Northern Mozambique has never experienced a storm as powerful and potentially damaging as Cyclone Kenneth.
Just six weeks after Cyclone Idai devastated the country and sparked a cholera outbreak that has so far resulted in over 6,300 cases and 8 deaths – despite a successful mass vaccination campaign that reached 900,000 people in four districts – Cyclone Kenneth threatens a repeat of those events in an area that, in many ways, is just as large, complex and vulnerable to disaster.
Cholera has not yet emerged as an active threat in the area likely to be most affected by Cyclone Kenneth. However, particularly in the wake of recent events in Beira after Cyclone Idai, it is clearly a risk. Flood waters and high winds lead to the loss of homes, displacement of population, damaged health infrastructure and, perhaps most seriously from the standpoint of waterborne diseases, damaged sanitation systems which can quickly spread the bacteria which cause cholera.
Can we already know enough precisely about the factors which may lead to cholera outbreaks, including the combination of those factors which may specifically exist in northern Mozambique, in order to focus attention, prioritize key geographic areas and possibly begin planning now for the scale, form and location of health emergency response which may be required?
Modeling Cholera Risk for the Aftermath of Cyclone Kenneth
To help identify the areas in Mozambique at greatest risk from the storm and its aftermath, a team of researchers led by Dr. Caroline Buckee, Dr. Ayesha Mahmud and Rebecca Kahn from Harvard University School of Public Health, in collaboration with Direct Relief, Nethope Crisis Informatics, Facebook Data for Good and Jen Chan from Northwestern University School of Medicine, developed an initial model-based estimation of likely cholera in the region.1,2
The model highlights several key areas primarily in the Cabo Delgado province, as well as areas of Tete, Zambezia and Sofala provinces, at the highest risk for potential cholera outbreaks following Cyclone Kenneth. Collectively, the five districts considered by the results of this model to be most at risk represent a population of nearly 1.2 million people.
Key factors included in this spatial disease-risk model include the previous cholera incidence for this area, estimated severity of flood impacts, and the likelihood of increased cholera incidence during El Nino years. Previous cholera incidence was based on modeled estimates derived from cholera outbreak data and ecological data from Lessler et al. on cholera hotspot detection for Africa, published in The Lancet in 2018.3 Flooding impact estimates were based on the most recent weather information available, with the highest severity in the northernmost districts. Sensitivity to the effects of El Nino was based on work from Moore et al published by the Proceedings of the National Academy of Sciences (PNAS) in 2017.4
Details of the Cholera Risk Model
Risk scores for each variable were scaled between 0 and 1, and maps were produced which show both the averaged effect for all variables and the individual impact of each variable in isolation. For Cyclone Kenneth, the projected overall cholera risk is an average of the flooding index, El Niño sensitivity index and previous cholera incidence.
Additional model outputs have also been produced for the Beira area in central Mozambique which is the location of the current cholera outbreak. In addition to the factors which have been identified for the Cyclone Kenneth area, the Beira model includes a “gravity model” which estimates the likelihood of population movement from the area where most infected individuals are located out to areas where those individuals may travel. The gravity (diffusion) model assumes that travel from Beira occurs based on the population size of Beira, the population size of the receiving district and the geodesic distance between Beira and the receiving district according to the formula:
The goal of this additional model output for Beira is to determine the likelihood that despite what appears to be the containment of the current outbreak, the disease may move along with travelers who leave that region and arrive elsewhere in the country. The “gravity model” simulates human movement, in the absence of detailed mobility data, and has been used previously in epidemiological models (for example, in Xia et al in the American Naturalist in 2004).5 High resolution population data was deployed from Facebook.6
Focusing Attention and Planning on Emerging Health Risks
As additional flooding and cholera case data becomes available, in the event of actual cholera outbreaks, these models can and will be updated to reflect changing predictions based on new information and new circumstances.
Direct Relief, Nethope and colleague organizations involved in health emergency response activities in Mozambique and other potentially affected countries will be continuing to pay close attention to any signs that communities may be seeing outbreaks of cholera in the days and weeks to come, as southern Africa copes with what is already the most serious sequence of storm-based disaster impacts for this region in recorded history. The modeling work performed by our colleagues at Harvard is an extraordinarily valuable guide to the risks which may still lie in the future. Additional modeling analysis is also in the works for areas of southern Malawi which have already been affected by Cyclone Idai and which lie in the inland path of Cyclone Kenneth.
The data, code and methodology which drives these models will be posted in the coming days to GitHub so that other researchers and interested parties may use these models, reproduce their results, and help us to improve our collective focus on and response to the enormous set of health risks faced by communities in Mozambique and elsewhere.
- The R-Shiny web application containing output of the cholera risk model
- Cyclone Kenneth Cholera Modeled Maps
- Lessler et al, “Mapping the Burden of Cholera in Sub-Saharan Africa and Implications for Control: An Analysis of Data Across Geographic Scales,” The Lancet, vol. 31, issue 10133, May 12, 2018
- Moore et al, “El Nino and the Shifting Geography of Cholera in Africa,” Proceedings of the National Academy of Sciences of the United States of America, 114 (17) 4436-4441, April 10, 2017
- Xia et al, “Measles Metapopulation Dynamics: A Gravity Model for Epidemiological Coupling and Dynamics,” The American Naturalist, vol. 164, no. 2, August 2004
- High-resolution population data set from Facebook
- GitHub Repository: Cholera and Mozambique