Background: while the effects of temperature have been extensively evaluated for mortality and hospital admissions, the association with other outcomes has received less attention. Aims: to investigate the relationship between summer temperature and emergency ambulance dispatches in a multi-city study from Italy, using novel statistical approaches. Methods: Time series data for the period 2002-2006 (May-Sept) were collected in the major 9 cities of the Emilia-Romagna region for people aged 35+ (168,180 emergency calls in total). A two-stage time series analysis is performed, with city-specific models including bi-dimensional splines of apparent temperature with a distributed lag non-linear parameterization on lag 0-15. Estimates are reduced to summaries of overall cumulative and predictor-specific effects, which are then pooled using a multivariate random-effect meta-analysis fitted through restricted maximum likelihood. Results: apparent temperature ranges homogeneously across cities between 10°C and 34°C. The exposure-response relationship is flat for mild summer temperature, with an evident non-linear increase above 27°C. The relative risk for 32°C versus 23°C, corresponding approximately to the 99th and 50th temperature percentile, is 1.14 (95%CI: 1.11-1.19) at lag 0 and 1.52 (1.36-1.71) when cumulated across all lags. The analysis of the lag structure suggests that the effect of high temperature lasts 11-12 days, with no evidence of short-term harvesting. Discussion: this analysis emphasises the strong association between high temperature and emergency ambulance dispatches, highlighting how the health effects are not limited to the increase in mortality and hospital admissions. The novel analytical approach adopted here includes city-specific and meta-analytical models which allow a flexible estimation of nonlinear and lagged exposure-responses. This effect indicator may be useful for defining real-time surveillance systems.
展开▼
