Уникальный идентификатор сессии в Python

Это вариант проблемы групп и островов. Вы можете использовать разность номеров строк для описания группы:

select t.*,
       dense_rank() over (partition by patient_id order by first_contact) as location_name
from (select t.*,
             min(contact_date) over (partition by patient_id, location_name, seqnum - seqnum_2) as desired_result
      from (select t.*,
                   row_number() over (partition by patient_id order by contact_date) as seqnum,
                   row_number() over (partition by patient_id, location_name order by contact_date) as seqnum_2
            from t
           ) t
      ) t;