1. Epidemiology and Global Health
Download icon

Spatio-temporal associations between deforestation and malaria incidence in Lao PDR

  1. Francois Rerolle  Is a corresponding author
  2. Emily Dantzer
  3. Andrew A Lover
  4. John M Marshall
  5. Bouasy Hongvanthong
  6. Hugh JW Sturrock
  7. Adam Bennett
  1. Malaria Elimination Initiative, The Global Health Group, University of California, San Francisco, United States
  2. Department of Epidemiology and Biostatistics, University of California, San Francisco, United States
  3. Department of Biostatistics and Epidemiology, School of Public Health and Health Sciences, University of Massachusetts Amherst, United States
  4. Division of Epidemiology and Biostatistics, School of Public Health, University of California, Berkeley, United States
  5. Center for Malariology, Parasitology and Entomology, Ministry of Health, Lao People's Democratic Republic
Research Article
Cite this article as: eLife 2021;10:e56974 doi: 10.7554/eLife.56974
25 figures, 7 tables and 5 additional files

Figures

Average tree crown cover (%) in 2016 (left) and percent area that experienced forest loss between 2011 and 2016 (right) within a 10 km radius in northern (top) and southern (bottom) Lao PDR.

See Materials and methods for details on forest and deforestation metrics. Upper right indent maps northern and southern (Champasak province) Lao PDR regions.

Malaria incidence (per 1000) and test positivity (%) over time.

Upper left boxed indent zooms in malaria incidence in the North to better show the temporal variation (see y axis for scale).

Associations between malaria incidence and a 0.1% increase in the area that experienced deforestation within 1, 10, or 30 km (left-right) of a village in the previous 1–5 years (top-down) in Lao PDR.

Adjusted for the probability of seeking treatment, the spatio-temporal structure of the data, the environmental covariates selected in the model and forest cover within 30 km in the year before the deforestation temporal scale considered as well as for malaria incidence in the previous 1 and 2 years. See Materials and methods for details.

Associations between malaria incidence and a 0.1% increase in the area that experienced deforestation within 1, 10, or 30 km (left-right) of a village in the previous 1–5 years (top-down) in southern Lao PDR, differentiated by malaria species.

Adjusted for the probability of seeking treatment, the spatio-temporal structure of the data, the environmental covariates selected in the model and forest cover within 30 km in the year before the deforestation temporal scale considered as well as for malaria incidence in the previous 1 and 2 years. See Materials and methods for details.

Associations between malaria incidence and a 0.1% increase in the area that experienced deforestation within 30 km of a village in the previous 1–5 years (top-down) and within areas with tree crown cover density above 0%, 68%, and 87% (left-right) in Lao PDR.

Adjusted for the probability of seeking treatment, the spatio-temporal structure of the data, the environmental covariates selected in the model and forest cover within 30 km in the year before the deforestation temporal scale considered as well as for malaria incidence in the previous 1 and 2 years. See Materials and methods for details.

Map of study’s districts.
For every 30 m Landsat pixel within a buffer radius r (1, 10, and 30 km) of study’s villages, the tree crown cover density in 2000 and the year of forest loss were combined to derive the deforestation and forest cover variables.

The two upper plots highlight the raw data at two example pixels from the lower plot.

Conceptual model for our analysis showing how the raw input data (blue boxes) were combined via intermediate data (white boxes) and models (white diamonds) to produce our estimated outputs (red circle).
Appendix 1—figure 1
Treatment-seeking modeling plots. Note that treatment-seeking at public health facilities is implied all along the manuscript.
Appendix 1—figure 2
Residual temporal autocorrelation when malaria incidence in previous 1 and 2 months are included or not.
Appendix 1—figure 3
Relationships between malaria incidence and the environmental covariates in the multivariable model described in equation 2 (30 km radius and 1-year temporal lag), additionally adjusted for the probability of seeking treatment, the spatio-temporal structure of the data (f(t), f(Lat,Long) and village random intercepts) and malaria incidence in the previous 1 and 2 months.

See Materials and methods for details. Note that 95% confidence intervals (see Appendix 1—figure 13) have been hidden for better visualization.

Appendix 1—figure 4
Relationships between malaria incidence and the temporal trend in the multivariable model described in equation 2 (30 km radius and - year temporal lag), additionally adjusted for the probability of seeking treatment, the spatial structure of the data (f(Lat,Long) and village random intercepts) and malaria incidence in the previous 1 and 2 months.

See Materials and methods for details.

Appendix 1—figure 5
Adjusted relationship between deforestation and malaria incidence.

All models were adjusted for environmental covariates and forest cover on top of the probability of seeking treatment, the spatio-temporal structure of the data (f(t), f(Lat,Long) and village random intercepts) and malaria incidence in the previous 1 and 2 months. See Materials and methods for details. Note that scales are different between buffer radius for better visualization. Appendix 1—figure 14 shows the raw scatterplot between monthly village malaria incidence rate and deforestation. Appendix 1—figure 15 and Appendix 1—figure 16 show the raw time series of malaria incidence, forest cover and percent area that experienced forest loss.

Appendix 1—figure 6
Distribution of average tree crown cover density within 1, 10, and 30 km of villages.
Appendix 1—figure 7
Distribution of percent area within 1, 10, and 30 km of villages that experienced forest loss between 2011 and 2016.

Note that the scales are different for every panel for better visualization of the distributions.

Appendix 1—figure 8
Distribution and time series of environmental covariates (population, altitude, monthly day temperature and monthly total precipitation) at study’s villages.
Appendix 1—figure 9
Additional figures from malaria registries: malaria infections.
Appendix 1—figure 10
Distributions of socio-economomical variables of all patients recorded in the malaria registries.
Appendix 1—figure 11
Additional figures from malaria registries: matched vs unmatched SES variables.
Appendix 1—figure 12
Distribution of travel time (in hours) from surveyed households to closest health facilities.
Appendix 1—figure 13
Relationships between malaria incidence and the environmental covariates in the multivariable model described in equation 2 (30 km radius and 1-year temporal lag), additionally adjusted for the probability of seeking treatment, the spatio-temporal structure of the data (f(t), f(Lat,Long) and village random intercepts) and malaria incidence in the previous 1 and 2 months.

Dashed lines are for 95% confidence intervals. Note that the y scale has been trimmed a bit for better visualization.

Appendix 1—figure 14
Raw scatterplot between monthly village malaria incidence rate and the percent area within 30 km of villages that experienced forest loss in the previous 1, 3, and 5 years.

Note that scales are different between regions for better visualization.

Appendix 1—figure 15
Time series of deforestation (percent area that experienced forest loss around villages), forest cover (average tree crow cover around villages) and malaria incidence, averaged over study’s villages and for varying buffer radius around villages (1, 10, and 30 km).
Appendix 1—figure 16
Time series of deforestation (percent area that experienced forest loss within 30 km of villages) and forest cover (average tree crow cover within 30 km of villages), for a few randomly sampled study’s villages.

Each color represents one village.

Appendix 1—figure 17
Adjusted relationship between deforestation and species-specific malaria incidence in southern Lao PDR.

All models were adjusted for environmental covariates and forest cover on top of the probability of seeking treatment, the spatio-temporal structure of the data (f(t), f(Lat,Long) and village random intercepts) and malaria incidence in the previous 1 and 2 months.

Tables

Table 1
IRR between malaria incidence and a 0.1% increase in the area that experienced deforestation within 1, 10, or 30 km (left-right) of a village in the previous 1– 5 years (top-down) in northern and southern Lao PDR.

Adjusted for the probability of seeking treatment, the spatio-temporal structure of the data, the environmental covariates selected in the model and forest cover within 30 km in the year before the deforestation temporal scale considered as well as for malaria incidence in the previous 1 and 2 years. See Materials and methods for details.

SouthNorth
Time lagBuffer radiusBuffer radius
1 km10 km30 km1 km10 km30 km
Previous11.011.1611.031.01
1 year[0.99; 1.01][0.99; 1.04][1.10; 1.22][1; 1.01][0.99; 1.06][0.94; 1.08]
Previous111.0811.010.99
2 years[0.99; 1.01][0.98; 1.01][1.04; 1.13][1; 1.01][0.99; 1.04][0.95; 1.03]
Previous0.990.980.9311.010.96
3 years[0.99; 1][0.97; 1][0.90; 0.97][1; 1.01][0.99; 1.02][0.94; 0.99]
Previous0.990.980.94110.97
4 years[0.99; 1][0.97; 0.99][0.92; 0.97][1; 1.01][0.99; 1.02][0.94; 0.99]
Previous10.970.9411.010.96
5 years[0.99; 1][0.96; 0.99][0.91; 0.97][1; 1.01][0.99; 1.02][0.93; 0.98]
Table 2
IRR between malaria incidence and a 0.1% increase in the area that experienced deforestation within 1, 10, or 30 km (left-right) of a village in the previous 1–5 years (top-down) in southern Lao PDR, differentiated by malaria species.

Adjusted for the probability of seeking treatment, the spatio-temporal structure of the data, the environmental covariates selected in the model and forest cover within 30 km in the year before the deforestation temporal scale considered as well as for malaria incidence in the previous 1 and 2 years. See Materials and methods for details.

P. falciparumP. vivax
Time lagBuffer radiusBuffer radius
1 km10 km30 km1 km10 km30 km
Previous11.041.27111.07
1 year[0.99; 1.02][1.01; 1.07][1.18; 1.36][0.99; 1.01][0.97; 1.02][1.01; 1.13]
Previous11.011.15111.06
2 years[0.99; 1.01][0.99; 1.03][1.08; 1.22][0.99; 1.01][0.98; 1.01][1.01; 1.11]
Previous0.990.990.8510.991.02
3 years[0.98; 1][0.97; 1.01][0.80; 0.90][0.99; 1.01][0.98; 1.01][0.97; 1.06]
Previous0.990.980.8510.991.01
4 years[0.98; 1][0.96; 0.99][0.81; 0.88][0.99; 1][0.98; 1.01][0.98; 1.04]
Previous10.970.8310.991.01
5 years[0.99; 1][0.95; 0.98][0.80; 0.87][1; 1.01][0.98; 1][0.98; 1.04]
Table 3
IRR between malaria incidence and a 0.1% increase in the area that experienced deforestation within 30 km of a village in the previous 1–5 years (top-down) and within areas with tree crown cover density above 0%, 68%, and 87% (left-right) in Lao PDR.

Adjusted for the probability of seeking treatment, the spatio-temporal structure of the data, the environmental covariates selected in the model and forest cover within 30 km in the year before the deforestation temporal scale considered as well as for malaria incidence in the previous 1 and 2 years. See Materials and methods for details.

SouthNorth
Deforestation within areasDeforestation within areas
with tree crown cover density abovewith tree crown cover density above
Time lag0%68%87%0%68%87%
Previous1.161.321.281.011.041.34
1 year[1.10; 1.22][1.14; 1.53][1; 1.64][0.94; 1.08][0.96; 1.14][0.99; 1.81]
Previous1.081.181.350.990.980.94
2 years[1.04; 1.13][1.08; 1.28][1.15; 1.59][0.95; 1.09][0.93; 1.03][0.80; 1.11]
Previous0.930.930.890.960.960.86
3 years[0.90; 0.97][0.89; 0.97][0.81; 0.99][0.94; 0.99][0.92; 0.99][0.76; 0.98]
Previous0.940.940.870.970.960.87
4 years[0.92; 0.97][0.91; 0.97][0.80; 0.94][0.94; 0.99][0.93; 0.99][0.78; 0.96]
Previous0.940.930.830.960.950.83
5 years[0.91; 0.97][0.90; 0.96][0.76; 0.90][0.93; 0.98][0.92; 0.98][0.75; 0.92]
Appendix 1—table 1
Data used to parameterized the transition matrix with the travel speed between any two adjacent pixels of the map.
Data layerCategorySpeed (km/h)
Digital elevation (slope)0°(flat)5
5°(uphill)3.71
−5°(downhill)5.27
Land coverCroplandNo adjustment
Artificial and bare areasNo adjustment
Open deciduous forest0.8 * Hiking speed
Sparse herbaceous0.8 * Hiking speed
Closed deciduous forest0.6 * Hiking speed
Herbaceous0.6 * Hiking speed
Flooded0.5 * Hiking speed
Other forest cover0.4 * Hiking speed
Water0.2 * Hiking speed
Roads and riversMotorway/trunk80
Primary/secondary60
Tertiary/unclassified10
Major rivers5
Appendix 1—table 2
IRR associated with a 0.1% increase in forest loss.

Adjusted for the spatio-temporal structure of the data, the environmental covariates selected in the model and forest cover within 30 km in the year before the deforestation temporal scale considered and malaria incidence in the previous 1 and 2 months. See Materials and methods for details. Sensitivity analysis: village population unadjusted for probability of seeking treatment.

SouthNorth
Time lagBuffer radiusBuffer radius
1 km10 km30 km1 km10 km30 km
Previous11.011.1611.031.01
1 year[0.99; 1.01][0.99; 1.04][1.10; 1.22][1; 1.01][1; 1.07][0.94; 1.08]
Previous111.0911.010.98
2 years[0.99; 1.01][0.98; 1.01][1.04; 1.13][1; 1.01][0.99; 1.04][0.94; 1.01]
Previous0.990.980.9311.010.96
3 years[0.99; 1][0.97; 1][0.90; 0.97][1; 1.01][0.99; 1.02][0.93; 0.99]
Previous0.990.980.94110.97
4 years[0.99; 1][0.97; 0.99][0.92; 0.97][1; 1.01][0.99; 1.02][0.94; 0.99]
Previous10.970.9411.010.95
5 years[0.99; 1][0.96; 0.99][0.91; 0.97][1; 1.01][0.99; 1.02][0.93; 0.98]
Appendix 1—table 3
IRR [95% CI] associated with a 1% increase in average tree crown density.

Adjusted for the probability of seeking treatment, the spatio-temporal structure of the data, the environmental covariates selected in the model and malaria incidence in the previous 1 and 2 months. See Materials and methods for details.

SouthNorth
Time lagBuffer radiusBuffer radius
1 km10 km30 km1 km10 km30 km
Current11.071.060.991.011.10
year[0.99; 1.01][1.04; 1.10][1; 1.12][0.97; 1.02][0.96; 1.05][0.99; 1.23]
Previous11.071.0911.011.12
1 year[0.99; 1.02][1.05; 1.10][1.03; 1.15][0.97; 1.02][0.97; 1.06][0.99; 1.26]
Previous11.071.0911.021.10
2 years[0.99; 1.02][1.05; 1.10][1.03; 1.16][0.98; 1.03][0.97; 1.06][0.98; 1.25]
Previous11.071.101.011.021.10
3 years[0.99; 1.02][1.04; 1.10][1.04; 1.16][0.98; 1.03][0.97; 1.07][0.98; 1.24]
Appendix 1—table 4
AIC fit of univariate models when including each of the seven monthly climatic variation one at a time as unique covariate in equation 2, solely adjusted for the probability of seeking treatment, the spatio-temporal structure of the data (f(t), f(Lat,Long) and village random intercepts).

AIC selected are in bold.

Outcome model
SouthNorthSouth P. falciparumSouth P. vivax
Day temperature
Current month18,546167113,22614,575
Previous month18,556170213,22414,590
2 months ago18,578166913,24914,594
3 months ago18,559167213,23214,593
Over current and previous month18,556167013,23114,583
Over current and previous 2 months18,570167013,24814,588
Over current and previous 3 months18,573168013,24914,592
Night temperature
Current month18,413166913,12014,474
Previous month18,453167013,15514,520
2 months ago18,547167313,23114,576
3 months ago18,581167213,25114,596
Over current and previous month18,296166413,04414,397
Over current and previous 2 months18,263166913,01414,385
Over current and previous 3 months18,262166313,00714,385
Precipitation
Current month18,532169313,19814,593
Previous month18,520166913,18114,575
2 months ago18,538165813,20714,594
3 months ago18,579166413,24314,596
Over current and previous month18,570167213,23914,596
Over current and previous 2 months18,543167013,18714,590
Over current and previous 3 months18,555167413,21214,591

Additional files

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)