Machine learning approach for disaster-related trauma and subsequent functional limitations

Along with social and economic losses, large-scale natural disasters often have adverse effects on the mental, cognitive and functional health of those who survive them. A new study looks at vulnerability factors.
 

By Thystere Matondo Bantidi, Ph.D., Temblor science writing extern (@Thysterebantidi)
 

Citation: Bantidi, T., 2023, Machine learning approach for disaster-related trauma and subsequent functional limitations, Temblor, http://doi.org/10.32858/temblor.311
 

Every year, the global community experiences significant damage inflicted by various natural hazards like earthquakes, tsunamis, floods and tropical cyclones. In addition to physical damage to both people and property, survivors may suffer from mental and cognitive health problems arising from the shock of losing loved ones or the sudden disruption in their homes or communities. Often, the poorest and most marginalized members of society experience the greatest adverse impacts.

Decision-makers tasked with allocating resources for recovery following a natural disaster often prioritize immediate damage. That’s not wrong, but it does leave out those most likely to struggle in the long term. Therefore, investigating the different ways that these traumatic experiences affect survivors — and linking those effects to factors like age, marital status, education level and income — may provide insights into the relationships between disasters and health for vulnerable populations, says Koichiro Shiba, an assistant professor in the Department of Epidemiology at Boston University and lead author of a new study published in the American Journal of Epidemiology.

In this work, Shiba and his colleagues describe adverse health effects of disaster-related trauma in the aftermath of the 2011 Tōhoku earthquake and tsunami that occurred off the Pacific coast of Japan’s main island of Honshu. The work has implications for recovery after disasters anywhere.
 

The earthquake

The magnitude-9.0 Tōhoku earthquake, often referred to as the Great East Japan Earthquake, struck along the Japan Trench where the Pacific Plate subducts beneath the Okhotsk Plate. At such subduction zones, one tectonic plate slides under another, but the downgoing slab may sometimes become stuck, causing stress to build up. When some or all of that stress is suddenly released, an earthquake occurs.

The Tōhoku earthquake also caused a massive tsunami that crippled the Fukushima Daiichi nuclear power plant and rendered parts of this region unlivable for nearly a decade due to radiation contamination. This triple disaster — earthquake, tsunami and nuclear meltdown — resulted in at least 19,729 dead, including several thousand victims who were never recovered. More than 121,000 houses were destroyed, with damage to almost a million more. Even today, more than a decade later, many people are still living in temporary housing.
 

 

Losing home, losing health

“Mental health” relates to psychological and emotional wellbeing, and includes conditions like anxiety and depression. “Cognitive health” relates to brain function, and includes memory, language, problem-solving and decision-making. “Functional health” relates to the ability to carry out daily tasks, like bathing, dressing oneself and eating. These three types of health are interconnected, meaning that problems in one area can affect the others.

For instance, individuals with mental health conditions like schizophrenia may have difficulty making decisions due to impaired cognitive abilities, which can in turn impact their functional health by making it difficult for them to carry out tasks of daily living. Cognitive health problems like dementia can lead to behavioral changes that can impact both mental health and functional health, again making it difficult for individuals to carry out tasks of daily living. And, functional health issues such as chronic pain can impact mental health by causing depression or anxiety, which can, in turn, affect cognitive abilities such as concentration and memory.

Numerous studies have identified the varied ways in which individuals respond to trauma. These studies have yielded two primary outcomes. First, some people who experience trauma do not develop mental health issues; they’re referred to as resilient. Second, studies show a correlation between how much trauma someone has experienced, called trauma exposure, and mental and cognitive health.

Cognitive health studies tend to show a correlation between trauma exposure and functional health. The more trauma experienced by an individual, the more trouble that person is likely to have when trying to simply go about their day — functional health is impacted. The earthquake and tsunami resulted in trauma exposure to survivors, and though trauma can manifest as mental, cognitive or functional health problems, their interconnectedness means that disentangling their relationships is not straightforward. Shiba’s new study looked at the connection between trauma imparted by these disasters and functional health, as well as factors that exacerbated trauma exposure, including mental and cognitive health.

To that end, Shiba and his collaborators used data from before and after the Tōhoku earthquake collected from Iwanuma, a city in the Miyagi Prefecture, located about 80 kilometers (128 miles) from the earthquake’s epicenter. The city was part of a nationwide cohort study of Japanese older adults, called the Japan Gerontological Evaluation Study (JAGES).

In Iwanuma, the JAGES study began in 2010, just seven months before the disaster. During the subsequent five and a half years following the earthquake, JAGES surveyed those who survived the disaster to measure individuals’ functional limitations in 2013 and 2016 using three different indicators: certified level of physical disability (e.g., whether someone can turn over in bed independently) ; activities of daily living (e.g., whether someone is able to accomplish basic daily activities such as walking or bathing); and instrumental activities of daily living (e.g., using public transportation).
 

 

Machine learning methods

Machine learning approaches use algorithms that can help scientists detect otherwise hidden insights in the data. Then, the algorithms “learn” from these insights and make predictions. Shiba’s team used a machine learning approach trained on 55 predisaster characteristics from the baseline of the JAGES survey conducted in 2010, which included demographic characteristics, measures of socioeconomic status, health conditions, psychosocial factors, and behavioral factors.

The team compared these predisaster demographic characteristics of the studied population of elderly people with home loss status after the disaster, and estimated the average effects of home loss on functional health. They explored whether some demographic characteristics of people who lost their home correlated with their ability to conduct daily activities. In short, the answer is: yes.

Based on their analysis, the team estimated that individuals who lost their homes displayed an increase in functional limitations. Moreover, those who experienced these limitations tended to be from lower socioeconomic backgrounds. Those who had symptoms of clinical depression before the disaster also tended to experience functional limitations afterward.

The team also found that people who were more vulnerable were more likely to be older, unemployed single men who lived alone and had a relatively low level of education. Members of this vulnerable group also had baseline health problems, such as more symptoms of depression and poor self-rated health. Nevertheless, income wasn’t necessarily a good predictor of vulnerability; those who had a higher income but were less educated also appeared vulnerable to being affected by the disaster.

These results offer new insights that were potentially overlooked in previous research, which linked higher income alone to disaster resilience. The new study uncovers a more complex relationship resulting from interactions between multiple characteristics like education and gender. However, the mechanisms of such complex heterogeneity remain unclear, warranting future research, Shiba says.
 

 

Assisting the elderly

Elderly people are particularly prone to post-traumatic stress disorder and depression that may result from disruptions where they live. “This fact alone explains why we must be interested in their physical, psychological or mental wellbeing after a natural disaster,” says Emmanuel Kagning Tsinda, a postdoctoral research associate at the Center for Biomedical Innovation at MIT who was not involved in this research.

This new study not only highlights variables that might explain post-disaster resilience versus vulnerability among the elderly, but also provides a model that helps to predict which individuals are at greater risk of experiencing adverse consequences after a natural disaster, Tsinda says. Such information could help decision-makers allocate resources more equitably throughout the different stages of disaster — mitigation and preparedness before an event, response during a disaster, and post-disaster recovery, he says.

Natural disasters like the Tōhoku earthquake and tsunami can have long-lasting impacts on those who survive them. Nevertheless, policymakers can utilize the new findings to implement measures that prioritize the wellbeing and safety of both individuals and communities, Shiba says. Prioritizing resources to go toward those who will be most likely to struggle with long-term recovery can contribute to more targeted post-disaster public health interventions, and therefore preserve survivors’ functional health.
 

Further Reading

Shiba, K., Daoud, A., Hikichi, H., Yazawa, A., Aida, J., Kondo, K., & Kawachi, I. (2022). Uncovering heterogeneous associations between disaster-related trauma and subsequent functional limitations: A machine-learning approach. American Journal of Epidemiology, 192(2), 217-229.
 

Copyright

Text © 2023 Temblor. CC BY-NC-ND 4.0

We publish our work — articles and maps made by Temblor — under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International license (CC BY-NC-ND 4.0).

For more information, please see our Republishing Guidelines or reach out to news@temblor.net with any questions.