We conducted surveys to gather information about the characteristics, experiences, and perceptions of high-risk seniors in three of the case-study MCOs: Aspen, Kaiser Colorado, and Keystone East.7 These surveys focused on the three groups of elderly high-risk Medicare beneficiaries we are studying: (1) those whom the MCO’s enrolled in care management programs, (2) those with advanced age, and (3) those with a recent hip fracture or stroke. A total of 1,657 beneficiaries were interviewed by telephone between March and December 1999. We also reinterviewed 301 of the beneficiaries in the hip fracture and stroke subsample approximately seven months after their initial interview.
We conducted the survey using computer-assisted telephone interviewing (CATI). Surveys were administered to three types of respondents. Whenever possible, we spoke directly to the sample members. When that was not possible, we spoke to a proxy or representative proxy respondent. The proxy respondents were interviewed for sample members who made their own medical decisions but could not complete the interview at the time of the survey. The representative proxies were interviewed when they were the ones who made health care decisions for sample members who were unable to do so. The survey included questions and options that the interviewers could use to switch respondent type during an interview if it became clear that the respondent could not complete the interview or when the sample member became available to complete an interview that we had started with a proxy.
We did not interview proxy respondents for sample members who had died. Thus, our results reflect the characteristics of people who survived from the time they were selected for the survey until the interview. For the care management group, the elapsed time between selection and interview ranged from 1 to 20 months. For the advanced age sample, the elapsed time ranged from 6 to 15 months. Beneficiaries included in the hip fracture and stroke samples were interviewed approximately 3 months and 10 months after their vent.
The overall response rate for the wave 1 survey, which includes all three subgroups, was 76 percent. The response rate for the wave 2 survey of hip fracture and stroke patients was 89 percent. The response rates varied among the MCOs. For wave 1 the rates were 79 percent for Aspen, 74 percent for Keystone East, and 69 for Kaiser Colorado. Response rates for wave 2 were 88 percent for Aspen, 90 percent for Keystone East, and 85 for Kaiser Colorado. The low rates for wave 1 at Kaiser Colorado were caused by a requirement imposed by their institutional review board that we give all sample members a prepaid postcard that they could use to opt out of the survey (Stapulonis et al. 2001). If we exclude the postcard refusals in calculating response rates, the rate would be 77 percent. Thus, the survey performance among individuals we had a chance to contact and interview by telephone was very similar among the three MCOs. The survey data were weighted to reflect the probability of selection and to correct for survey nonresponse. The correction for nonresponse was based only on gender and age, which were the only two relevant variables available on the lists of names provided by the MCOs.
The beneficiaries included in the survey were sampled from lists provided by three of the case study MCOs. The sample selection was conducted in an effort to be representative of the three high-risk groups in these MCOs. Specifically,
Care Management Subsample. The three MCOs provided lists of their members who had been enrolled in care management. For Kaiser Colorado and Keystone East, the lists included seniors who had been in care management between January and August 1998. For Aspen, whose care management program had started more recently, the list contained seniors enrolled in care management from September 1998 through January 1999. We selected people at random from each list. The number of people selected from each MCO was determined to give us approximately equal confidence intervals for MCO-specific estimates.
Advanced Age Subsample. The MCOs provided lists of all their members who had attained age 85 by October 1998. We selected a random subset of them for the survey. Again, the sample size for each MCO was determined to give us approximately equal confidence intervals for MCO-specific estimates.
Hip Fracture and Stroke Subsample. The MCOs provided lists of elderly beneficiaries who had been hospitalized for hip fracture or stroke from November 1998 through August 1999.8 The lists were updated on a monthly basis so that we could interview these patients 3 months after their event (hip fracture or stroke) and then again 10 months after the event. Since the flow of cases was fairly small, we attempted to interview every hip fracture or stroke case that we could.
The final samples for each of the MCOs and subsamples are given in Table II.3. This sample is quite large for a case study and gives us a good indication of the experiences of the selected groups of high-risk seniors at each of the three MCOs. Nevertheless, estimates for subgroups, particularly for the hip fracture and stroke groups, are imprecise and give us a basis for detecting only very large differences between subgroups.
The lists from which the subsamples were selected were not mutually exclusive. For example, some of the beneficiaries on the lists of people in care management were more than 85 years old and therefore were also included on the lists of people with advanced age. The figures in Table II.3 include each beneficiary in the subsample from which he or she was selected. In the analysis, we weighted the sample data to reflect the actual probability of selection (Stapulonis et al. 2001).
The timing of the two waves of interviewing for hip fracture and stroke patients reflects several factors. We wanted to describe two phases in the treatment these people will receive. Immediately following the event, it will be mostly medical care. Later, it will shift to be mostly ongoing monitoring or, in some cases, long-term custodial support. The two waves of interviewing were designed to capture this change in the nature of care.9 In addition, the second interview, conducted 10 months after the event, should capture the full extent of recovery for most seniors (Magaziner et al. 1990; and Jorgensen et al. 1995).
| TABLE II.3. Number of High-Risk Seniors Interviewed
| Survey Sample