November 16, 2010 CBCC Conference Call

Community-Based Collaborative Care Working Group Meeting

Back to CBCC Main Page

Attendees

• Jon Farmer
• Suzanne Gonzales-Webb CBCC Co-chair
• Mary Ann Juurlink Scribe
• John Moehrke Security Co-chair
• Ioana Singureanu
• Richard Thoreson CBCC Co-chair
• Serafina Versaggi
• Tony Weida
• Craig Winter

Agenda

1. (05 min) Roll call, approve minutes November 9th, call for additional agenda items & accept agenda
2. (55 min) Draft Semantic Health Information Performance and Privacy Standard Project Scope Statement

Review of HL7 Project Scope Statement Data Segmentation Pre-Requisites Preview of scope statement

The EHRS functional model working group is trying to match functions to content standards. The impetus of this scope statement is to improve data quality measures as well as to define missing functionality. This is required to support segmentation and privacy e.g. define automation for consent directive. Generally the first sections describes the art of HL7 balloting. We are expecting to;

• make changes to the EHR functional model e.g. some of the behavioural functional profiles that exist
• refine the DAM describing the health information for performance and privacy for data segmentation in particular
• refine vocabulary domain and value sets where they exists

2. Project Intent

The project intent is to revise and supplement current standards

2.a. Ballot Type

Draft for comment informative ballot or DSTU ballot

3. Sponsoring Group

We don’t have a complete list of project participants; however we expect groups to identify themselves as they hear about the scope of the project. We will seek more involvement from other groups e.g.

• Public Health reporting
• The National Quality Forum may be interested in the performance and quality data that we are looking to record

We are expecting to make changes to the EHR functional model or functional profiles; we would like to enlist Jim Krentz from SAMSA to help submit these for consideration as enhancements to the EHR-S FM.. We expect the National Quality Forum (NQF) might also interested be as a co-sponsor, and we will reach out to Floyd Eisenberg (NQF) and the Structured Documents Work Group (Bob Dolin) to see if there is interest in collaboration.

4. Project Definition

Section 4: is about background – the fact that data segmentation has emerged as a requirement

4.a. Project Scope

There is a need to identify the Meta data required for data quality and for data segmentation. We are looking to define the level or quality the data collected by electronic system will have to meet in order for data segmentation to be automated. Clearly you can read a paper document and determine whether something is to be embargoed or not but if you can automate the process using electronic data this is the future. These issues and data perquisites are needed to support privacy and policies. In terms of quality we are looking at advance capability of different quality measures and quality data that is intended to evaluate performance of health care delivery organization. But even there we are struggling to identify precisely what of the multitude of data that is already being collected would be relevant to measure overall performance. Furthermore sometimes it will be information that is not collected today e.g. evaluating the access to health care information - is the information capture? This information could be captured in a disparate system, e.g. not integrated with information that is linked to other data related to a performance measure.

4.b. Project Need

Issues to be covered in the project analysis

• Relationship of data context to performance measure
• Not just talking about one provider EHR system, in the new interoperable world care is coming from multiple providers
• Information needs to be semantically interoperable

Different approach depending on quality of the data in the underlying systems, how to handle the data if:

• Structured data / unstructured data
• Standard encoded / not encoded

Richard - The state is the single public payor for the safety net population and the question is do they have the information they need to assess the quality of health for their entire population. How is the information shared? Quality is sometimes focused on the provider level and aggregated in a particular provider system. A larger question is to look the whole pattern of care for the entire population as that provides information about the needs, services and about the outcomes. If you don’t have that information you can’t make policy in an informed way. If information is important for someone’s health it may also be a factor for public health. Eventually we need a scale so people can equate the performance of an information system based on scope of information and the quality/detail of the information that is available for people to make policy into an administered program.

Ioana - Quality of information also depends on the completeness of the information. We want to make sure that we understand what is considered to be a complete set for segmentation for quality evaluation? That is quality evaluation across the population and giving decision makers the possibilities to look for patterns. Types of completeness

• Coding value sets
• How well the data is being collected

What are the criteria to evaluate the data e.g. access patterns

Richard - There are new privacy and security requirements brought forth by Subtitle D of the Health Information Technology for Economic and Clinical Health Act (HITECH). New concerns that need to be addressed also the ongoing concern - pay for performance, giving decision makers a way to evaluate programs and population health.

Richard – For the policy agenda we need to know what does the information tell you or doesn’t tell you

Ioana – These are brand new concerns, also pay for performance

Additional background link http://gforge.hl7.org/gf/download/docmanfileversion/5950/7714/11-1_SHIPS.doc

John – Concerned about scope creep e.g. talking about quality of any particular record – then quality from an operational environment provides quality treatment vs not quality treatment. There seems to be overlap with the Public Health and Emergency Response Working Group.

Ioana - We need to insure if there is overlap we are reusing

John - What is the relationship between the segmentation part and the quality part e.g. segmentation is needed whether you have good data or not good data

Ioana – Is quality of data sufficient for data segmentation? Quality of data collected during the health care delivery process e.g. the quality of data that is produced by the EHR system.

• Is this data sufficient so it can be used automatically for data segmentation and computing measures?
  • Subset of meta data that goes along for segmentation
  • Subset for quality measures
• How the data is used to measure specific quality measures, and this information becomes available to inform policy makers

Both uses of the data require specific data quality. We will identify what that quality of the information is and represent it for everyone to understand. If we are segmenting data what data should we have in our hands before segmenting it.

• What sort of data we need so we can apply policies to it– that is what segmentation is all about (smaller work item)
• What quality of measures do we need to address today to understand range of concerns we have in deriving those measures (extensive work item)

Add sentence to scope: Data quality in real time performance evaluation (initially for high priority Behavioural Health topics)

Richard – Segmentation for privacy in terms of

• Quality of a service
• Outcome of service

All interrelated EHR certification could be based on functionality (stage 1)

4.c. Success Criteria

Ioana - Data segmentation we think of it as privacy and security primarily, and then deriving quality measures from the data and what specific subsets are required and what level of quality or range of choices one might have. Keep segmentation tied in with its purpose for enabling privacy and security and refer to a real time quality measurement for the performance part. So we have 2 aspects to the project, privacy and performance. In order to support privacy we need segmentation in order to support performance we need real time quality evaluation. Right now quality is determined on an adhoc basis by extrapolating from a lot of data and function typically done by humans. We are envisioning real time quality measurement based on the actual data or subset by the data – automated process for extracting the information.

John - What does the work segmentation mean? I hear you using it to create data that is of high quality, medium quality and low quality.

Richard - We are interested in knowing about any particular condition and how treatment is being done. Overtime we want to be able to track what the outcomes are. I would like to be able to pull a segment from a record according to measures I am concerned about and be able to extract information in an automated way.

John – Totally different use of the word segmentation. According to HIT standards and the policy discussions I’ve been in, the word is used to mean segmented data that needs to be protected according to rules. If that is the way you are using the word segmentation I have fewer problems with this proposal. When you use the word segmentation for the purposes of privacy then it becomes really confusing. However if you want to pull a segment of the record that is related to a condition then that’s ok.

Ioana - For this scope statement we are keeping it separate:

• data segmentation for the purpose of identifying protected data
• and the automated data process used for the purpose of real time quality measurement

John - These should be two separate projects they are not related

Richard - They are related – in the sense if someone’s access control service has to pass judgement what information can in fact be shared

Jon - This is an important point because the ability to measure quality in some cases will depend on access to data that is private and you need to think about it as one picture in order to understand the information

Richard - You need to convince me to share my information to some extent. Think about people pulling together genomics research they want to do

John -That is not segmentation it is authorization. I will be authorizing the use of my genomics data for secondary use Richard - Not wedded to the term segmentation

Jon - Do we really mean functionally information classification? The way to delineate data of the patient’s perception of sensitivity

John - What you just said is a project we need to work on. (Reference to what Richard said earlier: I would like to be able to pull a segment from a record according to measures I am concerned about and be able to extract information in an automated way)

This problem is not unique to the quality side of this project. That same problem is necessary for genomic research which has nothing to do with quality reporting nothing to do safety report. I want to have the discussion around segmentation independent of any specific one use case otherwise we will have a segmentation that will only work for creating a data set that can be used for quality evaluation. It will be totally useless for imaging analysis. Why are these married together?

Ioana - Two separate concerns that are not interdependent. This project deals with privacy and performance as two different aspects that are basically measurable that rely upon certain meta data set. The two sets of meta data would be different depending on domain where segmentation would be a constant across the board. We made a decision to address them concurrently because both of them rely on high quality encoded structured data. This is the premise so we have 2 different uses of the data one for data segmentation to determine which data is protected and may have to be embargoed and the other use is the same data may be used to derive data quality measures

John - I disagree with the premise. Data segmentation is absolutely necessary for low quality data and non structured data Ioana - You will only be able to segment all of the data because you are not able to look at entries of discreet data, if all we have is segment data based on sickle cell anemia as part of a narrative report. If you have sickle cell anaemia you will have to give us the permission to disclose it as there is no way to tell which data elements relate to sickle cell and which don’t. You will have to handle the segments of abstracted data differently with different levels of control if you don’t have atomic data that you can work with

Ioana - The more structure the more encoded the more you can take out specific atomic elements that deal with the protected information otherwise you can’t determine what is in and what is out. This is what we talk about often but it does not land in our documents

Ioana - Health care and what allows a computer system to segment data and embargos data and put it in different compartments. So this is where the level of quality …if you are dealing with all unstructured data there are limitation of what you can do with the data in terms of segmentation and real time quality assessment. Similarly if you are using structured data but are using all your own codes there are other limitations there are other things you need to do to compensate for this issue.

Richard - What is useful for segmentation privacy protection has a lot to do with substance of information – value for quality of care, value for outcomes of care e.g. data with information

Link to [NCVHS Letter on Segmentation of Sensitive Health Information: |http://www.ncvhs.hhs.gov/101110lt.pdf]

Uses the word segmentation with respect to privacy sequestering certain categories of health information and describes categories. Basically the letter’s focus is privacy. We should consider referencing this letter with respect to segmentation. Walter expressed interest in participating in this project and said it would be helpful to address some of the issues requirements are there but we need to understand how we are going to satisfy requirements. Suzanne – Table this portion of the discussion till next week when the letter is published?

Ioana – Example of how we will derive specific data quality information from a measure.

• Type of information
• Attributes
• Codes constrained

Deriving information of a data model instantiated for specific values that correspond to a set of measures from which we can extrapolate what data will be required for a specific domain/condition. Example of how we specify vocabulary binding and the types of data that would be required to compute the measure. Similarly we have the composite privacy / security DAM that specifies the rules that apply to data for segmentation. From that we can derive the meta data that would be required to support these rules.

4.d. Project Objectives / Deliverables / Target dates

Draft for comment ballot – May 2011 Informative ballot – Sept 2011

Interim deliverable draft electronic health record system functionality including specific functions and conformance criteria to support the operations related to data segmentation and performance

We will have to have the scope statement approved before January 15, 2010 Will add in ballot reconciliation, we expect one draft for comment and one informative ballot in January 2012

Question about the process to track minor editing changes and any new requirements coming from this work to Security – Privacy DAM

• Refer to Security Privacy team
• Tracking for future versions of DAM done in GForge (queuing when DSTU comes up for ballot)
• What we are doing here is to identify meta data for the domain model, if changes are needed to the DAM the security and privacy group will need to make the changes.

Richard - Next week

• Want to think and revisit issues of scope that John raised
• How we are using the term segmentation - in US realm how is it being used
• We will review segmentation when we will have access to Walter’s letter
• Connections to different work groups
  • Structured Documents – sponsoring / reviewing project scope statement for the National Quality Forum effort for eMeasure automate quality measurement from the EHR data. Serfina keeping track of dates for the discussions
• Quality of information vs quality of service (if we don’t have the quality of information we don’t have quality of service). The word quality is overused and we could do something to clarify it

Meeting was adjourned at 3:00 PM Eastern. No significant motions or decisions were made.