How to test if gender impacts mortality rates when already considering age

[Mary12]

Hi, I am currently working on a research project where I'm analysing the impact of different factors on mortality.

I have the past 10 years of data from a specific hospital and I have graphed the mortality curve by age separately for males and females. I can see that the one curve lies above the other so there seems to be a difference caused by gender but I want to have some sort of statistical value showing whether the difference is statistically significant.

I have p values showing that if I consider gender on it's own then this is significant (used a chi squared test I found online). But given that I'm also considering age, I want to check if gender still has a significant impact on mortality when I have already taken account of age.

Any help would be greatly appreciated!

Thanks so much.

Mary

 

[hlsmith]

Is mortality binary variable, Yes/No?


If so, you need to use multiple logistic regression and include in the model:


mortality = age + gender + age*gender, the latter being the interaction term. If it is significant the log-odds slopes by gender significantly differ.


Can you provide more information about the study context. I have a feeling there are more things you need to consider. I am an epidemiologist and run data all of the time and per your description, I am wondering if there are biases you need to address or control for in your analytics.


Thanks!

 

[Mary12]

So I grouped the data into age bands and calculated mortality rates for each band. I then graphed this for males and females separately and wanted to see if the two graphs are statistically significantly different. I was considering just calculating chi squared p values for each age band separately but I would've preferred one overall p value.

There's actually a lot of other variables I'm considering (both dependent and independent) but I was just getting stuck on this principle of testing whether gender makes a difference or is it just that the females all happened to be older in my sample.

I haven't done stats in about 6 or 7 years so my knowledge here is lacking and I've been using mostly online tools to help me compute things. I also don't have access to (or know how to use) any statistical software. My knowledge basically consists of Excel which doesn't help much here.

Thanks

 

[hlsmith]

When doing your age stratifications, is there a linear trend based on age and the outcome?

 

[Mary12]

No it's more like cubic or quadratic even. I think it would help if I could find a way to completely strip out the effect of age and then test if what's left between male and female is statistically significantly different but I don't know if that's possible.