10 beds. However, an estimated two-thirds or more of all residents are found in larger facilities (e.g., 11 beds or more).
Decisions about sampling residents are intertwined with decisions about data collection, and they too are challenging. As noted above, research suggests that residents at highest risk for abuse and neglect are those with cognitive impairment and behavioral symptoms. Research is more mixed about whether greater levels of impairment in ADLs represent a risk factor. In nursing homes, the majority of residents receive assistance in three or more ADLs, so a random sample would produce adequate numbers of residents with significant physical impairment. Most residents also have moderate to severe cognitive impairment. Many (though not necessarily all) of those will not be candidates for interviews. So, one key issue is how one can collect valid information about the experience of these residents. However, it will not be difficult in even a random sample to secure an adequate number of residents with this risk factor.
Other potential risk factors for abuse or neglect are less common among nursing home residents. For example, the most recent data suggest that fewer than 10 percent of residents are African American and only 9 percent exhibit physically aggressive behaviors. Similarly, if one wished to have results that were generalizable to short-stay residents, one would need to oversample such residents.
Fortunately, there are two sources of data that can inform sampling decisions in nursing homes. The Medical Expenditure Panel Survey (MEPS) Institutional Component provides estimates of the prevalence of various conditions and risk factors among a national probability sample of residents (Krauss and Altman, 1998). Even more immediate data are available through the national database CMS maintains on all nursing home residents in every nursing home certified to participate in the Medicare or Medicaid programs. These data, taken from the Minimum Data Set (MDS), provide information on hundreds of characteristics for the universe of residents.
In residential care facilities, there is less current information available with which to make sampling decisions about residents. The most recent multistate study of residential care facilities is the 10-state study conducted for the U.S. Department of Health and Human Services, Office of the Assistant Secretary for Planning and Evaluation (Hawes et al., 1995a,b). These data are from interviews conducted during 1993, and the states were selected on the basis of their regulatory environment. Within those states, facilities were selected on a random, stratified basis (i.e., size and licensure status), and residents were randomly selected, as were staff members. How-