Geocoding and public health

I think Geocoding is one of the most central issues in GIS (science and systems) and yet is probably one of the less well understood issues for lay people and non-experts who use spatial data (along with map projections). As the authors mentioned, there has been research in public health and epidemiology on the accuracy of street addresses. In fact, a PhD student in my lab lead a research project on this very issue here in Montreal (Zinszer et al. 2010, cited below). The team found that address errors were present in about 10% of public health records, the same ones that were used to perform spatial analysis to look for space-time clustering of campylobacteriosis in Montreal. Geocoding has all kinds of repercussions in public health research; while errors are an issue, anyone who performs epidemiological research with administrative databases are prepared to have some amount of error. However, when the error becomes differential with respect to some factor of interest, this can result in a huge problem (bias). For example, as mentioned in the discussion, the accuracy of geocoding was differential between urban, suburban, and rural areas. There is a lot of spatial epi research done with the urban-rural health divide in mind, and differential accuracy in geocoded addresses like this could pose a huge problem. I think papers like this one by Roongpiboonsopit and Karimi are very useful for people outside of GIS because they help us understand the scope of the issue. I also think Roongpiboonsopit is a super awesome name.

Zinszer K, Jauvin C, Verma A, Bedard L, Allard R, Schwartzman K, de Montigny L, Charland K, Buckeridge DL (2010). Residential address errors in public health surveillance data: A description and analysis of the impact on geocoding. Spatial and Spatio-temporal Epidemiology. 1(2-3): 163-168.



Comments are closed.