Printable Handouts

PDF

Navigable Slide Index

1. Introduction
2. Trial design
3. Types of randomisation (1)
4. Types of randomisation (2)
5. From individual to cluster randomisation
6. Why cluster randomisation?
7. Contamination
8. Individual vs. cluster randomisation
9. EASYcare study - elderly with geriatric problems
10. EASYcare study - design
11. EASYcare study - individual randomisation?
12. EASYcare study - cluster randomisation? (1)
13. EASYcare study - cluster randomisation? (2)
14. Dilemma
15. Solution?
16. Pseudo cluster randomisation (rando) (1)
17. Pseudo cluster randomisation (rando) (2)
18. Addressing EASYcare problems by psedo cluster
19. Pseudo cluster rando: less contamination
20. Pseudo cluster rando: less selection bias
21. Pseudo cluster rando: better recruitment
22. Conclusion for EASYcare study
23. Efficiency and power of pseudo cluster rando
24. Estimators treatment results
25. Simple, unweighted mean
26. Minimal contamination mean
27. Compromise: weighted mean
28. Sample size cluster randomisation
29. Sample size pseudo cluster randomisation
30. D-pseudo cluster (minimal variance approach)
31. Sample size EASYcare study
32. Cluster randomised design
33. Pseudo cluster design
34. Individual randomisation
35. Sample size comparison
36. Efficiency: contamination 20%, cluster size 10
37. Efficiency: contamination 20%, cluster size 30
38. Efficiency: contamination 40%, cluster size 10
39. Efficiency: contamination 40%, cluster size 30
40. Conclusion
41. Analysis pseudo cluster trial
42. Inappropriate methods
43. Mixed models
44. EASYcare study: variables
45. EASYcare study: dataset
46. EASYcare study: analysis (1)
47. EASYcare study: analysis (2)
48. EASYcare study: analysis (3)
49. EASYcare study: specify fixed part of model (1)
50. EASYcare study: specify fixed part of model (2)
51. EASYcare study: specify random part of model (1)
52. EASYcare study: specify random part of model (2)
53. EASYcare study: request model statistics (1)
54. EASYcare study: request model statistics (2)
55. Analysis program - SPSS code
56. Output: independent variables
57. Output: test of fixed effects
58. Output: estimates fixed effects (1)
59. Output: within and between variance
60. Analysis program: SAS code
61. Literature

Topics Covered

• Individual randomization versus cluster randomization
• Contamination as a possible problem with individual randomization
• Possible problems with cluster randomization: selection bias and slow recruitment control arm
• Pseudo cluster randomization as an alternative method: less contamination, less selection bias, better recruitment and statistically efficient
• Formula to calculate power
• Analysis methods

Links

Series:

• Design and Analysis of Randomized Clinical Trials

Categories:

• Methods
• Pharmaceutical Sciences

Talk Citation

Publication History

• Published on October 1, 2007
• Archived on May 31, 2018

Financial Disclosures

• Dr. George Borm has not informed HSTalks of any commercial/financial relationship that it is appropriate to disclose.