Concept: Episodes of Care

Introduction

An episode of care is defined as a continuous stay in the hospital system irrespective of transfers.

The macro _EPISODE creates several new variables in the Hospital Discharge Abstract data to identify individual episodes of care. For more information, see the _episode macro in the MCHP SAS® Macros concept.

Process:

• Sort data by individual, date admission, date separation. This puts the separations for each individual in chronological order.
  
  
• Read the data
  
  
  • Keep the prior (lag) values of the following variables where there are multiple admissions/separations for an individual.
    
    
    - individual - used to determine if current record belongs to same person as prior record.
    
    - hospital - used to determine if hospital has changed.
    
    - separation date - used to calculate difference from separation to admission.
    
    - transfer to - used to test transfer to codes for verification.
    
    
  • On the first record for an individual initialize the following variables:

    - lag individual - there is no prior record.
    
    - lag hospital - there is no prior hospital.
    
    - lag separation date - there is no prior separation date.
    
    - lag transfer to - there is no prior transfer to.
    
    - set difference to 999 - indicates no prior value. This is used since missing (.) and 0 actually mean something in some of our data.

  • Test for transfer looking at difference in days of -1 to 1.
    
    
  • Based on the Transfer From variable mark records that should be a transfers.
    
    
  • Indicate transfer types based on difference in days, transfer and hospital codes.
  Transfer types include:

  Verified NON-Transfer	/* Episode = 1 */
  Algorithm ONLY: diff hosp	/* Episode = 0 */
  Verified: trto/trfrom match	/* Episode = 0 */
  Same Hospital	/* Episode = 0 */
  Overlap Transfer	/* Episode = 0 */
  Algorithm ONLY: same hosp	/* Episode = 0 */
  Verified: trfrom only	/* Episode = 0 */
  NON-Verified: Diff >1	/* Episode = 1 */
  NON-Verified: Prev Yr	/* Episode = 1 */
  NON-Verified: Unexplained	/* Episode = 1 */
  NON-Verified: Nursing Home	/* Episode = 1 */
  NON-Verified: Out-of-Prov	/* Episode = 1 */
  NON-Verified: Other;	/* Episode = 1 */

  
  
• Sort data in reverse chronological order for each individual.
  
  
• Read the data in this order to accumulate days and mark the first (last in this sort order) record for an individual.
  
  
  1. Count days in each episode.
     
     
  2. Mark first record (last in this sort order) as 1 - start of episode. This may only be one record if an episode is only one hospital separation long. Mark all other records as 0 - continuation of episode.
  
  
• Re-sort data in original order.