I have a cohort study where I have to estimate the incidence rate of clinical malaria and its changes over follow up period in STATA...from the years 2000 to 2002 875 infants were enrolled, study finished in 2004; observation ends at the time of first episode of clinical malaria or exit from the study.
I am asked to estimate the incidence rate and its variation over time.
Thanks!
Dear Maja,
I hope I'm able to help...
First you need to identify whether your cohort study is "prospective" or "retrospective". Then you need also to define whether the time from enrollment to "outcome" is important or it does not matter...
I asking you this because, if the time to outcome is not important, you can easily do a simple logistic regression where you get the odds ratio (rather than the Rate Ratio). Usually, in retrospective cohort studies, the factor time is more difficult to determine, so often the logistic regression is chosen compared to other regression methods.
If you're study is "prospective", you can easily define the time, and the time factor is important, then you have to choose between "Cox" or "Poisson" regressions. The first one is chosen when you expect that the rates rapidly vary overtime, the second when you expect that the rates are rather stable overtime.
Whatever you use (either Cox or Poisson) the first thing is to define the time period of interest. This can be time from enrollment, date of birth, or age at enrollment, depending on what you're looking for. Lexis expansion is also often used.
Then, you proceed with your regression analysis.
I hope this helps
Good luck
Rodolfo
Dear Rodolfo,
thank you very much...my thoughts went the same way; the study is "retrospective" although the time component is very important in it...detection of cases took place between year 2000 and 2002 and they were followed up for another 2 years. I also have information on first detected episode of malaria and data on reoccurrence of clinically detectable malaria during the follow up period.
I think regression should be used as baseline analysis method and Cox/Poisson for further analysis; as there are many possibly explanatory variables in the data I assume associations should be explored as well.
Thanks for help!
With regards,
Maja
Hi Maja,
Happy if I’ve been of some help.
For a disease like malaria, I would use Cox regression since you can expect rapid changes in rates (just imagine seasonality).
However, you use a regression method (e.g. Cox) if you want to compare various factor associated with the rates of malaria; for example, you want to see if children sleeping under a mosquito net are more or less likely to get malaria compare to children not sleeping under a mosquito net, etc. In practice, you obtain a Ratio measure. With Cox regression, moreover, you can add various variables in the model (e.g. mother’s education, child age, etc) and obtain “adjusted” rates ratios.
However, if your study is not “analytical” (i.e. you don’t compare factor that may be associated with the rate of malaria), you can calculate in STATA simple rates (not ratios), or you can do a survival analysis.
Good luck
Rodolfo
Dear Maja, there are couple of good epidemiologists in Nacionalni institut za javno zdravlie in Ljublana. They are training them in a field epidemiology program EPIET and (on my opinion) they may be able to assist you. Kind regards
Vladimir
I am interested in the same problem, so I would be grateful if you share your ideas about solution in stata. What analysis did you use in stata?
Dear Palve,
In order to better help, it would be good to understand your research question and the methodology you intend to use.
Cheers
Rodolfo
Ha Maya, funny to see you on here and asking questions about the assignment. Cox regression won't run for the second dataset, so I'm probably going to use a poisson random effects model :-)
Hi Lisa, I will use first dataset _all...so I guess this is suitable to fit in Cox.
Ah okay! If you're only looking at first episode of clinical malaria then Cox is fine I think :-)
- Badges
- Science topic
- Similar topics
- Medicine
- Epidemiology