Solving the cluster puzzle: clues to follow and pitfalls to avoid

Area of Science:

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

Background:

• Numerous methods exist for disease cluster identification, but few are consistently applied in research.
• Novel statistical methods for disease cluster detection often lack comprehensive evaluation of their power and performance.
• Method selection by users is frequently based on convenience rather than statistical rigor or specific epidemiological contexts.

Purpose of the Study:

• To extend a typology of disease cluster identification methods.
• To critically evaluate a wide range of existing methods, detailing their strengths, weaknesses, and unique attributes.
• To provide evidence-based recommendations for method selection in diverse epidemiological scenarios.

Main Methods:

• Systematic review and critical analysis of existing disease cluster identification methodologies.
• Development and application of an extended typology for classifying these methods.
• Comparative assessment of method performance, statistical power, and suitability for different data types and research questions.

Main Results:

• Identification of key strengths and limitations across various disease cluster detection techniques.
• Demonstration of how arbitrary method selection can compromise study validity.
• Highlighting the need for standardized evaluation protocols for new statistical approaches.

Conclusions:

• A structured approach to method selection is crucial for robust disease cluster identification.
• Understanding the statistical underpinnings and performance characteristics of each method is essential.
• The review offers guidance to researchers for choosing the most appropriate methods for their specific public health investigations.