I think the study is somewhat more robust than you're letting on, since it compares HCQ against an alternative rheumatoid arthritis drug - Methotrexate. It's possible that the choice of prescribed drug itself is correlated with some other confounding factor, but it's certainly less likely to be so compared to HCQ vs nothing.

Also, the study itself is clear about the amount of evidence needed to validate whether HCQ actually reduces Alzheimer's risk: "This hypothesis merits testing through adequately powered clinical trials in at-risk individuals during preclinical stages of disease progression."