Human Monkeypox in the Democratic Republic of Congo: Presentation, Prediction and Animal Exposures Associated with Rash Severity
- Author(s): Morier, Douglas Scott
- Advisor(s): Rimoin, Anne W
- et al.
These analyses come from the first major, active surveillance effort of human monkeypox (MPX) in its endemic region since the 1980s. These data represent more than 1,200 investigations over 14 health zones in the Sankuru Health District of Kasai Oriental Province, Democratic Republic of Congo (DRC). Suspected cases of human MPX were investigated by trained, local nurse supervisors in conjunction with the ongoing human MPX surveillance efforts of the DRC Ministry of Health (MoH). In the first chapter, techniques from the field of machine learning are applied to the sample in an effort to predict true cases of human MPX from "false positives" among those who present to investigators. In particular, random forest and conditional inference trees (ctrees) are used to identify variables from the survey instrument which are most predictive of an MPX outcome and ctrees are then constructed from the variables which rank highest in importance. Findings show that a prediction tree generated from the data can predict as well as the overall performance of trained investigators in this study. Signs, symptoms and factors identified as predictive (whether positive or negative) include pain when swallowing, swollen nodes and a recent history of a rash illness in the surrounding area. The second chapter applies epidemiologic techniques including directed acyclic graphs (DAGs) and stratified analysis toward evaluating associations between increasing exposure to animals and increasing rash severity among those with laboratory-confirmed MPX. Findings show that excessive consumption of rats to be associated with more severe rash outcomes. Furthermore, that association is observed in the presence or absence of other animals suggesting a complex ecology to animal exposures associated with human MPX that demands further study. Greater exposure to squirrels was not found to be associated with more severe rash outcomes. The third chapter evaluates the unique finding of 151 cases confirmed by laboratory analysis to be co-infected with MPX virus (MPXV) and varicella zoster virus (VZV), the virus that causes chickenpox. Hypotheses are presented and evaluated to consider if these unexpected observations are the result of a biological process, or simply the consequence of two independent events with independent probabilities in the source population being over-represented in a sample as a consequence of selection into that sample. Findings suggest the latter.