Concept: Expected Length of Stay (ELOS) in Personal Care Home (PCH)

Introduction

Expected Length of Stay (ELOS) is calculated from individuals admitted to personal care homes (PCH) over an extended period of time. The ELOS is calculated using variables on the PCH data that may be related to length of stay (e.g. age, dependency code at admission, gender (Shapiro & Tate, 1988) ).

Data

The most recent personal care home files are used. These files are cumulative over time and thus contain all of the history for each individual that has been in the PCH. The MCHP Long Term Care Utilization Data Description provides additional information about the data, while the information in the MCHP Metatdata Repository Long Term Care Utilization Data (internal access only) contains detailed technical information on the data and variables. When calculating ELOS the data is limited to admissions within a given time period using the date of earliest admission. The statistical records are limited to those with a level to care in the range 1 to 4. Other codes (5+ range) represent respite care or care in a long term facility.

Variables

The following variables have been used or created when calculating ELOS:


• Birth Year as recorded on the PCH header file. This variable is used to calculate the age of the individual at the time of admission. This variable does not contain month or date information therefore the age at admission may not represent the individuals actual age. Since broad age groups were used the error is expected to be relatively small. Linkage to the registry information would provide more accurate date of birth information if necessary. Age at admission is grouped into 4 categories (0-64, 65-74, 75-84, 85+).
  
  
• Individuals are defined using the individual identifier variable found on the PCH file. Individuals separated prior to 1984 may not have a unique coded value. If years prior to 1984 are used then the family registry number and gender can be used. This has the implicit assumption that only the family head and possibly spouse are in PCH at the same point in time. Linkage to the registry to obtain identifiers over time may be a more appropriate method to obtain unique individual identifiers.
  
  
• Date of admission to PCH found on the PCH file.
  
  
• Level of care as defined on the first admission to PCH in the statistical records. Only the level of care at the time of assessment, and the current level of care are available on the PCH header file. The level of care at admission is not found on the header file. Since the level may change from the time of assessment to the time of admission the value that corresponds to the first admission to PCH at level 1 to 4 on the statistical records is used. The level of care at admission is grouped into three categories (1, 2, 3+4). If this code is to be applied for admissions after 1995 the first date of admission at level 1-4 in the statistical file should be used instead of the earliest data of admission. The earliest admission may represent admission to one of the long term care facilities not a PCH.
  
  
• Length of stay in the PCH is defined as the difference in time from admission to the paid to date variable on the most recent (last) record on the PCH statistical file.

Analysis

The SAS procedure LIFEREG was used to model the expected length of stay. The data was censored at 2 years after the date period of interest (i.e. 85-95 was censored at 1997/03/31). The data was time censored in this way so comparisons could be made with prior studies, it is not necessary to do this except for the most recent range of years where there was no data available after that time.

The expected length of stay was modeled using: gender, level of care at admission, admitted from code, and gender*level interactions. Only the gender*level interactions were used in further analysis. Expected length of stay, variance and standard errors were calculated based on the model information from the LIFEREG procedure. The SAS code listed below was developed and provided by Robert Tate (1998).

Average ELOS was calculated for every year (1989/90 to 1997/98) by:


1. assigning the corresponding expected length of stay for each age, level of care and gender to every new admission in a fiscal year.
   
   
2. The mean ELOS was calculated across all new admissions in each fiscal year.

Results

Note: Earlier years of ELOS where calculated for doing comparisons with earlier studies.



See table of Expected Length of Stay Over Time