Climate change will lead to more intense and longer-lasting extreme weather events, such as floods, hurricanes, and severe storms, which will lead to more frequent power outage (PO). Significant gaps remain in our understanding of the impact of PO on human health: most previous studies were based on self-reported survey data, which is subject to reporting bias. In addition to weather factors that directly affect human health, other concurrent factors such as PO may also mediate with extreme weather on health, for which research is also lacking. Few studies assessed both power infrastructures and population vulnerabilities, and compared the effects of different PO causes. Furthermore, no health prediction models have been developed for PO. To fill these knowledge gaps, the proposed study will build upon multiple ongoing/ completed studies to: 1) assess the effects and mediating pathways of PO on electricity-dependent health outcomes, co-morbidities, and nursing home transfers; 2) identify infrastructure, environmental, and population vulnerabilities (individual and community); and 3) develop a vulnerability index and prediction model. We will effectively link the accessible New York statewide hospital admission and emergency department (ED) visit data with the existing data on PO, weather, air pollution, census, and nursing home transfer data. The associations between PO causes, frequency, duration, or area coverage of PO and electricity-dependent hospitalizations or ED visits due to asthma, chronic obstructive pulmonary disease (COPD), dialysis, water-/food-borne diseases, injury, and carbon monoxide poisoning will be assessed through Bayesian spatial-temporal model. This advanced technique will be able to control for both socio-demographic differences by regions and multiple temporal variables simultaneously. To improve scientific rigor, we will use control days (without PO and extreme weather) to separate PO from weather effects, and control diseases (e.g. appendicitis) to examine temporal changes of disease reporting. To understand PO?s natural direct and indirect effects, causal mediation analysis will be used. Furthermore, new variable selection methods, including Sure Independence Screening and Generalized Additive Model Selection will be used to screen and select predictors highly associated with outcomes. A composite vulnerability index weighed by risk factors identified and vulnerability maps will be developed. We will establish PO and PO-related health predictive models using the state-of-the-art data mining techniques, including Random Forest, Gradient Boosted Tree, and Ensemble Learning Decision Tree model. The excellent team with multidisciplinary and experienced investigators, numerous already collected and geocoded datasets, innovative data mining and analysis methods, continuation of student training, and successful prior partnerships with government agencies will maximize the probability of our success and feasibility. This project will also significantly enhance our institute?s environment and students? involvement in research, and our findings will identify evidence-based strategies for emergency management and public health preparedness.
The proposed study will evaluate whether frequency, duration, coverage, and certain causes of power outage (PO) are associated with increased risks of electricity-dependent diseases, including asthma, chronic obstructive pulmonary disease (COPD), dialysis, water-/food-borne diseases, injury, and carbon monoxide poisoning. This may be the first study to assess whether people with certain demographic characteristics (e.g. elderly or different sex), living in a neighborhood with certain vulnerabilities (e.g. high land coverage or low hospital density), in certain seasons, degree of urbanicity, or specific causes of PO are more vulnerable to the impact of PO. The findings, vulnerability index and maps, and the forecast models derived from this study will help guide the state or federal environmental and health agencies to plan interventions and climate adaptation programs.
