Article Text

Download PDFPDF

Geographic information systems and spatial analysis: a statistical commentary
  1. Molly P Jarman1,2,
  2. James Byrne3
  1. 1Center for Surgery and Public Health, Brigham and Women's Hospital, Boston, Massachusetts, USA
  2. 2Surgery, Harvard Medical School, Boston, Massachusetts, USA
  3. 3Division of Acute Care and Adult Trauma Surgery, Department of Surgery, Johns Hopkins School of Medicine, Baltimore, Maryland, USA
  1. Correspondence to Dr Molly P Jarman; mjarman{at}bwh.harvard.edu

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Geography is a critical determinant of outcomes following traumatic injury. Time to care is well established as a key modifiable variable to improve clinical outcomes, and geography is the most significant driver of prehospital time.1 Kumar et al highlight the role geographically defined policies play in determining trauma outcomes—specifically, how policies prohibiting or encouraging transportation of critically injured patients across state lines might hinder or improve access to trauma care.2 This study also calls attention to the need for thoughtful approaches to spatial statistics in trauma research. Incorrect application of spatial analyses could misclassify accessibility of trauma care for defined geographic regions or introduce bias when estimating the relationship between access and clinical outcomes. In that context, this commentary serves as a primer on geographic information systems (GIS) and spatial statistics for trauma researchers.

GIS is the framework for gathering, managing, and analyzing geographic data. The addition of geographic coordinates (latitude, longitude, and elevation) to data lets us visualize locations, generate variables such as travel time and distance, and link observations based on their proximity to each other. Geographic data are roughly categorized as points, lines, and polygons. Point data include the granular locations of observed exposures, events (eg, assaults, car crashes) and outcomes represented by geographic coordinates. In trauma research, lines usually represent transportation routes described in terms of time or distance. Polygons are enclosed geographic areas demarcated by line segments and described by attributes of the region, such as a census tract with associated population characteristics. Sometimes, point data are used to represent the geographic or population-weighted centers of polygons for the purpose of analysis. In their study, Kumar et al combine these forms of geospatial data to evaluate drive times (lines) between trauma centers (points) and the population-weighted centroids of census block groups (points representing the centers of polygons). In doing so, they were able to identify populations for whom the nearest trauma center was out of state, allowing for new insights not otherwise available. Fortunately for trauma researchers, many commonly used data sets incorporate spatial data, including motor vehicle crash location data in the Fatal Analysis Reporting System3 of the National Highway Traffic Safety Administration and neighborhood social determinants of health data collated by the Agency for Healthcare Research and Quality.4

Once data are imbued with geographic characteristics, scientists can use spatial statistics to examine relationships between both geographic and non-geographic measures. Geographically proximal observations are inherently correlated. This correlation can be an asset to research. For example, if we suspect severe car crashes consistently occur in the same location(s), we can use spatial statistics to determine whether they are geographically clustered (ie, occur more frequently in one location than would be expected by random chance), thus identifying highway segments in need of redesign.5 Unfortunately, spatial correlation also poses challenges for analysis by violating assumptions of independence underpinning many statistical models, and thus introducing bias into studies when not handled properly. For instance, trauma patients from the same neighborhood will experience similar prehospital transport times, and likely also have similar socioeconomic characteristics, which may amplify or obscure the observed association between prehospital time and clinical outcomes.6 Hierarchical regression models and other statistical approaches designed to account for correlation between observations can minimize these biases.

Kumar et al exemplify a thoughtful approach to geographic analysis and spatial statistics in trauma research. We encourage our research community to advance the science of trauma care by incorporating similarly thoughtful spatial methods into their analyses.

Ethics statements

Patient consent for publication

Ethics approval

Not applicable.

References

Footnotes

  • Contributors MPJ drafted the commentary. JB provided critical review of the commentary.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Provenance and peer review Commissioned; internally peer-reviewed.

Linked Articles