Malaria is an important cause of death for children under 5 years of age in sub-Saharan Africa. Although advances in prevention, diagnostic tools and treatment have led to a decrease in deaths, malaria is still responsible for more than one third of avertable deaths in children in endemic areas in Kenya. In this study we investigated the socioeconomic determinants of malaria mortality in the region. Our results highlight important determinants which need to be considered for preventing malaria deaths in children.
The mortality rate in the HDSS area of Siaya County has been steadily decreasing since 2003 yet a sudden increase is seen in 2008. This has been observed in previous studies from the KEMRI/CDC HDSS and can likely be linked to the widespread violence between tribes that followed the 2008 national elections [11, 13, 15, 17]. Many of Kenya's inhabitants were internally displaced and there were frequent stock-outs and disruptions in delivery of ACTs, RDTs and other medical supplies within the public healthcare system.
Decrease in healthcare accessibility and implementation of malaria control measures have also been seen elsewhere. A study performed in northern Uganda where the population and local government had endured political instability for more than a decade experienced similar results [18]. Similarly, war and civil unrest in other parts of Africa have led to the resurgence and periodic epidemics of sleeping sickness [19]. In Kenya, post-2008 data shows a further decrease in malaria mortality rates giving further merit to the theory linking post-election violence to the sudden increase in malaria deaths during 2008 and the following years.
Similar increases in malaria incidence and mortality have also been seen after our study period related to the covid-19 pandemic where health care services were disrupted once again [20]. Outpatient visits decreased significantly as well as malaria testing due to more frequent RDT stock-outs as well as a decreased willingness to seek health care. As a result, confirmed malaria cases were also decreased during the initial phases of the pandemic [20, 21]. A recent study from Zimbabwe shows increased malaria mortality rates during the covid-19 pandemic [22] as well as WHO estimates of up to 47,000 additional deaths worldwide from malaria in 2020 directly linked to the covid-19 pandemic [1].
The highest mortality rates were found among the youngest children which is consistent with what is known about malaria in children. The first 3 months of life, a child has the benefit of immunity acquired from the mother via birth and breastfeeding. Children aged 3–12 months are more vulnerable and therefore the progression of malaria symptoms can be more rapid and severe than for both younger and older children. According to WHO, a potential issue is also that most antimalarial drugs lack correct prepackaged dosages for infants which could lead to inaccurate doses being administered, affecting treatment success [2].
In our study, there was a slight but non-significant sex difference in malaria mortality, males had a 1.04 (95% CI 0.96–1.12) hazard ratio compared to females. This can be compared to the under-5 mortality rate for all causes in Kenya 2013 where a significant sex imbalance was observed (75 per 1000 person years for males and 66 per 1000 person years for females) [23].
There might be several reasons for the sex differences in mortality, the Kenya Malaria Indicator Survey from 2015 showed that overall, females under 5 years of age are slightly more likely to sleep under a bed net, with 59.1% of females sleeping under any type of bed net, compared to 56.7% for males [24]. A later survey from 2020 showed a more even distribution according to sex, yet even lower rates at 49.0% for males and 50.2%% for females [8]. The survey from 2015 also showed that females under 5 were slightly more likely brought to medical attention (72.2% females, 71.6% males) and to be tested via a finger or heel blood test for malaria when febrile (42.0% females, 36.7% males), and were more likely to receive ACT treatment on the same or the next day (65.1% females, 55.6% males) when febrile [24]. Another important reason for the difference in mortality rates for males and females could be found in the fact that males have a higher parasite prevalence rate, spontaneous rapid diagnostic testing programs have showed that between the age of 3 months to 14 years, 15.3% of males spontaneously tested positive for malaria, compared to 14.0% for females [24].
Thus, the reason for the slightly higher mortality rate among males is likely multifaceted, parasite prevalence rates are slightly higher and healthcare-seeking behavior tends to favor females slightly as shown in both our study and previous studies.
In the present study, there was a significant lower malaria mortality rate among children whose mothers had a completed secondary school or higher education. The Kenya Malaria Indicator Survey from 2015 concludes that the education level completed by the mother has a large impact on a variety of factors. The higher the general education, the higher the likelihood of sleeping under an ITN, seeking medical attention for a child’s fever as well as the likelihood of receiving the recommended ACT malaria treatment [24].
The United Nations 4th Millennium Development Goal targets the area of education specifically and states that any level of maternal education is beneficial in reducing child mortality, something that Kenya has achieved quite well when considering that merely 1.6% of the mothers included in this study had no completed education whatsoever. The data also shows the benefit of further education with significantly reduced mortality rates among the group with a secondary school or higher education.
Similarly, higher education level is also linked to higher socioeconomic status and lower frequencies of HIV infection compared to those with primary school [25]. It is not entirely clear how large of an impact HIV has on malaria severity as research shows mixed results [26]. Mothers with higher levels of education have also been shown to have higher levels of self-confidence and skill in gathering information as well as increased autonomy within the family in an otherwise male-dominated culture, which can all contribute to increased tendencies to seek medical care at a higher rate and before symptoms become too severe [24, 27].
Our results show a clear trend where increasing level of socioeconomic status gradually decreases the child mortality rate. The Kenya National Malaria Strategy published in 2009 showed differences in access to treatment according to socioeconomic status, with 17.3% of children in the lowest wealth quintile and 29.1% in the upper wealth quintile taking an antimalarial drug to treat febrility [7]. However, a follow-up in 2015 showed mixed results for both general antimalarial drug treatment and specific ACT treatment according to wealth quintile, showing lower rates for those in the highest wealth quintile and peaks in the next-lowest and middle quintiles [24]. Another study showed that 40% of children under 5 with a fever in 4 rural Kenya regions could not afford malaria treatment [28].
The Kenya National Malaria Strategy also includes data regarding use of bed nets nationwide and shows significant differences in usage according to socioeconomic status where 39.2% of children under 5 years of age in lower wealth quintile families sleep under a bed net. Those in the higher wealth quintile bracket are covered at 61.6%. The figures are 29.1% and 44.5% respectively for ITNs and LLINs [7].
In Kenya, national policy states that any child under 5 years of age shall receive free health care, yet many Kenyans are unaware of this policy and therefore do not seek medical attention for their children [29]. Another barrier to seeking healthcare is not being able to afford transport to a healthcare facility [29]. It is common to borrow money from relatives or neighbors, but there is also a reluctance to do so [17]. As previously mentioned, socioeconomic status is also closely linked to education level and HIV infections which may also play a role in the mortality rate differences.
When looking at the mothers’ age, mothers under the age of 20 tend to have fewer children dying of malaria compared to mothers aged 20–30 and > 30, however the differences are small. Several studies show the opposite, that mortality rates among first-born children to mothers 18–21 years of age is high. It is also shown that older mothers have lower child mortality rates [30, 31]. The results generated here are however difficult to generalize due to the very small differences.
The distribution of ITNs and LLINs in a controlled and planned manner is proven to be effective in reducing morbidity and mortality in malaria [1, 32, 33] and the Kenya National Malaria Strategy has set a target goal of at least 80% of the population in endemic areas sleeping under an ITN or LLIN every night [7]. A 2010 study performed during a peak transmission period from the Kisii district (150 km from Siaya district) showed that 95% of 670 surveyed households owned at least 1 bed net and owned on average 2.4 bed nets per household. However, only 59.2% of children under 5 years of age slept under a net. Additionally, the study showed that 40% of nets had one or multiple holes [34]. Compared to national data, which states that each household owns on average 0.8 bed nets, large differences appear between regions, however this figure also includes the rather large proportion of Kenyans not living in malaria-endemic areas. More recent data from the Kenya Malaria Indicator Survey finds that 75.3% of children under 5 in the lake endemic region slept under any type of bed net the night before surveying in 2015 [7, 24] and even lower in the follow-up survey of 2020 where only 60.3% of children under 5 slept under a bed net [25].
Our results show only a slight benefit of ITN ownership, yet due to the large number of missing data any conclusions are difficult to draw. According to another study, ITN coverage is higher than found in the HDSS data as the study area has been given more attention and has participated in several large studies regarding ITNs over the past decade, thus they have received free distribution of nets both from the government and as part of epidemiological studies at steadily increasing rates since 2003 [35]. The HDSS data records whether children under 5 years of age in the household slept under an ITN the night preceding the visit by the community interviewer, making any conclusions about how regularly the children sleep under an ITN hard to draw.
The distance from the household to a healthcare facility clearly impacts the mortality rates. The results from our study show a statistically significant advantage of living close to a healthcare facility with increasing hazard ratios of living within 1–2 km and 2 + km respectively as compared to 0–1 km distance. Inhabitants of rural Kenya often lack their own methods of transportation and ambulance services are sparse or nonexistent. Additionally, households are often distant from the major road connections. Often, they must walk or rely on motorbike taxis or friends and relatives for transportation in emergency situations. A study performed in four Kenyan counties looked at the proportion of inhabitants who travelled to a health clinic run by the government and recorded the distance traveled. The study shows that the closer a household lives to a clinic, the more likely they are to travel there for care and that 56% of the patients will travel to their nearest healthcare facility [36].
An interview-based study performed in 3 African countries, including Western Kenya, identified distance to healthcare facilities as a major issue in seeking treatment. Participants said that they most often had to travel by foot, carrying their sick child, as other methods of transport were too expensive. This often leads to delays in seeking treatment and the child worsening further during transport [17]. Another study performed in the KEMRI/CDC HDSS investigating distance of residence and clinic attendance found similarly that increasing distance to health care facilities acts negatively on the rate of clinic visits [36].
The Kenya Malaria Communication Strategy outlines the goals for improvement in educating the Kenyan population regarding malaria risks and symptoms as well as improving healthcare-seeking behavior and providing information and assistance regarding preventive measures. It identifies the treatment-seeking behavior as an obstacle, as many patients self-treat at home before seeking care at a healthcare facility, delaying treatment and in some cases receiving incorrect treatment when visiting a traditional healer or dispensary rather than the recommended government or faith-based clinics [37]. The average time before reaching health care was over 2 days after onset of symptoms in a study from 2009 which also shows deficits in public perception regarding correct treatment and treatment-seeking behavior [28]. The Kenya malaria indicator survey from 2015 found that in the endemic lake region, 65% of febrile children under the age of 5 sought medical attention, with 59% leaving a blood test, 55% receiving antimalarial treatment and 52% treated with an ACT [24].
The Kenya National Malaria Strategy summarizes that general knowledge about malaria is at a 95% level in Kenya, however only 10% of the population know about the increased risk of anemia and its associated symptoms and complications and the dangers of malaria infection to infants and pregnant women [7].
There were some limitations in the present study. The ITN data was available from 2007, therefore all data from 2003–2007 are listed as unknown when it comes to ITN status, there is a total of 68.5% of individuals listed with an unknown ITN status during the entire study period and there is additional data post-2007 with unknown ITN status. There was missing data in the categories of household size, socioeconomic status, maternal education, and maternal age. Data collection for sex, GPS location and age were however complete.
Even though the proportion of missing data was relatively high for some variables, we found significantly lower mortality rates related to maternal age, socioeconomic status, and household size.
There is a potential risk of recall bias as the verbal autopsy system has certain flaws compared to the golden standard physician-performed autopsies. A large study from the Nairobi, Kenya HDSS showed that the verbal autopsy results only corresponded to the physician-reported cause of death in 31.6% of cases involving children under 5, however the main discrepancies stem from indeterminate cases of tuberculosis, diarrheal disease and HIV/AIDS related deaths, the results are much more accurate compared to physician-performed autopsies for among others malaria, meningitis, and pneumonia. The study concludes that the InterVA-4 system for analyzing verbal autopsy results is a method effective for analysis of cause of death on a large public health scale, especially considering that there is no other available and effective measurement of cause of death on a community level [16, 21, 38, 39].
Progress has been made mainly through increased ITN coverage, intermittent preventive treatment in pregnancy, diagnostics and improved pharmaceutical treatments for manifest malaria. These steps taken to reduce malaria mortality are proving effective and have begun to reverse the incidence and mortality rate of malaria, however reaching the entire population of malaria-endemic areas with these measures is problematic. Many still do not have knowledge regarding symptoms and recommended preventive interventions as well access to correct diagnostic possibilities and treatment regimens. The HDSS study area houses several healthcare clinics due to the large amount of research that has been conducted here over the past decades and the region has enjoyed increased international attention and aid. Many other rural Kenyan regions do not have the possibilities nor the infrastructure for similar healthcare development and could very well be worse off than the population studied in the KEMRI/CDC HDSS.