The implication of selection bias in clinical studies of late life depression: an empirical approach

OBJECTIVES: It is supposed that selection bias precludes the extrapolation of results of studies carried out in a clinical setting to the general population. There is little empirical evidence demonstrating the degree to which those depressed in the community are different from those treated in clinical settings. This study compared elderly patients with major depression admitted to a psychiatric hospital with those living in the community. METHODS: All elderly (55 years and older) patients admitted between 1990 and 1992 to a psychiatric hospital with DSM major depression as the primary diagnosis (n=104), were compared with all elderly patients with the same diagnosis (n=59) who were participating in a large community study (Longitudinal Aging Study, Amsterdam). Data were gathered from the clinical sample using chart-reviews while the community-based sample was interviewed. The two groups were compared with respect to differences in demographic variables, presenting symptoms, risk factors and treatment. RESULTS: The following characteristics were significantly more prevalent in the clinical sample: late onset of the depression, threat of suicide, conflicts with significant others and use of antidepressant medication. Chronic physical illness was the only characteristic that was more prevalent in the community sample. CONCLUSION: The results confirm that elderly patients treated in clinical psychiatry represent a group with more threatening and more disruptive depressive illness. Major depression in the community was more often associated with chronic physical illness, which may hamper the recognition and treatment of depression. As the two samples were similar in all other respects, selection bias, hampering comparison of results of studies carried out across treatment settings, appears to have a very limited effect