RSI = ∑ (Actual LOS) ÷ ∑ (Expected LOS)
Each episode has an Actual LOS and an Expected LOS. The Expected LOS is the the calculated average for that type of patient.
The Expected LOS stratifies patients based on age group, DRG, admission source, destination, emergency / elective, and complex / non-complex.
The Relative Stay Index calculation allows hospitals to determine how quickly they are discharging patients in relation to their peers, whilst accounting for casemix. It works by averaging the length of stay for all cases in a particular category (including outliers). This average becomes the "Expected Length of Stay" for any episode in that category in the future. To calculate the RSI, we add up all the actual bed days, and compare that to the total expected total bed days. The resulting value determines whether they are using fewer days than 'expected', overall, or more.
The patients are placed into a number of groupings:
- 0 years
- 1-16 years
- 17-34 years
- 35-49 years
- 50-64 years
- 65-79 years
- 80+ years
- Diagnosis Related Group (DRG)
- Sameday Emergency / Overnight Emergency / Elective
- Transferred in / Not Transferred In
- Discharged home / Dead / Statistical Discharge / Transfer / Transfer from specialist to non specialist hospitals (Based on hospital level assignment - Hospitals are split into 6 'levels'. If the transfer is between one of the top 3 and one of the bottom 3 they are assigned a different RSI to other transfers because they are considered a down transfer .)
Complex case / Non-complex case
(three or more co-morbidities from separate ICD10 Chapters)
- Caretype (http://meteor.aihw.gov.au/content/index.phtml/itemId/584408)
Once the patient’s grouping in each of these categories has been noted, the length of their stay is documented.
As an example, assume that a patient, we’ll call him Bob Jones, has come in for surgery. When he is discharged, the visit is added to the database. His case has the following characteristics:
- Bob is 58, putting him into the “50-64 years” age bracket.
- Bob was in because he has a chronically obstructed airway, or DRG E65B.
- He arrived by ambulance to your Emergency Department at 4pm on Thursday - making him an "unplanned" but normal admission (rather than a transfer)
- Bob was discharged home to recuperate at 3pm on the following Wednesday.
- Bob has chronic pulmonary disease, diabetes, and congestive heart failure. Having three or more diseases coded puts Bob in the 'complex' category.
- The total time from admission to discharge for Bob was 5 days and 23 hours = 5.958 days (Note that inpatient stay is recorded using decimalised days, rather than whole days.)
All of this information is collated and stored for each patient the hospital treats. Then, if the hospital wishes to determine the expected time of visit for a given group, then the length of stay for all patients, except same-day patients, in that group is averaged. The advantage of the system is that the grouping can be expanded or narrowed depending on the information desired – one of those categories could be used, or all, or just some of them. In Bob's case, the expected length of stay is calculated to be just over four days.
The data is then collated on a larger scale. Each group’s expected length of stay is calculated, by averaging the time required by equivalent cases - excluding same days. (NOTE: Episodes that contain any HITH days are assigned an Expected LOS equal to the Actual LOS, so these cases have minimal affect on the overall RSI.) Once they’ve done so, an individual hospital’s expected length of stay can be determined by tallying the expected bed days for their casemix.
As an extremely simple example, let’s assume a hospital has three types of patient: A, B, and C.
- Patient type A has an expected length of stay of 4.2 days
- Patient type B has an expected length of stay of 2.5 days
- Patient type C has an expected length of stay of 3.1 days
Now, these are the patients that the hospital just discharged:
|Actual Length of Stay||3.9||4.1||4.3||2.0||2.6||3.0||19.9|
|Expected Length of Stay||4.2||4.2||4.2||2.5||2.5||3.1||20.7|
The actual total length of stay was 19.9 days, a little less than the expected total length of stay (20.7).
The Relative Stay Index is a measure of how quickly a hospital is discharging patients compared to their peers, adapted for casemix - hospitals with a bed day count lower than the expected value are assumed to be doing better. It is calculated by dividing the actual length of stay by the expected length of stay – the lower the number, the better (if the number is below 1, they are doing better than ‘expected’). Therefore:
|LOS ELOS||=||19.9 / 20.7||=||0.96|
This hospital’s RSI is 0.96; lower than expected.
Inclusions & Exclusions
Include episodes where caretype='A' (Acute), or Caretype = 'P' (Psychogeriatric) AND the left letter of the DRG is "U". Exclude episodes with hospital in the home time, exclude where the DRG is classified as a 'same day DRG' and exclude 'Error' DRGs (800, 900 series), and also exclude healthy babies (P68D). Exclude patients that were never admitted to ward (i.e. exclude patients who were discharged directly from ED.)