statistical analysis of audit data - healthcare related

#1
Greetings,
I am trying to discern the best way to represent some audit results for my dissertation.
The first set of data I am trying to reflect is regarding temperature control of patients post cardiac arrest. I have collected the numbers of hours (or times) that the temp of the patient violated outside the ordered temperature range, as well as outside the range that was set by the standards I am auditing against.
Two issues I am having trouble with are
1: There are incidences of no recorded temperatures (should be recorded hourly)
2: Not all patients received the same length in time of temperature control (should receive 24 hours but varies anywhere from 4-24).

Should I reflect the violations to the predetermined ranges as a percentage of the hours actually achieved?
Should I include the omissions of recording as violations to range or as a separate statistic? It seems faulty thinking to assume they are outside of range but impossible to know. Some reviews showed as many as 11 omissions in a 24 hour period.

I am really having trouble figuring out the most meaningful way to represent this data and have it accurately reflect compliance.

I haven't had a lot of luck with contacting my course tutor so would appreciate some advice

thanks
 

hlsmith

Omega Contributor
#2
You can run the analyses a couple of ways, such as intent-to-treat or actual treatment. You can also do sensitivity analysis for the missing data. I would try to parallel what others have done in the literature. I saw a well written comparable pediatric paper awhile back in JAMA (I believe) on patient cooling, where their protocol was also not completely adhered to.
 
#3
Part of my issue is know how far to go with the analysis. I am not trying to measure clinical outcomes only compliance so is a sensitivity analysis going too far? Can I represent the results with a basic mean +/- SD for violations occurring or hours achieved? really quite lost here. I'll definitely read a few more papers for comparison but as they are generally dealing with actual outcomes (for example mortality) I don't know how much parallel comparison I can make.