To the Editor: A recent article by Helen C. Kales, M.D., et al. R1574BCFDFJCA draws attention to the increased inpatient medical/psychiatric resources utilized by patients discharged with comorbid diagnoses of dementia and major depression. The article represents a significant step forward in view of the paucity of literature on the subject.

We would like to point out, however, some limitations in the study design used in addressing the question of prognosis, in terms of the utilization of health care services, in patients discharged with diagnoses of dementia alone (group 1), dementia and major depression (group 2), and major depression alone (group 3). Among the key aspects of the study design that were needed to answer the question, perhaps the most significant limitation was the lack of "a representative and well-defined sample of patients at a similar point in the course of the disease."

The discharge diagnoses from a routine administrative database such as that used by this study are likely to be of limited accuracy. Various biases are likely to have affected them (for instance, the comorbid diagnosis of major depression may have been more likely to be given during a psychiatric admission than during a medical one, and more diagnoses may be listed for longer inpatient stays).

Also, since the database could not provide information about the severity and duration of the dementia, the patients in group 1 might have been at a more advanced stage of illness. This possibility is supported by the fact that the patients in group 1 were older, had a higher mortality rate, and were nonsignificantly more likely to be discharged to nursing homes (1,315 of 5,060) than the patients in group 2 (56 of 265) (Pearson’s χ²=3.11, df=1, p=0.08). If so, they would be expected to have not only a different pattern of health care utilization but also a diminished ability to report their depressive symptoms. A partial improvement could have occurred in the period from the first recorded diagnosis of dementia to the follow-up.

Of the outcome measures on which the groups were compared, the only ones on which groups 1 and 2 differed significantly were those pertaining to psychiatric readmissions. The issues discussed could have substantially affected these findings but were not addressed in the study design, analysis, or discussion. We hope that Dr. Kales et al., having drawn attention to this important finding, will conduct a prospective study comparison of these groups that takes into account the biases that can operate in such a study of prognosis R1574BCFIJCAH.