A Comparison of Household and Provider Reports of Medical Conditions

Recognizing that household respondents have difficulty recalling past events, the National Medical Care Expenditure Survey incorporated a record check component. For 12,000 persons included in the Medical Provider Survey, two data sources exist for health care events: household survey responses and provider record abstractions. This investigation compared medical conditions associated with the 35,000 visits, to conclude that agreement between sources was weak. Households tended to report fewer conditions per visit and familiar conditions, whereas providers reported related but different diagnoses. Levels of provider agreement with household reports ranged from 1% for upper extremity and shoulder impairments to 86% for pregnancy care and delivery. Over all 63 categories, providers concurred with household condition reports 40% of the time. Broader categories created by collapsing conditions substantially increased agreement levels. Analyses of household-provider agreement suggest that questionnair...