Challenges and Opportunities in Epidemiology of a Natural Disaster: Lessons From the 2011 Great East Japan Earthquake and Tsunami

Abstract

A natural disaster causes not only immediate injuries and loss of lives but also long-lasting sequelae including changes in lives and health of survivors. Emerging evidence has suggested that catastrophic damages and traumatic experiences from a natural disaster can lead to worsening health of survivors even years after the disaster onset. To better inform public health policies to maintain health of the victims, it is crucial to accurately estimate the long-term causal effects of disaster-related damages on health and explore the underlying mechanisms. For instance, property damage due to a disaster and subsequent relocation to temporary accommodations may lead to drastic changes in living environments of the survivors, which may in turn affect their health. On the other hand, the post-disaster relocation results in exogenous changes in residential locations of survivors and offers an interesting opportunity for epidemiologists to assess the causal effects of neighborhood environment on health. In this dissertation, I analyzed a unique natural experiment data, in which the 2011 Great East Japan Earthquake hit participants of the prospective cohort study of Japanese older adults during the follow-up. In Chapter 1, I took advantage of the changes in home address and exposure to neighborhood environment stemming from the post-disaster relocation to/from temporary shelters to estimate the causal effects of neighborhood environment on cardiometabolic profiles. The fixed-effects regression indicated that the exposure to neighborhood environment with high urban density was associated with unfavorable cardiometabolic profiles. The results also suggested that the protective impacts of built-environment in the neighborhood with high urban density may be offset by proximity to unhealthy food outlets. In Chapter 2, I assessed the extent to which the types of post-disaster accommodations explain the adverse impacts of home loss due to tsunami on cardiometabolic profiles of the disaster survivors 2.5 years after the onset. The results from causal mediation analysis indicated that the post-disaster relocation to trailer-style temporary homes explained most of the adverse effects of home loss on anthropometric measures while the effects on serum lipid measures appeared to be explained by other pathways. In Chapter 3, I reviewed two under-recognized challenges in causal inference of long-term health impacts of a disaster exposure; selection bias due to selective attrition and analysis of time-to-event outcome when the effect is time-varying. After reviewing methods to deal with the problems, I showed applications of the methods to the Japanese earthquake data as motivating examples. By comparing with the results from the proposed alternative approaches for causal inference, I demonstrated the conventional analytical approaches may significantly underestimate the true long-term causal effects of disasters on health of survivors and lead to misleading conclusions.