The reasoning: matching design to intervention
| Design | Unit randomised | Use | Key issue |
|---|---|---|---|
| Clinical trial | Patient | Treatment of disease | Standard RCT |
| Field trial | Healthy individual | Preventive interventions (vaccines, chemoprophylaxis) | Large numbers, low event rates; Salk vaccine trial (1954) |
| Community / cluster trial | Group (village, school, PHC) | Interventions acting at group level, contamination risk, herd effects | Intracluster correlation: sample size × design effect |
| Crossover | Individual, as their own control | Chronic stable conditions | Carry‑over, washout, period effects |
| Factorial | Individual, to combinations | Two or more interventions at once | Assumes no interaction (Physicians' Health Study: aspirin × β‑carotene) |
| Pragmatic | Individual or cluster | Effectiveness in routine conditions | Broad eligibility, usual staff (PRECIS‑2) |
| Stepped‑wedge cluster | Clusters cross over at random times | Phased roll‑out of a programme everyone will get | Time trends must be modelled |