MALARIA EPIDEMICS UNDER CLIMATE CHANGE SCENARIOS IN THAILAND - THAISCIENCE
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
J. Environ. Res. 35 (2): 1-11 J. Environ. Res. 35 (2): 1-11
Malaria epidemics under climate change scenarios in Thailand
โรคระบาดของมาลาเรียภายใต้สภาวะภูมิอากาศเปลี่ยนแปลงในประเทศไทย
Chayut Pinichka1*, Kampanad Bhaktikul1, Saranya Sucharitakul1 and Kanitta Bundhamcharoen2
1
Faculty of Environment and Resource Studies, Mahidol University, Nakhon Pathom 73170 Thailand.
2
International Health Policy Program, Ministry of Public Health, Nonthaburi 11000, Thailand
ชยุตม์ พินิจค้า1* กัมปนาท ภักดีกุล1 ศรัณยา สุจริตกุล1 และ กนิษฐา บุญธรรมเจริญ2
1
คณะสิ่งแวดล้อมและทรัพยากรศาสตร์ มหาวิทยาลัยมหิดล, นครปฐม 73170, ประเทศไทย
2
สำ�นักงานพัฒนานโยบายสุขภาพระหว่างประเทศ กระทรวงสาธารณสุข, นนทบุรี 11000, ประเทศไทย
received : February 14, 2013 accepted : April 5, 2013
Abstract B2 = 1,301 DALYs/yr. The compared model with
The objective of this study was to estimate actual climate data to predict the incidence of
avoidable burden on disease of malaria in Thailand malaria in 2012-2020 found malaria incidence has
under climate conditions in the future. The study increased the incidence with trend line equation
was based on climate projection under 2 different Y = 312.55X + 2480.1, R2 = 0.74 average incidences
situations which included the regionally economic 79,703 persons/yr or 4,042.9 DALYs/yr. The scenario
development (A2) and the local environmental B2 has been decreased incidence of malaria with
sustainability (B2). 1991-2011 climate data collection trend line equation Y = 20.223X3 – 363X2 + 1801.4X
was used to create nonlinear mixed regression – 19.483, R2 = 0.57, Average incidence 40,407
model. The variables in monthly time step, which persons/ yr, or 2,042.8 DALYs/yrs. Scenarios B2
included maximum temperature, minimum could have been avoided by A2 = 1,119.5 DALYs/yrs
temperature, precipitation, humidity, average wind or 49.3 %.
speed.
Keyword: Malaria, nonlinear mixed regression,
The results were found the best fitting model, climate change projection data, DALYs
model 2, which adjusted R-Square = 0.818 and
RMSE = 763.27. The average disease incidence
in the year of 2003-2011 on B2 = 26,869 persons/yr, บทคัดย่อ
baseline = 28,521 persons/yr, and A2 = 30,734 การศึกษานีม้ วี ตั ถุประสงค์เพือ่ คาดการณ์ภาระโรค
persons/yr. These burdens converted to DALYs ทีส่ ามารถหลีกเลีย่ งได้ (Avoidable burden of diseases)
for international comparison which were, baseline ของโรคมาลาเรีย ภายใต้สภาวะภูมิอากาศในอนาคต
= 1,391 DALYs/yr, A2 = 1,500 DALYs/yr, and โดยน�ำข้ อ มู ล ของสภาพภู มิ อ ากาศของประเทศไทย
*
corresponding author
E-mail : yut_emblaze@yahoo.com
Phone : +668 3996 2558
1Pinichka et al., 2013
พ.ศ.2534-2554 มาคาดการณ์ ภ ายใต้ ส ถานการณ์ worldwide today. Climate changes in our
การเปลี่ยนแปลงสภาพภูมิอากาศ (A2) และ (B2) ใน region are important impact on human
ประเทศไทยของศูนย์เครือข่ายงานวิเคราะห์วิจัยและ
health (1). Especially, malaria, an infection
การฝึกอบรมการเปลี่ยนแปลงของโลก แห่งภูมิภาคเอเชีย
ตะวันออกเฉียงใต้ (SEA START RC)ได้ใช้การน�ำเข้า disease that is sensitive to the climate. The
ข้อมูลสภาพภูมิอากาศของประเทศไทยในปี พ.ศ.2534- infection diseases that are dependent on
2554 ของกรมอุ ตุ นิ ย มวิ ท ยาจ�ำนวน 5 ตั ว แปรได้ แ ก่ several factors, but the factor of temperature
เดือน อุณหภูมิสูงสุด อุณหภูมิต�่ำสุด ค่าความชื้นสัมพันธ์ and humidity are extremely important and
ความเร็วลมเฉลี่ย และ ปริมาณนํ้าฝน เพื่อสร้างแบบ the climate can affect human behavior and
จ�ำลองถดถอยไม่เป็นเส้นตรงแบบผสมพบว่าแบบจ�ำลอง
social impact on the spread of infectious
ที่ให้ค่า ความแม่นย�ำสูงสุดคือแบบจ�ำลองที่ 2 โดยมีค่า
adjusted R-Square 0.818 ด้วยค่า RMSE 763.27 diseases as well, so a minimal climate
change impact on human tolerance levels
การศึกษาในปี พ.ศ.2546-2554 พบว่าสถานการณ์
เปลี่ยนแปลงสภาพภูมิอากาศ B2 เป็นสถานการณ์ที่มี
can have a direct impact on human health
อุ บั ติ ก ารณ์ เ กิ ด โรคมาลาเรี ย น้ อ ยที่ สุ ด กล่ า วคื อ มี อุ บั ติ immediately (2).
การณ์เกิดขึ้น 26,869 คนต่อปี หรือคิดเป็น 1,301 DALYS Malaria is an extremely climate-
ต่อปี A2 = 30,734 คนต่อปี หรือคิดเป็น 1,500 DALYS
sensitive tropical disease, making the
ต่อปี เมื่อน�ำแบบจ�ำลองถดถอยไม่เป็นเส้นตรงแบบผสม
ไปท�ำนายอุบตั กิ ารณ์เกิดโรคทีจ่ ะเกิดขึน้ ในอนาคต ปี พ.ศ. assessment of the potential change in
2555-2563 พบว่าอุบัติการณ์มาลาเรียในแบบจ�ำลอง A2 malarial risk, caused by past or projected
มีแนวโน้มทีส่ งู ขึน้ ด้วยสมการแนวโน้มคือ Y = 312.55X + global warming, one of the most important
2480.1, R2 = 0.74 มีอุบัติการณ์เฉลี่ย 79,703 คนต่อปี topics in the field of climate change and
หรือคิดเป็น 4,042.9 DALYS ต่อปี ในขณะที่แบบจ�ำลอง health (3). The incidence of malaria varies
B2 ส่งผลให้อุบัติการณ์ของมาลาเรียมีแนวโน้มที่เพิ่มขึ้น
seasonally in highly endemic areas, and
จากนั้นจึงลดลงอย่างต่อเนื่อง ด้วยรูปแบบชองสมการ
ก�ำลังสามคือ Y = 20.223X3 – 363X2 + 1801.4 X malaria transmission has been associated
–19.483, R2 = 0.57 โดยมีอุบัติการณ์ 40,407 คนต่อปี with temperature anomalies in some African
หรือคิดเป็น 2,042.8 DALYS ต่อปีซึ่งน้อยกว่า แบบ highlands (4).
จ�ำลอง A2 อยู่ 1,119.5 DALYS ต่อปี หรือคิดเป็น 49.3%
The WHO has estimated the global
ค�ำส�ำคัญ: มาลาเรีย, การถดถอยไม่เป็นเส้นตรง burden of disease (GBD) that could be due
แบบผสม, ข้อมูลคาดการณ์การเปลี่ยนแปลง to climate change in terms of disability
ภูมิอากาศ, DALYS adjusted life years (DALYS). This measure
makes it possible to take into account
Introduction impacts that do not necessarily lead to death
Climate change is an emerging risk but cause disability. Climate scenarios are
factor for human health. It is now clear derived from the output of global climate
the global climate has been changed in models that are, in turn, driven by scenarios
2J. Environ. Res. 35 (2): 1-11
of future greenhouse gas emissions and 1. Selecting the scenarios and time
epidemiological models. These scenarios period
were used to estimate the degree to which 2. Climate change modeling
these climatic changes are likely to affect a 3. Health impact model
limited series of health outcomes (malaria, 4. Conversion to a single health
diarrheal disease, malnutrition, flood deaths, measure DALY (Disability adjusted life year)
direct effects of heat and cold). These
This research follows guidelines of the
measures of proportional change can be
WHO and set the study’s purpose to create
applied to projections of the burden of each
a Statistical Climate health model of Thailand
of these diseases in the future, to calculate
to predict and compare incidences under
the possible impacts of climate change on
climate scenarios projected with the actual
the overall disease burden (5).
incidence of the disease under real climate
Prediction malaria incidence facilitates conditions and improve the results by convert
early public health responses to minimize to DALYs to enable international comparison
morbidity and mortality. Climate variables possible.
such as temperature, precipitation, relative
humidity and wind (6) are potential predictors Data Collection
of malaria incidence have been examined 1. Climate data during 1991-2011
in time series studies. In this study, we used from the Department of Meteorology were
adapted nonlinear time series analysis (7) to used including rainfall (Rainfall Intensity),
determine the association between climatic average monthly temperature (Average
variability and the number of monthly Ambient Temperature), average maximum
malaria outpatients over the past 20 years temperature (Maximum Temperature), RH
and predicted the next 10 years burden in (Relative Humidity), wind speed and time
Thailand. (month).
Methodology 2. Reports of malaria incidences in
In this research, we used WHO Thailand during 1991-2011 were collected
Environmental Burden of Disease Series, from the Bureau of Epidemiology.
climate change guidance, estimating 3. The predicted climate data used
attributable and avoidable burdens of dataset from Southeast Asia START Regional
disease method (8, 9): The main step includes Center (SEA START RC) projected. These
as following. data are daily climate data (transform to
3Pinichka et al., 2013
monthly data) under two different GHG mixed- regression technique from Zhou (6)
emission scenarios; scenarios A2 (regionally was used. This is due to the existence of
economic development) and scenarios classification on the function type such as,
B2 (local environmental sustainability). autocorrelation, climate variability function,
Our goal is to f ind the malaria burden and seasonal function (10).
difference between the environmental focus The number of malaria outpatients,
(B2) and the economic focus (A2) in the Nt, at a given time is likely to be affected by
heterogeneous world (regional development)(1). the previous number of malaria outpatients
(autoregression), seasonality, and climate
Statistical association between climate variability. Thus, the dynamics of the number
variability and malaria incidence of monthly malaria outpatients (Nt) can be
In this step, the adapted nonlinear modeled as in Equation (1).
Nt (1)
where
f (NiJ. Environ. Res. 35 (2): 1-11
And the model evaluation method the death of the life lost prematurely.
used the Root Mean Square Error (RMSE) The basic formula for YLL (without including
(also called the root mean square deviation, other social preferences), is the following
RMSD) is a frequently used measure of the for a given cause, age and sex:
difference between values predicted by a
model and the values actually observed YLL = N x L
from the environment that is being modelled.
Where;
N = number of deaths
L = standard life expectancy at age of
Where Xobs is observed values and death in years
Xmodel is modelled values at time/place i.
YLD = I x DW x L
Estimate burden of disease
Where;
Disability-Adjusted Life Year (DALYs) I = number of incidence cases
are calculated as the sum of the years of DW = disability weight
life lost (YLL) due to premature mortality L = average duration of the case until
in the population and the years lost due to remission or death (years)
disability (YLD) for incidences cases of the
health condition (11). Calculation is Results and discussion
Correlation between variables with malaria
DALY = YLL + YLD incidences
The number of years lost due to The study showed that wind speed
premature (Year of Life Lost - YLL) or has maximum lag period of 5 months with
premature death is a component of the correlation of 0.282 while humidity and rainfall
disease burden and mortality indicators. have maximum lag period of 0 month with
The measurement is based on the time of correlations of 0.297 and 0.241 respectively.
5Pinichka et al., 2013
Table 1 Correlation between variables with malaria incidences
Variable Lag period, τ (Months) Correlation Significance
Case 1 0.842 0.000 **
Maximum temperatures 0 0.046 0.469
Maximum temperatures 1 0.288 0.000**
Maximum temperatures 2 0.427 0.000**
Minimum temperatures 0 0.250 0.000**
Minimum temperatures 1 0.297 0.000**
Rainfall 0 0.241 0.000**
RH 0 0.171 0.007**
Wind speed 0 0.073 0.245
Wind speed 1 0.061 0.337
Wind speed 2 0.161 0.011*
Wind speed 3 0.251 0.000**
Wind speed 4 0.258 0.000**
Wind speed 5 0.282 0.000**
* Correlation is significant at the 0.05 level (2-tailed).
** Correlation is significant at the 0.01 level (2-tailed).
Although spatial incidences data method. The model was tested in three
in Thailand were not used, signif icant assumptions as following.
relationship is found conforming to the results - Model No EV assuming the
of Zhou (7), as well as some climate variables environment factors having no affect to the
such as; RH and Rainfall are no lagged number of patients; (g(x) = 0).
period (Table 1). However, these data are
surveillance and spatial analysis was not - Model 1 assuming the environmental
used. factors having an influence on the number
of patients given g (x) ≠ 0 and assuming
Nonlinear mixed regression analysis the interaction between all climate variables
Variables were selected and tested = 0 due to no interaction with the environment.
with time lag period climate variable - Model 2 assuming the environmental
associated with malaria incidence. The next factors having an influence on the number
step is regression analysis by stepwise of patients given g (x) ≠ 0 and assuming
6J. Environ. Res. 35 (2): 1-11
the interaction between all climate variables The results are shown in Tables 2-4 below.
≠ 0 due to environmental factors related.
Table 2 Nonlinear mixed regression analysis between time series data with malaria incidence
Model Method R R square Adjusted R Square Significance
No EV Stepwise 0.842 0.710 0.708 0.000**
1 Stepwise 0.908 0.825 0.821 0.000**
2 Stepwise 0.907 0.823 0.818 0.000**
Table 3 Model fitting results and effects of autocorrelation and seasonality (f (Ni < t,t))
Model Type α d β b1 b2
No EV 596.37 1 0.84 - -
1 - 25,790.64 1 0.90 2,218.16 1,158.79
2 - 25,404.83 1 0.91 1,421.37 1,497.25
Table 4 Model fitting results and effect of climate variables (g(x))
Parameter Model 1 Significance Model 2 Significance
Tmin (τ = 1) 511.411 0.000** - -
Tmax (τ = 2) 201.81 0.004** - -
RH (τ = 0) 107.39 0.000** - -
Wind speed (τ = 5) - 0.000** - -
Rainfall (τ = 0) - - - -
SumTmax × RH - - 3.331 0.000**
SumTmin × Wind speed - - 22.629 0.000**
SumTmin × Wind speed × RH - - -0.263 0.000**
* Correlation is significant at the 0.05 level (2-tailed).
** Correlation is significant at the 0.01 level (2-tailed).
7Pinichka et al., 2013
Table 4 shows the fitting results Model 1. In fact, the regression model in
and effect on climate variables including this study can represent and predict malaria
parameter and significant value of Model burden with reasonable accuracy when
1 and Model 2. Significance on interaction there is sufficient malaria epidemiological
effect between maximum or minimum data (non under reported province data)
monthly temperature, RH and wind speed available for input to the model. In the
with the number of malaria incidences prediction period, although there may be
are concluded. some over estimated between the actual
The results show that Model 2 has incidences and predicted, The RMSE
the best accuracy, therefore the prediction (Table 5) is relatively small when compared
process model 2 is used. Both models are with overall malaria burden (Figure 1).
adapted by non linear mixed-regression
technique; Model 2 used interaction of Table 5 Residual analysis
climate variables, Model 1 only use Model Type RMSE
climate variables at max lag period (τmax). No EV 1,258.14
As a result, Model 2 which combined the 1 767.24
interaction between climate variables
2 763.27
shows potential better prediction than
Figure 1 Comparison models 1, 2 with actual incidence
8J. Environ. Res. 35 (2): 1-11
Figure 2 Results estimate of malaria incidences under climate scenarios (monthly)
Figure 3 DALYs of Malaria in Thailand (2) 2012-2020
The compared models with actual or 4,042.9 DALYs/ yr. The scenario B2
climate data to predict the incidence of predicted would decrease incidence of
malaria in 2012-2020 found malaria malaria with trend line, Equation Y = 20.223X3
incidences would increase with trend line, – 363X2 + 1801.4 X – 19.483 R2 = 0.57,
Equation Y = 312.55X + 2480.1 R2 = 0.74, average incidence 40,407 persons/ yr, or
average incidences 79,703 persons/ yr 2,042.8 DALYs/ yr.
9Pinichka et al., 2013
Conclusions fever diseases and other climate-health
Climatic factor and seasonal pattern impact (3) would be reduced as well.
are the most direct affect on malaria Acknowledgements
transmission in Thailand. However, other
factors are also influencing malaria I wish to express my sincere thanks
epidemiology. For example, socioeconomic to Miss. Sineenat Thaiboonrod (Manchester
condition, public health service, military University), Miss. Pensiri Duangpoonmat
conflict, migration and water resources (Faculty of Medicine, Siriraj Hospital), the
management may all modulate the suitability officers of BOD Thailand (http://www.thaibod.
for malaria transmission. The changes in net/contact.html.), Mr. Watcharapong
extreme climate such as extreme rainfall, an Noimunwai, and the officers of SEA START
average temperature increase of result in RC (www.start.or.th).
a greater incidence of malaria. Each factor References
could vary with lagged period while rainfall
(1) Intergovernmental Panel on Climate Change
causing a direct effect on the incidence (IPCC). 2001. Climate Change 2001: The
of malaria has lagged period = 0 month. Scientific Basis: Contribution of Working
Therefore, quoted in the research(2) “a minimal Group I to the Third Assessment Report.
climate change impact on human tolerance Cambridge University Press, Cambridge.
levels can have a direct impact on human (2) McMichael, A., Woodruff, R., Whetton, P.,
health immediately” is reliable. Hennessy, K., Nicholls, N., Hales, S., Woodward,
A. and Kjellstrom, T. 2003. Human Health and
In fact, the national monthly malaria Climate Change in Oceania: A Risk Assessment
cases can be modelled and predicted 2002. Commonwealth of Australia.
using nonlinear mixed regression. Our (3) McMichael, A.J., Githeko, A., Akhtar, R.,
Carcavallo, R., Gubler,D., Haines, A., Kovats,
results showed the global climate change
R.S., Martens, P., Patz, J. and Sasaki, A. 2001.
B2, The local environmental sustainability Human health. In: McCarthy, J.J., et al., (Eds.),
scenarios can prevent disease burden Climate Change 2001: Impacts, Adaptation,
around 49.3% from global climate change and Vulnerability. Press Syndicate of the
as Model A2 in 2012-2020. However, University of Cambridge, Cambridge, UK, pp.
reducing greenhouse gases emission by 451– 485.
(4) Patz, J.A., Hulme, M., Rosenzweig, C., Mitchell,
international agreement or national policy
T.D., Goldberg, R.A., Githeko, A.K., Lele, S.,
can reduce or minimize amount of diseases. McMichael, A.J. and Le Sueur, D. 2002.
There are the possibilities that those others Regional warming and malaria resurgence.
mosquito-borne diseases such as dengue Nature. 420: 627-628.
10J. Environ. Res. 35 (2): 1-11
(5) Haines, A., Kovats, R.S., Campbell-Lendrum, Environmental Burden of Disease Series No.14.
D. and Corvalan, C. 2006. Climate change World Health Organization.
and human health: impacts, vulnerability and (9) Campbell-Lendrum, C. and Woodruff, R.
public health. Public Health. 120:585–596. 2006. Comparative Risk Assessment of the
(6) Service, M.W. 1978. The effect of weather on Burden of Disease from Climate Change,
mosquito activity. In: Weather and Parasitic Environ. Health Perspect. 114: 1935-1941.
Animal Disease. T. E. Gibson (Ed.) Tech. World (10) Stolwijk, A.M. 1999. Studying seasonality by
Meteor. Org. 159: 151-166. using sine and cosine functions in regression
(7) Zhou, G., Minakawa, N., Githeko, A. K. and Yan, analysis. J Epidemiol Community Health.
G. 2005. Climate variability and malaria 53(4): 235-8.
epidemics in the highlands of East Africa. (11) Murray, C.J.L. and Lopez, A.D. 1996. The
Trends in Parasitology, 21: 54-56. Global Burden of Disease. Geneva, World
(8) Campbell-Lendrum, C. and Woodruff, R. 2007. Health Organization, Harvard School of Public
Climate change: quantifying the health Health, World Bank.
impacts at national and local levels.
11You can also read
