Meta-analysis of effects of African population health initiatives

There are many ongoing African initiatives focused on local impacts on population health, with the goal of reducing mortality rates and increasing access to essential medical services.   In the past several years, a number of these initiatives have published analyses of impacts of their programs.  While these initiatives are programmatically different, they use the same metrics for measuring impacts (e.g., under-five mortality; access to treatment for fever, diarrhea, respiratory infections; vaccine coverage).  A database of published results has been constructed, and will be used by the student researcher to better understand different models for accomplishing health related development goals and identifying gaps in the evidence base.

Mentor: John Drake
Type of Project: Quantitative/Computer-based

Who is at risk of getting infected in structured contact networks?

Human immunodeficiency virus transmits through networks of people linked through a
range of contacts, including sexual contact and intravenous drug use. SARS and Foot
and Mouth Disease are spread through long-distance movements of infected people
and livestock, followed by local transmission. These outbreaks demonstrate the
important role of networks to transmission of pathogens. Networks can be quantified in
many ways, and an individual’s “importance” to the population can be described with
node “centrality” statistics. Identifying which centrality statistics indicate an individual
has high vulnerability to infection could greatly enhance surveillance and prevention.
However, pathogen transmission routes and human social networks are highly variable
in their structure. For example, sexual contact networks for HIV tend to be assorted by
race. This project will investigate our ability to predict the vulnerability of individuals to
infection when networks are structured in space or social groups. These results could
help us understand when it is worthwhile to estimate node centrality for surveillance
and prevention systems.

The student selected for this project will work closely with Paige Miller (PhD student) to
write computer code (R and python) for disease simulations on networks. The project
will be supervised by Dr. John Drake (Odum School of Ecology, Director of the Center
for the Ecology of Infectious Diseases) and Dr. Chris Whalen (College of Public Health,
Director of the Global Health Institute). This is a quantitative and simulation-based
project; we will not be collecting our own data. Long hours of learning how to code and
manage data in R will be required. An interest in mathematical modeling of infectious
diseases, ecology, and human sociology is encouraged. The student is free to tailor
this project to their own interests by focusing on specific pathogens or populations!

Mentors: Paige Miller, John Drake, Chis Whalen

Who infected whom? Creating a database of transmission trees for comparative outbreak analysis

Juliana Taube, a student at Bowdoin College, worked with Paige Miller and Dr. John Drake.

Abstract: Transmission trees contain valuable details about who infected whom in infectious disease outbreaks. We created a database with 81 published, standardized transmission trees consisting of 12 directly-transmitted pathogens (mostly viruses). We also demonstrated how the database could be used to help answer research questions in infectious disease epidemiology. First, we analyzed overall and pathogen-specific patterns between tree parameters (Rand variation in secondary infections). We found that outbreak size is nonlinearly associated with Rand the dispersion parameter, but emphasize that pathogen-specific patterns and intervention efforts may alter theoretical relationships between these variables. Second, we examined how superspreader contribution to onward transmission, either directly or through their tree descendants, varies across pathogens. Superspreaders were responsible for most cases via their descendants and the number of superspreaders varied across pathogens. Additional database exploration matched theoryabout how the proportion of superspreaders increases at intermediate levels of dispersion, an idea that should be further explored. We hope that our database will assist both theoretical and applied infectious disease epidemiology research in the future. 


1. Lloyd-Smith, JO, Schreiber, SJ, Kopp, PE, & Getz, WM (2005) “Superspreading and the effect of individual variation on disease emergence.”Nature438(7066): 355.

Taube

Age-structured model for Tuberculosis intervention planning

Kennedy Houck, a junior from Ursinus College, worked with Paige Miller in the lab of Dr. John Drake to study age-based interventions for Tuberculosis.

Abstract:  Tuberculosis (TB) represents a widespread public health concern.  The World Health Organization’s “End TB Strategy” has set the goal for global TB eradication by 2050.  Previous studies have suggested that current public health intervention strategies may not achieve this goal in many parts of the world that experience high TB incidence rates.  The goal of this project was to determine whether age-based interventions could enhance current interventions, which are currently implemented.  A standard TB model, which includes five state variables (Susceptible, Latent, Infectious, Noninfectious, and Removed), was modified to include 16 different age classes, and parameterized with previously published information for India and South Africa.  The model was run for 500 years until equilibrium was reached.  Once equilibrium was reached, 18 different interventions, all simulating faster rates of testing and treating, or shorter infectious periods, among active TB cases, were tested by calculating the rate of decrease of TB cases in each population over time.  A “baseline” scenario where the rate of treatment was held constant was compared to interventions where the infectious period was reduced by 10, 50, 70, and 90% independently for either a specific age class or overall (i.e. a “blanket strategy”).  To test the validity of model predictions, we calculated the correlation between the stable age distribution of cases at equilibrium and WHO TB prevalence data.  In general, age-targeted interventions were found to be more effective at reducing TB cases than the “blanket” strategy.  In India, targeting 15-19 year olds was predicting to result in the greatest overall decline in incidence of both latent and active TB at all intervention levels.  In South Africa, targeting 10-14 year olds was predicted to result in the greatest overall decline of latent TB at all intervention levels; however, targeting 10-14 year olds at lower intervention levels and a blanket strategy at higher intervention levels, were more effective at reducing infectious TB burden.  These results suggest that age-based interventions may complement current public health interventions by further reducing TB burden to achieve WHO eradication goals.  Future studies should utilize a more detailed model for TB dynamics to generate a more realistic prediction.

Download (PDF, 1.67MB)

Parasite Sharing in Marine and Terrestrial Mammals

Keri-Niyia Cooper, a student from Savannah State University, worked with Drs. John Drake and Andrew Park over the summer to investigate parasite sharing in mammals.

Abstract: After a literature review was performed to compile a list of parasites that infect marine mammals, we created a database of parasite-host pairings found in the articles.  We then merged this database with information gathered from the GMPD and CLC Life Cycle, after subsetting certain traits.  We used the resulting database to examine three questions: (1) Is parasite generalism greater in marine environments or terrestrial environments? (2) Is parasite sharing greater when hosts are grouped by taxonomy (cetacean/ungulate v carnivore) or habitat (marine v terrestrial)? (3) Are parasites that infect hosts of both environments drawn disproportionately from some parasite taxonomic groups? It was noted that parasite generalism is greater in terrestrial environments. Hosts have a higher chance of being infected by the same parasite if they are found in the same environment. There is a higher chance of helminth species being more commonly found in a single environment.

Download (PDF, 396KB)

Differences in age distribution patterns in urban and rural counties of São Paulo state, Brazil

Magdalene Walters, a student from the University of Notre Dame, worked with Dr. John Drake to study the age distribution of a measles outbreak.

Abstract:  In 1997, São Paulo, Brazil experienced a measles outbreak with an unusually high average age of infection. It has since been hypothesized that this high age of infection was due to unvaccinated rural adults traveling to urban communities.1 This project tested this hypothesis through the use of descriptive statistics and nonparametric analyses of variance. Evidence was found for varying adult transmission patterns between urban and rural communities. Forty-nine counties display a multimodal distribution of age of infection, and the rest were categorized as moderately multimodal or non-multimodal. The average outbreak size was significantly different between the multimodal, moderately multimodal, and non-multimodal counties. Counties which were not multimodal, displayed a high modal age of infection. Small outbreak sizes consistently displayed patterns associated with spread of infection between adults and evidence suggests a correlation between outbreak size and proportion of children infected.

Download (PDF, 2.28MB)

Simple models for complicated situations: Ebola in West Africa

Sarah Rainey, a Biology major at Radford University, studied structural uncertainty in disease models with Dr. John Drake.

 

Abstract: Mathematical models are an idealization used for disease forecasting and decision making. There is a tradeoff between timeliness and detail for models produced in response to an epidemic. It is important to quantify the level of uncertainty in a model to mitigate inaccuracies that could arise from parameter uncertainty and structural uncertainty. The impact of parameter uncertainty has been heavily investigated; however structural uncertainty is scarcely addressed. The 2014 Ebola outbreak in West Africa was used as a case study to investigate structural uncertainty. Models produced early in the course of an epidemic tend to be simpler due to a lack of information with which to estimate parameters and an abbreviated development time. Complex models are more cumbersome to develop but take into account more complex transmission scenarios reported by in-country observers. The goal of this project is to determine when a simple model can capture the trajectory of an Ebola outbreak generated by a more complex transmission process. The statistical software R was used to solve three mathematical models: a branching-process model (Drake et al., 2015), a modified SEIR model (Legrand et al., 2007), and the classical SIR model (Kermack and McKendrick, 1927). During the 2014 Ebola outbreak in West Africa, the model produced by Legrand et al. was widely used because it included all of the key components to model an Ebola epidemic simplistically. The model published by Drake et al. (2015) included greater detail about contact patterns, interventions, behavior change, and other features that may be necessary to realistically model the Ebola outbreak. The complex branching-process model published by Drake et al. (2015) was used to generate an Ebola epidemic with different scenarios by varying parameter values. The two simpler models were fit to simulated data. The correlation coefficient was calculated to test the fit of the models to observe how well they were able to capture the trajectory of the outbreak. Our findings conclude that the modified SEIR model published by Legrand et al. (2007) was superior to the classical SIR model in representing the disease trajectory of multiple Ebola outbreaks simulated with unique circumstances.

Download (PDF, 3.02MB)

Mapping autochthonous transmission potential of Chikungunya virus in the United States

Nicole Solano, a student at Agnes Scott College, worked with Dr. John Drake to model factors that could influence the risk of Chikungunya virus in the United States.

Abstract: Chikungunya virus (CHIKV) is an arbovirus endemic to Africa and South and East Asia, that is transmitted to humans by the bite of an infected mosquito, primarily Aedes aegypti or Aedes albopictus.  Since its identification in Tanzania in 1952, CHIKV has spread around the globe, making itself a very prevalent infectious disease. To date (20 June 2015) there have been eleven reported cases of autochthonous transmission in the U.S. (in Florida). Since its introduction into the Americas, concerns have been raised about which areas in the United States are most vulnerable to importation of CHIKV. We examined the correlation between human West Nile Virus (WNV) cases and human Meningitis and Encephalitis cases. A strong correlation was observed (p-value < 2.2-16) informing us that Meningitis and Encephalitis is a good predictor of WNV infection. Given this, we wanted to know which socio-economic covariates were important to consider when thinking about exposure to a disease. A regression analysis helped us identify age and poverty level as the most important covariates. Presence of Aedes albopictus and relative exposure per county was mapped to depict which counties are most vulnerable for onward Chikungunya virus transmission.

 

Download (PDF, 1.46MB)

Visualizing the Effect of Interventions during the 2014-2015 West Africa Ebola Outbreak

Timothy Wildauer, a student from Bethany Lutheran College, worked with Dr. John Drake to test new methods of determining time of infection for Ebola patients.

Abstract: Data recorded during the 2014-2015 West Africa Ebola Outbreak indicates when patients presented themselves at a hospital for treatment. However, to know if interventions were successful, we need to know when patients became infected. The number of people who became infected on a given date is corrupted by the incubation period for us to see when patients begin showing symptoms. The current method for finding when people became infected is to shift the dates on the observed data set backwards by the average infection period. We developed two non-parametric filters to find when patients became infected with Ebola: multiple trials of randomly selecting incubation periods for patients, and a Ridge Regression. These methods were tested on simulated outbreaks of different difficulties to see which method worked best. We selected four interventions during the outbreak to see if there was an effect on the transmissibility of the disease. In Sierra Leone, the transmission rate dropped significantly after the country was shut down for three days in September 2014. After other events, the transmission changed significantly, but the change may be the natural course of the outbreak. Through testing we found that Random deconvolution achieves a high correlation with the infection curve.

 

Download (PDF, 566KB)

Protective Population Behavior Change in Outbreaks of Emerging Infectious Disease

Evans Lodge, a student from Calvin College, worked with Dr. John Drake to measure how human behaviors change during disease outbreaks.

Abstract: In outbreaks of emerging infectious disease, public health interventions aim at increasing the speed with which infected individuals are removed from the susceptible population, limiting opportunities for secondary infection. Isolation, hospitalization, and barrier-nursing practices are crucial for controlling disease spread in these contexts. Ebola virus disease (EVD), Severe Acute Respiratory Syndrome (SARS), and Middle East Respiratory Syndrome (MERS) are all zoonotic infections that have caused significant international outbreaks in the past. Here, we use patient-level data from the 2014-2015 Liberian Ebola epidemic, 2003 Hong Kong SARS epidemic, 2014 Saudi Arabia MERS outbreaks, and 2015 South Korea MERS outbreak to quantify changing removal rates, burial practices, contact tracing, and other measures of protective behavior change. Using the removal rate, γ, as a measure of protective behavior change allows direct comparison of health behavior development in different outbreaks and locations. Robust regression analysis and analyses of covariance are used to estimate the rate at which γ increases in each outbreak by epidemic week and serial interval. Measured interactions between models show that mean removal rates varied within a factor of three, falling between the 2003 Hong Kong SARS outbreak and the 2014-2015 Ebola epidemic in Liberia.

Download (PDF, 516KB)