Binomial Logistic Regression with repeated measures

#1
Hello everyone,

This may be a simple question for you all to answer but I am fairly new to the world of scientific research and statistical analysis as a medical doctor.

I want to review whether a blood marker (called CRP) can be a good predictor of sepsis. I have 157 patients, but each one had numerous episodes of suspected sepsis, therefore there are 547 CRP results to analyse.

I was going to run binomial logistic regression with my outcomes as Sepsis = 1, No sepsis = 0. And then have the blood results as categorical variables (e.g. CRP<5 = 5, 6-9 = 4, 10-19 = 3, etc), as well as other variables such as age and gender.

However as mentioned some of my results will be from the same patients, in both groups (e.g. one patient may have had an episode of no sepsis, but then a few months later had an episode of proven sepsis). Although these results would be taken at a completely different time and during a different episode of sepsis.

I heard about multilevel binomial regression analysis, and have inserted a sample of my data in SPSS, involving 4 patients, each that have had between 2-3 episodes of infection (therefore 2-3 CRP results each). I’ve coded each patient 1-4, and put each infection episode into separate groups. So for patient 1, their results were 36.7, 118.2 and15.5. There is a gap in cell 11 as that patient only had 2 samples.

Is this a more appropriate method to adjust for the repeat measures than standard binomial logistic regression?

Any advice would be much appreciated!

Screenshot 2020-05-13 at 23.32.58.png