Optimized probability sampling of study sites to improve generalizability in a multisite intervention trial

Area of Science:

• Implementation Science
• Public Health Research
• Clinical Trial Design

Background:

• Type 2 translation research requires representative study sites and staff to ensure external validity.
• Convenience sampling of study sites limits the generalizability of findings in real-world settings.
• Developing unbiased sampling protocols is crucial for accurate adaptation of evidence-based interventions.

Purpose of the Study:

• To employ an optimized probability sampling protocol for selecting an unbiased and representative sample of study sites.
• To prepare for a future randomized trial of a weight loss intervention by establishing a generalizable study site selection process.
• To improve the external validity of type 2 translation studies through improved site selection.

Main Methods:

• Invited 81 North Carolina health departments within 200 miles of the research center.
• From 30 eligible and interested health departments, all combinations of 6 were generated.
• One combination was randomly selected from those meeting inclusion criteria to form the representative sample.

Main Results:

• Out of 593,775 possible combinations of 6 counties, 15,177 (3%) met the inclusion criteria.
• The selected subset of study sites demonstrated similarity to all eligible sites in health department characteristics.
• County demographics in the selected subset mirrored those of all eligible counties, indicating representativeness.

Conclusions:

• Optimized probability sampling significantly improved the generalizability of study site selection.
• This method ensures an unbiased and representative sample, crucial for real-world intervention studies.
• The protocol provides a robust framework for selecting sites in type 2 translation research.