This is a conceptual question; I'm struggling to decide whether I should cluster my standard errors.

I have a cancer registry of all cancers of a particular type, and am looking to run a simple logistic regression to determine the odds of a tumour receiving treatment X versus treatment Y. The analysis is undertaken at a tumour level, but some tumours are nested within patients. As such, I'm inclined to cluster on the patient identifier (e.g. vce(cluster patient_identifier)), but seeking a second opinion.

Many thanks.