1

I am interested in profiling patients that has high number of hospital visits in a year. However, I noticed that not all hospital visits are equal. That is, an inpatient visit is more expensive than an outpatient visit. In addition, an inpatient visits that has longer hospitalization days are also more expensive. I would like produce a single weighted variable of hospital visit based on average cost. Is there any way for me to do this?

tatami
  • 855
  • 1
  • 9
  • 27
  • Did you find the answer below satisfactory? Alternatively, is there more you would need to solve your problem? – famargar May 19 '17 at 12:57
  • @famargar Unfortunately regression is not an option because cost is not normally distributed. As much as possible, I would like to avoid using cost directly to obtained the weights because it was found to be bias in some aspect (e.g. arbitrary pricing over the period of 5 years). Would PCA work? – tatami May 21 '17 at 12:25
  • If you want to still use cost, you can probably build a regression model using as dependent variable the log of the cost. On the other hand, if you changed your mind about using cost, then the original question should be rephrased, as at this point it is not clear to me anymore what you are trying to achieve. – famargar May 22 '17 at 08:54

1 Answers1

0

It seems like you would like to build a new variable "hospital visit cost" that is the result of a regression model whose independent variables are a categorical variable "hospital stay type" (inpatient, outpatient) and an ordinal variable "number of days hospitalised". And maybe more variables that come to your mind! Given the relative simplicity of the problem, I would expect a multiple linear regression to be a good start as a model.

famargar
  • 851