Hye-Chung Kum (PhD, MSW), Principal Investigator, Texas A&M Univ., Computer Science (information privacy), secondary data analysis (user)
Eric Ragan (PhD), co-Investigator, Texas A&M Univ., Information Visualization, user study for human-computer Interaction (HCI)
Ashok Krishnamurthy, UNC at Chapel Hill, Co-I on Mid-South CDRN
Jeff Baumes, Kitware, Open source health application. HCI
Daniel Basile, co-Investigator, Texas A&M Univ., Patient (chronic illness), Security Expert, IT support for research
Elmer Bernstam (MD, MSE), Principal Investigator for sub, UT Houston, Health Informatics, MPI, CER, Research Data Network coPI (user)
Jeffrey Curtis (MD, MS, MPH), Consultant, ,UAB, Clinical, Research Data Network PI (user), CER, PCOR, ELSI
Not able to attend. Potentially, will join the user committee meeting in September
Li Xiong, Emory University, PI of PCORI project on Privacy
Sean O’Brian, Duke University, PI of PCORI project on Privacy
Peter Yu, Texas A&M Univ., HIPAA Privacy Officer
User Study (Need Chrome browser)
Training & Practice Questions 1 (10 min)
Mode 1 (10 min)
Mode 2 (10 min)
Training & Practice Questions 2 (10 min)
Mode 3 (10 min)
Mode 4 (10 min)
Mode 5 (10 min)
Discussion & Feedback (20 min)
On the Interface Design Options
Since manual record linkage is a taxing task, the mental toll could be reduced by presenting just 1 pair of record at a time and not inundating the user with a lot of data pairs.
Visual markup helps
Logic that underpins the closing out of information is very reasonable and practical.
As expected Mode 5 with only symbols is very difficult
Depending on the mode, the thinking process for decision making varied
Optimal privacy (mode 4) is the easiest in terms of decision making but is not always best for accuracy.
Mode 3 is more information to process, so maybe more difficult. But has better accuracy
Might be useful to add a flag for people who might possibly be twins. Also for nicknames.
Frequency icons are helpful, but
Need to make the size of database very clear to give the icons some context
Need to inform the users that the icons represent the frequency of the name in the database it is coming from.
Meaning of icons are sometimes difficult to remember: Emphasize to users that they can use “mouse over” to see the meaning of icons
On the aims of this project
When full data is needed for linkage, researchers can get access. But the cost is high
There is now more national movement in deploying encryption based private record linkage without a full understanding of the validity of the results
More study is needed on the validity of the encryption based methods, as well as other options for privacy enhancements, such as this project.
It will be important to test validity of results, both false links as well as missed true links, from this method on real projects.
Planned in the last year with UAB & UT Houston data
Separate from the interface design, it will be important to make safe deployment of the prototype software easy so the other sites can test the software. For this purpose, developing the prototype software as a simple VM from the onset might be best to ensure security of real data during testing.
Project Update (10 min)
Sep 2017: User Committee Meeting
Sep 2017: Submit to SIG CHI
Nov 2017: NGT session with ELSI experts
Feb 2018: NGT session with patients via Webinar (Arthritis Power Patient Powered Research Network)
