I am working with cancer data which provides a number of different indicators whether a patient has died.
However, it does not provide a definite date of when that person died.
We do have a set of dates associated with the patient which may or may not be populated. For example, when the patient was discharged, when the patient was last seen etc.
Should this be used to approximate when they died, or should the death be excluded entirely?
Based on the knowledge of your data pick the best date you can. If it is missing leave it out. The sheer fact that a patient died might be important for clinical trial recruitment, for observational research on longitudinal data it has little use.