Help with my research question analysis

Hi everyone, I am a doctor working on a project investigating an intervention we have been using in the intensive care unit in my hospital for treating Covid patients.

I am collecting data on all the patients who have been admitted and look at whether they have had this intervention, whether it worked, how it worked and why it worked.

So I have my whole population and say about half have had the intervention. Some have had it early in their admission some have had it late.

I have made a record of each of these individuals who has had this intervention. They may have had it once or say 10 times or any number of times in between. Theres no limit to how many times we do it. I have say 30-40 individuals who have had this intervention, and they may, on average, have it 3 times.

For each time they have had it I have recorded 3 variables, lets call them A,B,C. And I have measured the variable before and after the intervention - so I can measure a change.

I have a few questions I want to try and find out:

1. Does my intervention, in general, causes changes in my variables?
2. Does the size of any effects (including becoming insignificant ie observing no change) change depending on how many times they have had it before?
IE is the size of any change less the 7th time you do the intervention compared to the first time.
3. Are there groups of patients who don't respond the first time? If so do they ever respond

Because each of individual changes have been done in the same patient, I think I can use some form of paired test. I also think I should break down my data down into groups 1 = intervention 1, group 2 = intervention2 etc. and then do some form of ANOVA to compare differences between my groups - but im not certain.

Any help would be much appreciated!


Less is more. Stay pure. Stay poor.
Did you not submit an analyses plan for all of these questions to your ethics committee a priori to engaging in this? You have a lot of clinical question, which puts you at risk for false discovery, that and you seem to be finalizing questions after seeing your data. You will likely need to make multiple corrections to your significance level to be cautious. Also, i imagine patients were not randomized to treatment, so you will need to address threats of residual bias. Before and after designs can be weak when interventions were not randomized since baseline values may be unbalanced. You need to directly control for those values in analyses. The facilitation of interventions does not seem well-defined. You will need to transparently report all of these threats with your result so your audience knows the results are conditional on them.

Also, you do not mention how many patients did not receive treatment.

Thank you.


TS Contributor
In addition, did you record variables A, B and C from the untreated group also? If yes, when?

With kind regards