*2.5. Limitations*

Firstly, it should be noted that no estimation of the incidence or prevalence of the events of interest (depression and dementia) could be performed because this study lacked a population base.

In addition, the design type involved constraints in establishing the temporary sequence of possible exposures (depression and others) and effects (dementia and AD, with GDS 5–7), since the onset (diagnosis) of chronic conditions/diseases is always uncertain.

By including prevalent cases (in addition to incident cases), we may have included the most surviving cases, so long-term cases may have been overrepresented ("Neyman fallacy").

For institutionalized older people, one type of selection bias, "Berkson's bias," should be considered, because a selection of cases from a nursing home or dementia-specific facility population could contain a higher proportion of older people with a secondary disease.

However, as an analytical observational study, the main limitation of this study is the possibility of confounding factors not covered by the design. However, efforts have been made to reduce the influence of these confounding factors by using multivariate data analysis techniques.
