Application of a time-space clustering methodology to the assessment of acute environmental effects on respiratory...

A new methodology is proposed for the identification of environmental events of health significance. Health indices measured on a daily basis at various location in a single geographical area are collected over time. First, the daily variations are examined to determine whether they reflect purely random variations or whether there are days on which there are extreme variations not plausibly explicable as random events. After such days are identified, the question of whether they occur only at a single location within the larger geographical area at one time, or whether the occur simultaneously at more than one location is investigated. Tests of statistical significance for both temporal and spatial clustering are propose. The methodology is applied to daily hospital emergency room visits for various respiratory complaints to several New York City hospitals situated in two geographically separated districts which however have population of similar socio-economic and ethnic composition.