Epidemiology for public health practice: The application of spatial epidemiology

Abstract

Spatial epidemiology is the description and analysis of geographic patterns and variations in disease risk factors, morbidity and mortality with respect to their distributions associated with demographic, socioeconomic, environmental, health behavior, and genetic risk factors, and time-varying changes. In the Letter to Editor, we had a brief description of the practice for the mortality and the space-time patterns of John Snow's map of cholera epidemic in London, United Kingdom in 1854. This map is one of the earliest public heath practices of developing and applying spatial epidemiology. In the early history, spatial epidemiology was predominantly applied in infectious disease and risk factor studies. However, since the recent decades, noncommunicable diseases have become the leading cause of death in both developing and developed countries, spatial epidemiology has been used in the study of noncommunicable disease. In the Letter, we addressed two examples that applied spatial epidemiology to cluster and identify stroke belt and diabetes belt across the states and counties in the United States. Similar to any other epidemiological study design and analysis approaches, spatial epidemiology has its limitations. We should keep in mind when applying spatial epidemiology in research and in public health practice.

Key Words: Diabetes mellitus, Spatial epidemiology, Diabetes belt, Public health practice

Core Tip: This is a Letter to the Editor on the article published in World Journal of Diabetes 2021; 12: 1042, entitled: Spatial epidemiology of diabetes: Methods and Insights. Spatial epidemiology is a new sub-field of epidemiology. We read with great interest this paper, and would like to further address the application of spatial epidemiology for public health practice.

TO THE EDITOR

Dear Editor, we read with great interest the recently published review paper, entitled “Spatial epidemiology of diabetes: Methods and Insights” by Cuadros et al[1], in World Journal of Diabetes 2021. The investigators reviewed spatial methods used to understand the spatial structure of the disease and identify the potential geographical drivers of the spatial distribution of diabetes mellitus. Their report serves as a good review on the concepts and methods of spatial epidemiology. In this letter we aimed to briefly address few examples of the historical public health practice in the United Kingdom and the application of spatial epidemiology in the recent decades in the United States, and to address the potential limitations when applying this technique in research and in public health practice.

The method used in “analysis of geographic variation in disease” could be tracked back to more than 150 years ago, for example, the renowned study of cholera epidemic in London, United Kingdom In 1854. John Snow, a physician, used mapping approach to trace the source of the Broad Street cholera outbreak (or Golden Square outbreak) in central London[1]. In the United States, an example is that a “stroke belt” or “stroke alley” was identified in early 1980s using spatial analysis approach and to define a 11-state region, where the states had age-adjusted stroke mortality rates more than 10% above the national average[2,3]. In 2011, a study by Barker et al[4] identified a geographically coherent region of the United States, where the prevalence of diagnosed diabetes mellitus is especially high. This area is also known as the “diabetes belt”. The “diabetes belt” consisted of 644 counties in 15 mostly southern states. A further analysis indicated that the prevalence of obesity and sedentary lifestyle (two modifiable risk factors for diabetes) was significantly higher in the diabetes belt than in the rest of the United States[5].

However, it should be noted that similar to the other analytical techniques, spatial epidemiology also has potential limitations[5,6]. First, the basic analysis approach of spatial epidemiology is based on ecological analysis design. Exposures and responses are measured only for aggregates rath than individuals. Therefore, findings from the analysis are subject to have ecological fallacy[5-8]. For example, results from an ecological analysis suggested that there was a significant correlation between increased state-level stroke prevalence and state-level stroke mortality. However, of the study states, several states that had higher state-level stroke prevalence rates did not have high stroke mortality rates, which led to a relatively weaker association than results from analyses using individual-level data[2]. Second, most spatial epidemiological studies apply age-adjusted rates to examine and map the variations in disease rates across geographic areas, such as neighborhoods, communities, districts, counties, states and countries at a global level. However, the calculation of age-adjusted rate is based on the proportion of age distributions across the geographics defined areas. If the proportions of age distributions vary widely between the comparison areas or regions, a simple weighted age-adjusted rate may be meaningless and may lead to an inappropriate comparison[5]. Third, data from disease registries, such as a small regional cancer registry, disease surveillance, or data from hospital electronic health records in a specific township is susceptible to information bias as a result of limited sources. Fourth, data protection and confidentiality should be kept in mind, specifically if mapping disease across small areas, such as small neighborhoods. It is likely that the number of persons at risk (i.e., denominators) and the number of cases (i.e., numerators) are too small to be used[8,9]. In the situation, a combined sample should be considered[10]. Lastly, confounding effects on the study association between exposures and outcomes should be considered and controlled appropriately in spatial epidemiological study.

In conclusion, the application of spatial epidemiology plays a pivotal role in advancing our understanding of the geographic distributions of specific disease and disease risk factors, which significantly contributes to disease control and prevention at population and community levels. However, the limitations of the study design and analysis approaches should be kept in mind when applying it in research and in public health practice.