An Evaluation of Hospitalization Data from the Health Interview Survey

Publisher Website
Google Scholar
Abstract
Research by the U. S. National Health Survey points to a net underreporting of discharges in its household interviews. Underreporting appears to be most influenced by 3 factors; there is (1) an increase in the underreporting with increase in time interval between discharge and interview; (2) an increase in underreporting with decrease in length of stay in the hospital; and (3) an increase in underreporting with respondent''s probable assessment of increasing social threat to himself from the diagnosis. Two record-check studies, one based on a sample of 1833 discharges and the other on 350 discharges, found that about 12-13% of the known hospital episodes were not reported in the household interviews. Although the major emphasis was on underreporting in these studies, an attempt was made to evaluate overreporting in the same geographic areas. A sample of approximately 700 households, drawn from the general population, and interviewed in the same manner as the households with known discharge cases, produced 187 reported hospital discharges. About 3% of these were not confirmed by the hospitals named in the interviews. Thus, the net reporting error is probably in the neighborhood of 8-10%. The magnitude of error was also reflected by an internal analysis of hospitalization data reported in the Health Interview Survey of the U. S. National Health Survey, based on tabulations of reported data by the interval between discharge and interview. A statistical adjustment of these reported data, taking into account the interval between discharge and interview, cuts the overall net error to 5% or less. Accumulating evidence is establishing the fact that statistics on the event of hospitalization and a number of associated phenomena can be collected through household interview with a degree of accuracy satisfactory for many purposes. Even though means have been devised to cut the reporting error to less than 5% the National Health Survey is continuing to study the problem to further minimize response flaws.

This publication has 1 reference indexed in Scilit: