Interpretation of logistic regression, continue variable

[annemadewaard]

Hi there,

In my study I look at the number of hospitalization (in the attachted file named 'aantalklinischeopnames') and the possibility to get a complication. The outcome of my binary logistic regression is attatched.
I don't know exactly how to interpretate this. There is a strong significance but the Exp(B) is negative. Does this mean that the less hospitalization you have, the lower the change that you get a complication? And if this is right, how do I say this in my results?

Thanks in advance!

 

[hlsmith]

Correct, for every prior hospitalization patients have a 0.719 times odds ratio for complication. However, it is up to you to understand your study context and whether or not you have some type of selection bias or confounding, if this is a paradoxical estimate.

 

[Karabiner]

Does this mean that the less hospitalization you have, the lower the change that you get a complication? And if this is right, how do I say this in my results?

Provided that in the SPSS analysis the coding was complication=1 and no complicaion=0 .

With kind regards

Karabiner

 

[annemadewaard]

Provided that in the SPSS analysis the coding was complication=1 and no complicaion=0 .

With kind regards

Karabiner

My coding is: complication = 1, no complication = 2.
Do I have to change this for my outcome?

 

[annemadewaard]

Correct, for every prior hospitalization patients have a 0.719 times odds ratio for complication. However, it is up to you to understand your study context and whether or not you have some type of selection bias or confounding, if this is a paradoxical estimate.

Thanks. I always thought when the OR = <1.0, there is no significant connection. So is it right when I say: with every hospitalization the change you getting a complication increases with 0.719 times? Thanks in advance.

 

[Karabiner]

My coding is: complication = 1, no complication = 2.
Do I have to change this for my outcome?

The software uses a 0/1 variable for a binary logistic regression analysis, 1/2 does not work.
Therefore, your 1/2 variable was recoded for the analysis by your software. You have to check your
output, whether for this logistic analysis the software has re-coded the outcome variable into
complication = 1 and no complication = 0, or into complication = 0 and no complication = 1.

With kind regards

Karabiner

 

[annemadewaard]

The software uses a 0/1 variable for a binary logistic regression analysis, 1/2 does not work.
Therefore, your 1/2 variable was recoded for the analysis by your software. You have to check your
output, whether for this logistic analysis the software has re-coded the outcome variable into
complication = 1 and no complication = 0, or into complication = 0 and no complication = 1.

With kind regards

Karabiner

Thank you very much. I changed it en the outcome is much different, I attachted it. So now I can say: with every hospitalization the change of getting a complication increases with 1.4 times? Is that correct?
Thank you in advance!

 

[Karabiner]

How should I know? The crucial part is missing again.
I do not know whether "1" is the code for "complication"
or for "no complication". Regarding the interpretation
of logistic regression coefficients and odds ratios (Exp(B)
is SPSS), please consult an approriate introductory book,
or website or the SPSS manual, if possible.

With kind regards

Karabiner

 

[annemadewaard]

How should I know? The crucial part is missing again.
I do not know whether "1" is the code for "complication"
or for "no complication". Regarding the interpretation
of logistic regression coefficients and odds ratios (Exp(B)
is SPSS), please consult an approriate introductory book,
or website or the SPSS manual, if possible.

With kind regards

Karabiner

I'm sorry, I said I changed it and I assumed you would understand. I apologize. I changed it in: 0 = no complication, 1 = complication.