Table 5.
Example: effect of increasing the cluster size (M) given a fixed number of clusters (k) under the three designs
| Design constraints |
CRT |
CRT-BA |
CRT-SW |
|||||||
|---|---|---|---|---|---|---|---|---|---|---|
| k | ρ | M | N | Power (%) | N | Power (%) | N | t | m = M/(t + 1) | Power (%) |
| 10 | 0.01 | 100 | 1,000 | 61 | 1,000 | 49 | 1,020a | 5 | 17 | 55 |
| 10 | 0.01 | 300 | 3,000 | 78 | 3,000 | 87 | 3,000 | 5 | 50 | 91 |
| 10 | 0.1 | 100 | 1,000 | 16 | 1,000 | 41 | 1,020a | 5 | 17 | 49 |
| 10 | 0.1 | 300 | 3,000 | 16 | 3,000 | 83 | 3,000 | 5 | 50 | 90 |
Abbreviations: t, number of steps in the stepped wedge design; ρ, intracluster correlation coefficient; CRT, cluster randomized trial; CRT-BA, cluster randomized trial with before and after observations; SW-CRT, stepped wedge cluster randomized trial.
Example relates to a trial requiring 786 (z-test) observations under individual randomization.
a
Total sample size is 1,020 due to constraints with m having to be a multiple of (t + 1).