Methodological problems in investigating disease clusters.

Increased reporting of disease clusters is stretching health department resources while failing to uncover many significant environmental hazards. By reviewing the methodology by which health departments respond to such reports, we argue that there is a strong bias against investigating the most hazardous situations. Failure to find serious hazards is more a reflection of the methodology used than an assessment of the existence of environmental problems. The biases arise from the selection of situations that dictate full in-house consideration, the selection of the methodology used to evaluate reported or registry-based data, and the interpretation given to anomalous observations. We argue for more rigorous evaluation of statistical cluster detection methods and for use of more objective criteria in cluster investigation protocols.