Please note that I also posted this same question in an SPSS specific forum, however if anyone here can also help, I would appreciate it.
I have a question regarding repeating time-to-events measurements and after a quick search I really did not find anything that directly answered my question. So here goes…
Context (an example of the model I am trying to create):
I have say 25 patients who have all undergone the same procedure and who meet eligibility criteria for a study. Date of discharge from hospital is "Day 0" and I am tracking multiple events following their discharge that cause these patients to be readmitted. The type of event is not necessarily important at this stage (could be anything from infection, to bleeding, to device failure, etc.), as any event causing readmission is classified as 1=true or 0=false. Patients who then receive a subsequent procedure have reached their end-point are censored from that point on. The issue I am having is that each patient may have multiple readmissions following treatment, thus:
Patient 1: day 0-->6* (readmission), day 7-->25* (readmission), day 26 -->65* (readmission);
Patient 2: day 0-->34* (readmission), day 35-->38* (readmission)
Patient 3: day 0-->50*(readmission), day 51-->53**(subsequent procedure), *Day 54-->60(readmission)*
..etc.
where * notes the day being readmitted to hospital. and text following the 'double asterisks' means any readmissions at this point forth are excluded because they have reached end-point. Incidence per patient-year at say day 30 was just a matter of calculation by hand, but if I wanted to perform some sort of Kaplan-Meier survival type estimate or a Cox Hazard regression for risk of readmission using $x$ and $y$ covariates, is that possible to do in SPSS? Even though with 25 patients, the study may be underpowered to come to any meaningful conclusion with further analysis beyond descriptives, is this possible and how?
Thus far, my previous works have been limited to Kaplan-Meier survival, log-rank analysis, and Cox-regressions with single definite event (i.e., death) that is continuous and non-repeating… so I'm treading into new water.
If anyone has any advice or can point to me to specific reading materials, that would be greatly appreciated.
Best Answer
There is a whole theory developed on recurrent events. For some reason these problems and solutions are not often mentioned in survival analysis or reliability courses or books. The same can be said for the theory of competing risks. Anyway let me refer you to the two books i know on this topic. Wayne Nelson's came out first and was published in the ASA/SIAM series. It was the first or one of the first books in that series. Here is the link to more information on it at amazon.com: http://www.amazon.com/Recurrent-Recurrences-Applications-Statistics-probability/dp/0898715229/ref=sr_1_2?s=books&ie=UTF8&qid=1343339216&sr=1-2&keywords=recurrent+events
The second book was by Cook and Lawless and was published by Springer-Verlag: here is an amazon link to it:
http://www.amazon.com/Statistical-Analysis-Recurrent-Statistics-Biology/dp/1441924159/ref=sr_1_1?s=books&ie=UTF8&qid=1343339260&sr=1-1&keywords=recurrent+events