Date of first occurrence DataFrame

[SpiroGanas]

Create a pandas dataframe where the rows are SUBJECT_IDs, the columns are ICD 9 diagnosis codes, and the values are the earliest data where the subject has a record of having that diagnosis.

If a patient never had that illness, the values shown will be NaN.

This type of table should make it easy to create an "observation period" dataset consisting of all diagnoses that were recorded in some time period prior to an event. For example, if we are analyzing patients with heart failure, this DataFrame would allow us to easily find all of a patient's diagnoses in the two years prior to the heart failure diagnosis.