SUSCEPTIBILITY TO HIGH TEMPERATURES: A MULTI CITY CASE-CROSSOVER ANALYSIS

Abstract

ISEE-355 Introduction: Several studies have identified a strong effect of hot weather on overall mortality, especially among the elderly. Only a few, however, have investigated the individual characteristics that increase the risk of dying during hot days. The case-crossover design was employed to study the different effects of high temperature on daily mortality among population subgroups defined by prior hospitalisations for various conditions. Methods: We studied the association between mean apparent temperature and all-cause mortality (1997–2003) among residents over 35 years of age living in four Italian cities (Bologna, Milan, Rome, Turin). For all deceased individuals, information on prior hospitalisations (two years) and principal and secondary diagnoses were collected. The analysis was performed in three stages: firstly, generalised additive models were used to evaluate the exposure-response relationship between temperature and mortality, with the aim of choosing a city-specific cut-point above which mortality increases linearly with apparent temperature. In this analysis, smoothing terms for time (at least 4 degrees of freedom per year), day of the week, holidays, influenza epidemics, PM10, and short-term population decreases were considered. Secondly, a case-crossover method, using the time-stratified approach was applied (control days were same days of the week during the same month). The same confounders were considered. Apparent temperature had two linear terms in the statistical model according to the city-specific cut off point chosen: the first indicating low temperature (lag 1–3), and the second indicating high temperature (lag 0). Conditional logistic regression was used to estimate percent increase in the risk of dying corresponding to an increase of 1°C above the city-specific cut-off point. For each city, we considered the potential effect modification of the following variables: age, gender, income, hospital admissions in the two preceding years (yes/no), place of death, and having been (or not) hospitalised in the two preceding years for a list of 29 medical conditions, chosen a priori. Thirdly, results of the four cities were pooled and potential heterogeneity across the cities evaluated. Results: In the pooled analysis of the four cities, we identified age and gender as the main effect modifiers. In the population aged 35+, the percent increase in risk (per 1°C) for females was 5.9% (95% CI=3.6–8.2) (males: 3.2%, CI=2.4–4.0). Among the pre-existing medical conditions investigated, thyroid diseases (7.6%, CI=3.5–11.8), fluid/electrolyte disorders (7.5%, CI=3.1–12.1), depression (7.6%, CI=3.0–12.4) and heart conduction disorders (6.7%, CI=3.3–10.2) were suggested as medical factors that increase vulnerability. Limiting the analysis to the population over 75, diabetics (7.6%, CI=4.1–11.3) were found to be more susceptible than the general elderly population. Conclusions: The availability of data regarding clinical characteristics of the subject may be useful to identify groups especially vulnerable to the effects of high temperatures. Public health strategies for handling heat waves may be more effective if they target susceptible groups such as those identified in this study.