Uncategorized

More PopHealth Lessons Learned

In the March HIT Standards Committee we highlighted 3 gaps in the standards needed to calculate quality metrics automatically from EHRs

1.  A longitudinal (not encounter level) patient summary format to transmit appropriate data elements from an EHR to a quality measurement entity

2.  A batch reporting format to transmit data elements for multiple patients to a quality measurement entity

3.  Although PQRI XML and QRDA have been suggested for reporting data between quality measurement entities and organizations that use this data for payment/compliance, there is not a widely adopted standard for quality reporting in production today.

As I wrote in a wrote in a previous post, ONC/MITRE/BIDMC/Massachusetts eHealth collaborative  worked together to evaluate the PopHealth tool with 2 million Continuity of Care Documents.

The full results of that analysis are now available and here’s the document for public circulation.

Key lessons learned include:

1.  The CCD is a  “post-encounter message” not a lifetime clinical summary optimized for quality measurement

The CCD/C32 was designed as an encounter level summary from a single organization.    Each patient will have multiple CCD/C32s but there are no well defined process for merging CCDs from multiple institutions and applications.   popHealth was expecting each C32 to contain the complete clinical history for one patient since quality measures are often focused on longitudinal treatment of patient, not a single encounter.

2.  Meaningful Use Stage 1 permitted multiple vocabularies (or did not specify a vocabulary) resulting in optionality/variability in CCDs

BIDMC’s CCD uses SNOMED for diagnosis, CPT for procedure coding and LOINC for vital signs.   popHealth expected SNOMED for procedure coding and vital signs.

Meaningful Use Stage 2 will correct this problem by specifying one vocabulary without optionality for each portion of the record

3.  Quality numerators and denominators are imprecisely defined

Since quality measures are not defined in precise “SQL” or e-measure form, humans have to read the text of the measure and decide how to implement it.  For example, does less than or equal to 84 years old refer to 84 years and 0 days verses 84 years and 364 days

Thus, to accelerate quality measurement in the US we should:

1.  Chose a clinical summary standard for transmission of longitudinal patient care data to quality measurement entities

2.  Specify a one vocabulary without optionality for each portion of the record

3.  Use e-measurements format to describe numerators and denominators in machine readable logic

If we do this, we’ll be able to widely deploy popHealth to automatically calculate quality measures on data exchange from EHRS.     We’ll also be able to more effectively use architectures like QueryHealth that submit questions to the data rather than aggregate the data into a central quality measurement entity.

The Standards Committee is already hard at work on all these standards.

John D. Halamka, MD, MS, is Chief Information Officer of Beth Israel Deaconess Medical Center, Chief Information Officer at Harvard Medical School, Chairman of the New England Healthcare Exchange Network (NEHEN), Co-Chair of the HIT Standards Committee, a full Professor at Harvard Medical School, and a practicing Emergency Physician. He’s also the author of the popular Life as a Healthcare CIO blog.