OCR Text |
Show The Sample The sample consisted of ten strata, one for Medicaid and one for non-Medicaid in each of the five HMOs. Systematic sampling with random start was employed to generate a representative sample from each stratum. The Medicaid sample, when combined and weighted to correct for disproportionate sampling, is representative of all Medicaid managed care enrollees in Utah. The non-Medicaid samples when combined is not representative of the state's HMO population. However, the five HMOs are the largest in the state and account for about 88% of the HMO population and about 82% of the managed care population in Utah. The sampling design called for apportioning the interviews as follows: 400 non-Medicaid enrollees per plan and 400 Medicaid enrollees per plan, for a total of 2,000 Medicaid and 2,000 non-Medicaid enrollees overall. This sample allocation provides enough analytic power to draw conclusions about each health plan's non-Medicaid and Medicaid enrollees alone. DataStat asked the plans to provide them with approximately four times as many records as the number of expected completed interviews with each non-Medicaid group of enrollees, or in this case 1,600 records. For Medicaid enrollees, the corresponding ratio was suggested to be ten to one, or 4,000 records for each 400 completed interviews. This allowed DataStat to do careful replacement sampling, using replicates, for non-working numbers, no such person known at the household (wrong numbers), unreachable respondents, etc. (One plan could not meet these requirements, and therefore has fewer total completed interviews.) Interviews were completed with 4,035 enrollees, both Medicaid and non-Medicaid, allocated as follows: Medicaid Non-Medicaid BOBS 401 404 FHP 603 407 (including IPA/Select) IHC 403 410 Intergroup 407 195 UHC 403 402 Total 2,217 1,818 procedures for sample selection, and the format of submission. This assured that the samples were drawn according to the same specifications, which enhanced comparability of results. For both Medicaid and non-Medicaid samples, only the following enrollees were selected: Utah residents, currently enrolled in the health plan as of May 1, 1996, with no member of the family employed by the health plan, and with phone numbers on record. Additional criteria for the non-Medicaid samples were continuous enrollment for at least twelve months, and not having Medicaid or Medicare coverage. For the Medicaid samples, additional criteria were: not in foster care, group homes or youth correctional facility; continuously enrolled for at least three months; not enrolled in Medicare or any commercial health plan. Implementation A pretest of the questionnaire was conducted prior to actual interviews. The pretest yielded valuable information that was used to refine the questionnaire and instruct the interviewers and supervisors on issues that might arise. Interviews were conducted during the period from July to October, 1996 using a Computer-Aided Telephone Interviewing (CATI) system from DataStat's facilities in Ann Arbor, Michigan. DataStat provided the HMOs and the Bureau of Managed Health Care detailed criteria and Respondent DataStat selected the eligible respondent with the help of automated sample management software. DataStat worked with HDA and each health plan to carefully define sampling requirements. For example, in the Medicaid population, many primary policyholders were under the age of 18. DataStat and HDA discussed whether or not young policyholders were to be interviewed, and how young they should be. There were a number of complications in the Medicaid population which were decided upon and planned for. Once a household was reached, DataStat asked for the named respondent from the sample. DataStat asked for the adult most familiar with their health care if the named person in the sample was younger than 16 years old. In a few cases, DataStat interviewed such an "adult most familiar with their health care" who was himself or herself under 16 years old. There is a possibility of bias introduced by the use of proxy respondents. To evaluate this, the proxy respondents and the enrollee- Utah's Health: An Annual Review 1996 17 |