(2008) who hypothesised: ‘[t]hat by exploring differences between schools, we may be able to determine school factors that are, for better or worse, having an impact on children’s risks of obesity.

At the same time, we may be able to highlight ‘hot’ and ‘cold’ spots of obesity so allowing better targeting of resources to those communities in greatest need. To test this hypothesis Procter et al. (2008) employed a ‘value-added’ Akt inhibitor drugs technique similar to those developed in economics and regularly used to assess the educational impact of schools (Amrein-Beardsley, 2008 and Rutter, 1979). In education, an individual’s value-added score is the change in outcome (e.g. test score) during the period of their schooling. In order to compare school performance the individual scores are aggregated, and it becomes necessary to adjust for differences in school composition which could bias the scores (Amrein-Beardsley, 2008 and Rutter, 1979). Procter et al. (2008) accounted for the ethnic and socioeconomic composition of 35 primary schools in Leeds, England, who were participating in the Trends study to rank schools according to their mean observed and expected residual pupil weight status and ‘value-added’ score. The authors found that there was little

similarity between the ‘value-added’ and expected residual INCB018424 order rankings and concluded that this lent credence to the hypothesis that differing school environments have differential impacts upon their Phosphatidylinositol diacylglycerol-lyase pupils (Procter et al., 2008). As a result they suggested that obesity prevention efforts be targeted rather than

population wide as ‘hot’ and ‘cold’ schools for obesity had been identifiable, and hence future research should focus on such schools. Acknowledging the fallibility of such ‘league tables’, Procter et al. (2008) also suggested that these analyses should be replicated across a number of years to test the validity of the findings (Goldstein and Spiegelhalter, 1996). This study evaluates and expands upon the technique proposed by Procter et al. (2008) using repeated cross-sectional data from a large routine data source (the National Child Measurement Programme (NCMP)) to examine the potential differential impact of primary schools on children’s weight status. The English NCMP was introduced in 2005 to monitor progress towards a public service agreement to reduce the prevalence of obese primary school aged children (Dinsdale and Rutter, 2008 and South East England Public Health Observatory, 2005). Unless individuals or schools are actively opted out, all Reception (4–5 year olds) and Year 6 (10–11 year olds) pupils in state maintained primary schools have their height and weight measured by a health professional (Dinsdale and Rutter, 2008). Five years of NCMP data (2006/07–2010/11, involving 57,976 pupils) from Devon local authority were used in this study.