Minutes from 9/15/22
4:00 - 4:05pm: Announcements and Approval of 8/11/22 Minutes (5 minutes)
Paul Jaskot is filling in as ITAC chair today.
The minutes from August 11, 2022, are approved.
Preston Nibley from the Duke Student Government is joining ITAC. Preston is a Duke undergrad majoring in Molecular and Cell Biology.
Congratulations to Brandon Le for his work with the Duke Graduate and Professional Student Government that helped to achieve stipend raises and a reduction in parking rates.
4:05 - 4:45pm: Medical Student Clinical Portfolios – John Haws, John Herr (20-minute presentation, 20-minute discussion)
• What it is: The Medical Student Clinical Portfolio application summarizes electronic health record (EHR) information from students’ clinical encounters to describe how students are progressing through required experiences and skills acquisition.
• Why it’s relevant: During medical school, students should have a variety of experiences in their clinical rotations and should see patients with a variety of issues. This will aid them in becoming well-rounded and versatile doctors and improve their clinical practice. The Medical Student Clinical Portfolio will allow faculty visibility into their students’ patient encounters, which will help ensure that students are seeing diverse types of cases as well as cases relevant to areas of concentration. Additionally, perhaps this kind of system and data analysis could be applicable to other sets of students or areas in general at Duke.
Tracy Futhey introduces John Haws who helped Duke with processing Covid data. John Haws along with John Herr and Mary Claire Thompson are providing data analysis internally for Duke.
John Haws begins by describing the application called the Medical Student Clinical Portfolio. Duke is a leader in having medical students work directly with electronic health records. At Duke, students log in and document their clinical rotations. Faculty want to understand what experience the medical students are having and what experience they are missing. So, John Haws’ team built an app to provide this insight. The high-level goal is to provide an overview of the student experience on clinical rotations. Faculty want this information in real-time. Faculty want to be able to view the notes that students are writing, want a summary of diagnoses that students are a part of, and want to know what students are seeing in their interactions with patients. Then, once this data is obtained, statistics on demographics, types of encounters, and what students are seeing with patients can be provided. Also, this data provides insight into whether some departments are doing great on writing clinical notes while other departments need nudging.
Q. John Board says this data is toxically sensitive.
Q. Tracy asks if there are ways this might be useful to other departments.
A. John Haws says there are applications for this with the nursing school and other clinical settings and they are looking to expand into other areas.
Q. Robert Wolpert – Faculty might want other information that is not on your list.
A. John Haws – We initially had a meeting to understand what the needs were. Then, as quickly as possible the team put something together for quick feedback.
A. Robert Wolpert – Over time, demographics may change, and new things may come up.
A. John Haws – Yes, things evolve. We recently learned that faculty want a more granular view of the student notes. A student writes notes, the attending signs off, and often the attending changes the notes. Faculty want to see these changes.
A. John Herr – There is a separate Application Program Interface (API) which we kept as a discrete repository from the web interface. This API could exchange other data sets with the web interface to make this applicable to other clients. This design provides for flexibility.
John Haws shows a diagram of the architecture. The production side involves interacting with the hospital’s electronic health system; this involves extracting data from Epic. The Protected Analytics Computing Environment (PACE) is where the application lives. This is approved for HIPPA data and is highly secure.
Q. Mark Palmeri asks about deidentifying data.
A. John Haws says it is very difficult to deidentify this data.
John Haws says a development platform has been built with fake data because doing everything in the production platform with PACE is difficult. John Haws demos the development platform with fake data. The data includes:
• Clinical notes
• Patient encounters
• Patient demographics
• Diagnoses
John presents a statistical summary of the data:
• 1,132 students for the past 4 years (2018-2022)
• 171,288 clinical notes
• Av of 151 notes per student
Q. Mark Palmeri – Not every place is hooked up to Epic. Is the Epic lens a good enough portal for everything? For example, rural North Carolina would be a blind spot.
A. John Haws – Currently, this is tailored to Epic but what we are ingesting is generic. If we could get formatting close to Epic, we could bring it in.
Q. Mark Palmeri – Is this for accreditation?
A. John Haws – I think that would align but right now, we are clinical notes-centric.
Continuing with the demo of the development application, John shows the student view. The student logs in with shibboleth and the ability to log in is assigned via grouper groups. Each student has a persona and will only see their own clinical notes and encounters.
Then, John Haws shows the faculty view. Faculty can log in and see a list of all the students. Clicking on a student presents the notes in raw form. Notes include who the attending was, the patient information, a summary of the encounter, and what the diagnoses were. This enables faculty to determine what broader problems the student could see and work with. Students have 30-35 problems that they are expected to see.
The next steps include:
• Collecting feedback from needed development partners
• Incorporating additional student development data
• Working with the MIDS Capstone project on structure and normalization of notes text
• Incorporating latent encounter information from notes text
Q. Mark Palmeri asks about Fire. We use Fire a lot for data links between protected and not protected data. Fire protects the data, so we don’t have to manually cleanse the data. Mark wonders if there is some infrastructure that exists already because all you care about is the collated data; you don’t care who the patient was.
A. John Haws – The Analytics Center of Excellence (ACE) team on the Duke Health side may use Fire. Fire did come up when supporting the University’s Covid response and we would like to learn more.
A. Mark Palmeri – Fire does the protection for you and is FDA-approved.
Q. Damaris Murry – Asks about using grouper groups.
A. John Haws – makes a distinction between reference groups and policy groups. For example, there is a reference group for all medical students and a reference group for all faculty but for use of this application, John’s team creates a separate grouper group with specific users.
Q. Michael Greene asks if something like this would be desirable for the learning management system. Ed Tech faculty say the learning management system is not quite doing what they want so maybe something akin to notes and providing analysis would be desirable.
A. John Board – Some institutions mine student data like grades, location, and the nutritional content of the food eaten, but Duke has not done this.
A. Mark Palmeri - Less on the course side but more is needed for lab experience, design projects, and CAD. There are zero common resources to say who worked on what. We have access to all this data, but no one knows how to work with this type of data. A framework of data exists but working with the data and analyzing the data in something like Tableau is what is needed.
A. Sunshine Hillygus – Don’t assume “build it and they will come.” Faculty are used to what we are used to. So, you need faculty input because if they don’t know how to use it, they won’t use it. Also, just because the university can and this is not covered by the IRB, doesn’t mean the university should. But data analysis is important and there is not sufficient expertise among staff.
Q. Colin Rundel – On the health side, everyone is using Epic but some use other data resources, and some data is on paper, some data is on GitHub, etc.
Tracy says John Haws has discretionary resources that can be contracted if someone has needs that they would like to see addressed. John Haws says reach out; there is an initial conversation and especially, if the need has usefulness for the overall institution, there are resources available. For example, there is a project underway right now to study the use of space on campus.
4:45 – 5:00pm: Duke Digital Signage Update on the Transition from Four Winds Interactive to OptiSigns – Steve Toback (10-minute presentation, 5-minute discussion)
• What it is: The Duke Digital Signage system provides a lightweight way for users to create signage and place those signs on one or more monitors/screens. There are close to 200 locations on Duke’s campus.
• Why it’s relevant: Due to a change in licensing and costs, Duke has begun to transition off the Four Winds Interactive (FWI) platform to the more economical OptiSigns platform. We will provide a brief update on this transition as well as a look at the challenges and opportunities that are present. Expansion of the Digital Signage system can increase general awareness of announcements, including recruitment and job opportunities and upcoming events and deadlines.
Steve Toback provides a Duke Digital Signage Update.
Four Winds Interactive software drive all Duke digital signs. Although Four Winds is very complicated to use, it could be used by all departments (even off-campus) and everyone could manage their own signs. The annual cost for Four Winds Interactive has been $15,000 per year. Recently, this has gone up to $120,000 per year. Because the Four Winds software is so powerful, a $700 computer is needed for each sign.
Because of this jump in cost, Duke is switching to OptiSigns. OptiSigns charges $135 per sign. Duke has 200 signs so this will cost Duke $27,000. Another advantage is that the hardware needed for an OptiSigns sign is under $100.
Steve talks about the Four Winds Interactive digital signage architecture. Jpegs were stored on Duke Network Attached Storage (NAS), or video lived on YouTube, or consisted of an RSS feed as was the case for Duke Today. Steve’s group contracted with Ryn Nasser’s group to create the DukeFlyer publishing system. Duke Flyer enabled the creation of a digital bulletin board without the need to learn a complicated system.
Going forward, OptiSigns along with inexpensive hardware can be used with Duke Flyer. A slide can be created with Canvas or Photoshop and uploaded to DukeFlyer. DukeFlyer Player has just been released and can even run on a Raspberry Pi. DukeFlyer is totally free to use. After meeting basic security requirements, you can run feeds. You can use a smart monitor or TV or you can use a cheap PC to run a set of slides. This setup is applicable for emergency communications, and you can even put the application on your cell phone. Finally, you can put your flyer into a pool to make it accessible all over campus.
Q. Lindsey Glickfeld – Can you advertise?
A. Steve – No, but you can call to ask someone to put your slide into their space. In the beginning, there was a button to publish your slide elsewhere and everyone was getting spammed so this feature was removed. If you want your slide published by another department, contact the department’s communication officer.
A. Tracy Futhey – It would be nice if there were metadata so you could flag your slide as being applicable to another department.
Q. Mark Palmeri – Is there a way to turn these signs off at night to save energy?
A. Steve – There was a way with Four Winds. We will need to check to see if OptiSigns can do this.
Q. Michael Greene – Can you take an RSS feed and put it into a Zoom room?
A. Steve – Yes.
Steve speaks to the OptiSigns challenges and opportunities.
OptiSigns Challenges
• Interactive wayfinding – can make change to Google doc or Excel
• Detailed integration with SOC status updates – working with other OIT groups to move to web or phone
• Emergency integration
OptiSigns Opportunities
• Lower cost of software and hardware along with ease of use
• Expanded social media and video options – with Four Winds, each feature was an added cost
• Happy Birthday – pertinent birthdays can be pulled from Excel based on month without the need to update continually and each month
• Live streaming cams
Steve concludes with the announcement that the Bryan Center Studios are now open. These studios provide opportunities to create podcasts and videos and the team is happy to provide tours.
5:00 – 5:15pm: Ending the Exceptional Use of Data During COVID-19 – A 10-15 minute discussion led by John Board
• What it is: In mid-2020, with full transparency, Duke began collecting and aggregating health data such as symptoms and inferred locations. The primary purpose of this unprecedented action was for contact tracing during the COVID-19 pandemic. (ITAC minutes from June 25, 2020.)
• Why it’s relevant: The collection and use of this data is winding down, and just as it was originally presented to ITAC as one of the initial steps, we are formally updating the Council of its cessation.
John Board gives an update on Covid data collection. During the Covid pandemic, it was institutionally important to monitor the movement of people so personally identifiable data from Wi-Fi and card swipes was gathered and collated with Covid health data. Normally, Duke would never gather data like this absent a subpoena. Tallman Trask worked with us on this at the beginning of the pandemic and a declaration along with enumerated types of data collected was published. There was an understanding that there would be an update each semester. John is here to report that there is no longer a need to gather this type of personally sensitive data and collection has stopped.
A discussion follows about this data still existing even though it is no longer used, but there are explicit retention policies and Charley Kneifel says most of this data is deleted after 30 days. Colin Rundel asks if the data was deployed and used. The answer is yes, to the point of being locked out of halls if testing was missed.
Mark Palmeri concludes that everyone can now uninstall the SymMon app.