Dear Mohammad,

I think you should take a look at the "survival analysis" methods (e.g. https://www.cscu.cornell.edu/news/statnews/stnews78.pdf).

Prediction modeling in hospital readmissions is discussed in this article:

http://www.sciencedirect.com/science/article/pii/S0925527315003655

Dear Alexander,

I am familiar with survival analysis and I have used it (competing risk survival) in my analysis to describe the predictors of readmission. However, I am looking for a way to develop a model (e.g. repeated time-to-event) that describes how clustered are the readmission visits. Thanks though for your help!

Dear Sanden,

Thanks for sharing this interesting paper! The use of machine learning is a novel approach for predicting readmissions. However, I have already conducted the analysis about prediction modeling using competing risk survival regression, but i still could not figure out a way to describe the readmission visits, whether they tend to happen together or scattered throughout the year. I was looking at models like multi-state modeling, and other repeated time-to-event models (e.g. Anderson-Gill), but all seem to be predictive models. In other words, all I need is to provide a description of the readmission visit clustering.

At the back ground of all prediction models is the concept of odd's ratio. Use that in the development of your own specific  model.