We read the paper entitled “Depression and associated variables in people over the age of 50 years old in Spain”1 with great interest. We would like to underline the importance of this study, as it elucidates the factors which are associated with suffering depression. It also covers a very large sample. Nevertheless, we would like to express some observations and questions.
Was the possibility of multiple colinearity evaluated? Several independent variables may be associated with each other, such as perceived physical health, ELA difficulties and chronic diseases, or marital status and household unit. This could affect the supposition of independence in logistic regression.2
It would also be good to know why the weighting was not considered for the analysis.
Likewise, the transverse analytical study used the odd ratio (OR) to measure association. As using the OR in a transverse analytical study overestimates the strength of association when there is a high prevalence of the result (the outcome),3–5 as in the study by Portellano which shows a prevalence of depression in Spain that is higher than 25%. Due to this, it would be ideal to use the prevalence ratio (PR).3–5
On the other hand, in Table 4 does it make sense to cross emotional disorders with depression?
Lastly, other published studies find a strong association between a history of alcohol consumption and depression.6 It may therefore be of interest to study this variable. Other important variables are patient history and negative events in their lives.7,8 In this case it would have been optimum to refer to these lacks as a limitation.
Please cite this article as: Gutiérrez-Pérez GL, Mesones-Tello JP, Medina-Conde BB, Alvarado GF. Sobre condiciones de aplicación de regresión logística. Rev Psiquiatr Salud Ment (Barc). 2017;10:216.