Medical student - statistical advice needed please!

#1
Hi everyone,

I'm a medical student currently completing a project and I would appreciate some guidance if possible.

I am looking at a population of around 2000 patients who arrive at the emergency department with chest pain. I am planning on splitting them into primary endpoint/not and secondary endpoint/not. The primary endpoint will be if they were diagnosed with a heart attack and the secondary endpoint will be a thing called a MACE score (this is used to predict patients who are likely to have adverse outcomes).

I have access to the population's cardiac risk factors and their mode of arrival to the ED. The mode of arrival to ED (e.g. ambulance/walk-in) is the main thing I am trying to look at.

What I am trying to figure out is how to compare these variables to see which are associated with the development of the outcomes.

I am not sure if it is best to use a summary odds ratio (with 95% clearance interval) and also include a P value?

Thanks a lot in advance
A confused medical student
 

AngleWyrm

Active Member
#2
The mode of arrival to ED (e.g. ambulance/walk-in) is the main thing I am trying to look at.
What I am trying to figure out is how to compare these variables to see which are associated with the development of the outcomes.
I can compare two properties and say if one is a predictor of another.

In order to use a category as a predictor of membership in another category, two proportions are needed. Since I'm not well versed in the medical field I'll use apples to illustrate:
  • The proportion of rotten apples in the population. Might be expressed as rotten apples per 100 000 apples
  • The proportion of rotten apples with the visible characteristic or category being examined, listed here as apples in wooden baskets
apples.png

When the two proportions are equal, the category 'apples in wooden baskets' has no additional information about the category 'rotten apples.' But if there is a significant difference in the two proportions, then that difference expresses a prediction in membership between the two sets.

For example, let's say I find 150/100k rotten apples in the population, but only 25/100k rotten apples in wooden baskets. Then I have stated a measured observation that the apples in the wooden baskets category are less likely to be rotten apples than what was found in the population.
 
Last edited:

hlsmith

Less is more. Stay pure. Stay poor.
#3
what is the proportions with a heart attack? Better options, but for you I would say slam them all into a logistic regression. Also read westreich and Greenland's paper on table 2 fallacy, so you can kind of understand how results will be biased. Also, whatever you put in the model report. Don't cherry pick them and remove non significant terms after looking at their results.
 

AngleWyrm

Active Member
#4
Hypothesis testing:
H1: There is a significant difference between ambulance and walk-in development of outcomes
H0: There's no significant difference
 
Last edited: