Although "neighbourhood effects" on health have been recognised in many studies1 very little is known about the underlying mechanisms and very little experimentation has been done to find out what kind of neighbourhoods might have the greatest influence on health. While researchers have been challenged to consider the geographical scales at which processes might operate and to compare alternative neighbourhood sets, few studies so far have done this. By analyzing health variations within a single set of neighbourhoods, almost all published studies have been blind to whether alternative definitions of neighbourhoods might produce different results. The question is important for two reasons. Firstly, knowledge of the characteristics of the set of neighbourhoods that produce the strongest relationship with health might help to identify any processes involved. Secondly, policy-makers advocating interventions that focus on disadvantaged areas to improve individual health need to be better informed about how to identify the areas most amenable to treatment.

We (Jones and colleagues) have recently examined the importance of alternative neighbourhood definitions on the strength of area effects on accident rates to preschool children in the ALSPAC cohort2. Alternative sets of subjective and automated zone design neighbourhoods which incorporated different boundaries and different scales were used.Neighbourhoods based on different criteria produced similar area effect sizes as did neighbourhoods at different scales, but slightly stronger effects were observed in areas with populations less than 4,000.Most of the characteristics of children and mothers associated with accident risk were themselves associated with neighbourhoods, and particularly with the neighbourhoods of smallest size.

Another health related outcome where it is of interest to investigate the possible existence of neighbourhood effects is that of physical activity. In his report At least five a week, the Chief Medical Officer notes that the scientific evidence for the health benefits of physical activity are compelling. Nevertheless, the majority of adults and children fail to meet government guidelines for activity. Reviews of the evidence of the effects of individual level interventions indicate that while positive changes in physical activity can be achieved, the effects are small and short term3. In children and adolescents there is no evidence for the effect of individual-based educational interventions4. Larger, more sustainable changes in physical activity are more likely to be achieved by a multi-level strategy combining environmental and individual level interventions4. In recent years the research focus has moved from individuals to the environment. It is hypothesised that the environment is strongly associated with physical activity yet its actual relationship with physical activity in adults and children is relatively unknown within a UK context5.

One of the unanswered research questions concerns the role which the neighbourhood might play in determining the activity levels of the children living within it. It could be, for example, that certain neighbourhoods are more conducive to activity than others. Such effects could be ascribed to their physical characteristics, or may be associated with shared norms and behaviours amongst residents. Understanding how activity levels vary between different types of neighbourhoods will guide further studies of the more detailed environmental determinants of physical activity in children. The aim of this preliminary study is to compare the strength of neighbourhood effects on physical activity levels in children aged 11 years in the ALSPAC cohort, using alternative, contrasting sets of neighbourhoods.

Methods:

Based on our previous research we will divide the ALSPAC study area into thirteen different sets of neighbourhoods. One set are enumeration districts, the smallest units in the 1991 population census. The other twelve sets all use enumeration districts as building blocks and group them into larger units at three spatial scales. Some of these neighbourhoods are based on "communities" defined by planners in Bristol whilst others have been generated by the automated zone design program "A2Z". A more detailed description and maps showing examples of the zones within Bristol are given elsewhere.6