I'm working on a clinical study to investigate the influence of an assortment of self-reported factors on depression in a specific condition, and have a question that extensive googling and reading has not yet been able to provide an answer to. Due to the restricted sample size (N = 96) and an abundance of questionnaires and measures per participant, I've done a Principal Component Analysis of the sum scores from all included questionnaires (selected for the study due to their previously established relations with depression), rather than perform a PCA with all items. Theoretically there is a rationale for exploring shared variance between predictors, seeing as studies often employ single measures (i.e. self reported loneliness or trait neuroticism) without considering related constructs investigated in the literature and overlap amongst them.

The question I have is whether running a PCA on sum scores from questionnaires, rather than all the items from the questionnaires, seems like a reasonable approach, or whether this is somehow problematic?

Thanks in advance for any and all help!