Spatial Errors in Automated Geocoding of Incident Locations in Australian Suicide Mortality Data

Area of Science:

• Public Health
• Geographic Information Systems (GIS)
• Epidemiology

Background:

• Spatial analysis of suicide data is vital for identifying public locations needing intervention.
• Accurate geocoding is essential for reliable spatial analysis of suicide incidents.
• Existing geocoding processes may introduce completeness and positional errors.

Purpose of the Study:

• To assess the extent of completeness and positional spatial errors in geocoded suicide data.
• To compare automated geocoding with an alternate multiphase geocoding process.
• To evaluate the impact of geocoding errors on suicide hotspot identification.

Main Methods:

• Utilized Australian suicide mortality data (2008-2017) from the National Coronial Information System.
• Compared custodian automated geocoding with an alternate multiphase geocoding process.
• Conducted descriptive and kernel density cluster analyses to assess data completeness and positional accuracy.

Main Results:

• The alternate geocoding process improved address matching from 67.8% to 78.4%, reaching 94.6% with manual additions.
• Nearly half (49.2%) of nonresidential suicide locations had coordinates revised by over 1,000m.
• Spatial misattribution and hotspot misidentification were most pronounced at smaller geographic levels.

Conclusions:

• Automated geocoding erroneously assigns nonresidential suicides to centralized locations, distorting cluster identification.
• Refining geocoding processes is necessary for reliable detection of suicide hotspots.
• Findings offer insights for improving spatial accuracy in suicide data analysis.