Subject and informant characteristics influence the reliability and validity of family history information: an analysis based on the generalized estimating equations approach

Abstract

Family history information is a necessary surrogate for unavailable interview information in family studies. However, the reliability and validity of such information has rarely been assessed during the conduct of family studies. This paper presents a reanalysis of data on the reliability and validity of family history information for dementia and depression using the general estimation equations approach. All available relatives of patients and controls were interviewed and questioned about the psychiatric morbidity of other family members. Interinformant reliability of this family history information was evaluated as well as factors influencing this information. The validity of family history was investigated by comparing the informant derived diagnoses with interview-derived diagnoses. To account for possible lack of independence of family history provided by several family members on other family members, the generalized estimating equations (GEE) were used in statistical analysis. The interinformant reliability for depression (kappa = 0.13) was low. It was fair for dementia (kappa = 0.34). The informants more frequently agreed about the diagnosis of dementia when the subject was older. The sensitivity of family history was 35.2% for dementia and 31.8% for depression. The specificity of family history information was generally above 97%. The sensitivity of family history increased significantly with the severity of both disorders. The sensitivity of the family history for dementia was higher when the informant was a first-degree relative, when he was younger and when the index subject of the family suffered from dementia. The specificity of the family history was slightly reduced with higher age. The observed low sensitivity of family history information leads to underestimation of psychiatric disorders. The informants provide more useful information on more severe disorders. The sensitivity of family history was higher in families with an affected index subject than in control families, so familial aggregation of dementia might be overestimated in studies using the family history method. Copyright © 2000 Whurr Publishers Ltd.