Some public health interventions cannot be delivered to individuals in isolation. Vaccination campaigns, water fluoridation, health promotion strategies, school-based programmes, and environmental modifications operate at the level of communities.

This chapter explores the methodological challenges of community intervention trials, also known as cluster randomised trials.

Unlike individual RCTs, the unit of allocation is a group - a village, school, workplace, or district. This introduces new statistical and practical complexities.

Key issues include:

* Cluster randomisation and its implications

* Intra-cluster correlation

* Reduced statistical power

* Contamination between intervention and control communities

* Matching and stratification strategies

* Baseline comparability

* Ethical considerations when entire communities are allocated

The chapter highlights the need for larger sample sizes due to clustering effects and careful analytic techniques to account for group-level variation.

Community trials often aim to evaluate real-world interventions. As such, they sit closer to policy implementation than laboratory experimentation.

They demand collaboration across sectors and sustained engagement with local populations.

While methodologically demanding, community intervention trials provide essential evidence for large-scale public health strategies.

They test whether prevention works not just in theory - but in society.

Key Takeaways

* Community trials randomise groups rather than individuals.

* Intra-cluster correlation reduces effective sample size.

* Contamination between clusters threatens validity.

* Statistical adjustment for clustering is essential.

* Ethical considerations extend to entire populations.

* Larger sample sizes are typically required.

* Real-world conditions enhance external validity.

* Community trials inform policy-level interventions.

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit drmanaankarray.substack.com/subscribe