The NSRCF sampling frame was constructed from lists of licensed residential care facilities (i.e., facilities that are licensed, registered, listed, certified, or otherwise regulated by the state) acquired from the licensing agencies in each of the 50 states and the District of Columbia. State data on the number of licensed beds for each facility and the licensure categories were used to determine the list of eligible facilities. The individual lists of residential care facilities from state licensing agencies were checked for duplicate facilities and concatenated to form a list of all residential care facilities in each state. The individual state lists were then concatenated to form the sampling frame. The sampling frame for NSRCF contained all of the state-licensed residential care facilities that were licensed for four or more beds and, based on information obtained during frame construction, appeared to meet the survey definition at the time of frame development. The rest of this section summarizes the sampling frame development and sampling design activities. More information on sampling frame construction can be found in ‘‘NSRCF Sample Frame Construction and Benchmarking Report’’ (11), which is available from: http://aspe.hhs.gov/ daltcp/reports/2010/sfconst.pdf.
Building the actual sampling frame of individual facilities involved several steps:
- Obtain state licensure lists.
- Convert the state files into a usable, standardized format.
- Assess the completeness of the data.
- Clean and merge the state-level files.
Because definitions and nomenclature regarding residential care facilities vary widely across states, the first sampling frame task was to identify the licensure categories of residential care facilities within each state that met the study definition. To make this determination, the following resources were reviewed: Residential Care and Assisted Living Compendium: 2007 (7); Assisted Living State Regulatory Review, 2009 (12); Inventory of Long-Term Care Residential Places (5); and the websites of each state and their associated regulations for the different types of residential care. Fifty different terminologies for these state licensure categories were identified. Screening for facility eligibility was also performed during data collection because determining whether the facility met the study definition was not always straightforward.
State lists, which varied in format and degree of completeness, were obtained from the state regulatory agencies or, in a few cases, downloaded from the Web. All states provided facility licensure type and the counts necessary to determine number of beds per facility. Data were available on more than 99% of facilities for street address, city, state, ZIP code, and telephone number. Ownership data were missing for 57% of facilities, and a facility contact name was missing for 28% of facilities.
Issues concerning combining facilities, converting units into beds, and identifying chains were also addressed. Facilities were combined that had multiple licenses or where close-by buildings with separate licenses appeared to be under the same management, according to the study’s predetermined criteria. Namely, when facilities had addresses within two digits of one another and the same phone number and administrator, they were considered combined facilities and, as such, constituted a single case on the sampling frame. Overall, 446 facilities (1% of all facilities on the frame) were flagged as combined facilities. Four states licensed their residential care facilities by units (i.e., rooms and apartments) instead of beds. For these states, the number of licensed beds was imputed. Moreover, 16,379 facilities (41%) were identified as being affiliated with a chain, defined as an individual or corporation owning two or more facilities. Of chain-affiliated facilities, 1,381 were associated with 84 chains identified as the largest and most recognizable chains in the United States; the remaining 14,998 chain-affiliated were associated with regional or smaller chains.
Since the sampling frame lists were obtained from states beginning in June 2009, no list was older than 6 months at the time the NSRCF sample was drawn in January 2010 from the sampling frame of 39,635 facilities.
Benchmarking the Sampling Frame
Because there is no ‘‘gold standard’’ list of residential care facilities, alternative estimates of the number of residential care facilities and beds were identified and compared with the number of facilities and beds on the NSRCF sampling frame.
Four data sources were used to make these comparisons.
- Using the 2002 Health and Retirement Study, the 2002 Medicare Current Beneficiary Survey, and the 1999 National Long-Term Care Survey, Spillman and Black (8) estimated the number of older people living in residential care facilities. These numbers were converted to an estimated number of residential care beds by using the median nursing home occupancy rates reported by the American Health Care Association and the proportion of nursing home residents who are aged 65 and over from the 2004 NNHS.
- The Inventory of Long-Term Care Residential Care Places (5), while almost 7 years old, contained a comprehensive listing of residential care facilities against which to check the NSRCF sampling frame.]
- The Residential Care and Assisted Living Compendium 2007 (7) was used to benchmark NSRCF.
The NSRCF frame was compared with Stevenson and Grabowski’s (13) data that utilized information in the State Residential Care and Assisted Living Policy: 2004 (14).
- Licensed, registered, certified, listed, or otherwise regulated by their state within their specific licensure category.
- Have four or more licensed, certified, or registered residential care beds.
- Have at least one resident currently living in the facility.
- Provide room and board with at least two meals a day.
- Provide around-the-clock onsite supervision.
- Provide help with activities of daily living (e.g., bathing, eating, dressing) or health-related services (e.g., medication management).
- Serve primarily an adult population.
- Do not exclusively serve the severely mentally ill, the MR/DD, or both.
Sampling for NSRCF used a stratified two-stage probability design. The first stage was the selection of facilities, where the primary strata of facilities were defined by facility bed size (i.e., 4–10 beds, 11–25 beds, 26–100 beds, and more than 100 beds) and geographic region (Northeast, Midwest, South, and West). Within primary strata, facilities were sorted by the following characteristics: MSA status, where MSA denotes a metropolitan statistical area, and state. MSA status could be ‘‘metropolitan’’ (a county or group of contiguous counties that contain at least one urbanized area of 50,000 or more population; an MSA may contain other counties that are economically and socially integrated with the central county, as measured by commuting), ‘‘micropolitan’’ (a nonmetropolitan county or group of contiguous nonmetropolitan counties that contains an urban cluster of 10,000–49,999 persons; surrounding counties with strong economic ties, measured by commuting patterns, may also be included), or ‘‘other,’’ according to information available on the sampling frame. For more information, visit: http://www.whitehouse.gov/omb/ inforeg_statpolicy/.
Facilities were then selected by using systematic random sampling. This stage yielded the final NSRCF sample of 3,605 facilities and included 1,184 facilities with 4–10 beds, 1,038 facilities with 11–25 beds, 1,051 facilities with 26–100 beds, and 332 facilities with more than 100 beds. Reserve samples in excess of the base sample of 1,184 facilities in the small stratum were also created in case projected ineligibility rates among small facilities were underestimated. These cases were never fielded, however, because targeted numbers were reached using the base sample alone.
The second stage was the selection of current residents. This stage was carried out by the interviewers during the facility interview, with the aid of an algorithm programmed into CAPI. The interviewer first collected the census list, usually from the facility director or designee. The census list of residents was the total number of residents on the facility rolls as of midnight on the day before the facility interview. The interviewer reviewed and cleaned the list to ensure there were no duplicates or missing residents that should have been on the list. The interviewer then numbered the list and entered the total number of residents into the CAPI system. The CAPI program, through systematic randomization procedures, generated three to six numbers, depending on facility bed size. The interviewer then identified which residents matched up with these numbers. Three residents were selected for facilities with less than 26 beds, four residents for facilities with 26–100 beds; and six residents for facilities with more than 100 beds.