Can slide positivity rates predict malaria transmission?

Malaria is one of the major public health problems in China, especially in Yunnan Province, which has significant mortality, morbidity and economic burden. Yunnan Province is a malarial hyper-endemic area and had the highest number of malaria cases and deaths for more than 10 years until 2005 in China [1, 2]. The outbreaks of malaria happen annually along border areas in Yunnan, China. The likelihood of imported malaria cases has been increased along the border areas between Yunnan and Myanmar, Laos and Vietnam over recent years, due to increased trade and tourism in these areas [2, 3]. In order to control malaria it is important to enhance disease surveillance and evaluation of malaria transmission [4, 5] in this endemic region.

The intensity of malaria transmission can be estimated using different indicators such as annual blood examination rate (ABER), annual parasite index (API), slide positivity rate (SPR) and the incidence of malaria [6‐10]. In China, the annual malaria incidence is commonly used. Malaria incidence includes numbers of laboratory-confirmed malaria cases and other cases diagnosed with clinic symptoms (e.g. fever) as a numerator and the local population as a denominator. The local population size might be under- or overestimate because census is only carried out once 10 years in China. Huge population movement is common due to economic development in China in the last three decades. Thus, malaria incidence might be inaccurate due to limited health care resources [7] or under- or overestimates of population size [11]. It is important to estimate the burden of malaria accurately for planning public health interventions. Slide positivity rate (SPR) has been used as a surrogate to measure the incidence of malaria [7, 9, 12, 13], to define the level of malaria endemicity [11], and to identify malaria high risk areas [14]. This is a principal monitoring indicator in the malaria elimination programme in China for the period 2010 and 2020 and it has been monitored since the 1980s [15, 16] through the malaria annual reporting system. The changes in malaria incidence can be estimated from the SPR trends [7]. Some studies have demonstrated that SPR has steadily decreased with the decline in malaria incidence [8, 12], while others found that the annual parasite index (API) increased, but SPR kept steady at the same level over 20 years [8].

The development of the malaria early warning system (MEWS) has been started based on the surveillance system in China over recent years [17, 18]. However, these studies are limited to climatic indicators and did not take advantage of monitoring indicators, which can help improve malaria prevention and control, especially in the early stage of malaria elimination. Moreover, the relationship between SPR and the incidence of malaria is not clear in the border areas of Yunnan Province, China. This study aimed to examine the role of SPR in monitoring malaria transmission, and improve the planning and implementation of malaria control and prevention programmes.

Figure 2 shows the annual pattern of malaria incidence and SPR in Mengla County. In this hyper-endemic region, a total of 8,962 malaria cases were reported and annual malaria incidence rates ranged from 23 to 648 per 100,000, while the SPR varied between 0.42% and 13.08% from 1993 to 2008. The scatter plot with regression line depicts the crude relationships between incidence rates of malaria and SPR (Figure 3). The plot reveals that incidence rates of malaria were positively associated with SPR.

Spearman correlations between malaria incidence and socio-environment variables show (Table 1) that SPR (r = 0.85, p < 0.01), income (r = −0.76, p < 0.01) and humidity (r = 0.57, p < 0.05) were statistically significantly associated with malaria incidence. However, there was no significant association between other climatic variables and annual malaria incidence.

Six models have been used to evaluate the association between malaria incidence and predictors (Table 2). Model 1 shows SPR (β = 1.244, p = 0.000) alone can explain as high as 85% of the variation in the response variable. This provides strong evidence that SPR is a very good surrogate measure for the malaria incidence rates. Models 2–6 show that the inclusion of the additional covariates of Tmax, income and humidity moderately improved the model fit with the increase of adjusted R² (88-95%) and DW value (0.57-2.11), and decrease of AIC (75.93-65.62) in these models. Model 6 (R²: 95%, AIC: 65.62) is chosen as the optimal model due to its best goodness-of-fit of the data. In summary, the best fitting model includes SPR, income, maximum temperature and humidity as the predicting variables for the annual malaria incidence.

Table 3 displays crude and adjusted results from linear regression analyses. In crude models, four predictors were tested individually. Their adjusted R² were −3.2% (Tmax), 45% (humidity), 47% (income) and 85% (SPR), respectively. In the multi-variable models, without SPR, only 54% of variation of malaria incidence was accounted for by the other three independent predictors (Tmax, humidity and income), whereas 95% of variation of the malaria incidence was explained after SPR was added to the model.

Table 3 also shows that SPR (β = 1.326, p < 0.001) is a significantly independent predictor of malaria incidence after adjustment for Tmax, humidity and income. Keeping other independent variables constant, every 1% increase in SPR corresponds to an increase of 1.76/100,000 (the squared malaria incidence) in malaria incidence rates.

Figure 4 shows the results of the regressive forecast chart in which Figure 4-A included SPR and Figure 4-B did not. Figure 4-A indicates that the predicted and the observed value of annual squared root malaria incidence rates matched well. The incidence rates in 1999, and in 2003, were theoretically predicted by the model and validated by the observed values. However, for Figure 4-B the predicted and the observed value of the malaria incidence cannot be matched well, especially in year 1997–98. The observed values of these two years are out of the confidence interval. There is a wider confidence interval in Figure 4-B than in 4-A. All results stated that the regressive forecast of annual malaria incidence with SPR is more accurate than that without SPR in Mengla County over the study period.

The results of this study indicate that SPR is a strong predictor of malaria incidence. SPR varied between 5.48% and 13.98% from 1993 to 2004 in Mengla. SPR under 2.9% is considered the absence of indigenous transmission [9]. Evidently, there is indigenous malaria transmission in Mengla [2, 20]. Less than 5% of SPR is considered the transition from the control stage to the pre-elimination stage [22] which implied that Mengla went through pre-elimination malaria after 2004. Five Anophiline species have been identified to be vectors of malaria in Yunnan Province [2, 23]. Anopheles minimus is the major vector in this endemic border area - Mengla [2, 20].

SPR has been used as a surrogate of malaria incidence [7, 9, 12]. In Ugandan, SPR provided a useful measure to estimate malaria incidence among children [7]. To measure malaria transmission at a pre-elimination stage, SPR was used as an indicator to evaluate a malaria control programme on the island of Principe [12]. In current study, the decrease in SPR corresponded to the malaria incidence decline. This result is consistent with the result of other studies in which changes of SPR provided an alternative method for estimating changes in the incidence of malaria [7, 12]. A downward trend in SPR in Mengla is in accordance with the decline of both P.vivax and P. falciparum malaria incidence in Yunnan [24]. After 2005, both SPR and malaria incidence sharply decreased in Mengla. This may be due to the impact of the Mekong Roll Back Malaria program (2002–2004) and the Global Fund (Round one) between 2003 and 2008, especially with the free treatment for malaria infection financed by the Global Fund in Mengla County since 2005.

Malaria transmission is greatly affected by socioeconomic conditions [25, 26]. Low-middle income was significantly associated with malaria transmission in Indonesia [27]. The disappearance of malaria in some areas of Europe was associated with economic development [28]. In this study, income was negatively associated with malaria incidence. The decrease in malaria incidence was consistent with the increase in income. This can be explained by the development of the general economy in Mengla County in the last two decades. Mengla is a poor region. Twenty-six ethnic minority groups accounted for 72% of the total population, and approximately 96% of Mengla is mountainous. The main income is from rice, rubber, cane sugar and tea (it is the place of origin of Pu Er tea) [29]. The local economy has been improved since the 1990s by planting rubber trees, tropical fruit trees, tea, and an increase in trade with Laos, Myanmar and other Mekong-river region countries. The incomes of farmers have gradually increased, which has led to better living conditions and improvements in sanitation and health. These improved socio-economic conditions may be one of the key reasons for the decreased malaria pattern in this region. Further investigation of the association between socioeconomic conditions and malaria transmission is warranted in this endemic area.

In this study, relative humidity has a significant positive association with malaria incidence. Relative humidity appears to have an effect on malaria transmission indirectly, as humidity may affect the development of the parasite, and the activity and survival of anopheline mosquitoes [27]. More humid and hotter than usual conditions may increase anopheline survival, thus resulting in an increase in outbreaks of malaria [30]. However, low humidity could reduce the numbers of the mature mosquitoes [31], therefore resulting in no malaria transmission [27]. As a tropical rain forest area situated just south of the Tropic of Cancer, Mengla has wet and hot weather, which provides mosquitoes with favourable breeding sites.

Temperature has an important effect on the transmission cycle of the malaria parasite and mosquito survival [25, 32]. Temperature is considered to play a crucial role in malaria transmission, which was identified by other studies [24, 33‐36] and is reported to be a predictor of malaria transmission. In a previous study in 2008, a positive association between minimum temperature, maximum temperature and malaria incidence based on monthly time series data was found in Mengla County [21]. However, the association between temperature and malaria incidence was not observed in current study. This may because annual weather variables are used for analysis. In multiple linear regression analysis, however, maximum temperature became a significant predictor of malaria transmission after adjustment for other factors. Maximum temperature and another three predictor factors together explained 95% of variance of malaria incidence.

Malaria transmission is influenced by various factors including climatic [37, 38] and non-climatic factors [39‐41]. The spatial and seasonal distribution of malaria is largely determined by climate [42], and climatic factors (e.g. rainfall, temperature and humidity) have been widely used and recognized in the MEWS [43, 44]. However, climatic factors are not enough for MEWS, which requires comprehensive and integrated indicators. To predict the timing and severity of malaria epidemics in MEWS, epidemiological surveillance indicators (for example SPR) should be considered [45, 46]. Blood examination of parasite appearance is a key indicator in the early detection of malaria transmission and it is compulsorily reported in China. The use of SPR can obviously assist the development of MEWS in malaria elimination program in China. It can also be used to evaluate the malaria surveillance systems in China.

This study has several strengths. Firstly, this is the first study to examine the role of SPR in monitoring malaria transmission at a county level in China. Secondly, the model developed fitted the data quite well. The SPR alone, and combination with other predictors can explain 85% and 95% of variation in malaria transmission, respectively. Finally, the results of this study may help plan and implement malaria control and prevention interventions in the field.

The limitations of this study should also be acknowledged. Firstly, SPR and income were collected annually, and we were unable to examine the seasonal pattern of malaria transmission and conduct any finer analyses (e.g. monthly). Secondly, some other factors (e.g. mosquito density, movement of the people across the border and vegetation coverage) may play a role in the transmission of malaria. Because of the lack of the data, these factors were not adjusted for in the model. Finally, the data for this study were only collected from Mengla, Yunnan Province. There was no detailed information on P. vivax and P. falciparum, and the recurrence of P. vivax infection was also not considered, hence they could not be analysed separately. Thus, caution is needed when the findings of this study are generalized to other locations.

In conclusion, SPR was significantly associated with malaria incidence and identified as a strong predictor of malaria transmission in Mengla County. The results of this study support the use of SPR. The multi-variable regression model developed in this study may have implications for the global malaria elimination campaign. The improved understanding of the relationship between SPR and malaria transmission will assist in the establishment of a malaria early warning system to predict this wide spread disease in endemic areas.