The uses of spatial analysis in medical geography: a review

Gesler presents a smorgasbord of applications of spatial analysis to epidemiological and public health issues.

Areas of study:

• description of spatial patterns of mortality and morbidity, disease diffusion, disease etiology;
• spatial distribution, location, diffusion and regionalization of health care resources;
• spatial aspects of interactions between disease and health care delivery.

• points,
• lines,
• areas, and
• surfaces.

• Monte-Carlo randomization
• Moran's I Autocorrelation statistic
• Nearest-Neighbor analysis
• Black-White (BW) join count statistics
• Quadrat Analysis and Variance/mean ratio test
• generalized and compound point patterns
• trend surface analysis
• diffusion models
• Knox's method
• many other point and area clustering techniques
• "map comparisons"

p. 968:

""...King...discussed the ... difficulties that arise in geographical epidemiology:

• the necessity for aggregating disease rates over space and time which gains data stability but loses information;
• accuracy of death certificates and diagnoses;
• choice of a suitable rate standardization procedure;
• choice of scale and data classes when constructing maps of disease rates;
• modifiable units; and
• ecological fallacies."

p. 969:

"Medical geographers have often pointed out that spatial patterns or variable associations show striking differences at different scales of analysis. Because of this situation, disease and health care investigations should be carried out at several different geographic levels."

"The problem is a dual one: some processes can generate many spatial patterns, and the same pattern may result from many different processes.... A good example of this, applicable to disease spread, is the difficulty of distinguishing between generalized and compound point patterns."

"In the area of disease causation, Mayer [60] has pointed out that epidemiologists use few spatial techniques and geographers know very little about pathogenesis or biological processes."

p. 970:

"More use might be made of Monte Carlo simulation techniques in assessing point patterns of disease. If the population which contains disease cases is not randomly distributed over space, then it is not statistically valid to compare the distribution of cases with known theoretical distributions like the Poisson, Neyman type A or negative binomial."

"A major aim of this paper was to acquaint the reader with potential use thus make technique more accessible. Utilization of technique, however, can only come with the study of specific procedures andhow they have been employed in research."

Gesler's glossary.