While in Chicago last Thursday, I was asked how we validated our quality measures when we moved from chart abstraction to automated computation of PRQS, Meaningful Use, Pioneer ACO, and Alternative Quality Contract measures via the Massachusetts eHealth Collaborative Quality Data Center (QDC). This is an important question because Meaningful Use Stage 2 enables easy use of modular components outside the EHR such that data can be captured in the EHR and sent to a cloud based analytics engine via standards such as CCD/C32 for content and Direct for transport.
Initially we did spot checks to validate the integrity of the Continuity of Care Document data flows from electronic health records to the normalized QDC schema.
When Mitre Corporation offered to test their popHealth tool against 2 million BIDMC patient records to validate the Meaningful Use quality measures computed by our QDC, we jumped at the opportunity.
First, we ensured appropriate business associate agreements were in place to protect the privacy of patient data. Next, we required all work to be done on site in the Quality Data Center to protect the security and integrity of clinical summary data.
Mitre ran the tool against 2 million BIDMC Continuity of Care Documents and compared the results to the reports generated by the QDC.
The results were enlightening.
The computations aligned well for most quality measures, justifying our early manual validation.
However, Mitre discovered ambiguities in the CCD specification itself that led to some differences in the calculations. This was despite our use of this CCD implementation guide which provides even greater specificity than the HL7 standard.
For example, the CCD does not specify an allergy vocabulary. At BIDMC we use First Data Bank to codify medication allergies. PopHealth expects RxNorm, the vocabulary standard required for exchanging medication history. Even the Stage 2 NPRM does not specify an allergy vocabulary and we recognized the need to enhance the Stage 2 to include RxNorm for medication allergies (Penicillin VK), NDF-RT for categories of medication allergies (all Penicillins and Cephalosporins) and SNOMED-CT for non-medication allergies (food and environmental agents).
I'll post other pertinent findings from the Mitre analysis after next week's debrief meeting.
Mitre demonstrated their work at HIMSS in the interoperability showcase as illustrated in the photograph above.
BIDMC and MAeHC were proud to participate in this event, which we hope provided lessons learned for other provider, payer, and government stakeholders wanting to compute quality measures in the cloud using popHealth.
John, thank you! BTW, while it's true that Stage 2 NPRM does not mention allergy vocabularies, it does specify Consolidated CDA, which contains allergy vocabularies within it. Within the C-CDA allergy observation (section 5.4) these vocabularies are specified: RxNorm for drugs, NDF-RT for drug classes, and UNII for non-med allergies*** (Tables 118-120). ***HOWEVER, UNII for non-med allergies is carried over from HITSP but not in agreement with the HIT SC vocab task force recommendation to use SNOMED-CT for non-meds. So that's a discrepancy where the HIT SC may want to comment to ONC.
ReplyDeleteThanks again,
David
John what tool or method are you using to convert your FDB allergy codes to the standards that support the CDA?
ReplyDeleteFDB offers an RxNorm/GSN mapping tool. We use that.
ReplyDeleteThanks John, any more clarification on the non-drug allergies and the allergy categories
ReplyDelete