The individual reports discussed in this Final Report (which are all available from ASPE) have extensive research methods sections that provide relevant detail. The discussion of study methods here provides a general overview, emphasizing the sampling strategy and is intended to allow readers to better understand the results discussed in this summary.
In order to obtain generalizable results, the project staff implemented a complex, multi-stage sampling design. At the first stage, project staff selected a random sample of 60 geographic areas across the nation, known as first-stage sampling units (FSUs). These 60 FSUs were comprised of 1,086 counties in 34 states. At the second stage of sampling, staff selected a sample of facilities in those FSUs.1
The sample design called for selection of a set of geographic areas or FSUs prior to selecting the facility sample for several reasons. First, in order to conduct the survey, staff had to construct a listing of ALFs. There is no national list that is comprehensive and exhaustive. Moreover, definitions of assisted living vary across the states. In some states, there are no limits on the types of facilities that may call themselves assisted living or advertise that they provide assisted living, regardless of the kind of services and accommodations they provide. Further, some states did not have a licensure category known as assisted living or included all types of residential care facilities in the category called assisted living (Mollica and Snow, 1996; Mollica, 1998).2 As a result, the study could not rely on state licensure lists to provide a comprehensive and exhaustive listing of ALFs. Some places meeting study criteria would have been missed, while other facilities licensed under the category of assisted living might not have met more commonly understood definitions of assisted living.
Lists from established trade associations were also insufficient as a sampling frame. First, while there are multiple trade associations, their combined membership accounts for an unknown proportion of the total number of ALFs in operation. Second,ALFA merged with the association that represented board and care homes (i.e., the National Association of Residential Care Facilities). As a result, the membership of ALFA was expected to include both ALFs and places that were more traditionally thought of as board and care homes, some of which would meet study criteria and some which might not.
Finally, the study could not rely solely on retirement directories or local advertisements, since they appeared to have differing definitions (or no criteria) for what should be classified as assisted living. Similarly, in many localities, there were no restrictions on the kinds of places that could call themselves assisted living.
As a result, a crucial aspect of the sampling design was the development of an enumeration strategy that would enable selection of a nationally representative sample of ALFs. However, because of the extensive level of effort involved, creating a comprehensive list at the national level (i.e., in each of the >3,000 counties in the country) would have been prohibitively expensive for this project. Thus, project staff decided to select a random sample of geographic areas across the country in which to enumerate an exhaustive list of facilities.3 This involved a two-stage enumeration and screening process to provide comprehensive coverage of the target population of ALFs.
Creating a List or National Sampling Frame of ALFs
In order to create a list or sampling frame of ALFs, project staff first had to define the relevant study population. These were the criteria used to determine whether a place was eligible for inclusion in the study. As noted, the definition of assisted living varied across the country. Thus, to define assisted living for this study, project staff specified selected features about which there was general agreement among industry and consumer groups that the feature was characteristic of assisted living.
To be eligible for this study, a facility had to serve a mainly elderly population. In addition, a facility had to have more than 10 beds.
There were several reasons for the size restriction. We expected that few small facilities would meet the study criteria related to services provided and population served. A 1993 survey funded by HHS/ASPE examined board and care homes in ten states (Hawes et al., 1995b). Two-thirds of the licensed board and care homes had 10 or fewer beds. Few of those facilities identified themselves as assisted living or provided personal assistance with two or more activities of daily living (ADLs). In addition, none of the states that had specific licensure categories known as assisted living had actually granted a license to an ALF that had fewer than 11 beds. As a result, project staff expected that few small facilities (i.e., <10 beds) would meet the service-related criteria. In addition, small homes tended to have significant numbers of non-elderly residents; they were more likely to have residents with mental retardation, developmental disabilities, and persistent and serious mental illness.
In addition to the two eligibility criteria based on population served (i.e., elderly) and size (i.e., >10 beds), the facility had to describe or represent itself as being an ALF or it had to be a place that offered at least a basic level of services thought to be consistent with assisted living. Those services were:
- 24-hour staff oversight;
- At least 2 meals a day; and
- Personal assistance, defined as help with at least two of the following: medications, bathing, or dressing.
To be eligible for the study, a facility had to:
1. Serve a mainly elderly resident population
2. Have more than 10 beds
3a. Be a self-described ALF
- 24-hour staff
- At least 2 meals per day
- Help with at least 2 of the following: medications, bathing or dressing
Additional details on the selection of the FSUs can be found in the full report on the results of a telephone survey of administrators of a national probability sample of ALFs. What is important to know is that the 60 FSUs were randomly selected, with higher selection probabilities for those FSUs anticipated to contain larger numbers of expected ALFs. These FSUs consisted of 1,086 counties in 34 states. They contained 40 percent of the U.S. population aged 65 and older and 43 percent of the initial sample frame of expected ALFs.
Source of Listings of Candidate ALFs
The next task, which was completed in September 1997, was to create a comprehensive list of candidate ALFs in each of the 60 FSUs. This involved the collection of multiple lists of places that described themselves as ALFs and other places that appeared to meet the studys eligibility criteria. For each FSU, project staff obtained data from the following:
ALFAs 1997 list of members.
AHCAs National Center on Assisted Living 1997 list of assisted living members.
AAHSAs 1997 list of assisted living members.
The 1997 HCIA Directory of Retirement Facilities (DRF, 1997).
State licensure agency lists for 1997 for all types of residential care facilities that had 11 or more beds.
Listings and advertisements of assisted living, residential care, and retirement facilities from telephone book yellow pages.
State directories of local ALFA members, in those states in which the state affiliate of ALFA had one.
Local retirement directories and Internet listings of facilities.4
Among the 18,298 places on the combined list of candidates:
- 7,578 (41%) were ineligible because of size
Among the remaining 10,720 candidates:
- 70% appeared on only one source listing
- 19% appeared on two source listings
- 11% appeared on three or more source listings
- Small facilities were more likely to appear on only one source list (e.g., 86% of the small facilities were on only one list), while about half of the larger facilities appeared in two or more source listings
- 46% of the candidates appeared on the state licensure lists
After collecting these multiple lists, project staff spent considerable effort making sure they were in one of the 60 FSUs5 and creating an unduplicated list.6
From all these sources, for the 60 FSUs, an unduplicated total of 18,298 candidate facilities were listed. Project staff then eliminated known ineligibles based on size. Project staff found that among the 18,298 candidates 7,578 places (41 percent) were ineligible because they had 10 or fewer beds. The degree to which such small facilities might otherwise have met study criteria (e.g., be a self-described ALF or provide the required services) is unknown.
Selecting the Sample for the Telephone Screening and Survey
The next task was to select a sample of facilities from the remaining 10,720 candidates on the combined listing (e.g., after eliminating the 7,578 places with 10 or fewer beds).
In selecting the sample to be screened by telephone, project staff oversampled the larger facilities (>51 beds) in order to increase the likelihood of encountering ALFs offering a high level of services. This was based on an earlier study of residential care that found that larger facilities were more likely to have nurse staffing and to offer more services (Hawes et al., 1995a and 1995b). Thus, project staffs assumption at this stage was that larger facilities would have greater capacity to offer more services. Further, this approach would improve the sampling efficiency at later stages (i.e., when the resident sample was selected, since most residents lived in large ALFs). This oversampling did not affect final estimates about the prevalence of various types of ALFs (including various sizes), since data were weighted to account for the oversampling.7 From this list of 10,720 potential candidate facilities, project staff selected a stratified (by size), random sample of 2,945 facilities for the telephone survey.
The administrators of a sample of these candidate facilities were then surveyed by telephone during 1998. If the candidate facility met the study eligibility criteria specified in a set of screening questions, then the administrator was asked to respond to questions about the facility, its size, occupancy, accommodations, services, price and basic admission and discharge criteria. A total of 1,251 facilities was contacted, found to be eligible, and interviewed. These interviews were the basis for the first project report, and data from these interviews were also used in subsequent reports.
Our second report presents data from a series of in-person and telephone interviews with administrators, staff, residents and family members in a subset of these eligible 1,251 facilities. The subset of ALFs involved in this data collection represented that group of ALFs that offered a mix of services and privacy thought to approximate most closely key aspects of the philosophy of assisted living.
To identify this subset of ALFs, the facilities in the original sample were divided into groups, based on the level of privacy (i.e., high, low or minimal) and the level of service (i.e., high, low or minimal) that they provided. An explanation of the way in which these groups were defined is discussed in detail in Report 1 ("Results of a National Survey of Facilities") and Report 2 ("High Service or High Privacy Assisted Living Facilities, Their Residents and Staff") of this report. Facilities that offered either minimal services or minimal privacy were combined with the ALFs that provided a combination of low services and low privacy. These ALFs were excluded from any on-site data collection. Facilities in three of these groups (i.e., high service & high privacy, high service & low privacy, low service & high privacy) were chosen for further study through site visits that included more detailed data collection. The site visits included interviews with facility administrators, staff, residents, and family members, as well as a walk-through evaluation of the facility environment. The 300 facilities involved in the site visits represented a population of 4,383 ALFs across the nation -- or all those that met the definition of a facility that offered either high services or high privacy.
| TABLE ES.1: Distribution of ALFs Nationwide and Those Included in Site Visits
|| High Services
|| Low Services
| High Privacy
|| 11% of all ALFs
| 18% of all ALFs
| Low Privacy
|| 12% of all ALFs
| 59% of all ALFs