Abstract
Knowledge on the reliability of family history information is essential for every family study. However, systematic analyses of interinformant reliability of family history information on individual relatives have not yet been published. Consequently, family history information on 1306 first-degree relatives and spouses of patients and of control subjects was collected from at least two other family members using questionnaires. Interinformant reliability was acceptable for dementia [Kappa = 0.58, 95% confidence interval (CI) = 0.48-0.68], but less so for alcoholism (Kappa = 0.41, CI = 0.23-0.59), depression (Kappa = 0.26, CI = 0.14-0.38) and anxiety disorders (Kappa = 0.19, CI = 0.05-0.43). Demographic variables of subjects and informants and their familial relationship did not influence diagnostic agreement on the diagnosis of dementia. Diagnostic agreement on depression was significantly reduced when information from siblings of index subjects was compared with information from spouses of index subjects. The interinformant agreement for the diagnosis of depression was higher in younger than in older subjects (relative risk for disagreement 1.08/additional year, CI = 1.02-1.15). Siblings of index subjects seem to provide different, but not necessarily less relevant, family history information in comparison with other relatives. Researchers should be aware of the problem that depression in the elderly can be easily missed by family history. It seems more important for the diagnosis of depression than for a diagnosis of dementia to get information from multiple informants.