Data Collection Frequency Effect in the National Medical Care Utilization and Expenditure Survey

Introduction The National Medical Care Utilization and Expenditure Survey (NMCUES) was conducted to meet the needs of government agencies and health professionals for more comprehensive data on the utilization, costs and sources of payment associated with medical care in the United States. A longitudinal survey design was adopted for the household component in order to provide accurate measurements of population characteristics which are sensitive to changes in time. Data collection for the core health care measures was to be applied to the same panel of sample households in five rounds of interviewing, with 1980 as the reference period. Short recall periods of two to three months in duration were generally implemented to minimize reporting errors of omission. A subset of sample participants, referred to as holdovers, ~ere not contacted for a particular wave of the survey and data were gathered at the subsequent round of interviewing for the two time intervals that were spanned. This naturally occurring study treatment provided a unique opportunity to investigate the effect of data collection frequency on reporting behavior. In this paper, national estimates for a representative set of health care utilization, expenditure and morbidity measures were derived from the sample of holdovers and compared with estimates derived from the respondents with five complete rounds of data collection. The analysis controlled for demographic characteristics that distinguished the two study groups. A more detailed round-specific level of analysis was then conducted to test for a data collection frequency effect, further controlling for length of recall period. The research focuses on the implications of a departure from a panel survey characterized by five waves of data collection. This study replicates the analyses that were conducted to test for a data collection frequency effect in the National Medical Care Expenditure Survey (NMCES) conducted in 1977 and on which the NMCUES was modeled. The NMCES study findings demonstrate the presence of a data collection frequency effect in a panel design similar to the NMCUES. The study results argue for consideration of four rounds of data collection as an alternative to five, in-survey designs which mirror the NMCES (Cohen and Burt, 1984). The findings identify a survey component that would significantly benefit by a redesign strategy to reduce cost without impairing the quality of survey estimates. Given the similarity in survey design and questionnaire wording between the NMCES and the NMCUES, coincident findings of a data collection frequency effect in the NMCUES would provide support for future survey redesign strategies, which considered a reduction in the number of rounds of data collection.