Canada Post community mailboxes: implications for health research.

Fuller, Daniel

It is recognized that where you live plays a small but important role in determining health. (1) Six-digit postal codes are commonly used in health research to identify individuals' location in space, usually their residential location, and derive environmental exposure measures based on census data or other administrative data sources (2) which are then linked to health data. Over the next 5 years, Canada Post will transition 5 million Canadians from door-to-door postal delivery to community mailboxes. (3) Eleven communities will pilot test the transition. These include neighbourhoods in Calgary, Fort McMurray, Winnipeg, Oakville, Ottawa, Rosemere, Lorraine, Bois-des-Fillion, Charlemagne, Repentigny, Halifax, Lower Sackville and Bedford, and represent a combined total of 86,950 addresses.

As Canada Post transitions to community mailboxes, 6-digit postal codes will be assigned to community mailbox areas instead of smaller postal code areas. Currently, 6-digit postal code areas correspond to approximately one city block in urban areas, while community mailbox areas are larger than postal code areas and can vary widely in size. It is plausible that over the course of the transition to community mailboxes, Canada Post may choose not to maintain 6-digit postal codes and transition to a unique community mailbox number instead. In both cases, the transition to community mailboxes may have important implications for health research. We highlight these below.

Individual's residential location will be misplaced

In comparison to assignment based on 6-digit postal codes, assigning people to a geographic location based on their community mailbox location will increase positional error. This error stems from the difference between an assigned location and one's true location in space. (4,5) A recent Canadian study estimated that when using 6-digit postal codes, positional errors ranged from 109 metres to 1363 metres in urban and rural areas respectively. (4) Reliance on community mailboxes will increase positional errors. Using a single community mailbox to represent multiple 6-digit postal codes will reduce precision when health researchers assign individuals to a home location (Figure 1A).

Linking residential location to the census or national health survey

Health researchers often use postal code information to link health data with census data. Currently, Statistics Canada develops Postal Code Conversion Files (PCCF & PCCF+) to link 6-digit postal codes to census geographic areas (e.g., dissemination area [DA]). The transition to community mailboxes is of concern for Statistics Canada, as noted in the 2013 Postal Code Conversion File Reference Guide: "community mailboxes are a growing source for multiple records per postal code on the PCCF. In newer urban delivery areas, postal codes are assigned to a community mailbox that may cover partial dissemination blocks, both sides of a street, and different streets within 200 metres of the community mailbox. These situations often result in multiple links being established between a postal code and block-faces, unlike the more traditional urban postal codes, which correspond generally to a block-face." (6, p.9) (Figure 1B) Linkage errors from postal codes to census geographies will make studies using these linkages biased in unknown ways.

Examples of the problem

Positional and linkage errors described above will result in greater misclassification of environmental exposures in health research. Researchers will not be able to use 6-digit postal codes to locate, as precisely as possible, individuals' homes, and derive exposure measures based on this location.

[FIGURE 1 OMITTED]

There are three major research implications resulting from the move to community mailboxes. First, using residential location as a proxy for socio-economic status (SES) in epidemiological research will be biased. There is evidence that residential SES is related to health (1) and that aggregate measures of SES at the postal code level are a reasonable proxy for individual-level SES in urban areas. (7) It is not known how, for example, the results of a study by Roos et al.,8 who used individual health data linked to the Canadian census to estimate the odds of premature mortality by neighbourhood SES, would change given positional and linkage errors. Second, studies examining access (or proximity) to health-enhancing (e.g., green space, grocery stores) and health-diminishing (e.g., fast food outlets) resources will have greater measurement error, which will bias results. (9,10) Finally, researchers will struggle to define environmental exposures by linking postal code information to census geographies. (11) For example, a study by Hoek et al. (12) used residential location to define exposure to air pollution and showed that cardiopulmonary mortality was associated with living near a major road (relative risk 1.95, 95% CI 1.09-3.52). If residential location is not correct based on postal code, the results of this study could be biased and erroneous. We have presented here three illustrative examples of the problem but there are potentially many other examples, depending on the research area.

What will the future look like?

In many places in Canada, particularly small towns and rural areas, community mailboxes or post office boxes are already in place. Examining research comparing linkage and positional errors in urban and rural areas sheds light on the future of this type of research in Canada. For example, Pampalon et al. (13) showed that survival inequalities in small towns and rural areas are lower than elsewhere when area-based measures of socioeconomic status are used. It is plausible that the different results for urban and rural areas are due to positional and linkage errors in the data. Healy et al. (4) show that in rural areas, the mean distance error for access to the closest hospital is 3285 metres compared to a mean of 414 metres in urban areas. In the context of health policy planning, an error of over 3 kilometres in the estimate of the distance between someone's home and a hospital's location is not acceptable.

Moving forward with health geographic research in Canada

The upcoming changes to Canada Post have the potential to bias studies using postal codes and census or national health survey data. We believe these changes have broad implications for the health of all Canadians and should be addressed by the research community, Statistics Canada, and Canada Post. Canadian researchers should study potential positional and linkage errors due to the implementation of community mailboxes. We believe that the 11 pilot communities can and should serve as an important case study to evaluate potential measurement error and biases in health research that could accompany the transition to community mailboxes.

Conflict of Interest: None to declare.

REFERENCES

(1.) Meijer M, Rohl J, Bloomfield K, Grittner U. Do neighborhoods affect individual mortality? A systematic review and meta-analysis of multilevel studies. Soc Sci Med 2013;74(8):1204-12.

(2.) Cummins S, Macintyre S, Davidson S, Ellaway A. Measuring neighbourhood social and material context: Generation and interpretation of ecological data from routine and non-routine sources. Health & Place 2005;11(3):249-60.

(3.) Canada Post's transition to community mailbox delivery will begin in fall 2014 in 11 communities across Canada. Canada Post. 2014. Available at: https://www.canadapost.ca/cpo/mc/aboutus/news/pr/2014/2014_cmb.jsf (Accessed March 6, 2014).

(4.) Healy MA, Gilliland JA. Quantifying the magnitude of environmental exposure misclassification when using imprecise address proxies in public health research. Spatial and Spatio-temporal Epidemiol 2012;3(1):55-67.

(5.) Duncan DT, Kawachi I, Subramanian SV, Aldstadt J, Melly SJ, Williams DR. Examination of how neighborhood definition influences measurements of youths' access to tobacco retailers: A methodological note on spatial misclassification. Am J Epidemiol 2014;179(3):373-81.

(6.) Postal Code Conversion File (PCCF), Reference Guide. Catalogue no. 92-154-G: Ottawa, ON: Statistics Canada, 2013.

(7.) Pampalon R, Hamel D, Gamache P. A comparison of individual and area-based socio-economic data for monitoring social inequalities in health. Health Reports 2009;20(3):84-94.

(8.) Roos LL, Magoon J, Gupta S, Chateau D, Veugelers PJ. Socioeconomic determinants of mortality in two Canadian provinces: Multilevel modelling and neighborhood context. Soc Sci Med 2004;59(7):1435-47.

(9.) Fuller D, Cummins S, Matthews SA. Does transportation mode modify associations between distance to food store, fruit and vegetable consumption, and BMI in low-income neighborhoods? Am J Clin Nutr 2013;97(1):167-72.

(10.) Shareck M, Kestens Y, Frohlich KL. Moving beyond the residential neighborhood to explore social inequalities in exposure to area-level disadvantage: Results from the Interdisciplinary Study on Inequalities in Smoking. Soc Sci Med 2014;108:106-14.

(11.) Pampalon R, Hamel D, Gamache P, Raymond G. A deprivation index for health planning in Canada. Chronic Dis Can 2009;29(4):178-91.

(12.) Hoek G, Brunekreef B, Goldbohm S, Fischer P, van den Brandt PA. Association between mortality and indicators of traffic-related air pollution in the Netherlands: A cohort study. Lancet 2002;360(9341):1203-09.

(13.) Pampalon R, Hamel D, Gamache P. Health inequalities in urban and rural Canada. Comparing inequalities in survival according to an individual and area-based deprivation index. Health & Place 2009;1-21.

Received: August 1, 2014

Accepted: August 31, 2014

Daniel Fuller, PhD [1] Martine Shareck, PhD [2]

Author Affiliations

[1.] Department of Community Health & Epidemiology, University of Saskatchewan, Saskatoon, SK

[2.] Departement de Medecine Sociale et Preventive, Universite de Montreal, Montreal, QC

Correspondence: Daniel Fuller, Department of Community Health & Epidemiology, Health Science Building, 107 Wiggins Road, University of Saskatchewan, Saskatoon, SK S7N 5E5, Tel: 306-491-1232, E-mail: [email protected]