Developmental deficits related to malaria are complex and varied. In a systematic review of the relation between P. falciparum infection and cognitive function, Kihara et al. reported that specific parts of the brain may not always be impacted more than others by malaria as the effects can vary and be diffuse [10]. Impairments are driven by severity of malaria, symptomology, and pre- and post-disease exposure [11]. Malaria infection has been found to be significantly associated with a variety of impairments in areas such as language [12,13], motor [14], memory [14,15,16], executive function [13], and attention [15,16,17] in prospective case-control studies of children in Sub-Saharan Africa. Results vary, however, largely due to heterogeneity in the timing of malaria onset compared with outcome assessments, sample sizes, and measurements of risk factors, malaria infection, and child development outcomes [10,11].

The path to cognitive impairment may result from many interacting biological and social risk factors [12]. Risk factors for cognitive impairment associated with malaria include fetal exposure to malaria, repeated malaria infections, malaria symptoms (seizures, coma duration, psychosis, impaired consciousness, hypoglycemia, high fever, inflammation, lethargy, and compromised play/learning time), antimalarial medication, malnutrition, and low socioeconomic status [5,18,19]. In endemic areas where children are prone to repeated infections and other risk factors, malaria can be considered a syndrome with enhanced longer term brain and behavior effects [11].

The functions that are impaired may depend on the stage of brain growth at which the infection is acquired and impairment may not manifest until a child is older [5]. Age at malaria infection is a critical factor influencing cognitive impairment [5,6] and the pathways connecting malaria and child development change with age [20,21,22,23]. Although most research is on school-age children, malaria attacks are more common and severe in younger children and cognitive effects might be worse in children under five years of age [6,9]. For instance, in cross-sectional studies in Zanzibar, Tanzania, malaria among children under two years was associated with lower motor activity, lower language scores, and later onset of walking controlling for socioeconomic status, length-for-age Z-scores, anemia status, and season [20,24,25]. In Uganda, two longitudinal studies found that 12 months after having malaria, children under five years had low cognitive ability, attention, memory [17], gross motor, fine motor, visual perception, and language scores [14]; however, these studies only include children who had one cerebral malaria episode.

Malaria may impact early childhood development through a pathway of anemia/low hemoglobin concentrations. It is well-established that malaria causes anemia, especially in children under five years in Sub-Saharan Africa [26,27,28,29]. Malaria leads to anemia via two pathways: through the rapid destruction of red blood cells and as a result of virus-induced systemic inflammation that triggers a release of the iron-regulatory hormone hepcidin from the liver [26]. Hepcidin effectively blocks the absorption of iron in the intestine and leads to sequestration of iron in macrophages [30]. This hepcidin-induced hypoferremia can lead to iron-deficiency anemia, if prolonged. Recent evidence also highlights the role of hepcidin-mediated effects of inflammation on brain iron metabolism with cognitive implications [31,32,33]. Children 6 to 24 months of age with iron-deficiency anemia are at risk for poorer cognitive, motor, social-emotional, and neurophysiologic development in the short- and long-term [34]. Therefore, iron-deficiency anemia may be an important mechanism in the association between malaria and child development.

Although there is evidence that iron-deficiency anemia can lead to poor child development, there is a research gap in assessing malaria, anemia, and child development. Higher iron concentrations can increase malaria risk among children [35]. However, iron supplementation can increase hemoglobin concentrations in young children and in the WASH (water, sanitation, and hygiene) Benefits trial in Kenya, lipid-based nutrient supplements (LNS) increased hemoglobin concentrations, reduced anemia, and increased iron status [36]. In a randomized controlled trial of malaria prevention among children under three years of age in Uganda, anemia mediated the negative association between malaria infection and cognitive performance in all treatment arms [37]. In Tanzania, hemoglobin concentrations played a significant role in the pathway between malaria and motor abilities among children as they started to walk [24]. More well-designed studies longitudinally evaluating the pathways (including anemia) between mild and severe malaria infection, and a variety of child development outcomes among children under five years are needed to elucidate causal mechanisms and account for seasonal and age-dependent nuances.

Malaria is widespread throughout Kenya and a leading cause of morbidity and mortality in the country [38]. About 74% of the Kenyan population is at risk for malaria and malaria accounts for 30% to 50% of all outpatient clinic or hospital visits [38]. In western Kenya, malaria is endemic and transmission is perennial with seasonal peaks in May to July and October to November [39]. Malaria is estimated to cause 20% of all deaths in children under age five years in the country [38] and the disease is common in this vulnerable population of young children [40]. Malaria and anemia are concurrently high in many parts of Kenya [41]. In 2016, the prevalence of anemia among children under five years in Kenya was 41% [42]. The prevalence of other indicators of poor child health related to nutrition and child development in Kenya is high. Information about early childhood development in Kenya is limited however, and most of the program and policy efforts in the country focus on preschool-age children despite the importance of development in younger years [43].

This study aimed to describe the association between malaria infection and early childhood development. The objectives of the research were to: 1) investigate associations between malaria infection and three child development outcomes, and 2) assess the role of anemia as a potential mediator in the pathway between malaria infection and child development outcomes. The following two connected hypotheses were tested: 1) malaria infection is associated with increased odds of being at risk for gross motor, communication, and personal social developmental delays, and 2) anemia is a significant mediator along the pathway between malaria infection and being at risk for gross motor, communication, and personal social developmental delays.

These data were collected during the WASH Benefits trial, which is a cluster-randomized controlled trial assessing the impact of WASH and nutrition interventions individually and combined on child health outcomes [44,45,46]. The study was conducted in Bungoma and Kakamega counties in western Kenya. The region is rural and the majority of the population are subsistence farmers. The WASH Benefits trial took place from November 2012 to July 2016. Households were selected to participate in the trial if they were in a rural community (defined as <25% of residents living in a rented home (most families in rural areas own homes compared with urban households in Kenya), with <2 gas stations, and <10 shops in the area) and had limited access to water and sanitation (defined as communities where >80% of residents did not have access to piped water). Female participants were eligible for enrollment into the study if they were pregnant, if they or their partner owned their house, and if they were not planning to move within the next 12 months. The infant that was in utero during the village census was the index child.

Households were visited at enrollment and two follow-up periods, one and two years later. A sub-study assessing environmental enteric dysfunction, malaria infection, and hemoglobin concentrations from biological samples enrolled 2,304 households from four arms of the WASH Benefits trial (WASH, WASH and Nutrition, Nutrition, or Active Control) [36,44,47]. The nutrition intervention entailed nutrition education and all children between 6–24 months in the participating household receiving LNS with iron. In the second follow-up, 1,449 children were included in the sub-study [36,44,47]. This sample size was calculated based on having the power to detect differences in environmental enteropathy biomarkers [48] and this sample size was sufficient to determine the association between malaria infection and child development outcomes. An equal proportion of children were selected from each intervention arm.

Caregivers brought study children to a central location for sub-study data collection between August 2015 and April 2016. Blood samples for malaria and anemia diagnoses were taken by trained phlebotomists and health surveys were completed. Caregivers provided written informed consent. Consent for blood collection and a sufficient blood sample was obtained for 699 children and therefore malaria and hemoglobin measurements only existed for this subset of children. An average of two to three months later, caregivers brought study children to a central location for main trial data collection, including a child development assessment. Local enumerators conducted most surveys in Kiswahili, the regional language. Surveys and data collection tools were developed from validated measures [49] and locally adapted.

Venous blood samples were drawn from sub-study participant children in a central location within villages for biomarker analyses and drops of blood were used to assess malaria parasites and hemoglobin concentrations. Rapid diagnostic tests (RDTs) for P. falciparum, P. vivax, P. ovale, and P. malariae were conducted using rapid diagnostic kits (SD Bioline, Malaria Ag, P.f, P.f, P.v, Alere). The kits can detect parasite concentrations up to 28 days after infection. All children with a positive test result were defined as having malaria in these analyses. Axillary temperatures were taken for all children and if the malaria parasite test was positive and the child’s temperature was 37.5 degrees Celsius or higher, he/she was referred to the nearest clinic for malaria treatment. Hemoglobin concentrations were measured using a Hemocue (Hb 301) machine (Angelholm, Sweden). Anemia was defined as having a hemoglobin concentration below 11.0 g/dL [50]. A medical history form was also completed with caregivers about the child’s recent experiences with illness, prophylaxis, and treatment related to a variety of health conditions, including malaria.

Household demographics, including asset ownership and maternal education, were assessed using standardized questionnaires administered at enrollment. Socioeconomic status was defined with an asset index using enrollment survey results for all sub-study participants. The asset index was created using the first principal component from a principal components analysis, which has been shown to be an appropriate proxy for household wealth and is correlated with consumption expenditures [51,52,53]. Items included in the asset index were housing structure, electricity, and ownership of a radio, television, mobile phone, clock, bicycle, motorcycle, gas stove, and car. Maternal education was categorized into the following categories: incomplete primary school, completed primary school, and any secondary school.

Three domains of child development were assessed: gross motor, communication, and personal social. The corresponding subscales of the Extended Ages and Stages Questionnaire (EASQ) were used to measure each developmental outcome among all WASH Benefits trial participants during the last data collection round (at approximately age 24 months). The EASQ is a modified version of the Ages and Stages Questionnaire (ASQ), which is widely used for the developmental screening of children under five years of age [54,55,56] and has been used in LMICs [57]. The questionnaires contain a series of age-specific items assessing the achievement of developmental milestones and tracking the child’s progress. The EASQ was piloted to assess respondent bias, maternal accuracy in reporting children’s abilities, cultural appropriateness, and feasibility according to standard procedures [49]. The Gross Motor subscale evaluates body and muscle movement, including tasks like standing, walking, and balancing. The Communication subscale assesses language development and the use of words or sounds to express feelings. The Personal Social subscale reflects emotional responses and social interactions. The EASQ was translated and adapted for this study by using local culturally appropriate items, examples, and tasks, but no substantial changes were made to the original questionnaires. Continuous scores for each subscale were calculated and converted to age-adjusted Z-scores based on the control group within the larger WASH Benefits trial. Children were considered at risk for delay in this population if their Z-scores were below the 25th percentile of control group Z-scores for each domain. This method of defining risk for delay was used as a conservative strategy because there was no prior research informing a culturally appropriate clinical cut-off in this study population. Binary variables were used to incorporate a quantitative definition of poor development and to be consistent with analyses used in the main WASH Benefits publications [46,58].

All statistical analyses were conducted in Stata 14 (StataCorp, College Station, TX, USA, 2015). Three sets of logistic multivariate analyses were undertaken to test the hypotheses corresponding to each child development outcome (gross motor, communication, and personal social). Two regression models were built for each child development outcome to assess the pathways between malaria infection (binary) and being at risk for gross motor, communication, and personal social delays (binary). The first model included malaria infection (binary) and the child development outcomes (three separate binary outcomes). Anemia status (binary; anemia defined as hemoglobin concentration <11.0 g/dL) was added to the second model. The interaction between malaria status and participation in the nutrition intervention was also analyzed in multivariate models to assess whether being in a nutrition treatment arm differentially affected malaria infection. All models were adjusted for child characteristics (sex [binary] and age [continuous, months]), household demographics (asset index [continuous] and maternal education [categorical]), study cluster, and whether the household received the nutrition intervention (binary). Analyses controlled for arms that received nutrition interventions because they were likely confounders.

Mediation analyses were then undertaken to assess whether anemia was a mediator along the pathway between malaria infection and being at risk for gross motor, communication, and personal social delays. The binary_mediation method in Stata [59,60,61] was employed. Like the Sobel–Goodman test for continuous outcomes, the binary_mediation method shows that mediation occurred if: 1) malaria infection was significantly associated with the mediator (anemia), 2) malaria infection was significantly associated with developmental delay in the absence of the mediator (anemia), 3) the mediator (anemia) was significantly associated with developmental delay, and 4) the effect of malaria infection on developmental delay decreased upon the addition of the mediator (anemia) to the model. The proportion of the total effect of malaria infection on the risk for gross motor, communication, and personal social delay that is mediated by anemia was also calculated and reported.

Children with missing malaria data or child development assessments were removed from analyses. Nine percent of children had missing data from one of these measurements and thus were not included in the final analyses. The final sample size was therefore 592 children. Differences between those who provided consent for blood collection and had a sufficient blood sample, and those who did not were assessed due to the low response rate and no significant differences or systematic attrition were found [36].

The WASH Benefits trial is registered as a clinical trial with the U.S. National Institutes of Health (https://clinicaltrials.gov/ct2/show/NCT01704105). Ethical approval was obtained from the University of California, Berkeley Center for the Protection of Human Subjects and the Kenya Medical Research Institute Ethical Review Committee. Adult participants provided written consent for themselves and their children prior to enrollment.

Maternal education in this population was low: 53% of mothers did not complete primary school (Table 1). The mean age of the children was 24 months, with 29% of children stunted and 7% underweight. The prevalence of anemia among children included in these analyses was 31%, which is similar to the prevalence of anemia among children in the Nutrition intervention arm (36%) and WASH + Nutrition intervention arm (27%), and lower than the prevalence of anemia in the control group (49%) [36]. Eighteen percent of children had a positive malaria blood test suggesting that they were infected with a malaria parasite in the previous month. Of these, the majority (81%) were infected with P. falciparum. The percent of children that were at risk for gross motor (20%), communication (21%), and personal social (23%) delays in this population (had Z-scores in the lowest quartile based on control group domain Z-scores) were similar across domains.

Having a positive malaria test was significantly associated with increased odds of a child being at risk for gross motor (OR 2.22; 95% CI 1.35, 3.64), communication (OR 1.92; 95% CI 1.19, 3.09), and personal social (OR 2.86; 95% CI 1.81, 4.55) delays (i.e., having domain Z-scores in the lowest quartile) adjusting for child characteristics, household demographics, study cluster, and participation in the nutrition treatment arm (Model 1, Table 2). The odds of a child being at risk for developmental delays (i.e., having domain Z-scores in the lowest quartile) across all subscales decreased with the addition of anemia (Model 2, Table 2) to models, but remained significant. The strength and significance of the associations were greater for personal social delays across Models 1 and 2. Anemia was only significant in personal social analyses. Interactions between malaria and participation in the Nutrition intervention arm were not significant as the treatment arm did not play a role in this association.

Anemia was a significant mediator in the pathway between malaria infection and risk of gross motor, communication, and personal social development delays (i.e., having domain Z-scores in the lowest quartile) (Table 3). Across all subscales, malaria was significantly associated with being anemic and risk of developmental delay, anemia was significantly associated with risk of developmental delay, and the effect of malaria on risk of developmental delay decreased upon the addition of anemia to the models. The proportion of the total effect of malaria on risk of developmental delay (i.e., having domain Z-scores in the lowest quartile) that is mediated by anemia in this population was greatest with the communication subscale (16%), however, all proportions were small (Table 4).