-1
$\begingroup$

I have a data set for patients visiting emergency departments containing following features:

enter image description here

The output variable in this data set is "disposition" - whether a patient becomes admitted or discharged. I would like to predict the probability that a patient comes back as well as how long the patient stays.

I am wondering whether this is possible since the only output variable I have, is the disposition.

Do you guys think it is possible to predict the re-admission probability and the length of stay for a patient based on the features mentioned above?

$\endgroup$
  • $\begingroup$ 'Disposition' is not in your list of variables above. Do you also have data on length of stay and whether they were re-admitted? If not, you don't have any information with which to build a model. $\endgroup$ – mkt Feb 24 at 11:38
  • $\begingroup$ The first variable is Disposition. No I don't have data on that. I was thinking whether it is possible to build some unsupervised model. What do you think? $\endgroup$ – cherryp Feb 24 at 13:30
  • $\begingroup$ Ah, I missed that. No, you cannot answer the question you are interested in without relevant data. The most you could probably do is identify whether there are clusters in your data. $\endgroup$ – mkt Feb 24 at 14:24
  • $\begingroup$ Alright. Thanks for the input. Appreciate it! $\endgroup$ – cherryp Feb 24 at 14:39
0
$\begingroup$

If you don't have data on length of stay or readmission, how could you possibly model them?

I mean, you could make some guesses - e.g. older patients are more likely to be readmitted - and you could look at other studies of this (there are lots) and base it on that, but for your data, you wouldn't even be able to tell if you were correct or not.

Also, it's hard to imagine that you have only the features listed. What about the reason they were admitted, the reason they were discharged, whether they died ... and so on?

$\endgroup$
  • $\begingroup$ Thanks for your response. I was thinking whether it is possible to build some unsupervised model since we don't have the target value. yes there is a bunch of more features - binary features representing diseases. I don't have the data whether they died or not. $\endgroup$ – cherryp Feb 24 at 13:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.