An Evaluation of a Method Used to Impute Residence Data¹

Issue title: Health Related Statistical and Methodological Analyses

Article type: Research Article

Authors: Potter, D.E.B.^a | Cunningham, Peter J.^a

Affiliations: [a] Agency for Health Care Policy and Research, 2101 E. Jefferson St., Suite 500, Rockville, Maryland 20852, USA

Note: [1] The views expressed in this paper are those of the authors. No official endorsement by either the Agency for Health Care Policy Research or the Department of Health and Human Services is intended or should be inferred. The authors acknowledge the invaluable computer processing support of Maria Nicholson of Social and Scientific Systems, Inc; the secretarial support of Kelly Carper; and review comments by Mamatha Swamy.

Abstract: Data from the 1987 National Medical Expenditure Survey, Institutional Population Component were used to characterize partial and complete respondents with respect to outcome measures based on event history data. In this analysis, data for a sample of persons who used a nursing or personal care home (NH) anytime during 1987 were used to examine the 1987 residence profiles of persons with complete residence data and contrast them to persons with missing residence history data. The results indicated that after controlling for sample type (resident on Jan. 1 verses 1987 admission), persons who spent any time in a nursing or personal care home during 1987 and for whom incomplete residence data were collected were significantly different from their counterparts with complete data. Most importantly, persons with incomplete data were more likely to move from place to place, and to have more stays in NHs during 1987. Differences were also noted on the dimensions of health status, insurance coverage and facility location. The results further provided some evidence that a minimum distance function imputation technique could be used to impute residence data for multiple events to persons missing some event history data, and that in some circumstances the imputation reduced the nonresponse bias of survey estimates. Evidence is also provided that the imputation improved the precision of some NH use and expenditure estimates. The paper also provides a brief description of the methodology used to construct and impute the event history profiles of missing residence data.

DOI: 10.3233/JEM-1994-20306

Journal: Journal of Economic and Social Measurement, vol. 20, no. 3, pp. 263-283, 1994