Categorisation of built environment characteristics: the trouble with tertiles

摘要: In the analysis of effect built environment features on health, it is common for researchers to categorise exposure variables based arbitrary percentile cut-points, such as median or tertile splits. This categorisation leads a loss information and lack comparability between studies since choice cut-point sample distribution. this paper, we highlight various drawbacks adopting variables. Using data from SocioEconomic Status Activity in Women (SESAW) study Melbourne, Australia, alternative approaches which may be used instead order assess effects health. We discuss these using an example examines association number accessible supermarkets body mass index. show that categorisation, transformations variable factorial polynomials, can implemented easily standard statistical software packages. These procedures utilise all available data, avoiding power experienced when adopted.We argue should retain by continuous exposure, where necessary.