Malaria in Southern Venezuela: The hottest hotspot in Latin America

Malaria elimination in Latin America is becoming an elusive goal. Malaria cases reached a historical ~1 million in 2017 and 2018, with Venezuela contributing 53% and 51% of those cases, respectively. Historically, malaria incidence in southern Venezuela has accounted for most of the country's total number of cases. The efficient deployment of disease prevention measures and prediction of disease spread to new regions requires an in-depth understanding of spatial heterogeneity on malaria transmission dynamics. Herein, we characterized the spatial epidemiology of malaria in southern Venezuela from 2007 through 2017 and described the extent to which malaria distribution has changed country-wide over the recent years. We found that disease transmission was focal and more prevalent in the southeast region of southern Venezuela where two persistent hotspots of Plasmodium vivax (76%) and P. falciparum (18%) accounted for ~60% of the total number of cases. Such hotspots are linked to deforestation as a consequence of illegal gold mining activities. Incidence has increased nearly tenfold over the last decade, showing an explosive epidemic growth due to a significant lack of disease control programs. Our findings highlight the importance of spatially oriented interventions to contain the ongoing malaria epidemic in Venezuela. This work also provides baseline epidemiological data to assess cross-border malaria dynamics and advocates for innovative control efforts in the Latin American region.

For example, 1) "We suggest that a source-sink pattern of Plasmodium sp. dispersal account for the reemergence and progression of malaria transmission in the last 4 years across the country due to the internal migration of infected people to and from the hotspots and other malaria-prone ecosystems." The authors do not present any data on migration to support this. As a result, the depicted spatial patterns could be a result just of the collapse of the health system with a large increase in malaria incidence in areas more suitable for malaria and its vectors. 2) "We observe a similar pattern explaining the spillover of cases across international borders affecting neighboring countries." While the authors present some data from the Brazilian Ministry of Health, this is not the primary data that are the focus of their analysis. 3) "This study provides baseline epidemiological data and guidance for malaria control to further assess the dynamics of cross-border malaria, the role of asymptomatic carriers, drug-resistant evolution". There is little in this article about the role of asymptomatic carriers and drug-resistant evolution.

R0 calculations
Lines 111-120: As the authors point out in the discussion, there are a number of limitations associated with their calculation of R0. Therefore, it is not entirely clear to me that adding the R0 results contribute significantly to this article

Continuous surfaces
Line 138. "Finally, we used kriging, a local geostatistical interpolation method [29], to generate an estimated continuous surface from the scattered set of points (i.e., municipality centroids) to better capture the local spatial variation of malaria spread across the country during the 2014-2017 period". There are much better methods to estimate this continuous surface, such as GAMs (generalized additive models) and geospatial regression models.

Chi-square test
Lines 193-201: It would be useful to create a figure displaying the proportion of people in each occupation that were tested for malaria that obtained a positive test outcome (e.g., (#miners diagnosed with malaria)/(#miners tested)). The same could be done for age groups and gender. In my perspective, this would be more informative than a chi-square test because, given the sample size, this test is likely to detect very small deviations from the null hypothesis. These deviations, even if statistically significant, may or may not be epidemiologically significant.

Increase in population vs mining activities
The significance of the results in Fig. 7 depends on the size of the population at each locality.
Can the clustering of localities > 1000 malaria cases be just an artifact of greater population in these areas? Fig. 8. Perhaps the results shown here are again just an artifact of population size. In other words, tree cover lost (%) might just be a proxy for population density. In this case, perhaps it would be more accurate to state that: "Increase in malaria cases is associated with increased population in these malaria prone regions, most of which were attracted by the gold mining activities"

Forest loss vs mining
Line 133. "Additionally, to identify illegal mining operations in the study area, we obtained data on the deforested land cover distribution areas in the parishes classified as hotspots across the study period via Global Forest Watch". Is illegal mining always the reason for deforestation in this region? This is an implicit assumption in this article that may or may not be true.

Other details
Lines 62-72: It would be nice if the map showed the names of the municipalities. That would allow the reader to follow the description of where is the state capital, where is the main urban center, etc. etc.
Line 119 "then estimated R0 as the calculus of the rate of". What do you mean by calculus? Derivative?
Line 129-130. "from the surveyed population for malaria in Bolivar state". Do you mean the people that were tested for malaria? I am assuming that these data arise from health facilities. If that is the case, I think that it is misleading to talk about these data as coming from a "survey".
Line 36 Clarify what you mean by "Route 10". For example, replace "(Route 10)" by "(e.g., through the road to Brazil called Route 10)". Fig. 1 It is hard to see the river (it should be blue in the map but it is not) and to see route 10 (make line thicker) Line 156. "sharp (exponential) and significant increase since 2014 (R2 = 0.92, P < 0.05)." Not sure what type of model was used here and that generate these results Line 158: confusing phrase "Bolívar contributed about half (~47%) of the total cases in Venezuela during 2017, but this proportion represented from 60 to 80% in the previous years," Fig. 3 It is not possible to see the name of the parishes. As a result, it is hard to follow the description of the results in lines 169-175 "spatial heterogeneity was also detected for P. falciparum" Do you mean "there was also substantial spatial heterogeneity for P. falciparum"?
Supplementary Fig. 1: this is not a useful figure to me because it is influenced by the number of individuals (regardless of malaria status) in each age group. For example, if more people from the 21-30 age group are tested, it is more likely that we will see a greater number of cases in this age group, even if incidence per 1,000 in each age groups.
Line 203: what do you mean by "localities"? My understanding was that data were aggregated at the parish level. Are the data available at an even finer geographical scale?