Odds ratios for multilevel categorical variable


I have a categorical variable with three levels that represent disease severity (0=no disesase, 1=mild, 2=severe disease) and I have put it into logistic regression model as independent variable. I have obtained significant result for both dummy variables "mild" and "severe disease" and odds ratios of 5 and 8. Model was adjusted for several other variables.
My question would be how to properly interpret these odds ratios? Do they indicate risk in comparison to absence of disease (baseline) or in comparison to other two levels of variable combined? For example, are patients with mild disease 5 times more likely to have tested outcome in comparison to healthy ones or in comparison to non-mild disease patients (healthy + severely ill)?


Omega Contributor
I believe if it was a fixed effect, it gets the typical OR interpretation, but you say when controlling for blank fixed effects and blank random level effects. So the effect holds across groups. If you had a cross level interaction term, then there would have been an interpretation based on group.
I'm sorry, but I did not understand what you mean by typical OR interpretation.

To make things simpler, how could I interpret these ORs if there were only two dummy variables (mild and severe disease) in the model? Does this mean that patients with mild disease are 5 times more likely to have metabolic derangement than healthy and those with severe disease are 8 times more likely to have it than the healthy ones?

Do you mean I should present my data in YES/NO form regarding disease to obtain such conclusion?
Last edited:


Omega Contributor
Yeah. Patients with a mild classification of disease X had a 5 (95% CI: ?.?, ?.?) times greater odds of derangement than patients without disease X.