Effect of earthquake-related losses and post-earthquake events on morbidity and mortality: causal mediation analysis of the prospective cohort data of the 1988 earthquake survivors in Armenia
- Author(s): Khachadourian, Vahe
- Advisor(s): Arah, Onyebuchi A
- Armenian, Haroutune K
- et al.
Earthquakes are one of the most common types of natural disasters. They remain unpredictable and often result in substantial damage and destruction. Earthquakes have been found to be associated with adverse mental and psychological conditions such as post-traumatic stress disorder, depression, and generalized anxiety disorder. Despite the significant negative health and social impacts of earthquakes, to date, only a limited number of studies have explored physical health outcomes and long-term survival among populations exposed to earthquakes. Most of these studies have methodological limitations such as lack of generalizability due to population specificity; limited sample size and power; non-temporal design; and the use of aggregate-level exposures and outcomes. In this dissertation, we presented a few causal mediation analysis frameworks and discussed their potential in exploring causal mechanisms in disaster research. Subsequently, we used data from a prospective cohort of surviving adults with differential exposure levels to the 1988 earthquake in Armenia to investigate the impact of earthquake-related exposure (housing damage, death of a family member, and serious injury) on all-cause mortality. Various data adaptive approaches, including super learner and random survival forests algorithms were applied to simulate and impute the outcome for the subpopulation with a shorter follow-up time. Furthermore, we applied the four-way mediation analysis framework and decomposed the effect of housing damage on all-cause mortality with respect to receiving permanent housing in the aftermath of the earthquake. Similarly, we assessed the impact of earthquake-related exposure on risk of developing diabetes. The interventional approach for path-specific effect estimation framework was used to decompose the effect of housing damage on diabetes with respect to permanent housing aid and job loss. Finally, a simulation study, incorporated the estimated effects to assess the impact of hypothetical interventions on risk of all-cause mortality and diabetes. The hypothetical interventions were based on providing permanent housing, or preventing job loss in the aftermath of the event. The dissertation sheds light on effects of earthquake related exposure and potential post-earthquake interventions on risk of all-cause mortality and diabetes. Furthermore, it provides accessible evidence for decision making and prioritizing interventions and optimal resource allocation in the aftermath of an earthquake. Future studies can expand the current work by investigating the effect of earthquake related exposure through other pathways. Such studies would also benefit from the inclusion of social support and psychological outcomes in their framework analysis. Future studies should also consider more extensive sensitivity analysis for uncontrolled confounding and incorporate sensitivity analysis for measurement error and selection bias.