Malaria is a disease that is not solely transmitted by itself, instead, it requires a specific vector to successfully inject itself to the host body. Theoretically, this is based on a causal concept of epidemiology that the occurrence of a diseases depends on three primary factors: (i) the host, (ii) the agent, and (iii) the environmental factors. Studies in the Amhara region, the northwestern part of Ethiopia, suggested an insight into critical factors comprising malaria risk. In general, three factors have been recognized to be the key factors deriving malaria risk: climate variables, entomological parameters, and human population dynamics. The impact of climate variables is inevitably accounted for as a potential circumstance in malaria transmission, as it known to be an enhancement factor of malaria transmissibility due to increasing vector capacity by providing potential sources of breeding places, increasing mosquito longevity, and feeding rates. ^{1 – 11} It is plausible that the interconnection of co-factors such as environmental constraints affecting the entomological and parasitological constituents enhance the transmissibility of malaria. However, a broad range of influences could drive the transmission pattern of malaria from either humans or the vector. ^{10 , 12} Therefore, instead of controlling the impact of climate change, upgrading public health interventions and socioeconomic conditions might better affect malaria. ¹³

Besides climate change, whose alterations on various aspects have been well documented, another essential variable driving malaria transmissions are the entomological parameter and human population dynamics. Biological and behavioral variations of mosquito species, such as vector-habitat relationships; factors affecting vector population abundance; host-seeking behavior; and the emergence of vector resistance to insecticides, have been known to affect the transmission pattern of malaria. ^{7 , 14 – 16} Endophilic mosquitoes tend to feed and rest indoors, thus poorly constructed dwellings and close proximity with vector breeding sites along with human’s behavior attracting mosquitoes such as unprotected sleep and placing their livestock in the house intensify the chance of mosquito contact. ¹⁷ However, there is an observed behavioral change in the feeding habits of mosquitoes from endophilic to exophilic and its feeding time from late evenings to early evenings. ^{18 , 19} It is also possible that vector control intervention could change the natural behavior of mosquitoes from endophilic to exophilic which is caused by avoiding control strategies such as insecticide exposure which are usually utilized inside human dwellings. ^{18 , 20 , 21} On the other hand, the human population dynamic also becomes a potential source of malaria transmission by increasing the likelihood of spreading the disease mainly through import cases. The imported cases can either be from recent migration or short-term travel. ^{22 , 23} A study report from Ethiopia in the 1980s found that a large population movement affects high transmission rates of malaria. ²⁴ This large movement of the human population might have a close relationship with agricultural work, ²⁵ and these immigrant workers are more likely to live in non-permanent houses, which are vulnerable to mosquito bites, and occasionally sleep outside, as well as having inadequate information about malaria risk. ^{22 , 26} A difference in practical agricultural activities as suitable habitats of vector mosquitoes may be the predisposing factor of malaria transmission. ²⁰ Additionally, an untraceable small number of mobile sub-population groups might delay the malaria elimination strategy due to its higher risk of infection or might even re-introduce malaria in previously eliminated areas. ²²

Moreover, besides evidence from Ethiopia, plenty of studies have associated several other risk factors influencing malaria transmission, mostly published in recent years. ^{27 – 38} In the late’90s, it was known that the older the patient, the less the incidence of malaria as well as less knowledge of malaria prevention. Several other associated factors were included such as exposure to forests and receiving previous antimalarial treatment. ^{27 , 39} Bed nets could not be a very effective protective measure in a setting such as the environment in which this study was done; environmental intervention may be better applied. ²⁷ In pregnant women, the associated factors of malaria infection are lack of education, and non-possession of insecticide-treated nets (ITNs) followed by a decrease of parasite density as age increased. ²⁸ Children under the age of five years were also particularly at risk of being infected by malaria parasites, especially in sub-Saharan Africa. ²⁹ The associated risks of this particular at-risk population are mostly sociodemographic related factors such as the main floor and main wall material of the house and availability of electricity. However, indoor residual spraying (IRS) significantly reduced a child’s risk of malaria, with additional information that older children have a higher risk of malaria, notwithstanding that their risk decreases with increases in cluster altitude and their caregiver’s education level. ³⁰ Another study showed an exciting method to discover the associated factors of malaria infection. Pinchoff et al., ³¹ used a case-control approach based on positively detected incidence by a rapid diagnostic test (RDT) with a sophisticated statistical method. They found that, in multivariate model generalized by generalized estimating equations (GEE), the odds of being RDT positive are highest in five-17 years old (8.83 odds compared to 18 years old (or more)) and do not vary between seasons. Additionally, there is an interaction between age and report of symptoms, with an almost 50% increased odds of reporting symptoms with decreasing age category. Instead of using a case-control approach, Elijah Chirebvu et al., ³² uses a more convenient method over which the history of malaria infection is an independent variable and found that the correlated factors of malaria are household income, late outdoor activities, time spent outdoors, travel outside of the study area, non-possession of ITNs, hut/house structure, and homestead location from bodies of water. In addition, the proximity of a health facility and low vegetation cover are advantageous protective factors.

Another interesting study by Kazembe et al., ³³ used a spatial regression analysis to estimate risk factors. These findings based on regression estimation found that the children who visited rural areas have six times the risk of being infected by malaria parasites, as with previous findings the higher the age of the children, the more the likelihood of being infected notwithstanding that the risk reduces as individuals gain a higher sociodemographic status. Proximity to a garden, river, or standing water are not associated but act as a cofactor of increased risk. Furthermore, this study showed that a spatial cluster of households of the infected patients affects the risk of transmission which may be explained by the variability of the environmental factors. A group of researchers, ³⁴ using a secondary database on a nation-wide scale with a regression model, found that wealth status is the first socio-economic factor which mostly contributed to the difference of malaria risk among African children. They did not find any demographic factor among the associated variables. On the other hand, sex of child and river or water body proximity are not associated with the risk of being infected with malaria. The country of resident and temperature could be a cofactor in the analysis with supplementary information of negative associations between population density and malaria incidence. One thing that should be noted, is that the study completed a comparison study of differing malaria risk and found there are several differences in associated variables between low and high-risk countries. ³⁴

In Indonesia, such risk factors have not been extensively discovered. Several studies were attempting to determine the associated factors of malaria. Based on active and passive surveillance assessing three common species of malaria in Aceh, a study ³⁵ found that the related factors are male (AOR 12.5), adult (OR 14.05), visiting the forest within the previous month regardless of the reason (OR 5.6), and working place located in the forest with overnight stays (OR 7.9). In Papua a study ³⁶ adopting the Bayesian hierarchical logistic model found that rural Papuans, as well as those who live in poor, densely forested, lowland districts, are at higher risk of being infected of malaria with the additional information of nine areas on the island having higher-than-expected malaria risks. Environmental factors such as the distance of the resident to forest areas, altitude, and rainfall are also associated with malaria. These environmental factors were also found to be strongly varied spatially in different regions. ³⁷ Additionally, a case-control study in the Purworejo district has found that not sleeping under a bed net and not closing doors and windows from 6 p.m. to 5 a.m. are associated with higher risks of malaria. ³⁸ With that limited information on malaria risk factors and the fact that there are such varying associated variables, the current study has an objective to uncover the risk factors with a broad categorical variable including individual and environmental factors. To strengthen the differences between low and high-countries as well as spatial effect of malaria, this study also included a comparison of risk factors between different endemicity areas. Additionally, to prove the risk of the sociodemographic factor, a series of entomological observations that include house type materials and condition was also included.

This research had a study duration of four months in Jambi, the western part of Indonesia, and another four months in Sumba Island, the eastern part of Indonesia, from February to October 2019. The total of 157 cases of both locations were successfully collected during the field sampling time. ⁸⁰ The proportion of case and control was following a 1:2 ratio. Therefore, out of 158 cases, there were 328 controls with a percentage of 32.3% and 67.7%, respectively. The basic demography of each location is presented in Table 1. The proportion of sex between Jambi and Sumba have a slightly similar pattern of male and female. The age strata from the two sites are identical at six-24 years. However, Sumba has more cases in children (0-5 years).

Several individual factors from both Sumba and Jambi have been associated with malaria incidence ( Tables 2 and 3). In Jambi, night activity outdoor (OR = 0.32; CI: 013-0.79), history of visiting forest areas in the previous month (OR = 0.35; CI: 0.15-0.84), and working place is located inside the forest (OR = 0.17; CI: 0.07-0.43) were protective factors against malaria infection. The individual risk factor for malaria infection in Jambi were not having a bed-net for sleeping (OR = 2.09; CI: 1.04-4.18), low level of education (OR = 1.01; CI: 0.29-3.45), occupation ( P value = 0.000), and contact with malaria-infected patient (OR =. 3.37; CI: 1.62-7.01). The observed environmental factors that are associated with malaria in Jambi are the existence of shrubs around house areas (OR = 28.00; CI: 6.45-121.59), the existence of puddles or stagnant water around the house area (OR = 2.49; C I: 1.05-5.98), the presence of livestock nearby the house area (OR = 6.36; CI: 2.94-13.79), and the proximity of houses to forestry areas (OR = 10.84; CI: 3.97-29.58). There were eight associated individual variables with malaria from Sumba. The risk factors of malaria in Sumba are not having a bed net for sleeping (OR = 2.55; CI: 1.52-4.29), low level of education (OR = 6.09; CI: 2.12-17.47), never consumed antimalarial drug (OR = 4.16; CI: 2.09-8.28), if they ever had contact with malaria person (OR = 17.33; CI: 8.04-37.32), frequent traveling outside of the residential area (OR = 5.22; CI: 2.32-11.74), if they ever visited the forest in a previous month (OR = 1.96; CI: 1.03-3.73), and requiring an overnight stay in the forest (OR = 2.88; CI: 1.22-6.81). Additionally environmental risk factors associated with malaria are existence of shrubs surrounding house (OR = 20.99; CI: 8.24-53.46), existence of puddles or stagnant water surrounding the house area (OR = 39.98; CI: 13.86-115.32), existence of livestock inside house (OR = 3.24; CI: 1.70-6.18), existence of livestock nearby house (OR = 9.44; CI: 2.87-31.07), non-permanent house wall (OR = 5.22; CI: 1.81-15.06), non-permanent floor construction (OR = 20.79; CI: 2.81-153.79), house is in a close proximity to rice fields (OR = 14.69; CI: 0.75-286.96), and not having a ceiling of the house (OR = 19.72; CI: 1.18-330.29). Since Jambi and Sumba have different endemicity level, the generalized linear model (GLM) was applied to discover if any difference in risk factor variables from both sites. Due to any difference in the number of cases of both locations, only prospective cases from Sumba were included in the analysis. Prospective cases are those who enumerated within five to seven days after being confirmed by a rapid diagnostic test or the result from two independent microscopists or combination of both. The result of GLM indicates that there is a difference in risk factor variable from both Jambi and Sumba ( P value = 0.002) ( Table 4).

In order to discover the effect of house type with malaria infection, a series of entomological observations were carried out as described in the methods section. A total of 2,435 Anopheles mosquitoes were successfully collected from both sites. Out of the total collected Anopheles, 2.9% (71) is from Jambi, and the rest 97.1% (2,364) is from Sumba. Jambi was dominated with Anopheles balabacensis (79%), followed by other species; An. Maculatus (18%), An. barbirostris (1.41%) and An. sinensis (1.41%). Two species accounted to 40% and 58% of the total mosquitors catched in Sumba, An. aconitus and An. Sundaicus, respectively. The other species found were An. maculatus (1.06%), An. subpictus (0.17%), An. barbirostris and An. vagus (0.084%) and An. farauti and An. leucosphyrus (0.04%). No plasmodium was detected either from salivary nor abdominal part of the mosquitoes over 250 randomly selected mosquitoes from both Jambi and Sumba.

There is a significant difference between the total number of mosquitoes collected from Jambi and Sumba ( P value ≤ 0.0001) ( Figure 1). As explained in the methods section, this entomological observation characterized the houses into three types: malaria houses, non-malaria houses, and permanent houses. There is no significant difference in each house types from Jambi ( P value = 0.1856). Although malaria houses from Jambi have the highest mean collected mosquitoes (0.64) than the other house types, permanent house types (0.34) have a higher mean of collected mosquitoes than the non-malaria house (0.29) ( Figure 2). On the contrary, there is a significant difference between malaria-houses vs non-malaria houses ( P value = 0.0143) and permanent houses ( P value = 0.0351) in Sumba as presented in Figure 3. However, no difference was observed between non-malaria houses and permanent houses ( P-value ≥ 0.9999). Additionally, if both sites are combined ( Figure 4), only malaria-houses and non-malaria houses have a significant difference in the number of collected mosquitoes ( P value = 0.0301). Permanent house type is slightly higher in the mean number of collected mosquitoes compared to non-malaria houses (5.6 and 5.032, respectively).

During the entomological observation, weekly malaria screening was conducted to ensure the presence or absence of malaria in each house types as well as to discover the incidence rate of each house types. Malaria has detected only one in the first week from Jambi. Otherwise, 10 cases were detected during three weeks of observation from Sumba ( Table 5). If the number of cases is transformed into an incidence rate per collection method per year, then malaria houses from Jambi have 1.4 incidence rate per year and null for other types of houses. However, considering the difference in the endemicity level in Sumba, malaria houses have 8.7 incidence rate per year while non-malaria and permanent dwellings have the same rate of 2.9. Additionally, based on the calculated odds ratio, the odds of malaria houses compared to other house type is 3.77 (CI: 0.76-18.81) while non-malaria versus permanent houses have the odds of 1 (CI: 0.14-7.30).

Based on the Indonesian basic health profile, Jambi was categorized to have low cumulative incidence, while Sumba as part of Nusa Tenggara province has high cumulative incidence. ⁴⁰ By transforming the total number of collected cases in each location, Jambi and Sumba, then both sites have a high cumulative incidence of malaria (5.4 and 15.7, respectively). There is a discrepancy of classifying endemicity level between national data and the collected data from the current study. This phenomenon partly can be explained by the different denominators of the data, as national data considers a total number of populations in a provincial level rather than each sub-district level. Considering the fact that malaria varies greatly between sub-district and district and is not uniformly distributed, this may be the case. ^{43 , 44}

One of the interesting findings of the current study is the different patterns for the source of infection between Jambi and Sumba. Night activity outdoor, history of visiting forest areas from the previous month, and working place is located inside the forest are protective factors in Jambi, while the history of visiting forest areas and requiring an overnight stay inside a forest are risk factors in Sumba. This finding suggests that most of the case from Jambi was infected in the residential areas and forest areas were the source of malaria infection from Sumba. This finding underline that visiting forest areas are not always a risk for malaria infection as previous research has found. ^{32 , 35 , 36 , 39} The phenomenon may be explained by the relationship of human and mosquito infection over which may be caused by uneven distribution of mosquito bites across the human population. For example, a study showed that in several areas a core group of the human population receive a substantial proportion of mosquito bites. ⁴⁵ Additionally, another finding indicates that a group with more individuals experiences a lower rate of mosquito bites. ⁴⁶

Jambi and Sumba share the same individual risk factors, namely no possession of bed net for sleeping, low education level, and if they ever contacted a malaria infected patient. It is common that bed nets and low education levels are a risk factor for malaria as previously discovered. ^{28 , 32 , 38} The effective impact of bed nets has been extensively described in previous studies. ^{47 , 48} Although defining the coverage of the bed net use is problematic. ^{49 , 50} The effect of education on malaria infection has also been found from the previous study. ²⁸ Studies have demonstrated that the poor performance of children at school is a risk of malaria, and if knowledge of prevention is adequately elevated it will lower the incidence of malaria. ^{51 , 52} While other researchers found that the performance of education may be temporary and not prolonged. ⁵³ Additionally, most of the cases had contact with the other cases being compared to control suggesting that they may have an infection during the interaction process. Considering the proximity of the distance of houses as neighboring, cases may also sleep in the same house or be involved in a late conversation or another way of interaction in which mosquitoes could bite them simultaneously.

There are several differences in individual risk factors between Jambi and Sumba. Occupation is statistically significant to be the risk factor of malaria in Jambi. This finding is in line with the previous discussion where most of the controls are a farmer that require them to go to the forestry areas. Most of the cases have an occupation that needs them to reside in the housing area such as a midwife, workshop worker, or odd jobs. On the other hand, in Sumba, having never consumed antimalarial drugs and traveling outside the residential area are the risk factors of malaria infection. Considering the effectiveness of the current antimalarial drugs, the higher number of antimalarial drug consumption in cases is since the majority of the cases may have re-infection that requires them to be prescribed with frequent antimalarial drug. It was described that re-infection is a common situation in a high transmission area. ⁵⁴ The risk of traveling outside residential areas with malaria infection is in line with a previous study. ³² The participant of the current study may have traveled in neighboring villages in which infection rate are high. There are several studies that have demonstrated the risk of traveling into a high infection rate area. ^{55 , 56} A reporting system need to be established to identify import cases from neighboring villages or areas. ^{57 , 58}

There are several same environmental factors between Jambi and Sumba, i.e., the existence of shrubs and puddles/stagnant water surrounding the household area and the presence of livestock near the house area. Studies have found that bushes are a risk factor in a densely forested area and can encourage mosquitoes to breed. ^{31 , 59 , 60} Since case and control resided in a densely forested, basic biological attributes of the vector may play a role. It was shown that shrubs promote the malaria transmission capacity by providing an abundant source of sugar for the male mosquito while lack of sugar source contributes to lower insemination rate to females. ^{61 , 62} Moreover, as observed in An. gambiae, Anopheles mosquitoes are distributed among dense growths of brush. ^{63 , 64} They may rest in the shrub prior to get ready to bite. The existence of puddles/stagnant water was found to be a potential breeding place where Anopheles mosquito could oviposit. ^{65 – 68} Although, evidence has suggested that no or low Anopheles larvae density is found when water is identified as turbid as puddles, drains, or swamps. ⁶⁹ As previously described, the existence of livestock nearby the house area increases the chance of mosquito contact. ¹⁷ It was previously found that the presence of livestock at the household level can significantly alter the local species composition, feeding and resting behavior of malaria vector. ⁷⁰ However, the net impact of livestock-associated variation in malaria vector ecology on malaria exposure risk was unknown. ⁷⁰ In addition, the pattern of host attraction and biting behavior of Anopheles mosquito in Indonesia has not been yet extensively studied and only limited to one locality. ⁷¹ Anopheles mosquito can be attracted to livestock even with the primary vector of malaria, because of their biting preference as zoo-anthropophilic species. ⁷² Furthermore, placing livestock inside the house is also significantly correlated with malaria in Sumba, suggesting a different cultural behavior of tethering livestock inside. Our study demonstrates the importance of controlling malaria using livestock-based intervention or using any zoo-prophylactic agent as described elsewhere. ^{73 , 74}

House construction has been associated with malaria in Sumba such as non-permanent house walls, non-permanent floor construction, and not having a ceiling of the house. Such factors are in line with the previous report regarding the associated demographic factor of malaria infection underlying the importance of human dwelling construction. ^{17 , 29} However, this finding may not be the case since there is a difference with entomological finding as discussed below. Additionally, the proximity of the house to forest areas and rice fields are the risk factor for malaria in Jambi and Sumba, respectively. As discussed previously, housing location in proximity to lower vegetation cover is the protective factor for malaria. ³² Jambi and Sumba have different agricultural activity. Most of the people from Jambi work on rubber and palm plantations that require a large area of land, a single person could only handle five-10 hectares. On the contrary, Sumbanese people are mostly working on cashew or rice which requires a relatively limited space of land. Agricultural activities have been shown to be a predisposing factor for malaria in which suitable habitat of the vector may take place. ^{20 , 25} It was previously described that rice field agro-ecosystems contributed significant vector populations. ^{75 , 76} However, with the same densely forested areas, the source of infection is different between the two sites as noted in the above discussion.

Previous studies have found differences in associated variables between low and high-risk countries. ³⁴ As well as environmental factors, these are also varied across spatially different regions. ³⁷ In order to strengthen this fact, in the current study, we selected a different annual parasite index area for comparison. Based on GLM, there is a significant difference in the risk factor between Jambi and Sumba. The GLM analysis was set to equate starting from the number of case and control and the only associated variables that the sites share the same manner. It suggests that the frequency of risk factor variables between Jambi and Sumba is in a different state following its differing annual parasite index (API). Additionally, considering the fact of the different number of associated individual and environmental variables between Jambi and Sumba suggests that a high API area like Sumba has more diverse risk factors than low API area.

Finding from the current study and the other indicates that housing construction is associated with malaria infection. ⁷⁷ Others recommended that improved housing is a promising intervention for malaria. ^{78 , 79} However, our entomological observation found that housing construction does not necessarily lead to decreased risk of Anopheles bites. Only malaria houses were found to be significantly different with non-malaria or permanent and non-malaria houses. Permanent house types had a higher mean number of collected Anopheles mosquito than non-malaria house regardless of the sites. This finding is also supported by malaria infection rate of the house types that malaria house type is higher than the two types of houses while non-malaria and permanent houses share the same number of infection rate. This phenomenon can be best explained by the existence of risk factors other than only housing types such as the presence of livestock, shrubs, puddles/stagnant water, or housing localities. As long as the other environmental risk factors are not controlled then the housing improvement program may not be effective as stated in the previous findings. ^{78 , 79}

Despite the findings given from this study, there are some limitations that need to be addressed in the future. The number of samples is not equal between the two locations where Jambi is substantially low. Due to an extremely low number of cases, a convenience sampling technique was performed, thus leaving inadequate analysis for Jambi. It is also necessary for future research to increase the number of samples of houses for mosquito density in each house type. Finally, to find the best association model, future research needs to consider matching analysis when performing a case-control study for malaria risk factors.

In the current study and the others have demonstrated that risk factors play a notable role in the malaria infection. In summary, there is information on our findings for malaria control strategies ¹ ; visiting forested areas is not always a risk factor for malaria as a source of infection may differ between location, ² livestock-based intervention or using any zoo-prophylactic agent is inevitably effective to avoid mosquito attraction regardless of the area, ³ improving dwelling strategy may not be successful before controlling other environmental factors, and ⁴ risk factors are site-dependent suggesting that applying risk factor management need to consider the endemicity status of an area.