CLI - Identify improvements to CLI tool

[nathanbunker]

Minimal solution 1: Excel spreadsheet with extra header columns that include:

• ID1, ID2 FN1, FN2, LN1, LN2 including all the other fields used to match to allow the user to visually compare the patients. Ordered in their matched order.
• Add the mismo signature including the matching score (zero through F) to identify how they matched
• Add a Feedback Column (which IIS responses will be either Agree/Match, Disagree/Not a Match)
• Add a Reason Column - which user will need to indicate why they determined that match response. Free text field.
• Since output file will contain PHI then AIRA will need to support jurisdiction by training and guidance on how to return the file appropriately.
• Proposing the need for a second (input/feedback) File which is returned to AART to inform AART/AIRA what the IIS's final decision was on the matches found to improve the tool.
• Initial (output) file used for getting the jurisdiction the matched data, but the second (input/feedback) file will contain the Signature column, Feedback and Reason column in it. This will allow jurisdiction to return feedback to AIRA on the matches.
• Need to identify a way to have the IIS focus on prioritized matches (i.e., number of times signature appears, etc.) since they may not have the time or bandwidth to review the entire data set.

Goals of this proposal:

1. improve the output file
2. Aim to get feedback from the jurisdiction on their analysis of matches and reason why they determined the match via that second (input/feedback) file.
3. Jurisdiction can use the output file to upload in their system to improve/use for deduplication as an incentive.

Minimal solution #2: Command Line Interface in the file

• Shows the match side by side, allows the user to determine if they do or don't agree with the match and document the reason for the decision.
• This solution would allow the user to save their progress and user friendly.
• Pros: allows for receiving feedback from users.

• Cons: Different environments on computers may have the application appear differently. System security may not allow for running the file in system.

Dream Solution: Interface where it shows the record side by side and be able to see if match or not. Need human validation to validate the matches and verify signatures.

• Document mismo match score
• Findings category on the top of Real Match, False Match and Reason
• Left side navigation that contains the tree of matches
• Could use the filter/category of Match to auto merge clients in bulk versus one at a time

Day 2 Group Brainstorm:

Create an HTML output that would show the graph form of counts of repeated Patient ID. Output that says you have number of patients that match to 1 person. Use hash output to allow us to visualize.

[adelacruz84]

Smarty Integration:

• Cons : uses an API call (open network), network latency, upcoming change to Smarty API, batch versus sequential.
• Pros:
• Feedback: Smarty is not aligned with the current data. have the jurisdiction run Smarty first. Run the extract with Smarty and without Smarty to see if there is a difference.

Placeholder Names

• This would be a pro for having the second (input/feedback) file to see the if the reason was due to placeholder names.
• Generate an separate output file that detects a patient falls under placeholder name for IIS to review results.

[adelacruz84]

Final Decisions:

Modify the existing output file to add more columns:

• Column - Match Status: with the response of Match, Not a Match, Unable to determine. Match and Unable to determine mean that Mismo was correct in flagging it as a possible match.
• Column - Fields(s): This will be the fields or data points they used to determine the decision of No Match clients. Free text field to allow users to say what data point they used to determine it was NOT a Match. Example: Middle Name
• Column - Details: free text field to allow the user to add more details about the field used to determine the patients were not a match and the reason for not a match. Example: Patient 2 had a different middle name.
• Sort the output file to have the largest count of signature repeats shown first (on top). We would need to communicate this to the IIS for awareness on how to work the file.

IIS would return a modified version of the file that contains only the following columns with filled in response that would be emailed back to AART@immregistries.org.

• Signature
• Match Status
• Fields
• Details