Relative risk estimates from spatial and space-time scan statistics: are they biased?

Area of Science:

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

Background:

• Spatial and space-time scan statistics are established methods for detecting disease clusters.
• While effective at identifying clusters, the accuracy of estimated relative risk within these clusters has not been thoroughly evaluated.
• Understanding potential biases in relative risk estimates is crucial for accurate public health assessments.

Purpose of the Study:

• To evaluate potential bias in relative risk estimates derived from spatial and space-time scan statistics.
• To investigate how statistical power influences the bias in these estimates.
• To compare bias patterns between purely spatial/space-time scan statistics and prospective space-time scan statistics.

Main Methods:

• Analysis of bias in relative risk estimates from purely spatial and space-time scan statistics.
• Evaluation of bias under varying levels of statistical power.
• Assessment of bias in prospective space-time scan statistics.

Main Results:

• Purely spatial and space-time scan statistics show an upward bias in relative risk estimates for low-power clusters, which becomes negligible with medium to high power.
• Prospective space-time scan statistics exhibit a consistent downward bias in relative risk estimates, increasing with the power to detect clusters.

Conclusions:

• The choice of scan statistic method impacts the accuracy of relative risk estimation.
• Researchers must consider potential biases, particularly the upward bias in purely spatial/space-time methods and the downward bias in prospective methods, when interpreting cluster relative risks.