Malaria has been the major public health concern in many tropical and subtropical countries but the interventions have greatly reduced both morbidity and mortality cases [1]. Despite of the observed disease burden reduction, sub-Saharan still bears 90% of all malaria cases and 92% of mortalities [1]. Asymptomatic infections in Africa have been halved and clinical incidence of the disease reduced by 40% between 2000 and 2015 [2]. Similarly, malaria outpatient consultations in Kenya have dropped from 25 to 35% to 18% and from 20 to 6% of all hospital admissions [3, 4]. The increased access, ownership and use of the long-lasting insecticidal nets (LLINs) have greatly contributed to the decrease of morbidity [2]. The use of indoor residual insecticide spray (IRS) in combination with LLINs resulted to the observed reduction of disease burden [5, 6]. The LLINs ownership and use in Kenya have been consistently increasing since year 2000 [7, 8]. However, the massive application of IRS in western Kenya started in 2005 and by 2010 only 38% of households in epidemic prone areas were covered and even less in the recent years [4, 8]. Prompt diagnosis and management of malaria using efficacious anti-malarial drugs of choice remains one of the three important interventions [9–11]. The adoption of artemisinin-based combination therapy (ACT) recommended use of artemether 20 mg–lumefantrine 120 mg as first-line treatment for uncomplicated malaria since 2004 [12]. About 92.8% of children with fever in 2015 were given ACT and only 1.4% used sulfadoxine–pyrimethamine (SP) [7].

Despite of the interventions described above, some areas in western Kenya successfully controlled the disease while others experienced changing dynamics [13]. Similar observation of sustained high transmission despite of the available interventions has been also observed in other parts of sub-Saharan Africa in recent years [14, 15]. However, reasons contributing to this observation remain not clearly known in the midlist of the reported increasing vectors insecticide resistance and shift of vector populations in western Kenya [16, 17]. The known factors attributing to malaria resurgence and changing transmission dynamics [18] were used to explore the reasons contributing to the contrasting outcome of the available malaria interventions in western Kenya.

Sub-Saharan Africa still carry the highest malaria disease burden with 92% of the global mortalities [1]. However, there is a significant reduction of death by 40% between year 2000 and 2015 [2]. Similar success has been also observed in Kenya where the asymptomatic infections have been halved and significant decrease of clinical outpatient consultations as well as admissions [3, 4, 7]. The global malaria action plan (GMAP) of 2008–2015 targeted to reduce malaria cases by 75% but the infection in some areas in western Kenya was still increasing (Fig. 2). Existence of disease resurgence despite of the intensified interventions in the recent years has not only being seen in western Kenya but also in other countries of sub-Saharan Africa but with inadequate explanation on the cause [13, 15, 27].

This study found a significant reduction of indoor resting malaria vectors since the start of interventions in some sites (Fig. 4). Marani experienced an increase of indoor resting vectors from 2014 to 2016 which was also along with an escalation of asymptomatic parasitemia among primary school pupils (Figs. 2, 4). In conjunction with increased parasitemia at Marani, An. funestus s.l. became the main vector as they constituted three quarters of the total indoor female vectors and the rest being Anopheles arabiensis. Before interventions An. gambiae s.s. covered about 80% of all indoor resting vectors thereafter An. funestus s.l. took over. Elsewhere in East Africa studies have reported an increasing importance of An. funestus s.l. in malaria transmissions with an increase of not only abundance but also sporozoites rates [28, 29]. A study in western Kenya showed consistent findings as An. funestus s.l. was also found to have a raising abundance with highest sporozoite rate as compared to other vectors [30]. The LLINs ownership and use scale-up also changed the composition of malaria vectors at Kombewa, the population of An. gambiae s.s. reduced by 22% while that of An. funestus only decreased by 6%. Iguhu has the highest composition of An. gambiae s.s. with the least populations of An. funestus s.l. where there was a sustained low transmission and controlled vector population (Figs. 2, 4). One factor that clearly differentiates these areas is that sites with non-improving outcome of interventions (Marani and Kombewa) have An. funestus s.l. as the major vector (Fig. 5). The vector have shown to be the main malaria transmitting agent as a result of an increased sporozoites rates as well as abundance while exhibiting highest insecticide resistance to the widely used pyrethroids [30, 31]. Studies on insecticide susceptibility of this vector in East and South Africa found very high resistance to both deltamethrin and permethrin [29–32]. Populations of An. funestus s.l. from Kisii showed similar susceptibility upon deltamethrin exposure. Other studies in western Kenya also found as low as 10% mortality upon exposure to deltamethrin [30, 31]. The sustained control of malaria infection at Iguhu could be due to presence of insecticide susceptible An. gambiae s.s. as the major malaria vector (Figs. 5, 6). Whereas as Kombewa had a composition of all insecticide resistant vectors (An. gambiae s.s. as well as An. funestus s.l.) which could have limited the benefits of LLINs and IRS. Generally, study areas with high composition of insecticide resistant vectors experienced infection resurgence or sustained high transmission which is contrary to the global technical strategy for malaria 2016–2030 [1]. Moreover, having higher composition of highly anthropophilic An. funestus s.l. with such high resistance levels amid of changed biting behaviour increase chances of more malaria transmission [17, 30, 31, 33]. Nevertheless, the change in biting time from midnight to earlier or late has been reported from all the three study sites and, therefore, this could contribute to even more infection transmission potentials to areas with increasing or sustained high indoor vector densities [17, 33, 34].

Precipitations and air temperature among others factors significantly affect the breeding and population growth of malaria vectors [35, 36]. In East Africa highlands for example, climatic warming has been associated with malaria epidemics [37, 38]. The increasing monthly mean minimum and maximum ambient temperature from 2012 to 2015 was seen to all study sites. In conjunction with this, highest peak of rainfall at Marani was noted in September 2014. The combination of increased rainfall and air temperature increase at Marani gives the possible explanation of malaria resurgence in this area (Figs. 2, 4, 12). The increase of mean minimum temperature plays a major role on mosquito breeding in cool highland areas such as that of Kisii (mean annual temperature of 21.13 °C in 2016) than lowland warm areas [37]. The raise of vector populations at Marani was preceded by an increase of the mean minimum air temperature by 2.5 °C and rainfall (Figs. 2, 12). The climatic warming at Marani has resulted to an increase of the mean minimum temperature to 16.16 °C which might shorten the larvae stage and also gametocyte cycles in adult mosquitoes [38]. Along with this, increasing land use as a results of population growth have also contributed the increased suitable breeding sites and survivorship of An. funestus populations as it has been reported elsewhere [39].

The efficacy of ACT in western Kenya remains high despite of the reported increase of polymorphisms of specific key codons [40–42]. Availability of SP in drug dispensing outlets remains high but the use of this drug remains low (Table 2 and Fig. 9). Areas experiencing infection resurgence had lower use of SP and therefore drug resistance looks unlikely to explain the incident. The continued use of SP for presumptive treatment and self-prescriptions is consistent to the sustained SP specific codons polymorphisms while those of chloroquine diminishes [41, 43, 44]. Elsewhere in Africa, the sustained use of already resistant anti-malarials was associated with persistent malaria high transmissions [45]. Moreover, poverty has been associated with malaria morbidity and mortality for long but these three communities in western Kenya have similar social economic status and economic inequalities [46, 47].

The school-based surveillance of asymptomatic malaria demonstrates to be a better metric for monitoring transmission intensity and intervention effect size than the hospital based (Figs. 2, 3). For example at Kombewa, the number of positive cases in 2011 among primary school aged children was the same as in 2015 but the hospital survey shows higher number of cases in 2011 than 2015 (Figs. 2, 3). This could be due the fact that the later surveillance system may be affected by number of factors like case management rate, reporting rate and case confirmation. In western Kenya, malaria diagnosis and treatment used to be based on blood slide as well as clinical judgment, therefore some cases in 2011 could be clinically diagnosed and reported as confirmed [48]. However, malaria case detection following the introduction of rapid diagnostic tests and blood slide microscopy training over the recent years has greatly improved [49]. Moreover, the hospital-based surveillance system is often interrupted by frequent strike of physicians and nurses. Asymptomatic malaria prevalence data among school age children was collected systematically and thus more reliable than hospital-based malaria case data (Figs. 2, 3). This study also found of an increased LLINs ownership to over 80% in all areas, studies however shows lowest use among the 5–14 age group (school age) [25]. The unpublished data from the study sites shows 72 and 58% of this age group slept under LLINs a night before survey at Marani and Kombewa, respectively. Whereas at Iguhu (areas showing sustained low transmission), only 50% of the 5–14 age group slept under LLINs. The use of LLINs among the school age was highest at Marani (malaria resurgence site) and lowest at Iguhu (an area with controlled transmission). One would expect to see highest use of LLINs in an area that has attained sustained transmission control but the opposite is true. This means that other factors like increase in vector population and insecticide resistance could be the likely major drive of infection transmissions in these populations. The over 80% LLINs coverage could have provided community wide protective effect [50] to all study sites despite of low use among the school age but the explaining reasons for the observed variation in response to interventions are likely to be type of the vector, population density and insecticide susceptibility (Table 1). In western Kenya, suitability of asymptomatic malaria surveillance in schools has been evaluated and found to be representative of the general population [51]. Therefore, the described trend of malaria transmission which also correlates with the indoor vectors populations represents the true infection transmission dynamics in the study area (Figs. 2, 4). This study however lacks information of the long-term malaria case management rates, site specific An. funestus s.l. insecticide susceptibility and site specific vector behavior. This information would have improved the analysis on the cause of the observed changing dynamics of malaria infection in western Kenya.

The sustained highest composition of the highly anthropophilic An. funestus s.l. and also availability of pyrethroids resistant An. gambiae s.l. could be the cause of the sustained high malaria transmission at Kombewa. The increase of the mean minimum air temperature and precipitation have led to an increased abundance of insecticide resistant An. funestus s.l. population at Marani which may have subsequently caused the observed infection resurgence. At Iguhu where there was a sustained control of infection, the pyrethroids moderately resistant An. gambiae s.s. had the highest composition. Climate variability, insecticide resistance and vector population shift are likely the cause of the contrasting outcome of malaria interventions n western Kenya. To meet the GMAP 2030 targets there is a call for use of carbamates and organophosphates for indoor targeted interventions and expansion the integrated vector management [16, 31]. Along with this, the continued surveillance of vectors and clinical and subclinical infection is highly recommended for changing infection transmission risks. Monitoring of insecticides resistance should be done along with use of air temperature and precipitation data to predict vector and parasite dynamics especially in highlands where slight changes in these parameters could lead to devastating infection outbreaks. Malaria transmission competence and biting behaviour of re-emerging Anopheles funestus complex should be also studied.