The best way to reduce error in mapping is to convert from the use of square grids to hexagonal grids. This reduced the 40%+ error generated by apices in the grid to 14%, a more than 25% reduction in total grid mapping and analytics error (this is allowing for the natural +/-3-5% error we often rely upon).
The popularity of this technique is demonstrated by the numbers of downloads my site is getting for the excel sheet I produced in winter of 2003/4. There are also SQL formulas for this for limitless cells, but the traditional method uses a hexagon overlay for your analyses, calculated using the lat longs for your area, as defined by this page.
To state bluntly what this method corrects for, if we are using grid analysis for intervention or surveillance (determining causality), there is error in our point-area relationships. 41% of the area mapped and assigned a point value for a findings may be assigned to the wrong centroid due to edge-centroid areal trigonomic or geometric features. This math rule applies to all mapping techniques on a surface (flat or curved) with square grid monitoring techniques.
Hex grid approaches to flat surfaces and sphere more accurately represent the area below, and produce results at least 25% more accurate than traditional grid mapping techniques.
We haven’t used hex grids in the past due to ease of producing square grids in GIS. Manually or even pseudo-automatically calculating grids, before excel and the PC days, was a time (in not labor and thought) intensive process. These limits no longer exist today. So why do we keep using this method capable of generating so many errors?
Another question to ask is do you wish to produce contour or isoline related risk analyses products? If you do, then hexagonal grids produce more accurate and understandable contour images (the lines are smoother and more real).
In general, I have asked – which cultures or countries excel in Medical GIS and especially the use of innovative ways to produce your results?
The stats for my site indicate US visitors are the most frequent, but based on feedback from emails, comments, etc., Great Britain and Canada are most engaged in the use of this new GIS technique.
Moreover, based upon my years of researching geographic medicine in general, the support of geographic approaches to health and disease, in a statistical spatial sense, is mostly a skillset adhered to and frequently used by geographers in Great Britain. I blame this on the lull in interest in "Geography" that this country had through the mid to late 20th century.
The largest support of this method at the professional level, from Canada, is perhaps the result of an offshoot of British culture and traditions into Canadian academic patterns and behaviors. Fortunately, Canadians make excellent use of this more accurate technique, the respondents tell me.
There have been very few outcomes of hex grid work in general in the refereed, academic publications on GIS. I have also found it hard to convince the companies to accept my help in making this a standard tool available to GIS technicians. (But I admit I am also very brazen about this result of professional jealousy, now 20 years into this profession.)
It is important to note–that two changes related to GIS need to happen in medical GIS to make it more accurate, and useful at the professional level.
1. Agencies, organizations, insurance companies, PBMs have to become fully engaged in GIS and spatial analyses–not just for that occasional research project (which is now the case), but at the 1000s of analytics per year, for each program they attend to. Managed Care insurance agents and facilities have to catch up with this technology. The current outbreaks are happening due to the failure of these programs to evolve over the past decade, change with the time, initiate new thoughts, recruit new thinkers (us GIS pros are under-employed and not at all respected).
2. These companies etc. also have to slow down, hire and make better use of the skills of better trained GIS individuals, check their data and work for errors, change to spatial analyses, and go beyond just "the experimental stage" (a status now 10-15 years old).
It doesn’t help that some major GIS businesses or industries are non-supporting of the new ideas, methods, skillsets, and technology.
It’s been well over a decade since I began promoting this use of GIS at the statistical and administrative-clinical level.
Think of this as getting the wrong address for an emergency call–We don’t want to send our services, products and staff to the wrong address. We need to reduce that 41% error the traditional form of spatial grid analysis has, as much as possible.
My information sheets on Grid Analysis in general, and the hex grid technique I developed more than ten years ago, are as follows:
Grid mapping health and disease in the United States –
2. Grid Cell Analysis (and Ecology) –
3. Sequential Series analyses using Hexagonal Grid Cell techniques –
4. Applying Grids to Managed Care programs Medical GIS –
6. Downloads Page for Hex Grid formulas/calculation sheets (the stats for which are cited above) –