Modeling Heat Stroke Related Hospitalizations in the Elderly Residents in Boston, USA

• Activity Number: 452
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Elena Naumova and Ruiruo Wu*+ and Alexander Liss
• Companies: Tufts University and Tufts University and Tufts University
• Keywords: Time series ; Poisson regression ; heat wave ; heat stroke ; hospitalizations ; elderly

Abstract

Generalized log-linear modeling was applied to study the direct effect of elevated ambient temperature on health of elderly residents in the Metro-Boston area between 1991 and 2006. We abstracted 701 hospitalization records due to heat stroke (HSHs, ICD-9-992.0-9) from the Centers of Medicare and Medicaid Services data set. We built daily time series of hospitalizations and weather records for the same period. This study found that when maximum daily temperature reaches and exceeds its long-term 95th percentile (30.4°C) the rate of HSHs rapidly accelerated. We defined the local heat wave accordingly. We proposed exponential transformation of daily temperature to reflect non-linear behavior of HSH above the heat wave threshold. We estimated seasonal peak in HSHs on 191st day, 14 days before the annual peak in ambient temperature, and on average, within 7 days of the first seasonal heat wave. This indicates that the most severe health effects occur within the first heat wave periods. The final model explains 56% of variability in HSHs. It includes terms for the transformed ambient temperature, the indicator variable for the presence of heat waves, and seasonality adjustments.