Hello,
The researcher plan to conduct a (RCT) to compare a control group with an intervention group. They have provided a (MCID) for the length of stay (LOS) of one day. However, the standard deviation is extremely high at 120, resulting in an unreasonably large sample size requirement of approximately 500,000 participants, which is not feasible for a prospective RCT.
The issue arises from the high variability in LOS.I have set the power at the conventional level of 0.8 and the significance level (α) at 0.05.
could you please help me how to address this issue?
Thank you!
The obvious way is to include a number of covariates that predict length of stay. I'm not an expert on your domain but I imagine things like age, severity of illness and so forth. The beauty of an RCT is that the expectation of the outcome is not altered by including these covariates so they just reduce the error term. This is why ANCOVA is often preferred for these designs.
Thank you for your answer, yes narrowing down the characteristics of the trial group to bring down the SD might be an option, but This would reduce the generalizability and could make recruitment difficult.
I suggested running a pilot study instead of a full RCT because a power analysis is a necessary requirement for a full RCT. Given the high variability in LOS, the required sample size would be extremely large, making a full RCT impractical at this stage.
- Badges
- Science topic