Spatial filtering using a raster geographic information system: methods for scaling health and environmental data

Abstract

Despite the use of geographic information systems (GIS) in academic research, it is still uncommon for public health officials to use such tools for addressing health and environmental issues. Complexities in methodological issues for addressing relationships between health and environment, investigating spatial variation of disease, and addressing spatial demand and supply of health care service, hinder the use of GIS in the health sector. This paper demonstrates simple spatial filtering methods for analyzing health and environmental data using a raster GIS. Computing spatial moving average rates reduces individual affects and creates a continuous surface of phenomena. Another spatial analytical method discussed is computation of exposure status surfaces including neighbors’ influences weighted by distance decay. These methods describe how health and environmental data can be scaled in order to better address health problems. Spatial filtering methods are demonstrated using health and population surveillance data within a GIS that were collected for approximately 210,000 people in Matlab, Bangladesh.

Introduction

An important component of understanding disease is the interaction of humans with their environment. Human–environment interaction can help inform the emergence, resurgence, and distribution of infectious diseases. Environmental phenomena vary in space and therefore when studying environmental determinates of health one should consider the spatial variation of these phenomena (Briggs and Elliott, 1995). Health is affected by a variety of lifestyle and environmental factors including where people live (Scholten and de Lepper, 1991). Characteristics of where people live, including socio-demographic and environmental exposure qualities, offer valuable insight into epidemiological investigations. Because health and environmental issues are complex, it is essential to choose appropriate methods to adequately address these types of problems. Geographic information systems (GIS) are important tools for addressing human–environment health problems because diverse data sets can be analyzed and related to one another. Because of this capability, GIS has reinforced and strengthened an old theme in the medical sciences—it is important to know where a disease occurs. A GIS provides the utility to map health indicators and disease incidence, analyze spatial patterns of disease, and investigate health delivery systems (Garner et al., 1993). A GIS can be used to describe the distributions of social and environmental factors that affect human health (Loslier, 1996).

Despite the use of GIS in academic research, it is uncommon for public health officials to use a GIS to analyze health data (Rushton, 2000). Defining the relationship between health and environment is complex because it is difficult to choose a suitable spatial scale to represent the spatial variation of disease, health, and health seeking behavior (Glenn and Burkett, 1999; Elman and Myers, 1999; Egunjobi, 1993; Matos et al., 1990: Pringle, 1986; Cook et al., 1999; Tursz and Crost, 1999; Kirscht et al., 1976). Various methods have been developed for addressing health and environmental problems at different spatial scales (Gatrell et al., 1996; Bailey and Gatrell, 1995; Haining, 1990; Cliff and Ord, 1981). However, the use of GIS in public health has been hampered by a variety of conceptual and methodological problems (Onsrud and Pinto, 1991). In order for public health officials to benefit from such tools, the methodology needs to be made more accessible to users (Rushton, 2000). Also, quick, reliable, and scientifically valid methods of rapid assessment are needed to assist in health research (Scholten and de Lepper, 1991). Raster GIS, which divides space into discrete units called cells, can be a simple and useful tool for analyzing spatially referenced data. Raster GIS is efficient for managing and integrating diverse data sets including satellite image data. It can be used to scale data by means of spatial filtering methods commonly used to enhance satellite imagery. Spatial filtering can be used to create smoothed maps of health data (Talbot et al., 2000; Rushton and Lolonis, 1996; Kafadar, 1996).

There are several reasons to use spatial filters on data in health-related research. Filters can remove random noise caused by inaccurate records or mislocated cases. Filters can also be used to address influences of neighbors on disease processes. For instance, behavioral practices of a household may be influenced by neighbors (Twigg et al., 2000; Munshi, 1996), which can affect human health. People are exposed to the affects of poor sanitation in their neighborhoods, not just within the household in which they live. Addressing these issues by using simple raster spatial filtering techniques can make a valuable contribution to future health research efforts. This paper demonstrates the use of scaling health and environment data by using several spatial filtering techniques.