Epidemiological risk factors for clinical malaria infection in the highlands of Western Kenya

The study was conducted in Iguhu and Mbale, two highland areas in Western Kenya (Fig. 1). Iguhu (34°45′E, 0°10′N, 1430–1580 m above sea level) is located in Kakamega County and Mbale (34°74′E, 0°07′N, 1530–1690 m asl) in Vihiga County. Study participants were recruited from several clusters of villages in these two sites. Rainfall (annual average of 1977 mm) is seasonally bimodal, with the long rains occurring from March to June and the short rains from October to November. Average annual minimum and maximum temperatures are 13.8 °C and 28 °C, respectively. Malaria transmission is seasonal with peaks occurring 2–3 months after the peak rainy season in April–May, although the extent of the malaria burden varies considerably from year to year [17]. Vector breeding sites are known to be mostly found at the valley [18, 19]. Plasmodium falciparum is the predominant malaria parasite species in the study sites and it is transmitted by Anopheles gambiae, Anopheles arabiensis and Anopheles funestus [2, 18, 20, 21].

This was a matched case–control study with 302 clinical malaria cases and 604 matched controls to explore association between participants/household risk factors and malaria infection undertaken from March 2012 to July 2013. There were 192 and 110 cases from Mbale and Iguhu, respectively. Confirmed malaria cases were recruited from patients who reported to the Iguhu and Mbale District Hospitals during the routine diagnosis of the out-patients. A malaria case was defined as patient confirmed with fever and malaria parasites identified in thick or thin blood smears who presented to the two local health centres with other malaria symptoms (chills, headache or vomiting). A finger prick blood sample was taken from each study participant from which a drop was placed on a microscope slide. Both thin and thick blood smears were prepared from the drop of blood on the slide. These were taken to the laboratory of the Kenya Medical Research Institute (KEMRI), in Kisumu and stained using Giemsa for microscopic examination. All participated patients were cross-checked microscopically for malaria parasite infection by the laboratory technician at KEMRI. Parasite density was scored against 200 leukocytes when the slide was positive; otherwise, the whole slide was carefully scanned before being declared negative. Parasite densities were converted to number of parasites per microliter of blood, assuming a leukocyte count of 8000 cells/μL [17]. Participants were recruited to the study regardless of age, gender, and socio-economic status. Enrollment of participants in the study was voluntary following a written consent/assent form. Infants below 5 months and adults above 45 years were excluded from the study due to the difficulty in obtaining a case and age-matched controls within the same locality.

Controls were recruited from the community and were matched to a case by age and neighborhood. Controls were individuals living in the same village as the cases and of almost the same age. All controls had no fever and other symptoms of malaria during the past 48 h prior to survey. These control individuals were also screened for parasite infection by microscope as described for case individuals.

Recruited cases and controls were followed up to their homes and a questionnaire was issued to them or their guardians in the case of children to collect socio-economic data. The altitude, longitude, and latitude of the households of all study participants were determined using a hand-held Trimble GeoExplorer Global Positioning System (GPS). Data on the type of household construction including the presence of eaves, screens, roof type, floor type, wall type and number of rooms in the house were recorded. A pre-tested questionnaire was administered to the household head (or spouse) of both cases and controls to obtain socio-economic and behavioral data such as number of occupants, level of education and occupation of both household head and spouse, travel history, health-seeking behaviour and household assets.

Pyrethrum spray catches were done in the homes of both cases and controls to quantify the adult malaria vector densities in the houses. Mosquitoes were identified morphologically as An. gambiae sensu lato (s.l.), An. funestus, other Anopheles, and non-Anopheles [22], and characterized by gonotrophic stata (empty, blood-fed, gravid and half gravid female mosquitoes).

Ethical approval was obtained from the Institutional Review Boards (IRB) of the Kenya Medical Research Institute and the University of California at Irvine, USA. Written consent for adults and assent for minors were obtained from all participants after explaining the objectives and the methodology of the study to them.

Data analysis was restricted to a comparison of cases and controls. Univariate analysis of the risk factors was conducted using logistic regression to estimate odds ratios (OR), adjusted to account for within-household clustering of cases. In multivariate analysis, conditional multiple-logistic regression was employed, here matching variables including matched group (one case and its two matched controls) and case versus control group (a covariate with case and control as values) were used as effect modifier. Student t-test was used to determine the differences in vector densities between case and control groups. Data analysis was conducted using SPSS.

In this study, 48% of the surveyed malaria cases were males while 52% were females. The proportion of the controls matched to these cases was 46.7% and 53.3% males and females, respectively. Majority of the cases (60.9%) included in this study were aged below 5 years old. They were matched to 62.6% of the controls aged below 5 years old (Table 1).

In terms of education and occupation status, families where the mother in the house was employed (logistic analysis, OR = 0.43; P = 0.04) and educated up to secondary school level (OR = 0.27; P < 0.0001) had lower chances of contracting malaria. However, the occupation status of father of the house (P = 0.54) and their corresponding level of education (P = 0.76) were not significant risk factors of malaria (Table 1). Both the cases (61.3%) and controls (61.4%) have similar household size with more than five family members living in the same house. The household size was not identified as a potential risk factor for malaria (OR = 0.99; P > 0.05; Table 1).

Most families of malaria cases and controls in the study sites lived in iron-roofed houses (89.7% and 94.7%, respectively) while the rest lived in grass-thatched houses. Compared to the grass-thatched roof, the iron roof was associated with a decrease in the risk of malaria (OR = 0.46; 95% CI 0.29–0.82; P < 0.005; Table 2). The majority of malaria cases (93.0%) and controls (84.9%) lived in houses constructed with mud walls, while the rest living in permanent houses made of cement blocks or bricks. Living in a permanent house made of cement blocks or bricks was significantly associated with lower malaria infections (OR = 0.42; 95% CI 0.26–0.69; P < 0.001; Table 2). In addition, most families of malaria cases and controls in the study site lived in houses with earthen floors (95% and 86.9%, respectively) while the rest had cemented floors. Families who lived in houses with cemented floor were well associated with a decrease in the risk of malaria (OR = 0.35; 95% CI 0.20–0.61; P < 0.001; Table 2). Living in a house with closed eaves was significantly associated with lower risk of getting malaria compared to living in houses with open eaves (OR = 0.46; P < 0.0001; Table 2). The size of houses and number of rooms per house were not malaria risk factors (OR = 1.07; 95% CI 0.68–1.70 and P = 0.76; Table 2).

There was no significant difference in vector density in the houses of malaria case groups (0.38 ± 0.058 vectors/house) compared to the control group (0.34 ± 0.078 vectors/house; t = 1.96; df = 895; P = 0.71; Fig. 2). In both study groups, a higher density of An. gambiae s.l. (0.36 ± 0.057 vectors/house) than An. funestus (0.02 ± 0.009 vectors/house) was observed.

Insecticide-treated bed nets usage was higher in control group (64.6%) as compared to case group (48.3%), understandably, malaria risk was lower in families using ITNs compare to those who did not use ITNs (OR = 0.51; 95% CI 0.39–0.68; P < 0.0001). The proportion of families that use alternative mosquito prevention measures such as aerosol sprays, herbs, and mosquito coils was higher among the control groups (63.1%) compared to those of the cases group (44.0%). Those who used other mosquito prevention measures had lower malaria risk compared to those who did not use the prevention measures (OR = 0.46; 95% CI 0.35–0.61; P < 0.0001).

There were more individuals in the control group (45.7%) used prophylactic anti-malarials compared to the cases group (22.2%). The use of prophylaxis was significantly associated with lower malaria risks (OR = 0.34; 95% CI 0.25–0.46 P < 0.0001; Table 3). Most families of both malaria cases (98.7%) and controls (92.7%) sought diagnosis and treatment from government health facilities. However, the percentage of families that resorted to private health facilities was lower among malaria cases (1.3%) compared to control groups (7.3%; Table 3). The health facility from which therapy was sought was associated with lower risk of malaria infections (OR = 0.17; 95% CI 0.06–0.48; P < 0.001; Table 3).

The risk of malaria was significantly reduced for families that use malaria prophylaxis and those where the mother of the houses occupation was farming. The risk of malaria was high in families where the mother was not educated or whose income was less than 2000 Kenya shillings (USD 25 at the time of data collection) per month (Table 4).