Monday, July 25, 2011

Distributed population queries - a national priority

 By Rich Elmore, Initiative Coordinator
Office of the National Coordinator for Health Information Technology

This August, ONC will be sponsoring the “Summer Concert Series” on the state-of-the-art of distributed population queries. The “Summer Concert Series” webinar presentations will provide tremendous insight from a real-world perspective to interested stakeholders on practitioners’ experience with distributed queries.     

Distributed population queries can be applied to public health surveillance, research, quality measures, chronic disease stratification and many other purposes.  These queries return aggregate information from a standard clinical information model, leaving protected health information safely behind the health care organizations’ firewalls. 

The use of distributed population queries for clinical insight has been hindered by organizational limitations on data sharing, inconsistencies in the expression of clinical concepts and lack of standards for distributed queries.  This has ultimately limited their use, with a few notable exceptions, to large healthcare organizations with more sophisticated IT infrastructures and deeper research budgets. 

One may ask “Why not just centralize the data?”  Well for one, for non-regulated purposes many healthcare organizations will choose not to loosen their control over their data.  Also, as a matter of principle, patient level data shouldn’t be exchanged when aggregate level information will do.  Finally, keeping data close to the source may improve adaptability to dynamic information needs and responsiveness to patient information sharing preferences. 

As we approach critical mass of deployments of certified and standardized Electronic Health Record (EHR) systems, we may for the first time be able to consider broader use of distributed population queries.  EHR certification standards require greater adherence to standardized vocabularies and information exchange structures.  While challenges with comparative data will remain, Meaningful Use Stage 2 will likely further improve vocabulary, message and transport standards.  

As an example of distributed queries, Wes Rishel recently explored the changing public health reporting needs during an epidemic.  He suggested use of distributed population queries to “send questions to the data.”  

For another example, until recently the most extensive research to identify comorbidities for autism patients was a multi-year study on 163 patients.  Zak Kohane’s new work on autism patients and comorbidities used distributed queries across the i2b2 SHRINE network linking 4 Boston hospitals.  By tapping into a much larger patient population, his analysis provided greater insight into the comorbidities.  The time required to conduct the research was 3 weeks.   Distributed queries delivered these deeper insights  by reaching an order of magnitude more patients and did so in two orders of magnitude less time. 

Query Health, a public/private partnership project, will be launched in September to take some practical first steps towards this national priority.  The ultimate goal of the Query Health initiative will be to define the standards and services to better enable the use of this distributed approach. 

Please mark your calendars for the “Summer Concert Series” in August and the Query Health project launch on September 6th at 3 PM ET.  Visit the Query Health web site for more information on these events.   

Thursday, July 21, 2011

The July HIT Standards Committee Meeting

by John Halamka, Life as a Healthcare CIO

Farzad Mostashari, national coordinator for healthcare IT, began the meeting with a discussion of the issues we have always faced while harmonizing standards.  Standards that are widely adopted by the marketplace and are well tested make harmonization easy.  However, many standards are mature but not widely adopted or novel but not well tested.   We want to encourage innovation and use market adoption as a measure of our success.   It's clear that at times we'll have to consider new standards that seem very reasonable for the purpose intended and test them in real world scenarios before forcing top down adoption through regulation.   Bottom up adoption of standards that are implemented and improved by stakeholders is a better approach.  

Per the Standards Summer Camp schedule, the July HIT Standards Committee meeting focused on

Vocabulary recommendations 
ePrescribing of discharged medications recommendations
Patient Matching recommendations
Syndromic Surveillance recommendations

Jim Walker, chair of the Clinical Quality workgroup, presented an overview of the vocabulary work done to support all our clinical coordination and quality measurement activities.  The charter for the group was to select the minimum number of vocabulary standards with the minimum number of values to meet the requirements of meaningful use stages 2 and 3.   Reducing the number of standards makes mapping between different vocabularies much easier.    The workgroup used SNOMED-CT and LOINC wherever possible and tried to select one vocabulary per domain (allergies, labs, medications etc).   Examples of their selections include

Adverse Drug effect - RxNorm for medications, SNOMED-CT for non-medication substances, SNOMED-CT for severity of reaction

Patient characteristics - ISO 639-2 for preferred language, HL7 for administrative gender, PHIN-VADS (Centers for Disease Control) for Race/Ethnicity

Condition/Diagnosis/Problem - SNOMED-CT

Non-lab Diagnostic study - LOINC for name, SNOMED-CT for appropriate findings, UCUM for Units

A rich discussion followed.  Points of concern included:
*Using RxNorm for all medications including vaccines, even though CVX is the required vaccine vocabulary for Meaningful Use stage 1.   We clarified this with an example from Beth Israel Deaconess Medical Center:

BIDMC uses First Data Bank as the medication vocabulary for its internal systems.  However, when BIDMC sends clinical summaries, it maps FDB to RxNorm for all drug names.  When BIDMC sends immunization records to public health, it uses CVX codes.   Thus, the HIT Standards Committee will not specify the vocabularies used within enterprise applications, just those vocabularies that are needed for specific purposes when data is transmitted between entities.

Next, Doug Fridsma began a discussion of our Summer Camp items, noting the many projects of the S&I framework are proceeding according to plan.

Scott Robertson presented the work of the Discharge Medications Power Team.  They recommended HL7 and NCPDP script as reasonable standards for sending discharge medication orders to hospital pharmacies and retail pharmacies.

Discussion followed regarding two specific points - their recommendations did not include a specific version of HL7, since existing Medicare Part D regulations do not specify an HL7 version.  The power team will make additional more specific HL7 recommendations.   There was discussion about the  specific aspects of RxNorm that constrain the way dose and route are specified.   The HIT Standards Committee members felt additional work was needed before mandating this level of specificity, so our recommendations will include RxNorm for medication name, but not additional specificity for dose and route vocabularies at this time.

Next, Marc Overhage presented the recommendations of thePatient Matching Power Team.   The scope of the Patient Matching work is to provide guidance to implementers who want to understand best practices for the use of demographics in machine to machine matching of patient identity.   Per the RAND Report, use of different fields results in variation of specificity and sensitivity.   Some fields such as social security number (or a subset of it) greatly increase specificity, resulting in fewer false positives such as matching the wrong patient.   However, social security number is controversial because of the potential for identity theft and the fact that immigrants may not have one.   The final report will take into account all these observations.

Chris Chute presented the recommendations of the Surveillance Implementation Guide Power Team, which aims to specify one implementation guide for each public health transaction.  They are studying the difference between HL7 2.31 and 2.51 as well as considering the potential for public health entities to use CDA constructs.

Dixie Baker presented a project plan for the NwHIN Power Team,which aims to specify a set of building blocks for secure transport of data in multiple architectures.

Finally, Judy Murphy and Liz Johnson presented their plans for the Implementation Workgroup, collecting lessons learned from  certification and attestation.

We're on track with Summer Camp.  Our next meeting in August will include the final recommendations for

Simple Lab Results 
Transitions of Care 
CDA Cleanup
Patient Matching

Every meeting with the HIT Standards Committee (this was our 27th) brings us closer as a working team.   We're transparent and passionate, openly sharing all the issues and concerns about the standards we're selecting.  Coordination with all the moving parts (ONC, Policy Committee, S&I Framework)  keeps getting better and better.

Thus far, Summer Camp is a winner and I am confident we'll meet all our September deadlines for offering recommendations to ONC in preparation for Meaningful Use Stage 2 regulations.

Thursday, July 7, 2011

Testimony to the HIT Policy Committee

 by John Halamka, Life as  Healthcare CIO

Doug Fridsma and I were asked to brief the HIT Policy Committee about the current activities of the HIT Standards Committee, ensuring coordination as we all work to finalize Meaningful Use Stage 2.

We used this presentation, which covers three major themes:

*Meaningful Use Stage 2 gap analysis work
*The Standards Summer Camp Activities
*The Standards and Interoperability Framework activities

Doug started by reiterating the guiding principles of the HIT Standards Committee  which essentially translate into

"We will select no standard before its time"

Doug reflected on the way we select standards, assigning each requirement to one of four "buckets"

a. Functional criteria only – no standards are needed
b. Sufficient standards and implementation guides are available
c. Existing standards are available but not implementation guides
d. No standards or implementation guides are available

He discussed early work to place proposed Meaningful Use Stage 2 policy goals into these 4 buckets.

Since many of the Stage 2 goals will not have supporting standards and implementation guides in time for the regulations, it's like that some of stage 2 will be described using functional criteria and not standards.   What does that mean?   Here's an example

Electronic medication administration records (EMAR) are unlikely to require a specific bar code format.  Instead, it will be sufficient to require that a certified application be capable of 5 functions

*Generate alert for wrong patient
*Generate alert for wrong medication
*Record the dose and route
*Record the provider administering the medication
*Record the time/date the medication was administered

Implementation details will be left to the creativity of the marketplace.   

I described the Standards Summer Camp schedule and offered updates on

• Metadata
• Patient Matching
• ePrescribing
• Surveillance Implementation Guide

Doug completed the presentation by describing the Standards and Interoperability Framework Projects

*CDA consolidation
*Transitions of Care
*Lab Results reporting
*Provider directories
*Distributed query (Using a web browser to query multiple databases such as is done with Shrine/I2B2 )

We took questions from the committee including how best to develop drug/drug interaction standards, how the policy committee can best work with the standards committee by providing requirements early and often, and how we can all plan for the future of stage 3.

A great meeting!