Influence of spatial resolution on space-time disease cluster detection

Area of Science:

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

Background:

• Precise spatial resolution (e.g., patient addresses) is ideal for disease outbreak detection but raises privacy concerns and data availability issues.
• Aggregating spatial data to broader scales (e.g., ZIP code, county) can mitigate privacy risks but may obscure localized high-risk areas.
• Understanding the trade-offs between data precision and privacy is crucial for effective public health response.

Purpose of the Study:

• To experimentally evaluate the impact of spatial data resolution on space-time cluster detection for disease outbreaks.
• To compare the statistical power, sensitivity, and positive predictive values of outbreak detection using address-level data versus aggregated ZIP code and county-level data.
• To determine how spatial aggregation affects the accuracy of identifying outbreak locations and sizes.

Main Methods:

• Extracted administrative medical claims data for 122,500 viral lung episodes in Tennessee (2007-2010).
• Generated 10,000 address-level spatial datasets and corresponding ZIP code and county-level aggregated datasets (30,000 total).
• Utilized the space-time permutation scan statistic and SaTScan™ software to analyze outbreak detection performance across different spatial resolutions.

Main Results:

• The power to detect disease outbreaks was not significantly diminished by using spatially aggregated data compared to precise address information.
• Spatial aggregations negatively impacted the ability to accurately pinpoint the exact spatial location of outbreaks, especially for smaller clusters (<800 km²).
• The likelihood of over-estimating outbreak areas increased with data aggregation.

Conclusions:

• Spatial data aggregation does not inherently lead to a loss of detection power or sensitivity for disease outbreaks.
• The effectiveness of spatial aggregation depends on a complex interplay between relative risk within the cluster and the cluster's size.
• Public health surveillance systems must consider both cluster size and relative risk when interpreting results from aggregated spatial data to avoid misrepresenting outbreak extents.