Right Heart Catheterization Dataset

This dataset was used in Connors et al. (1996): The effectiveness of RHC in the initial care of critically ill patients. J American Medical Association 276:889-897. The dataset pertains to day 1 of hospitalization, i.e., the "treatment" variable swang1 is whether or not a patient received a RHC (also called the Swan-Ganz catheter) on the first day in which the patient qualified for the SUPPORT study (see above). The dataset is suitable for use in papers submitted in response to the call for papers on causal inference, by the journal Health Services and Outcomes Research Methodology. The original analysis by Connors et al. used binary logistic model to develop a propensity score that was then used for matching RHC patients with non-RHC patients. A sensitivity analysis was also done. The results provided some evidence that patients receiving RHC had decreased survival time, and the sensitivity analysis indicated that any unmeasured confounder would have to be somewhat strong to explain away the results. See Lin DY, Psaty BM, Kronmal RA (1998): Assessing the sensitivity of regression results to unmeasured confounders in observational studies. Biometrics 54:948-963 for useful methods for sensitivity analysis, one of which was applied to the RHC results.

Here is a .zip containing the original SAS code used to do the published analyses, from Charles Thomas. A reverse chronological directory of the SAS code files is here.

The S-Plus dataset is easy to use if you have the Hmisc library in effect. If you don't, you need to define the following S-Plus functions.

ddmmmyy <- function(x)
{
	y <- month.day.year(trunc(oldUnclass(x)), attr(x,"origin"))
	yr <- y$year
	m <- c("Jan","Feb","Mar","Apr","May","Jun","Jul","Aug","Sep","Oct",
		"Nov","Dec")[y$month]
	ifelse(yr<1900 | yr>=2000, paste(y$day,m,yr,sep=""),
		paste(y$day,m,yr-1900,sep=""))
}

"[.labelled"<- function(x, ...) {
  lab <- attr(x, "label")
  x <- NextMethod("[")
  attr(x, "label") <- lab
  if(!inherits(x,'labelled'))
    attr(x,'class') <- c("labelled", attr(x,'class'))
  x
}

as.data.frame.labelled <- function(x, ...) {
  y <- x
  cy <- attr(y,'class')
  cy <- if(length(cy)>1) cy[cy!='labelled'] else NULL
  if(length(cy)==0) cy <- NULL  # handles wierd case e.g. class=rep('lab..',2)
  attr(y,'class') <- cy
  d <- data.class(y)
  methodname <- paste("as.data.frame", d, sep = '.')
  if(exists(methodname, mode = "function"))
            (get(methodname, mode = "function"))(x, ...)
  else {
	if(options()$check)
		warning(paste("no method for coercing",d,"to data.frame"))
	as.data.frame.AsIs(y, ...)
  }
}

August 11, 2000

Division of Biostatistics and Epidemiology
Department of Health Evaluation Sciences
University of Virginia Health System

This documents the SAS transport data file and the S-Plus data file for The Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments (SUPPORT).

1. First unzip either of the two files
2. The SAS file may be transferred to another computing platform using SAS PROC CIMPORT.
3. The S-plus data file was created using S-plus function data.dump, so it may be restored by using function data.restore to put the data into your local _Data directory.
4. The variable descriptions are presented in the following table.

Table 1: SUPPORT Variable Description