For the seventh Public Health Twitter Journal Club, a short list of three recent longitudinal studies on public health topics was put to a vote. The favoured paper topic was neighbourhood environments and obesity.
Full paper citation:
Timperio, A. et al (2010) ‘Neighbourhood physical activity environments and adiposity in children and mothers: a three-year longitudinal study’, International Journal of Behavioral Nutrition and Physical Activity 7: 18 http://www.ijbnpa.org/content/7/1/18 doi:10.1186/1479-5868-7-18
Question 1: Is the study design appropriate for the aims?
Study aim was identified as:
‘..to add to the current evidence base by examining cross-sectional and longitudinal associations between objectively –measured physical activity-related neighbourhood characteristics and adiposity among children and their female carers.’ (p.2)
Discussion participants agreed overall, that the study design was appropriate for the aims. One person raised the question of whether a comparison group should have been used. Another commented that :
Question 2: Was the sampling technique clear and appropriate? What about follow-up?
It was noted that readers had to refer to other publications from the wider study, to garner full details about recruitment and selection of participants. Some participants found this frustrating. There was some discussion around the reported socio-economic status of the sample at follow-up, which was more homogeneous than might be expected given neighbourhood differences (almost half of the female carers had higher or tertiary education. It was suggested that either, more middle class participants may have been willing to take part in the follow-up study, or the original study sought homogeneity. The description of the sample that PHTwitJC posted into the ‘discussion points’ post from another journal article, indicates that the latter was not the case.
Decisions about the chosen age cohorts at baseline (5-6 years and 10-12 years ) were not explained within the paper.
It was also noted that there was substantial ‘fall-off’ of the sample between baseline and follow-up , a possible limitation of the study. Study respondents who had moved address during the three-year period, and those on whom data were incomplete, were excluded. Another aspect that was not fully discussed during journal club is that the statistical investigations reported in this study included only ‘female carers’, although ‘parents’ were the original target of the overall study. This decision was not explained, although we might conjecture it was because of a paucity of response data from fathers/male carers.
Question 3: Were the measurements used clear, relevant and valid (BMI/z- score)?
A number of readers / journal club participants were not familiar with z-score as a measure . It was noted that this was not explained within the paper – perhaps it was assumed knowledge for the journal’s readership. The authors’ assumptions about what level of z-score indicated overweight or obesity were not clearly defined , which was another frustration for readers.
Use of BMI as an ‘adiposity’ measure was questioned by participants, given there are significant challenges to this. Current best evidence appears to be pointing toward waist circumference offering a more reliable measure:
It was felt that the authors should have acknowledged this debate and explained the decision to use BMI as the adiposity indicator for the adult sample.
Q4: were the ‘objective’ neighbourhood phys activity environment measures clearly explained & valid?
It was mentioned that the GIS system used to facilitate the physical activity environment measurements sounded very sophisticated . While interesting, this also raised the possibility that too wide a range of measures was used–because they were possible, rather than because of relevance. However, we affirmed that this was put forward as an experimental set of measures and a novel approach. The study, to some degree, was testing the usefulness of the measures, as well as their association with adiposity.
Readers found the discussion of some of the measures in the paper to be counter-intuitive. For example, cul-de-sacs were negative indicators from a ‘walkability’ point of view (the main criterion used in the analysis), although they might also be viewed as positive in enhancing outdoor play opportunities for example. On the other hand, having a large intersection in proximity to the home was a ‘positive’ walkability aspect on the grounds that it opens up access for pedestrians as well as motor traffic, in different directions. We felt major junctions might also be indicative of heavy traffic, which could be off-putting for walkers. However, we accepted the ‘walkability’ criterion as interesting, and a good challenge to our common-sense ideas about neighbourhoods, roads and pedestrian access.
One limitation was that the paper, in its reliance on set measurements, did not allow for nuances in the impact of different designs in ostensibly similar road /footpath/cycle ways ; or between similar designs in different types of neighbourhood.
Readers considered the claim to originality on the basis that the study analysed ‘real’ neighbourhoods (the area proximal to the home) as opposed to administrative boundaries. It was agreed that this was an interesting aspect of the study.
There was a little discussion of how the objectively measured neighbourhood presented in the paper, although having some novelty, adhered to a conventional view of place and geography. A highly cited paper by Cummins et al on a relational approach to ‘place’ in health research was linked. The relational view of geography sees locales not in terms of fixed, measured distances but as fluid, and context dependent, incorporating complex interactions between individual actors, social networks and spatial features.
Question 5. The results were inconclusive: are there any that are particularly noteworthy; are the researchers’ explanations valid?
The paper is very clear that the results were inconclusive and sometimes inconsistent. We considered whether this meant that neighbourhood environments do not affect physical activity levels (/adiposity) . It was felt more likely that more sensitive understanding of neighbourhood nuances; and of families’ approaches to physical activity and their neighbourhoods, was required. The wider study included a questionnaire on children’s and adult’s perceptions of their local neighbourhoods; it would have been informative for those results to be discussed alongside the geographical measurement results, but they were not to any great degree. Where in this paper, the authors speculate on some of the results and the differences between groups, it would have been useful to back up the speculation with data on participants’ actual reported use of or perceptions about, their neighbourhoods.
6. Are there implications for further research?
Journal club suggested that a qualitative study on the aspects mentioned above would further enhance and help to explain the geographical measurement findings. Different combinations of data as measures of the local environment should also be tried in further research; the authors themselves suggest for example, the inclusion of food environments.
7. Are there implications for policy /practice?
Some participants felt that the finding that the number of busy roads in the neighbourhood was associated with overweight for older children was particularly policy-relevant, as the authors also indicated:
However, others felt that the results reported were too limited to be useful on their own, as wider issues needed to be considered for obesity prevention (including diet and physical activity in schools).
For readers, the chosen paper presented some structural limitations. Reporting of study as a whole (CLAN) seemed to have been carved up into different journal papers (for understandable reasons), but this had an impact on the paper’s readability. Understanding of some aspects of the paper really required the reader to seek out related publications (not all of which were open access). Also, key measures (e.g. z-score) and their significance were not explained for the reader, which made the data more difficult to interpret than was necessary.
Overall, the paper’s data strategy was interesting and different from standard epidemiological studies – this could be refined for use in further studies incorporating neighbourhood effects.