2020 03 25 Minutes

Agenda

Attendance

• Keith Boone (Ai)
• Katie Iser (Ai)
• Michael Donnelly (Epic)
• Christina Caraballo (Ai)
• Chris Lindop (GEHC)
• Josh Mandel (Microsoft)
• Dale Evanchak (NextGen) - Product Manager for Connected Health Suite (Mirth, etc. Not on Health Data Hub (repacement for results CDR))
• Joe Wall (Meditech)
• Ankit Agarwal (New Wave)
• Aaron Seib (New Wave)
• Mark Scrimshire (New Wave)
• Samir Mahapatra
• Steve Nichols (GEHC)
• Gino Canessa (Microsoft)
• Dave Pyke (Ready Computing)
• Jenni Syed (Cerner)
• Bill Mehegan (Carequality)
• Dongwha Kim (New Wave)
• Fola Soyoye (New Wave)
• Carl Anderson (Microsoft)
• John Moehrke (ByLight Consulting)

2. Signup
3. Announcements/Current Work
4. Hot Topics

Hot topics:

• Measure Report vs Group Several folks have suggested using Measure Report instead of group. There are some benefits, as well as some challenges. Let's discuss.

  • Some industry discussion among EHR vendors; some DaVinci work too. In r4, MeasureReport is FMM Level 2, where Group is FMM Level 1. Group hasn't seen as much attention from the community.

• Classifying Beds Josh notes: It's going to be challenging to get facilities to classify their beds to support automated reporting. How can we resolve this?

  • We're looking at ~4 facets, and 3-10 values per facet, which would make it hard to convince health systems to do automated reporting across this whole space. Care providers are already way overloaded, and focused on near-term logistics like making virtual visits work smoothly.
  • So, reporting on data that exists in the facility, but maybe not neatly coded in a master file, that's still manual work for someone.

• Granular Source "ADT A20" type info, about beds (occupied, unoccupied, contaminated, ready for use, etc) -- this sort of info could come out of existing bed management systems, existing outbound v2 feeds.

• [jmandel] Understanding the data supplier perspective. Who's going to populate data for this IG? What sources will they pull together to do so? Will they have enough information without roping human staff into the job of data entry? What granularity of information can be conveyed most readily from existing bed management systems?

Meeting Notes

• Keith: Welcome! Please sign up if you haven't yet

• IG content is in github including fsh source!

• High-level: what's the problem we're trying to solve? We should have a spec that people can quickly implement, using different solutions and data sources to communicate the data to some system that can aggregate and report to public health. Goal: shift the burden of reporting and aggregation to resources outside of the healthcare facility.

• David: Are we talking with public health? Do we know what they need?

• Keith: Many of us have, independently. Bed availability is a top concern, with a common list of asks very similar to HITSP 2006 requirements (utilization of beds, equipment, faceted by ICU vs ED vs Other, Isolation status)

• Josh: We talked a bit yesterday re: white house covid-19 task force, helping federal policymakers understand what's easy/possible, and the associated timelines.

• Keith: We've seen proposals from Epic, Cerner -- with customer-driven reporting requirements. A lot of the same stuff in terms of faceted bed counts, devices most especially ventilators, other equipment that might be substituted for ventilators, some discussion about staffing (though many organizations manage staffing through platforms distinct from their EHR systems, tied into scheduling, badge access, payroll, etc).

• Jenni: Cerner has supported CDC biosurveillance requirements for some time. [Syndromic Surveillance IG(http://www.hl7.org/implement/standards/product_brief.cfm?product_id=503).

• Michael: Epic has been working on a series of phased metrics -- posted at https://chat.fhir.org/#narrow/stream/226195-Covid-19-Response/topic/Metrics/near/191739816, and copying here:

Open ICU Beds  ["ready to receive patients"]
Open Acute Beds
Total Number of Ventilators 
Number of Ventilators Available 
Stats from the last week

Communicable disease screening compliance

COVID-19 tests ordered
COVID-19 tests with a presumptive positive result, by age group
COVID-19 tests without a final result yet
Percentage of all visits that have a presumptive positive test result, by age group

• David: Will isolation bed counts be added to Epic's phase 3 metrics?

• Michael: There's active work on defining 3rd and 4th phases; we'll bring this feedback to our clinicians

• Keith: shows Cerner customer survey, including "With your consent, and on your behalf, Cerner will submit the following data from your Cerner Millennium® and/or HealthIntent® platforms to the White House Task Force, CDC and other public health authorities" with data elements including:

Emergency Room visits
Inpatient Room census
Isolation room census
ICU admissions/census
Hospital Beds
Disaster/Expansion beds
ICU beds disaster/conventional
Ventilator usage
Respiratory Isolation Room capacity and usage
Rooms containing PUI and COVID-19 positive test patients
Positive COVID-19 test
Negative COVID-19 test
Pending COVID Test
COVID Patient Age
COVID Patient sex
COVID patient zip code
Respiratory illness visits
COVID lab test ordered
COVID lab tests on hand
PPE inventory
Medical Supplies relevant

• Keith: These details provide some "signal" that give you an idea of real-world capacity, but you need to layer on information about staffing, etc.

  • "Census" is a countable number that isn't really a matter of opinion
  • "Licensed capcacity" is publicly available information, so orgs shouldn't need to report these independently
  • "GSP Coordinates" for a facility are publicly available, and could help support this effort

Topic: Group vs MeasureReport

• The Cerner list focused on counts rather than ratios. It's nice that MeasureReport can deal with facets, built-in. But there's a challenge when there's an ED system, and a seprate Hospital system, and a separate ICU monitoring system. And Group lets you roll things up into a granularity of choice; though MeasureReports could be submitted indepently by source system.

• Josh: we see some common measures for sure, but also very active new development about what needs to be counted -- so we want to provide an IG that's flexible and can handle new measures even if they're defined after the IG is published.

• Keith: that's fair. Looking at Group, it seems robust to all kinds of counts; I didn't think about ratios initially. If ratios are being reported at a facility level, they're hard to roll up unless you know the numerators and denominators separately. CDC definitely understands the challenges of working with ratios, and there's a preference for numerator + denominator counts.

• Chris Lindop: what's the criticality for this information? How soon does it need to be available?

• Keith: I think hours not minutes. But will people be using this info for deciding where to route patients, or longer-lead-time tasks?

• Michael: People will use it for anything they can think of, if it helps

• Keith: Even if you use a report to guide your choices and next steps, you're still picking up the phone before you send a patient somewhere. After the Boston Marathon bombing, there were similar routing decisions that needed to be made very quickly.

• Keith: So what do people think?

• Michael: Agree with Josh's point that Whatever we do, it should be flexible to accommodate measures that might be define later

• Gino: When we look at bed counts, should we be able to express debt (negative numbers, patients present but without beds)?

• Mark: We should make data submission as simple as possible, and let reporting do the heavy lifting.

• Josh: Agree with the principle, but what approach would be the simplest?

• Keith: We should look at examples comparing how a facility would report data out of different systems (ED, ICU, Acute) beds, captured as a MeasureReport, or 3 MeasureReports, and see what they data would look like that way vs in >>3 Groups.

• Keith: For putting institution-wide numbers together, some systems like major EHR vendors might have panoptic views; others might need to break reporting down into counts from several source systems.

• Meditech: We've been providing guidance for our customers to help them use existing tools.

• Gino: we're working on a tool to generate examples, https://github.com/microsoft-healthcare-madison/learning-spike-erp has work-in-progress

TODO: Gino will build examples, per discussion above

Topic: Classifying Beds

• Keith: We'll want to understand a minimum set of facets, to help reduce the combinatorics. Are there thoughts on saying "ED Bed" vs "ICU Bed", or "Supports Isolation" vs "Doesn't support Isolation", etc.

• The challenge here might be classifying individual beds according to these taxonomies. Can this be automated? How accurately?

• Mark: could these details be deduced based on low-level data?

• Keith: Yes, that's promosing

• Josh: But those models can only be run by systems that have the low-level data; if we push sumamry reports along, the organization receiving the summaries wouldn't be able to make these deductions.

• Josh: My uninformed intution would be that for a bed management system, passing low-level events through to a centralized reporting backend might be easier and faster to build, vs trying to create active/correct summary-level information and reports within the bed management system.

TODO: Michael will try to bring in someone from Epic's bed management team

TODO: Joe will try to bring in someone from Meditech's bed management team

Discussions

Measure Report vs. Group

• Some industry discussion among EHR vendors: some DaVinci work here too.

• In r4, MeasureReport is FMM Level 2, where Group is FMM Level 1.

• Measure Report vs. Group is priority

• What does success look like for a guide like this?

  • What is the problem we are trying to solve? We want people to be able to quickly implement this using a variety of sources and start communicating data so it can be aggregated for public health and shifting the aggregating and reporting to OUTSIDE the facility.

  • What does Public Health want? Their asks are about beds and equipment: ICU beds, isolation beds, ED beds, and other

    Would like a finer grain set of detail

• COVID TF - lots of top down looking for information from government agencies.

• Path for this work? A whole Ai team working on a path for this.

• Reporting on facility staffing a concern but probably at a later date because this will call into other factors such as key cards and staffing systems

• Reviewed Cerner Data Elements

  • Not seeing ratios, seeing more counts.
  • Challenge: facilities using two different systems to manage ED and rest of hospital, also something else that monitors ICU data. Reporting data from separate systems and aggregating it up.
    • Group works well here
    • Also argued you could take one large report then fragment it into separate parts
    • Group approach has more simplicity
  • Clearly a core set of measures that are of interest though we aren't going to know the more specific groups
  • Whatever we choose we want to be able to handle bed counts but want something flexible enough for account for measures we did not know about at creation
  • Beds, ventilators, and people counts
    • Whatever we do it should be something that is flexible enough to handle measures we aren't thinking of right now.

NOTES: Quantities on groups are unsigned, Facilities at one point may be reporting negative beds

System handling the reporting also computing and communication would be best, keep it as simple as possible and let the reporting do the heavy lifting

On the reporting side you can get into the analytics, computations, etc

• Start to put together some examples and start to put together data about what people want out of reporting
• Gino working on a tool to generate these examples and will share on GitHub

Classifying Beds

• If people are going to report about beds and we have multiple facets and values per facets which would make it hard to convince health systems to do automated reporting across the whole space. Care providers are overwhelmed and trying to make virtual visits work.
• How can we simplify this?
• How can we classify bed around some number of facets? What are the key facets? What is the minimum value set for faceting this?
  • Proposals with value sets with codes in existing IG
  • Classifying according to those facets, someone has to pick the core values
  • Concern is that will be hard for people to do on the ground but also because we have to make it as easy as possible
    • Interfacing with bed management systems? No recent experience. Would be good to have someone with a BMS system involved.
    • Anyone at Epic who could join and help us with this? Michael D: Will try but they may be involved with another effort right now.
    • Joe Wall: can try as well, but unsure.

Granular Source

• ADT AD20 type info