It's correct the interpretation is different. The Odds Ratio or Opportunity Ratio is interpreted as follows: Consider the following logistic model y = married age schooling, where the odds ratios are as follows:

Married 0.0736466; Age 0.9901606; schooling 1.962392

The model indicates a woman's likelihood of being in the labor market. If you are married you have (0.0736-1) x100 = -92% fewer opportunities to be in the labor market than that of the woman who is single. For the case of age and schooling you can do the same interpretation but you should consider that being age and schooling discrete and non-binary variables changes in a unit of age and schooling in the probability that the event will occur or not will be smaller when The change occurs at the extremes (example 40 to 41 years or 25 to 26 years in age) and the effects will be greater when the variation is in data in the center (example 35 to 36 years) because they do not follow a linear distribution, its logarithmic

I hope my comment helps you.

The UCLA web-page linked below may be helpful.

http://www.ats.ucla.edu/stat/mult_pkg/faq/general/log_transformed_regression.htm

I agree with Stephen. He explained great!!

in logistic regression, dependent variable is binary ( for example 1 for selecting the label of assurance and 0 otherwise) 

interpretation of coefficients in the logistic model :

when variables are logged :

1 unit change in explanatory variable( logged explanatory  variable)  is associated with a B (coefficient) percent change probability of occurrence of a phenomenon (in our example assurance selecting) ----> unit-percent in log case

when variables are Not logged:1 unit change in an explanatory variable( non-log)  is associated with a B (coefficient) unit change probability of occurrence of a phenomenon ------> unit - unit in non log case

I hope it be usefull,

bests,

Ehsan.