ADVERSE HEALTH CONSEQUENCES OF EXPOSURE TO AMBIENT TEMPERATURE IN A CHANGING CLIMATE

Climate change with its associated elevated climate variability has the potential to substantially increase the number and intensity of frequent and rapid fluctuations in ambient temperature that, in turn, has a detrimental effect on susceptible and fragile population. Increased life expectancy in the last half-century and overall better treatment of life threatening conditions such as HIV and cancer has led to communities with a high proportion of elderly who are highly vulnerable to thermal exposure. Extreme weather events, such as heat waves and cold spells, can cause excess mortality and morbidity of this affected population due to hypothermia and heat strokes. Other highly vulnerable groups include children, pregnant women, and people with chronic conditions such as diabetes, kidney failure, cardiovascular and respiratory diseases, electrolyte disorders, and patients undergoing immunosuppressive treatment.

Public health professionals, economists, and environmental scientists, apply two main designs for statistical modeling to study these issues. One design usually employs annual or monthly panel data methods, using mixed effects model, controlling for differences over large geographic areas, coarse time scales, seasonality and temporal trends. The exposure to ambient temperature is estimated as a single set of exposure gradients for a study area. Another design is usually based on a repeated cross-section or time series study of daily measures of temperatures and health outcomes at a city or community level. This design typically explores the associations between meteorological parameters (most commonly, ambient temperature and its derivatives) and adverse health outcomes on a refined spatio-temporal scale.

In this work we examined the relationship between meteorological conditions and spatial-temporal patterns of hospitalizations due to hypothermia and heat strokes, regional risk rates and adaptation effects, in Medicare recipients 65 years old and older in the United States. We applied both designs to 124,900 records of hypothermia and heat stroke related hospitalizations abstracted from 220 million hospitalization records collected during time span of 16 years (1991-2006). We studied relationship between ambient temperature and hospitalization rates, adjusted for patient's age, gender, seasonal and annual fluctuations in temperature and population. We also evaluated differences in hospitalization rates due to adaptation to prevailing climate conditions in eight predefined climate regions. These climatic regions were constructed using satellite remote sensing data with machine learning techniques. We applied generalized mixed effects models with log-link and Poisson distribution to fit both study designs . We found, that both designs agreed in the magnitude and direction of estimated effect, however, the second design provided better resolution in cases when the spatio-temporal scales of exposure and effect match.

We found that hospitalizations during study period in the USA due to excessive cold are more frequent (74,030 cases versus 41,927 cases ) and deadlier (13,146 cases (17.76%) versus 1,526 (3.64%) than those due to heat exposure. The annual average rate of hospitalizations related to hypothermia and frostbites was 123.6 ± 15.2 cases per one million Medicare enrollees. The annual average rate of hospitalizations due to exposure to heat was 79.5± 70.1 cases per one million Medicare enrollees. Relative risk associated with 10 degree Fahrenheit changes in ambient temperature for exposure to cold is 1.35 (CI: 1.33-1.36), and for heat 1.1245 (CI: 1.123-1.126). We also found significant difference in regional risk rates, suggesting presence of a strong adaptation effect to prevailing climate conditions. Specifically, subjects residing in a relatively warm climate exhibit stronger sensitivity to lower temperature than those living in colder climates. On the other hand, those who live in continental climate with a large range of annual variability of ambient temperature, exhibit less sensitivity to temperature fluctuations, suggesting a better adaptation to both cold and hot environments.