Adjusting for nonresponse bias in a health examination survey

There is a potential for nonresponse bias in most population studies using health examinations. This is true of the Mexican American portion of the Hispanic Health and Nutrition Examination Survey (HHANES), conducted by the National Center for Health Statistics, in which unit nonresponse to the examination accounted for 24 percent of the sample. Even though the full effect of nonresponse can never be really known, ancillary information from the interview sample can be used in an attempt to adjust for bias in estimates. Two techniques for nonresponse bias adjustment are presented and illustrated using health status level and hypertension status from published studies based on the HHANES of 1982-84. The first approach uses conditional probabilities and the second approach uses direct standardization. The examples examine whether or not an adjustment for socioeconomic status, sex, and age--variables related to both response status and the conditions under study--changes the prevalence estimates of (a) Mexican Americans who report poor, fair, or good health status and (b) hypertension among Mexican Americans.