Person matching

Hi Emily, I recall that this has been a topic in the past on how different states use what algorithms in their surveillance systems to do matching. I'm not sure I have general notes from those past calls. In CTEDSS-Maven, there is a MPI. We are using the default settings, and if the match threshold is not reached (I think it is set to 85 or 90%), the system will kick out a notice for person deduplication. The match is based on last name, first name, DOB, address  in that order but these are weighted. Unlike other states, we do not have a single person or group of people ensuring that John Smith is the same across all disease areas; we let each disease area do their own person deduplication (historical decision), so I know we have people who have moved who are actually the same person in the system more than once. Unfortunately, we are not allowed to collect SSN, and we don't capture vital records ID, so we have no "gold standard" unique ID to use for matching. 

Hi Emily - we are a Maven-using jurisdiction and we have a complex algorithm that uses exact match, transposition, fuzzy match, heuristics and other tools such as non-conflict bonuses to give a score to any incoming record and decide whether it should automatically match (no human intervention needed) vs. partially match (human review required).  We match on first name, last name, street address, SS (if available), alias, and date of birth.  We also use previous addresses and names.  I might be missing some details!  We have penalties for common names and addresses for large congregate living facilities such as prisons.  I don't know if that is helpful but we do feel that our matching is fairly robust, even though we have some major challenges such as many short and often transposed Asian names for patients with hepatitis B.  I could have one of my staff provide more detail if you are interested.