CTN-0138: adaptation, implementation, and cluster randomized trial of a Community Pharmacy-Based Prescription Drug Monitoring Program Opioid Risk Assessment Tool—a protocol paper

Table 3 Power calculation with different assumptions for intra class correlation and effective size

ICC	Design Effect^a	Total N	Effective N per arm^b	Effective Size^c	Power
0.01	2.24	6600	1471	10%	100
0.01	2.24	6600	1471	8%	100
0.01	2.24	6600	1471	5%	100
0.01	2.24	6600	1471	3%	100
0.01	2.24	6600	1471	1%	97.1
0.03	4.72	6600	698	10%	100
0.03	4.72	6600	698	8%	100
0.03	4.72	6600	698	5%	100
0.03	4.72	6600	698	3%	99.6
0.03	4.72	6600	698	1%	75.5
0.05	7.20	6600	457	10%	100
0.05	7.20	6600	457	8%	100
0.05	7.20	6600	457	5%	100
0.05*	7.20	6600	457	3%	96.3
0.05	7.20	6600	457	1%	57.3
0.1	13.43	6600	246	10%	100
0.1	13.43	6600	246	8%	99.6
0.1	13.43	6600	246	5%	94.7
0.1	13.43	6600	246	3%	78.3
0.1	13.43	6600	246	1%	34.9

^aDesign effect: this is a correction factor that is used to adjust the required sample size due to the clustered randomized design. If a simple randomization is used and the required sample size is N, then to achieve the same power using the clustered randomized design, the sample size should be N times Design effect
^bEffective N per arm: the required sample size to achieve the same statistical power if a simple randomization design is used instead of cluster randomized design. Effective sample size per arm * 2 * design effect = Total N for clustered design
^cEffective size: the anticipated difference between the 2-intervention arm in responder rate

ISSN: 1940-0640