Current malaria elimination guidelines are based on the concept that malaria transmission becomes heterogeneous in the later phases of malaria elimination ^[1]. In the pre-elimination and elimination phases, interventions have to be targeted to entire villages or towns with higher malaria incidence until only individual episodes of malaria remain and become the centre of attention ^[1]. With increasing evidence of clustering of malaria episodes within villages, we argue that there is an intermediate step. Heterogeneity in malaria transmission within villages is present long before areas enter the pre-elimination phase, and identifying and targeting hotspots of malaria transmission should form the cornerstone of both successful malaria control and malaria elimination.

Variation in the risk of malaria between villages in endemic regions has long been recognized ^[2]–^[4]. This variation is common for many infectious and parasitic diseases where a small number of human hosts are most frequently or most heavily infected while the majority of a local population carry few or no infections ^[5]–^[8]. In malaria, this is exemplified by a study in Dielmo, Senegal, where children were monitored daily during their first 2 years of life. Some children suffered only one episode of clinical malaria, whilst others suffered up to 20 episodes ^[9]. In Kenya, researchers noted that malaria exposure could not be homogenous as malaria incidence did not follow a Poisson distribution, a phenomenon they describe as over-dispersion ^[10]. Over-dispersion is commonly recognized in other infectious diseases, where a small proportion (20%) of the population is responsible for the majority (80%) of transmission, the so-called “20/80 rule” ^[8],^[11]–^[13].

Micro-epidemiological variations in malaria exposure are most easily recognized in areas of low or moderate transmission intensity where a considerable proportion of the population may remain malaria free for several years while others experience multiple episodes ^[8],^[11],^[14]. In areas exposed to intense malaria transmission, heterogeneity in exposure is also present ^[15],^[16], but may be obscured because the majority of the population experiences at least one infection per year and many infections are carried asymptomatically. At present, the factors underlying the micro-epidemiology of malaria are not fully understood but include variation in distance to the nearest mosquito breeding site, water body or vegetation ^[14]–^[17], household structural features ^[14]–^[17], and both human behavioural ^[15],^[17] and genetic factors ^[15],^[17] that may also result in differential attractiveness to mosquitoes ^[18]. These factors differ at global and local geographical scales and lead to different and confusing definitions of foci of malaria transmission and hotspots of malaria transmission. Entire countries or islands have been classified as malaria hotspots ^[19],^[20], or the term hotspots of malaria transmission may be reserved for smaller geographical areas ^[14],^[21]–^[23], sometimes smaller than 1 km^2[22],^[23].

Two related but distinct geographical units in malaria transmission can be defined: (1) The World Health Organization defines a focus of malaria transmission as a defined and circumscribed locality situated in a currently or former malarious area containing the continuous or intermittent epidemiological factors necessary for malaria transmission. Foci of malaria transmission can be classified as residual active, residual nonactive, cleared up, new potential, new active, endemic, or pseudofoci ^[1]. In more academic terms, an active focus of malaria transmission is a geographical area that supports malaria transmission, where the local Anopheles population sustains the basic reproductive rate (R₀; average number of secondary infections arising in a susceptible population as a result of a single individual with malaria over the course of their malaria infection) at a level above 1 ^[16]. Its size depends on the mosquito breeding site that forms the centre of the focus and the effective dispersal range of vector mosquitoes, which is several kilometres. The border is the furthest location where malaria is still supported by the breeding site. (2) A hotspot of malaria transmission is defined as a geographical part of a focus of malaria transmission where transmission intensity exceeds the average level. Several hotspots of malaria transmission may be present in a single focus of malaria transmission. Micro-epidemiological conditions for malaria transmission are favourable in a hotspot of malaria transmission, resulting in R₀ estimates that exceed the average for the focus of malaria transmission. The size of a hotspot of malaria transmission is variable but typically <1 km² and smaller than the maximum dispersal range of vector mosquitoes; its borders are defined by the distance from the centre of the hotspot where transmission intensity is no longer (statistically significantly) higher than the average for the focus of malaria transmission ^[14],^[21].

Heterogeneity in mosquito exposure is key to understanding the differences between foci and hotspots of malaria transmission and their implications for malaria control. Individuals who are bitten most often are most likely to be infected and can amplify transmission by transmitting the malaria parasites to a large number of mosquitoes. Estimates of R₀ are very susceptible to variations in mosquito biting behaviour. R₀ may increase considerably as a consequence of heterogeneity in this behaviour ^[8],^[12]; the susceptibility of R₀ to heterogeneous bitingthis is illustrated by Table 1 where estimates of R₀ increased 1.5- to 4.5-fold as a consequence of introducing heterogeneous biting into a mathematical model of malaria ^[24] in four villages exposed to moderate transmission intensity in northern Tanzania ^[14].

There are two reasons why hotspots are relevant for malaria control ^[8],^[12],^[25]. Firstly, if interventions are untargeted, hotspots are likely to be the areas where residual malaria transmission will persist. This hypothesis is supported by observations that hotspots of malaria transmission remained unaltered after overall malaria transmission is reduced ^[22],^[23],^[26]. Hotspots of malaria transmission can thereby form a major stumbling block in efforts to eliminate malaria ^[25].

Secondly, hotspots of malaria transmission are likely to play a catalysing role in areas of stable transmission. Figure 1 shows a schematic of the hotspot theory whereby a few households maintain higher transmission at all time periods. In the dry season the hotspot supports continuing transmission; in the wet season it acts as a source of infection for the rest of the village. This exemplifies the difference between hotspots and foci of malaria transmission: hotspots fuel transmission within transmission foci, whereas foci form independent malarious areas that may contain hotspots. Only the emigration of human parasite carriers or transportation of infectious mosquitoes can result in a spread of parasites beyond the borders of a focus. Interventions targeted at transmission hotspots, but not foci of malaria transmission, therefore have the potential to reduce community-wide malaria transmission. Using the same mathematical model ^[24] and the same dataset from northern Tanzania ^[14], we show that targeting hotspots with long-lasting insecticide-treated nets (LLINs) and indoor residual spraying (IRS) could lead to malaria elimination while untargeted interventions with the same resources would lead to more modest reductions in malaria parasite prevalence (Figure 2A). In areas of higher endemicity, targeted interventions alone are unlikely to result in malaria elimination. Nevertheless, also in these settings, targeted interventions have a markedly larger impact compared to untargeted interventions with the same resources (Figure 2B).

Having argued that hotspots should be targeted, the next obvious question is how can they be identified? Spatial patterns in malaria transmission have been described using (combinations of) micro-epidemiological elevations in malaria incidence ^[11],^[14],^[21],^[22],^[27], asymptomatic parasite carriage ^[21],^[22], reported fever ^[28], drug use ^[28], serological responses to malaria-specific antigens ^[14],^[29],^[30], mosquito abundance ^[14],^[30], and exposure to infected mosquitoes ^[14],^[30]. Environmental models are very valuable in defining (larger) foci of malaria transmission ^[31], but currently have limited resolution in identifying small-scale hotspots of malaria transmission within foci of malaria transmission ^[14],^[21].

The most direct evidence of hotspots of malaria transmission is gained by finding an increased exposure to infectious mosquito bites. However, this gold standard measure for defining transmission intensity is extremely laborious and has low sensitivity at low transmission intensity. Furthermore, mosquito sampling strategies for outdoor biting and resting mosquitoes are poorly standardized despite their increasing relevance for transmission ^[32]. These limitations render an entomological detection of hotspots logistically unattractive.

Clustering of asexual parasite carriage and malaria-specific immune responses currently appear to be the most robust indicators of hotspots of malaria transmission. Incidence of clinical malaria episodes is frequently used as an indicator for increased malaria exposure. However, it should not be used for detecting hotspots unless in an age group defined by low immunity, such as infants or young children, because the higher malaria exposure in hotspots leads to a faster acquisition of immunity against clinical malaria and high density parasitaemia ^[33]. The likelihood of developing symptoms upon infection may, therefore, be lower in hotspots of malaria transmission. Clustering of asexual parasite carriage forms a more stable indicator of hotspots of malaria transmission than clinical malaria episodes ^[21], since immunity that prevents malaria infection is acquired later in life, if at all. Antibody responses to malaria-specific antigens can also be used to define small-scale variations in malaria exposure ^[14],^[29],^[34]. Because antibody responses are acquired with cumulative exposure and are relatively long-lived, serological markers of malaria exposure are most suitable for detecting stable hotspots of malaria transmission ^[14] in areas of lower endemicity and can be derived from simple health facility-based surveys ^[14],^[35]. Antibody responses can be analysed as age-dependent sero-conversion rate ^[14],^[36], individual antibody prevalence, or (age-adjusted) individual antibody density ^[21],^[29],^[36]. The most suitable approach will depend on the study setting, notably the average level of transmission intensity, and the resolution at which hotspots can (or need to) be detected.

Compared to settings of moderate to low transmission intensity, little research has been done on operational ways to detect hotspots in areas of more intense transmission intensity. Spatial heterogeneity in malaria exposure is common in high endemic settings ^[15],^[26],^[36],^[37]. Hotspots of malaria transmission, as defined in this manuscript, have been identified by geographical clusters of parasite carriage ^[26],^[36]–^[38] and malaria incidence ^[15]. Serological markers of malaria exposure have been used in high endemic settings ^[36], but their value for detecting hotspots of malaria transmission against a background of intense transmission intensity remains to be confirmed.

Three important arguments on hotspots need to be addressed. Firstly, are hotspots stable over time? This is important for practical reasons. Some consistency in the geographical location of hotspots would make implementation of control methods much easier. The predominant observation is that hotspots are remarkably stable even when the intensity of transmission declines ^[14],^[21]–^[23],^[26],^[39],^[40]. However, clusters of higher clinical incidence may vary with time ^[21],^[39], especially in settings where outbreaks are related to movement patterns of infected human parasite carriers ^[22]. In coastal Kenya, evidence was found for the presence of stable and unstable hotspots within the same study population ^[21].

Secondly, do hotspots seed transmission to the rest of the focus of malaria transmission? The theory behind hotspots fuelling transmission (Figure 1) is supported by several entomological studies that show very focal mosquito exposure in the dry season and more wide-spread mosquito exposure in the wet season while the same households experience the highest relative mosquito ^[14],^[41]–^[44] and the highest parasite prevalence in the different seasons ^[14]. In some areas of low transmission intensity, these persisting hotspots form the only likely source of parasites for seasonal or epidemic increases in malaria transmission in the wider community ^[22],^[40]. Against this are observations that suggest that movement of some vector species is highly localized ^[45], thereby limiting the spread of malaria from a hotspot to the rest of the village. A study in Tanzania where mosquitoes were captured, marked with fluorescent powder, released, and recaptured observed that 68% of mosquitoes returned to the same household from where they were initially captured ^[46]. In Papua New Guinea mosquitoes appeared to have a “memorized” home range and limited dispersal range in the focus of malaria transmission they are accustomed to ^[47]. This nonrandom mixing could have important epidemiological consequences for strategies to control hotspots and would lead to overestimations of impact of hotspot-targeted interventions. In the extreme scenario where mosquito populations do not mix, there would be no community benefit of hotspot targeted interventions. This issue should be addressed in formal evaluations of the community effects of targeted interventions on malaria transmission.

Thirdly, at what geographical resolution can hotspots be detected? The scale at which hotspots are present will greatly influence the feasibility of their identification. Hotspots that are present as geographically clustered groups of households can be more readily identified than smaller hotspots such as individual households. Hotspots may be also more complicated to detect in high endemic settings where the prevalence of malaria parasites and malaria-specific antibodies are high. In these settings, alternative approaches may be needed to determine small-scale variations in transmission intensity. These may include contact tracing of individuals with clinical malaria in the youngest, least immune age groups through health surveillance data or school surveys. More intensive surveillance systems may be capable of combining parasite prevalence and antibody prevalence or sero-conversion rates in young age groups, or examine the number of parasite clones acquired over a certain time period, i.e., the molecular force of infection, once tools are optimised ^[48].

Spatially targeted interventions will not replace the current practice where LLINs and intermittent preventive treatment (IPT) are preferentially provided to young children and pregnant women, groups that are at the highest risk of severe disease. Rather, it will supplement this approach that aims to reduce severe morbidity and mortality with an approach that specifically aims to reduce malaria transmission. Following scaling up in moderate and low transmission settings where malaria transmission is highly heterogeneous, hotspot-targeted interventions form a logistically attractive alternative to untargeted interventions that may need coverage levels nearing 100% to drive transmission lower ^[8],^[12],^[14]. To be financially attractive, the costs of detecting hotspots need to be outweighed by the savings made by targeting only a proportion of the total population. For low transmission areas such as those in pre-elimination or elimination phases of malaria control (i.e., malaria incidence below 5 episodes per 1,000 person-years at risk) and in areas that have succeeded in elimination and are preventing re-introduction, the outcome of this equation is very likely to support hotspot-targeted interventions ^[25]. Hotspot-targeted interventions will also accelerate malaria control in areas of higher endemicity but will require a low-cost and operationally attractive detection system to be financially attractive.

The nature of malaria transmission in hotspots, intense mosquito exposure, and high levels of (asymptomatic) parasite carriage in the human population, will require a combination of interventions that target both the human and vector hosts. In addition to scaling up conventional vector control tools such as LLINs and IRS, several less commonly used tools may be particularly suited for hotspots.

Malaria hotspots appear to maintain malaria transmission in low transmission seasons and are the driving force for transmission in the high transmission season. Targeting the hotspots would mean the most infected and most diseased households would be prioritized with the added benefits of reducing transmission to the whole community. Identifying the hotspots is possible by mapping asymptomatic carriers or using serological tools. Treating hotspots by ensuring high coverage of interventions for a few households is likely to be easier and much more efficient, and may allow for more complicated interventions than using untargeted approaches. The recent successes of scaling up interventions for impact on malaria have revealed the policy gap of what to do afterwards when coverage is good yet malaria transmission continues. In this paper we have argued that the next evidence-based step is to tackle malaria hotspots. Although knowledge gaps exist, we argue that hotspot-targeted interventions should take place at all transmission levels where resources are sufficient and rapid reductions in malaria transmission will be seen.

AG was paid by GSK to attend a single advisory board meeting in December 2009 on how to develop models for the economic evaluation of RTSS. Currently she has a collaborative agreement with GSK to re-analyse RTSS Phase II trial data, which is managed via Imperial College but does not involve monetary exchange and under which the company does not have final decision on publication. None of these activities relates to this article on hotspots. All other authors have declared that no competing interests exist.

This work was supported by the Bill & Melinda Gates Foundation 51991 to Teun Bousema. Roly Gosling is supported through grants from the Bill & Melinda Gates Foundation and ExxonMobil to the Global Health Group at the University of California, San Francisco. Chris Drakeley is supported by the Wellcome Trust (UK), Jamie Griffin by a fellowship from the UK Medical Research Council, Azra Ghani by the Bill & Melinda Gates Vaccine Modeling Initiative and the UK Medical Research Council, Tom Churcher by the European Commission FP7 Collaborative project TransMalariaBloc (HEALTH-F3-2008-223736), and David L. Smith by the RAPIDD program of the Science & Technology Directorate, Department of Homeland Security, and the Fogarty International Center, National Institutes of Health (http://www.fic.nih.gov). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Provenance: Not commissioned; externally peer reviewed.

Graphic design was supported by Kirsten Svendsen at the Global Health Group.