Introduction

Conservable bed days for day surgery were calculated for each hospital "if their day surgery rate for a particular procedure was below the benchmark rate. The day surgery rate difference (benchmark minus actual day surgery rates) was multiplied by the number of cases. This yielded the excess inpatient cases - i.e. the number of inpatients who might potentially have had their procedure conducted on a day surgery basis" ( ICES, 1996 :199).

E.g. If a hospital had 100 cases annually, 30 of them being day surgery, and the day surgery benchmark was 50, "the LOS of the 20 shortest inpatient stays (minus the average day surgery LOS for the peer group) would be summed to yield the day surgery conservable bed days"( ICES, 1996 :199).

Benchmark Levels for LOS ( ICES, 1996 :198):

• Outcomes within each hospital were initially adjusted for "patient characteristics such as age, distance from home to hospital, and comorbidity. We used the Deyo modification of the Charlson index..."
  
  
• Predicted outcomes were calculated (LOS or rate of day surgery "for the entire population within a peer group after adjusting for the influence of patient, but not hospital, factors".
  
  
• Benchmark levels were defined as "quartiles (25% for LOS, 75% for day surgery) of the predicted population values for these variables", i.e., the benchmark "or 25th percentile for LOS for a given diagnosis indicates that 25% of patients admitted to hospitals within the grouping were at the benchmark LOS or below (adjusted for patient factors)".
  
  
• Benchmark levels for day surgery were similarly calculated, i.e., "the benchmark-adjusted linear predictor in a logistic regression model was determined analogously to the benchmark predictor in the linear regression model for LOS".

PROGRAMMING NOTES

Truncation of LOS:


ICES has noted that excessive length of stay was truncated for their analyses in Chapter 8: "any LOS above the 97.5 percentile for a clinical grouping was considered excessive resulting in the assignment of a LOS value truncated at the 97.5 percentile" (1996:198).

This method does not exclude alternate level of care (ALC) days. Note also how they calculate fractional lengths of stay.

Episodes of care:

The most relevant LOS is "the total length of the hospitalization, regardless of the number of institutions within which this occurred." "An acute hospitalization episode was defined as either an admission to an acute care setting from which the patient is discharged, or a continuous sequence of hospital stays in different hospitals to which the patient is transferred. The LOS for a hospitalization episode is the sum of the lengths of stay for all component hospital stays within the hospitalization episode. We attributed the hospitalization episode to the hospital initially recording the most responsible procedure or diagnosis" ( ICES,1996 :199).

Exclusions:
• "Deaths that occurred in hospital" ( ICES, 1996 :198)(on initial admission if looking at episodes of care).
• For episodes of care:
  • Records for whom a transfer was coded, but no corresponding record was found ( ICES, 1996 :200).
  • Patients were excluded if they were "transferred into the initial institution from another acute, chronic or rehabilitation institution" ( ICES,1996 :200).