of 4%, but much lower than a study in Tanzania (38%) [33]. It is likely that a combination of factors influence the low ACT rate. For example, a 2008 study showed only 36% of PHC health workers in Sokoto had adequate knowledge of malaria case management [34]. In addition, another study showed that only 49% of public facilities in Nigeria have at least one quality assured first-line dose of ACT [35].

Given the fact that PMVs are the most commonly accessed treatment location, interventions to strengthen their role in malaria case management may result in improved outcomes. In Kenya, an intervention that subsidized ACT for retail outlets, trained retail outlet workers, and strengthened community awareness of case management standards, saw a 25% increase in ACT treatment for children under five. In addition, a USAID report recommends engaging the private and even informal sector to improve coverage and case management of child illness [24]. Policy and logistics planning should address the lack of training and shortages of essential malaria commodities in recommended and frequently accessed treatment locations.

Given the current stock-outs of ACT and RDTs [18] in Sokoto State, commodity procurement and logistics need to be improved to bridge the gap between available and needed RDTs and ACT. Further research may be needed to understand weaknesses in the supply chain and training of health professionals. Future programmatic and policy activity should focus on strengthening these aspects of health care infrastructure.

While the majority of caregivers (77%) seek treatment for children under five presenting with fever, when care-seeking pathways are studied, only 1% of children under five with fever receive treatment according to WHO and Nigeria’s NMCP standards: prompt care-seeking, diagnosis with an RDT, and treatment with an ACT. This leaves a staggering discrepancy between the Nigeria NMCP and RBM goal coverage rate of 80% for prompt RDT use and subsequent ACT treatment if indicated [4, 6, 8] and the actual rate among children under five with fever in Northern Nigeria (Figure 3).

Policy and caregiver education strategies are important, but will have limited effect on improving malaria case management if health systems remain weak. In Nigeria, current health system weaknesses include stock-outs due to poor procurement and supply chains for malaria commodities in the public health sector, weak delivery of health services in public facilities, lack of necessary and adequately-trained human resources for health, and nearly non-existent diagnostics [4].

The poor state of malaria care in Sokoto and Bauchi could be improved by strengthening national policy and regulatory power in the informal sector. Since informal sector regulation and systems strengthening is often a difficult a lengthy process, incentives such as vouchers or conditional cash transfers may be used to direct patients toward the formal sector until the private and informal sector is can correctly deliver malaria case management.

Child’s age was a significant predictor in the regression for seeking treatment versus not seeking treatment. However, our modeling was limited by the number of variables on the malaria questionnaire. Thus, other factors may be as, or more, predictive of seeking treatment. However, cultural practice could plausibly contribute to the age differences. Several studies have shown that physical confinement is common in Nigerian postpartum practices—often lasting 40 days [36‐38]. Confinement has implications for the ability of the mother to seek care for her febrile newborn and may have affected the decreased odds of seeking care for 0–6 months. However, this information may be inconclusive due to the limitations in variables and sample size. Further research should be done to assess factors that facilitate recommended malaria case management.

There are some limitations to using a study based on LQAS methodology to assess care-seeking behaviour of parents. LQAS methodology is designed to cost-effectively compare programme coverage rates with set thresholds to evaluate programme performance. Thus, LQAS surveys typically have relatively small sample sizes (low power) and a relatively limited number of questions. This makes it difficult to perform more complex statistical analyses to establish causal and associative relationships. However, with over 800 respondents, there was good power to assess care-seeking, with a margin of error of only 2–5% for each question, but a limited number of explanatory variables.

Similarly, although polygamy is common in the study area, the survey did not assess whether each household was comprised of a polygamous or monogamous family; a study in Ethiopia found that mothers in a monogamous marriage had 3.4 greater odds of seeking treatment for a child under five with fever than mothers in a polygamous marriage [39]. The survey did not specifically ask if medications were procured at location of first line treatment or elsewhere. Thus, it is assumed that the medication was obtained from the first place of treatment. Lastly, the study did not assess the result of the diagnostic blood test. This is a limitation to assessing quality of treatment, since the proportion who received a medication cannot be compared to those who were diagnosed with malaria. However, malaria prevalence rates are so high in these areas, it is assumed the majority of fever cases were malaria.

Several types of bias may have been introduced in the study, but should not affect the conclusions of the analysis. Recall bias is a risk given this was a retrospective cross-sectional survey, compounded by the fact that largely uneducated women were asked to recall complex information. However, this bias should be minimal since the maximum time that had to be recalled was two weeks. Misclassification bias was minimized by only classifying a child as receiving treatment if the mother answered both yes to the child receiving a medication and listed what the medication was, even if it was “other” or “unknown”.