Programming the statistical procedures from SAS

Complication versus Length of Stay

Reply
Super Contributor
Posts: 1,040

Complication versus Length of Stay

Hello Team,

 

I have the following information on a cohort of patients:

1)Actual Length of Stay in the hospital(days)

2)Expected Length of Stay(days)

3) Complication date

 

 

Want to know wether for the cases Length of stay is a result complication or

if complication is occuring becasue of longer Length of stay

 

What kind of analysis do i need ot perform? and how should I proceed?

Should I group patients in to two?

group1: (complication date-Admit date) less than Expected Length of Stay

group2: (complication date-Admit date) greater than Expected Length of Stay

 

Thansk

Respected Advisor
Posts: 2,655

Re: Complication versus Length of Stay

What does a plot of actual vs. expected look like?  Before I start divvying up patients into categories, I'd look at the plots, and try to figure out what is going on.  You might also want to plot out days to complication (complication date minus admission date) as a function of both of the previous two variables.

 

Steve Denham

Super User
Posts: 768

Re: Complication versus Length of Stay

As @SteveDenham says, look at the data distributions and plots first.

 

But as a conceptual matter, why can't both be true?   As a matter of common sense, wouldn't some types of complications cause increased length-of-stay?  Especially if the complication occurs near the end of "expected" length-of-stay.  And as a stay lengthens, isn't the per-patient probability of iatrogenic or other-genic complications increase due to a simple increase in exposure?

 

Can this question be reconceived as a study of daily hazard rates for complications (i.e. hazard_rate{I}= probability of complication on day I, having no complications prior to day I).  If the hazard rate goes up, especially for days near and beyond the expected stay, would that suggest a strong effect of length-of-stay on complication liklihood?

Ask a Question
Discussion stats
  • 2 replies
  • 108 views
  • 0 likes
  • 3 in conversation