2
$\begingroup$

I am looking for most suitable model for count data in the following case:

  • we collect number of patient's visits in a hospital for $t_i$ days ($t_i$ varies across subjects)
  • some patients died during the study time ($t_i$ is the number of days since we started observing them till their death), some did not ($t_i$ is the number of days since we started observing them till their last update)
  • number of patient's visits in a hospital may be equal to 0 (this is not 0-truncated case)
  • there is no explicit information that patient is / is not undergoing i.e. some particular disease which could make him visit hospital at least once or is in a condition that would make him not visit at all

This is not a randomized clinical trial. I have demographical variables in my dataset (age, race, gender) as well as some health-related variables (MBI, has kidney disease, has hypertension, others).

My goal is to examine factors associated with increased hospital visits.

What count data model should be used to model number of visits in a hospital?

$\endgroup$
3
  • $\begingroup$ What questions would you seek to answer/what information would you seek to obtain using this model? $\endgroup$
    – Glen_b
    May 5, 2016 at 4:38
  • 1
    $\begingroup$ There is a lot of modeling of such data in heart failure trials when it is about comparing two treatments for their effect on hospitalisation and death. Lots of research and publications on how to do coming out, but I did not watch this closely. A joint frailty model for hospitalisations (marginally negative binomal) and time to death was one of the options I thought was quite logical. $\endgroup$
    – Björn
    May 5, 2016 at 5:07
  • $\begingroup$ Thank you for your answers! I updated my post with some more details about the dataset and I specified the goal I want to obtain from the analysis. (BTW, I have just came across the article Björn probably meant: Analysing recurrent hospitalizations in heart failure: a review of statistical methodology, with application to CHARM-Preserved) $\endgroup$ May 5, 2016 at 14:44

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.