Choice of test to assess correlation???


New Member
Hi everyone,

I am considering doing some research into a recent educational initiative that the hospital I work at has put into place to reduce the rate of central line infections.

The initiative (the intervention in this case) was simply to give monthly lectures to the medical residents with instructions on some do's and don'ts when dealing with patients with central lines. These lectures were started in early 2017 and the infection rate (central line infections/patient hospital days) has fallen markedly compared to years before.

I recognise that the intervention has only been in place for a short while but the fall in the rate of infection since its implementation has been impressive.

I wanted to know if there is a way of testing the significance of this relationship?

Thank you very much in advance!


Not a robit
This is harder than it seems - I speak from experience, since i am asked to run similar analyses all of the time.

Are there providers who don't get education (nonteaching service patients)? Do residents rotate so you have different residents on service in different months and years so you can't untangle if it's the residents or intervention?

Also what are the rates, pretty low, since it should be a rare event? Also do you need to control for infecting organism and patient characteristics or is it solely considered of iatrogenic origins?
Last edited:


New Member
Thanks for your response.

The lectures were only given to internal medicine residents (not to residents in other specialties). They were given every month because half of the rotating residents in the medical department were new every month i.e. the lectures were frequent enough that all the medical residents should have attended at least one lecture.

The infection rates were rates for the entire hospital, not just the rates from medical floors. However, I think I could find out how many infections occurred on medical floors and how many occurred on floors covered by other departments.

The rates were high in the baseline year (2016). I don't know what the exact infection/1000 patient-day rate was, but the incidence in that year was in the 30s. In 2017, when the lectures were implemented, the incidence dropped to about half that of 2016. This year, the lectures actually ceased for 3 months (Jan-March), and in that time 6 infections were reported. Lectures were then reinstated from march onward and zero infections have been recorded since then.

A central-line blood stream infection (CLABSI) is reportable when any patient who has had a central line in for 2 calendar days, has positive blood cultures (any organism). This can mean that another source of the infection (say a urinary catheter) or contamination of blood samples with skin flora can lead to a reportable CLABSI. The lectures that were given did instruct on measures to avoid contamination of the central line as well as those to avoid contaminated blood samples in an effort to avoid reporting CLABSIs that were, intact, not CLABSIs.

I hope I have answered all of your questions. Please let me know if there is something I have missed.

Thanks, again.


Not a robit
I would confirm attendance by all residents. If you can break the infection rates into teaching and non-teaching patients, then I would check out an interrupted time series model with non-teaching service patients as the control. If the rates and patients are a little different in non-teaching service that can be fine, since what you are looking for in those patients is for there not to be a drop at the same time there is a drop in the teaching service patients. Parallel drops may mean contamination of the intervention to none targeted physicians or and underlying change in infection risk at the hospital.


New Member
Thanks so much for your advice.

The direction I'm going to take in this project has changed a little. As it seems the change in CLABSI rates was the result of a multifactorial effort in which nursing protocols were better enforced, more thorough nursing simulation training was carried out, and the lectures to house staff, which I've described before, were given.

My research mentor thinks I should first analyse whether the fall in infection rates was significant. How could I go about analysing the fall in rates and what the p-value is? I'm not sure if this is a more or less complicated analyses?

Thanks again!