Introduction

In the latest World Malaria Report 2020, an estimated 229 million malaria cases occurred worldwide in 2019 compared to 238 million patients in 2010 [1]. Although there were an estimated 19 million fewer malaria cases in 2019 than in 2010, data for 2015–2019 highlighted that no significant progress in reducing global malaria cases with incidence declined by less than 2%. Most of the patients were in the WHO African Region. Malaysia had no cases of human malaria in 2018 and 2019 [1].

As part of its country-owned and nationally-funded malaria strategy, Malaysia has committed to eliminating indigenous human malaria transmission by 2020. In 2017, Malaysia reported a total of 500 cases (local and imported) of the human type of malaria, down substantially from 6141 cases in 2010 [2]. Overall, malaria transmission in Malaysia is mainly in Sabah and Sarawak, two states located on the island of Borneo, where a significant proportion of the population is at risk of the disease. P. knowlesi malaria, a parasite typically found in monkeys, now accounts for most local cases since its significant discovery in 2004. P. knowlesi, to date, remains a zoonotic disease without documented sustained human-to-human transmission. In the state of Sabah, P. knowlesi notifications increased from 2% (59/2,741) of total malaria notifications in 2004 to 98% (2030/2,078) in 2017 [3]. P. knowlesi cases are increasing over the years and can cause severe disease and death [4]. Increased detection capacity had contributed to the increase in P. knowlesi prevalence.

Literature has shown several risk factors over the years [5–20]. However, there is a gap regarding P. knowlesi acquisition risk factors related to practice specifically in working age group. Targeted interventions had varying degrees of success, including insecticide-treated clothing and hammocks (for use in the forest), toxic mosquito baits, and personal insect repellents [21]. The main objective of this study was to identify the risk factors for acquiring P. knowlesi infection in Sabah among the working age group.

Methods and materials

Statistical analysis

Statistical analysis was performed using IBM SPSS software version 23.0 (SPSS, Chicago, IL, USA). Raw data were cleaned and recoded according to produce two categorical data sets for logistic regression analysis (Table 1). The extent of missing data was assessed, and the sensitivity of results to alternative approaches to imputing missing data was explored, wherever necessary. Univariate analysis was conducted using central tendencies and dispersion measures according to the normality of data (mean and standard deviation for normally distributed data, median and interquartile range for non-normally distributed data), and percentages for categorical data. Bivariate analysis for the quantitative stage was conducted using binary logistic regression to compare a categorical independent variable with the dependent variable. Variables correlating with a p-value of less than 0.25 and clinically important variables were selected for subsequent multivariate analysis. Multivariate analysis was conducted using multivariable logistic regressions with the statistical significance set at a p-value less than 0.05. Hosmer-Lemeshow Test (p>0.05) and Classification Table (>70%) were used to measure the performance of the models.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Coding of variables.

https://doi.org/10.1371/journal.pone.0257104.t001

Results

The mean age for cases was 40.4 (standard deviation [SD] 14.72) years, and mean of 37.9 (SD 13.04) years for the control group. The median household income for cases was RM900 (interquartile range [IQR] RM300) and RM870 (IQR RM500) for the control group. Based on Table 2, males constituted 214 (80.5%) of the cases and 322 (60.5%) of controls. In both cases and controls, native Sabahan (>90%) dominated the study population. The majority of the population were literate (>95%), having completed minimum primary education than those without formal education. Three out of four cases were currently married, n = 216 (81.2%), while more than two-thirds of control were currently married, n = 386 (72.6%). The majority of the study population were regular income earners (>80% for both groups). Around four out of five of the study population had forest-related occupation, n = 230 (86.5%) for case and n = 405 (76.1%). More than two-thirds of the study population had families who own empty land. The majority of the study population had telecommunication signals at home, n = 234 (88.0%) for cases, and n = 515 (96.8%) for control. More than 90% of the study population heard malaria health messages previously, n = 260 (97.7%) for cases, and n = 484 (91.0%) for control. More than 60% of the study population used LLIN, n = 165 (case), n = 335 (control). Although nearly 80% of cases used mosquito repellant, n = 212, the percentage was lower compared to 87% of control, n = 463. About 48.1% of cases (n = 128) reported spending overnight in the forest compared to only 26.3% of control (n = 140). A bigger proportion of cases reported going into the forest during dusk hours, n = 185 (69.5%) than 48.1% of control (n = 256). A bigger proportion of cases reported contact with macaque n = 160 (60.2%) compared to 45.9% (n = 244) of control. Less than 10% of study population had pet macaque, n = 20 (7.5%) for cases, and n = 18 (3.4%) for control. A total of 84.2% of cases had a history of previous malaria infection, n = 224 compared to only 10% n = 53 in the control group.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Characteristics of study respondents, summary of bivariate analysis and selection of variables for multivariate analysis.

https://doi.org/10.1371/journal.pone.0257104.t002

The bivariate analysis revealed that male gender, not currently married, having a forest-related occupation, no telecommunication signal at home, going to the forest during dusk hours, spending overnight in the forest, having pet macaque, having contact with a macaque, the family owning empty land, not using mosquito repellent, and history of previous malaria infection, and never heard or seen malaria health message were found to have a crude association with P. knowlesi infection. Variables with a p-value of ≤0.25 and other clinically important variables were included in multivariate logistic regression analysis (marked with *) (Table 2). Forward and backward logistic regression methods applied. The odd ratio (β) for the significant factors shows the increase (or decrease if the ratio is less than one) in odds of being in one outcome category (positive or negative for Pk infection) when the value of the predictor increases by one unit. From Table 3, individuals who were male gender were 2.71 times more likely than individuals who were female gender to acquire P. knowlesi infection, all other factors being equal (AOR = 2.71; 95% CI: 1.63–4.50). Individuals who practiced spending overnight in the forest were 1.92 times more likely than individuals who do not practice spending overnight in the forest to acquire P. knowlesi infection, all other factors being equal (AOR = 1.92; 95% CI: 1.20–3.06). Individuals who did not use mosquito repellent were 2.49 times more likely than individuals who used mosquito repellent to acquire P. knowlesi infection, all other factors being equal (AOR = 2.49; 95% CI: 1.36–4.56). Individuals with a history of previous malaria infection were 49.34 times more likely than individuals without a history of prior malaria infection to acquire P. knowlesi infection, all other factors being equal (AOR = 49.34; 95%CI: 39.09–78.32). The model was fitted with the Nagelkerke R Square = 0.634 and Hosmer–Lemeshow test (p = 0.333), with the overall correctly classified percentage of 88.1% (>70%).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Summary of multivariable logistic regression analysis of risk factors for acquiring P. knowlesi infection.

https://doi.org/10.1371/journal.pone.0257104.t003

Discussion

Age was a risk factor of P. knowlesi infection in bivariate analysis, but was not an independent predictor in the final multivariable model. Previous studies showed that all ages were susceptible to infection, with cases also reported in Malaysian children [22], and Vietnamese children [23,24]. Patients with knowlesi malaria demonstrated a wide age distribution (median 33, IQR 20–50, range 0.7–89 years). The presence of familial clustering of cases had been demonstrated, indicating transmission is probably now occurring peri-domestically, which may be linked to deforestation or land-use change in these environments [5], or other yet to be determined factors. Another study reported mean age to get P. knowlesi infection is 44.9 [6]. The scope of this study was 18 years and older because the evidence of peri-domestic infection has not been widely explored and is different than the more extensively studied risk predominantly linked to exposure activities [14]. Respondents with history of previous malaria infection were more likely than those without a history of malaria infection to acquire P. knowlesi infection as found in this study. This factor was also an independent predictor in the multivariable model. This was most likely due to their activities and practices. Moreover, as of date, there is lack of evidence supporting any form of immunity from previous malaria infections. Current available control measures which were effective against human indigenous malaria have shown little success in controlling P. knowlesi malaria due to the outdoor biting nature of its vector [25]. A clinical parameter study may be helpful to explore this factor. Many of the respondents (more than 15%) could not give an estimated monthly income because their income was not regular. They supported their daily lives with natural resources and financial assistance from the government. Although a significant percentage had a steady income, many (>80%) still earned less than RM2,000 per family per month, which was below the Poverty Line Income (PLI) per month, put at RM2,537 [26]. Despite this fact, having no regular income was not a statistically significant risk factor for developing P. knowlesi infection. We may relate income to the ability to procure mosquito repellent, which the Ministry of Health does not provide, Malaysia, under its Vector Borne Diseases Control Program Sector. This factor can be explored in future research.

There was a correlation between income and level of education obtained [27]. In this study, a proportion of illiterate respondents had never attended formal education but was not significantly associated with P. knowlesi infection. Siri found that educating women could enhance their essential life skills, trust in public health systems, and their ability to become more conscious about the various factors that affect the health of their children [28]. Poor knowledge and understanding of malaria could subsequently lead to adverse attitudes and systems towards malaria prevention [19]. This evidence can be applied to advocate improving the educational level of the local population for better absorption of malaria health message and empowerment of the local community to build resilience in preventing P. knowlesi infection despite the fact that the variable never hearing malaria health message was not statistically significant to be associated with P. knowlesi infection. Having an internet signal at home is an essential factor in ensuring the timely conveyance of health messages. Not having an internet signal at home caused many individuals to enter the forest at night for a good internet signal, which led to them being infected with P. knowlesi. Although there was a statistically significant association between not having a telecommunication signal at home with P. knowlesi infection in the bivariate analysis, it was not an independent predictor in the multivariable model.

The male gender presented the more significant proportion of the two genders in both groups of control and cases. This factor was a statistically significant factor for P. knowlesi in bivariate analysis and was also an independent predictor in multivariable analysis. This confirmed the findings reported in a previous study, which stated that men had higher risks of P. knowlesi exposure, especially adult men working in agricultural areas, had the highest risk of P. knowlesi infection [13]. Another study conducted in Aceh also confirms similar findings [14]. As this study was conducted in Ranau District in Sabah, with more than 90% of its population being Native Sabahan, it was not surprising that being of Native Sabahan ethnicity was not a statistically significant factor for P. knowlesi infection. We believe that many of Native Sabahan in this study acquired P. knowlesi infection due to their respective residential area or occupational group. The poor were concentrated in Bumiputera, with a higher incidence of poverty, especially in Sabah, where the poor were concentrated in rural areas where P. knowlesi incidence is high [26].

Changes in land use such as shifts in agricultural practices, deforestation, and forest fragmentation have been proposed as key drivers in the emergence of P. knowlesi infection [29]. This complex interaction could be related to the proximity of vectors and include conflict around the availability of food for macaques from human agricultural practices such as cultivated fruit orchards, rice paddies, or cornfields, in addition to traditional or changing techniques such as trapping or hunting monkeys for pets or food. Despite spending most of their time in the forest, macaques invade farmland in search of food. The percentage of the total macaque population in daily contact with the farm is unknown. In order to understand more on P. knowlesi transmission, we require timely information on degree of changing forest cover and land use and the efects on the distribution of vectors and hosts of P. knowlesi and parasite transmission rate through increased in surveillance activities utilising latest technology [30]. Vythilingam et al. found that the prevalence of P. knowlesi infection of macaque in the wild (forests) was exceptionally high at 97%., compared to urban macaques, which were infection-free [31]. Lee et al., also found a prevalence of 87% in Sarawak among long-tailed macaques [32]. Many studies have described the association of P. knowlesi with occupation. Occupations such as forestry, agriculture, and hunting had been identified as risk factors for getting P. knowlesi infection [7,13]. A prospective study of malaria patients from a referral catchment area of north-western Sabah described 92% (119/130) of PCR confirmed P. knowlesi malaria cases having forest or plantation exposure, including living within a 20 minute of unclassified forest type or a plantation (p = 0.001 and 0.015, respectively) [9]. Apart from occupation, some reports of P. knowlesi infection from traveling activity in a popular low-risk travel destination [33]. Ranau district, a frequent travel place for local and international tourists alike, had this risk to consider among travelers. An. Balabacensis, the vector implicated in Sabah for P. knowlesi, prefers to breed in-ground pools formed in fruit orchard, rubber, and palm oil plantations [34]. With this precedence, there was no surprise that both having forest-related occupation and contact with macaque were statistically significant risk factors for P. knowlesi infection in this study, albeit not being an independent predictor in the multivariable model. With human populations increasingly encroaching on and supplanting macaque habitats, there is intense selective pressure for the parasite to switch its natural mammalian host. Human-to-human transmission of P. knowlesi has been shown to occur under experimental conditions [35]. A study in Singapore showed more than 80% of wild macaques sampled were infected with Plasmodium parasites [36]. Keeping all these in mind, there is an urgent need for risk mapping of macaque density as well as surveillance of wild macaque.

P. knowlesi infection reported in two clusters in Aceh, Indonesia, revealed a history of spending overnight in the forest, with or without protection, in the area where macaques were abundant [10]. A local study found a strong association between traveling in the forest area at night [13]. Individuals with a workplace location in or near the forest and requiring overnight stays and individuals who visited the forest in the previous month for any reason had higher odds of P. knowlesi infection compared to those who don’t [14].

Vector mosquitoes, members of the An. leucosphyrus group are found throughout forested regions of Malaysia and are particularly associated with dense jungle and forest fringes [37]. An. leucosphyrus mosquitoes are exophagic, typically feeding and resting outdoors after dusk. To this extent, exploration was also conducted to find an association between spending overnight in the forest and going to the forest during dusk hours. Although both these factors were statistically significant during bivariate analysis, only spending overnight in forest was an independent predictor in the multivariable model. Further study needs to be conducted to determine the duration of time spent in forests with forest-related occupation or activities to produce a more reliable model that fits both variables.

The Leucosphyrus mosquito complex has been predicted to occur in areas with high forest cover but no intact forest cover [34]. The sparse data and low volume of data for this complex mean that these predictions should be interpreted cautiously. This complex is responsible for P. knowlesi in the region in which knowlesi malaria occurs most frequently in Borneo of Malaysia. Deforestation (loss of intact forests) occurs [11,12]. The conversion of intact forest to disturbed forest and the resulting impact on the likelihood of members of the Leucosphyrus complex could be a factor in to increase of P. knowlesi cases. A study in the northern part of Sabah in Borneo, Malaysia recently found an association between two forest variables, namely forest loss and total cover within two km of a village, and the estimated incidence of knowlesi malaria at the village level in the two districts studied, but this does not differentiate between intact and disturbed forest [38]. In our study, there were individuals with family empty land opening activities infected with P. knowlesi. Although the bivariate analysis stage of this study elicited the association between family owning empty land and P. knowlesi infection, it was not an independent predictor in the multivariable model.

In Malaysian Borneo, studies found that M. fascicularis populations were declining due to logging activities. Still, their local abundance was higher in areas logged ten years previously than in unlogged forests [39]. M. fascicularis populations can occupy a wide range of habitats. More importantly, the distribution of this species encompasses many locations close to human habitation (urban areas) or activity (disturbed forests, orchards, croplands, etc.) [40]. Apart from that, there is evidence of zoonotic transmission of diseases harbored by companion animals about the close contact with their owners. There is evidence that Macaca fascicularis and M. nemestrina monkeys are kept as pets on Sulawesi [41]. The same practice was found in Ranau District, whereby investigation of a suspected peri-domestic knowlesi malaria infection of a young child has led to the discovery of the practice of having macaque as a pet which was subsequently tested and was found to have P. knowlesi parasite (unpublished). However, in this study, although there was a statistically significant association between having macaque as a pet and P. knowlesi infection, it was not an independent predictor in the multivariable model.

With malaria elimination at the grasp of the hands for Malaysia, one of the main challenges is the prevention and control of zoonotic knowlesi malaria in rural areas where there is a large pool of parasites in the macaque reservoir hosts. A reduction in these macaque reservoir hosts would be one option, but implementing such a measure would be impractical and burdensome. The most common vector control strategies for malaria are Long Lasting Insecticidal Nets (LLINs) and the Indoor Residual Spraying (IRS) of houses; both focused on protecting against human malaria. The mosquitoes’ outdoor biting and resting behaviour hinder the effectiveness of these methods due to the zoonotic reservoirs in the macaques. Preventing humans from being bitten by vectors would be the primary prevention and control of knowlesi malaria in these settings. It would involve the use of insect repellents. Still, these are prohibitively expensive for subsistence farmers, hunters, logging camp workers, and other rural people whose daily activities take them to the forest and forest fringe. The use of insecticide-impregnated hammocks had been successful in controlling forest malaria in Vietnam [42]. Although some people stay overnight in the forest of Malaysian Borneo, hammocks were not traditionally used here. So much a control measure would be challenging to implement successfully.

In this study, non-users of LLIN were not shown to have higher odds than users to acquire P. knowlesi infection. Non-users of mosquito repellant, on the other hand, were at higher odds than users of mosquito repellant to develop P. knowlesi infection and was an independent predictor in the multivariable model. Insecticide-treated clothings (ITC) was reported as effective against bites from Anopheles mosquitoes [43,44]. However, ITC is not easily accessible, thus was not assessed in this study [45]. Further research is warranted to explore the potential of ITC for reducing zoonotic and human malaria, particularly regarding its scalability, cost-effectiveness, duration of durability, safety under variable environmental working conditions, resistance to UV light exposure, and washing [45–47].

The possibility of eliminating knowlesi malaria is limited because forests that harbor Anopheles mosquitoes and macaque monkeys will remain a reservoir for the zoonotic transmission of P. knowlesi. The risk for P. knowlesi infection in humans is highly variable. There is a sense of urgency to be on top of controlling and preventing Pk infection. The problem is compounded further because there are numerous vector species with highly variable anthropophilic and many simian host species [48,49]. Increasing knowledge on mosquito bite prevention has been shown to reduce Pk incidence effectively [50].

Conclusion

In summary, male gender (AOR = 2.71; 95% CI: 1.63–4.50), spending overnight in forest (AOR = 1.92; 95% CI: 1.20–3.06), not using insect repellent (AOR = 2.49; 95% CI: 1.36–4.56) and ever diagnosed with malaria (AOR = 49.34; 95% CI: 39.09–78.32) were found to be independent predictors of P. knowlesi infection. This study showed the need to strengthen the strategies in preventing and controlling P. knowlesi infection specifically in changing the practice of spending overnight in forest and increasing the usage of personal mosquito repellent.

Acknowledgments

The authors would like to thank the Director-General of Health Malaysia for permitting the publication of this article. Gratitude is also extended to Sabah State Health Department, Tuaran Area Health Office, Ranau District Health Office, and all participants for their utmost support for completing this study.