Concept: Time To / From Surgery

Introduction

This concept describes how cases were defined and the variables used (surgery date, preoperative LOS, postoperative LOS, hospital type) to calculate pre-operative and post-operative length of stay for hip fracture surgery. It also describes how the output was generated.

A. Case Definition

1. Diagnosis Codes
   
   Primary diagnosis of fracture (ICD-9-CM=820-820.99). Using only primary diagnosis eliminates most of the in-hospital fractures, which would be recorded in other diagnosis fields. In-hospital fractures are usually different from primary diagnosis fractures, in that they tend to occur in older, sicker people. Adjustment strategies can partially account for this, but because they are so different they should either be excluded or included as a special category.
   
   
2. Procedures Codes
   
   Usually up to discretion of PI, but the main ones are ICD-9-CM procedure codes: 78.55, 79.05, 79.15, 79.25, 79.35, 79.65, 81.61*, 81.62*, 81.51**, 81.52#

   *	Discontinued 1990
   **	Revised 1990
   #	Added 1990

B. Variable Definition

1. Surgery Date
   
   
   Using an array, search for the position of the above procedure codes and record the value found in the corresponding surgery date variable.
   
   See Time To / From Surgery SAS Code below for more information.
   
   NOTE: If surgery date is missing or zeros, then admission date is used instead. I have found that date of hip surgery is fairly reliably coded, so admission date is rarely (if ever) needed.
   
   
2. Pre-Op LOS (Delay)
   
   
   This involves calculating the difference between date of admission and date of surgery. The SAS code for this is:
   
   
   delay = surgery date - input(admit date, yymmdd6.);
   
   Note: If admit date is used for surgical date in B.1) above, then DELAY will always equal zero (which could be completely wrong). For this reason you might want to exclude the small number of cases with missing surgical date.
   
   This definition of pre-op LOS only counts days spent in the hospital where surgery ultimately took place. In the case of transfers, there could be one or more days spent in another hospital prior to the surgery stay.
   
   About 24% of hip surgery claims record a Manitoba hospital in the transfer from variable. Not all of these prior records could be found, probably because in some cases the patient was seen in an emergency room for such a short time that the "stay" was never recorded. The receiving hospital, however, records the patient as having been transferred.
   
   
3. Post-Op LOS
   
   
   Calculating the difference between surgical date and date of discharge. The SAS code for this is:
   
   
   pstoplos = input(discharge date, yymmdd6.) - surgical date;
   
   These can get very large. If the patient is sent to long-term care or paneled for Personal Care Home, you could see values in the hundreds. Means can be greatly affected by huge outliers, so trim your data at 90 days (or whatever number you prefer) or use medians.

C. Generating Output

• Proc Univariate
  
  Univariate calculates means and medians fairly rapidly. The SAS code for this is:

  proc univariate data=hips;

  var delay pstoplos;
  title1 'Pre-op LOS, post-op LOS by discharges C90-94';

  run;