Hello Md. Saiful Islam ,

Yes, you can obtain the adjusted odds ratios in spss.

1. Run logistic regression with all of the relevant IVs included, and the DV of interest.

2. The "adjusted" odds ratio for a given IV is, simply, exp(B). That is, the exponentiated logistic regression coefficient for that IV. This gives the expected/predicted change in odds of a case being in the target outcome level per unit change in that IV. holding all other IVs constant. Many people just report exp(B) as "OR" or "Adjusted OR."

Good luck with your work.

Hello Md. Saiful Islam ,

Generally, unadjusted odds ratio (sometimes called "crude" odds ratio) represents the influence of a single independent variable (IV) on the outcome of interest (dependent variable, or DV). For example, are men and women (the single IV) equally likely to have high blood pressure (the DV) when presenting at a physical exam?

An adjusted odds ratio represents the influence of a given IV when controlling for other possible IVs or influences. Using the previous example, if you were concerned that age, body mass index, and level of activity were other variables that could possibly explain the likelihood of having high blood pressure, then a model that incorporated all four IVs would allow you to determine the influence of sex on high blood pressure status with age, BMI, and activity level taken into account. The two values (unadjusted, adjusted) may be quite different, especially if there is overlapping explanatory power for the other variables used in generating the adjusted OR value.

Good luck with your work!

Hello @David Morse,

Thanks for your excellent answer. I am now clear about adjusted & unadjusted odds ratio. But I can't calculate adjusted odds from spss. Please mention the procedure.

The unadjusted odds are without any other covariate, the adjusted odds are after including other variables.

For example if you have Dead or Alive (after 20 years ) as your outcome and three categories of smoking as your key exposure variable (None, Moderate, Heavy) then a model with Age of the respondent as another predictor variable (either as categories or continuous) then the exposure odds are adjusted (taking account of) Age. Statisticians talk about this as conditioning on Age. In a logit model you exponentiate the estimated differential logit to get the relative odds.

Two other things: it makes a lot of sense to choose the base category of the odds with care and not rely on software defaults (eg some alphabetical order). So for me it would seem natural to choose the Non-smoker as the base, so that for example the Mod smoker is 3.2 times more likely to die and the Heavy smoker is 6.1 times nor likely to die than the Non. People have difficulty with odds less than 1, (an odds of 0.33 is of course equivalent to an odds of 3) so I often choose those at lowest risk as the base category, so their odds are set automatically to 1, and all the other relative odds are larger than 1.

Including 'controls' or predictors (to adjust) need to be done with care by thinking about the status of the variables eg concepts like supressor distorter, confounder, mediator etc .

see https://www.researchgate.net/publication/236671039_Epidemiology_An_Introduction

and the chapter on analysis in https://www.researchgate.net/publication/234015155_Health_disease_and_society_a_critical_medical_geography For some debate see http://andrewgelman.com/2013/05/26/how-to-understand-coefficients-that-reverse-sign-when-you-start-controlling-for-things/

Thanks Kelvyn Jones .

Outstanding answer. But I can't analyze adjusted odds from SPSS software.

Md. Saiful Islam , can you say why the binary logistic regression procedure in SPSS can't do this? A quick google search comes up with sites like: https://www.scalelive.com/unadjusted-odds-ratio.html , which is for SPSS, so I assume whatever search you did before asking the question on RG came up with sites like this.

Daniel Wright

Thanks for your suggestion. Actually, I have already searched on Google or Youtube. But these are really confusing.

Hello Md. Saiful Islam ,

Yes, you can obtain the adjusted odds ratios in spss.

1. Run logistic regression with all of the relevant IVs included, and the DV of interest.

2. The "adjusted" odds ratio for a given IV is, simply, exp(B). That is, the exponentiated logistic regression coefficient for that IV. This gives the expected/predicted change in odds of a case being in the target outcome level per unit change in that IV. holding all other IVs constant. Many people just report exp(B) as "OR" or "Adjusted OR."

Good luck with your work.