Difference between revisions of "Main Page/Research/Papers/fast food and arterials/AJPM/comments"

From phurvitz
Jump to: navigation, search
(No difference)

Revision as of 01:06, 27 January 2009

Reviewer #1: This study investigates whether measures of neighborhood socioeconomic status (SES) and arterial street and freeway density are associated with fast food restaurant density.

The main problem I have with this paper is related to the rationale for considering measures of arterial street and freeway density in relation to fast food restaurant density. It is interesting to correlate fast food restaurant density with low SES neighborhoods because low SES populations constitute vulnerable populations already cumulating a number of risk factors. Thus it is interesting to assess the extent to which low SES populations are exposed to unhealthy food in order to attempt to correct this "environmental injustice". The same is true for the association between schools and fast food restaurants: school children constitute vulnerable populations, because of their lack of awareness of problems caused by unhealthy eating. However, I don't see why it is interesting to correlate road network variables with fast food density. It is obvious that fast food franchises attempt to locate fast food restaurants in busy roads, where they will be seen and used by many people, rather than in places where no one goes. Moreover, fast food restaurants are particularly suited for people traveling by car, far from home, or without enough time to have a full lunch at home. Thus, it is rather logical to expect higher densities of fast food restaurants along busy streets.

Why is it interesting to investigate the association between road network characteristics and fast food restaurant density? Why is it interesting to demonstrate that "arterial/freeway density [has] a stronger relationship with fast food counts than SES"?? What are the implications? My guess is that it is not particularly relevant to focus on this issue. It is also obvious from the fact that this association may be demonstrated for almost all types of services, that tend to be located along arterial streets. For example, would not we find the very same association between arterial streets and traditional (non fast food) restaurants?

Lines 163-172: If it is not really interesting to investigate per se the associations between road network characteristics and fast food density, at least it could be argued that it is necessary to take these variables into account as adjustments factors when focusing on the SES - fast food restaurants associations. Did it appear particularly necessary in the present study to control for arterial / freeway density?

As least an interaction between arterial streets/freeways and low SES in relation to fast food restaurant density may be interesting. Was it observed in the data?

As noted in the discussion, measuring fast food counts at the census tract scale is rather problematic, as many fast food restaurants are located on streets at the side of the tract, and are therefore accessible to populations of at least two census tracts. In this GIS study, it would have been extremely easy to consider buffers around the boundaries of each tract in order to capture fast foods located on the other side of the street just at the limit of the tract.

It would need to be explained why, in the model, the offset was defined in terms of area size rather than in terms of population size. Is it really more relevant to consider densities of fast foods defined as counts per 1 km² rather than as counts per 1000 inhabitants? At least a sensitivity analysis would be useful.

Perhaps it would be more convincing to define the environmental variables as categorical variables introduced into the model with different dummy variables. It would allow the readers to assess whether fast food density regularly increases with decreasing area income.

Minor issues: - The last sentence of the abstract, as well as the last paragraph of the discussion, are rather disconnected from the empirical analyses.



Reviewer #2: American Journal of Preventive Medicine Manuscript Number:07-1029-746 Title: What drives fast food location? Road density and area socioeconomic status as correlates of fast food restaurant density.

Review:

I enjoyed reading this manuscript and thought it was well written. From the background section, it appears that there are two lines of research in this area - the density of a food service type (e.g., fast food, supermarket) and how it compares between areas (e.g., high and low income areas ) and a comparison of the availability of "healthy" and "less healthy" food service types in an area (choice or lack of choice). This research represents the former and I think it does a good job in that the hypothesis is clearly stated and the data/methods were appropriate for conducting the research. The discussion and conclusion sections are informative, do not reach beyond the data, and present the major limitations of the study while providing some suggestions for future work to lessen the limitations.

Specific comments: · A particular strength of the study is the inclusion of the density of arterial streets and freeways to capture and measure one aspect of decisions that are made by businesses in siting food sources. As noted by the authors, this measure can be readily introduced in studies in other areas looking at correlates of fast food density (lines 181-182). · The assignment of the fast food restaurants to a single census tract is a limitation of the manuscript (as noted by the authors on starting on line 188). Thus, I think the point that other methods are needed to generate density measures is critical (lines 196-197). Also, in future work, the use of network analyst rather than straight line distance might be more useful as a measure of fast food accessibility. · Although I liked the inclusion of parcel data to measure area SES (through property value) I was somewhat unsure if it was appropriate to divide the value of a multifamily unit by the count of units to calculate a value per residential unit. Also, I think differences between areas of owner-occupancy and renter-occupancy is left out (which can be related to the SES of an area) (lines 74-77) · The manuscript would be strengthened by a few specific examples in the discussion about how the findings can help to shape public health interventions (lines 169-170). · An important point to make is that as the location as fast food franchises can be explained by vehicle accessibility and visibility - these same factors may inhibit residents from walking and other forms of physical activity (lines 206-209). These might also be less desirable neighborhoods (because of traffic, noise, exhaust fumes) and thus more likely to be populated by residents of lower income (lines 208-209) · King County has a relatively small minority population (Census 2000 indicates the Asian alone population is the largest minority group (percent)) and is a relatively affluent county. The manuscript notes that the study may not be generalizable (line 181). To what extent do the authors think the county demographics influence the results and limit the generalizability?



Reviewer #3: This is an interesting study that is revealing with regards to the retail geography of fast food outlets using a case study are in King County WA. Whilst many recent studies in various settings have demonstrated that locational access to, and density of, fast food outlets is stratified by neighborhood SES/deprivation (with better access and more outlets in lower SES areas), few of these studies have identified the mechanisms that explains this social stratification. This paper potentially makes a valuable contribution to the understanding of the link between neighborhood SES and the neighborhood food retail environment. However, there are some issues that need to be addressed before the paper is suitable for publication.

1. Background section, p1: It would be useful in the background section if the authors could (briefly - couple of sentences) develop the theoretical contribution of the work - i.e. elaborate a little on the recent resurgence in interest in neighbourhood effects on health, role of place, context v composition etc. See, for example the work by Diez Roux and colleagues.

OK will do


2. Background section, p1: The literature review seems fairly comprehensive and includes almost all of the international studies on neighbourhood fast food access. The exception is a New Zealand study; the authors should include this reference, especially as it was published in the Am J Prev Med: Pearce J, Blakely T, Witten K, Bartie P. Neighborhood deprivation and access to fast-food retailing: a national study. Am J Prev Med 2007;32(5):375-82. Further, this study makes the important point that access to other forms of food retail outlets, including 'healthy' food stores, are patterned by neighborhood deprivation in a similar way.

got it


3. Background section, p2, lines 25-33: The authors could usefully mention the role of geodemographics in determining location of food outlets. In other words, urban and transport planning issues are importing in determining location, the social and demographic characteristics of the population are too (an issue discussed in 'Fast food Nation' for example). This is an issue that the authors already explore in their subsequent analysis so a mention here would strengthen their rationale.

maybe, will look into this


4. Methods, p3, lines 54-5. I think this sentence could be rephrased as it suggests they created an index - better to simply say a 'measure of neighborhood wealth'? Further, and more importantly, what is the justification for two separate 'wealth' measures - median income and property values?

drop property value. it is not significant anyway


5. Results, p 5, lines 101-5: would this short section be better placed in the methods section - they are not really results?


6. Figure 1. The cartography of this map would be improved if the water bodies were given less prominence. Further, I am not quite sure what the authors mean when they imply that Figure 1 demonstrates 'positive skew'? If the density measures are positively skewed then does Figure 1 show this, or would it be better to separate this point from discussion of Figure 1?

7. Results, p7, lines 138-51. I am a bit confused by the percentage values reported in the text. If I am understanding the results correctly, then the figures reported in the text do not correspond with the figures reported in table 3 (e.g. 5.9 should be 6.1, 29.1=25.6 etc.)? Further, rather than just report whether results are significant or not, I'd have thought the key point is the size of any effect? See: Sterne et al. BMJ 322 (7280): 226: 2001.

make sure that it is clear that because these are logistic coefficients the betas need to be exponentiated for calculation of effects


8. A key point is the authors' interpretation of the findings - Discussion section, lines 155-61. The authors state "However, this study found that arterial street and freeway density was a stronger determinant of fast food restaurant density at the census tract level than selected area-based measures of SES." This is the key finding of the authors' study. However, I am not totally convinced that this is necessarily what the authors have found. The authors base this assertion on comparing the coefficients of each variable in Model 3 in Table 3. The concern is that the parameter estimates may not be directly comparable in the way the authors think they are. If I understand the regression modeling strategy correctly, each of the variables that the authors compare uses different units - they are not standardized. If the units were changed then the size of the parameter estimates would too, and it may be that the substantive conclusion changes as well. It is important that the authors consider this issue and clarify the point within the text and/or re run their analysis with standardized coefficients. If, on the other hand, I am misunderstanding the modeling approach then the statistical analyses section needs to explicitly address this point.

coefficients have been standardized


9. Regardless of issue above, the authors have slightly over simplified the interpretation of their findings. To some extent some previous studies of neighbourhood fast food accessibility have either taken into account population density in their models, a useful proxy for land use. I agree that the present study more rigorously examines this issue, but it is misleading to suggest that earlier studies have completely ignored it.

10. Discussion, p9, lines 199-97. The authors are quite right to draw attention to the way they have constrained their analysis to a single census tract - this is a limitation in their study. However, they should also recognize that some studies of fast food accessibility have used more advanced GIS methods to better address this issue. For example, a study that the authors already reference (Block et al) is not constrained in the same way as the present study as it uses neighbourhood buffers: Block JP, Scribner RA, DeSalvo KB. Fast food, race/ethnicity, and income: a geographic analysis. Am J Prev Med 2004;27:211-7. Further, the previously mentioned New Zealand study offers an alternative approach based upon network distances, and hence is not constrained by arbitrary census areas. See also a related methodological paper by the same research team: Pearce J, Witten K, Bartie P. Neighbourhoods and health: a GIS approach to measuring community resource accessibility. J Epidemiol Community Health 2006;60:389 -95. There are also other examples from the literature on access to other forms of food retailing that the authors could usefully reflect upon.