Difference between revisions of "January 4, 2011 CBCC Conference Call"
|Line 133:||Line 133:|
Rob: '''QDS spreadsheet'''
Rob: '''QDS spreadsheet'''
D = standard_concept
D = standard_concept
E = standard_category
F = QDS_data_type
E = standard_category
H = standard_taxonomy
F = QDS_data_type
H = standard_taxonomy
Revision as of 23:47, 10 January 2011
Community-Based Collaborative Care Working Group Meeting
- Jon Farmer
- Suzanne Gonzales-Webb CBCC Co-chair
- Michelle Johnston
- Mary Ann Juurlink scribe
- Rob McClure
- Richard Thoreson CBCC Co-chair
- Ioana Singureanu
- Cliff Thompson
- Serafina Versaggi
- (05 min) Roll call, approve minutes December 14th, call for additional agenda items & accept agenda
- (05 min) Action Item update
- (05 min) CDA R2 Implementation Guide for Consent Directives update
- (45 min) High level overview of SHIPPS timelines and approach
1. Action Items
Action: Regarding SHIPPS outreach - Serafina to report back on MITRE Corporation
Action: Regarding SHIPPS outreach - Richard and Serafina to present to EHR agenda working group
Action: CBCC to discuss format of SHIPPS publication prior by February 2011
Action: In order to send PPS to Steering Committee; approve PSS as complete e.g. accept all revision markers. Outreach still ongoing
Action: Touch base with Ken Salyards (Floyd) requesting copies of QDS for NQF measures (retooled or otherwise)
Action: The need to make the process, value add explicit (Mary Ann)
Serafina made a motion to approve minutes from Decemeber 14, 2010, which was seconded by Ioana
Informal meeting January 11, 2011 as WGM in Australia next week
The document DMAG_Contribution_to_HL7_S&P_Ontology to be discussed in Security prior to CBCC. Possible meeting date 20110118. Contribution to the Security and Privacy Ontology background information about research group (DMAG) at the Universitat Politècnica de Catalunya (UPC) in Barcelona (Spain)
CDA R2 Implementation Guide for Consent Directives
All necessary votes completed e.g. ballot reconciliation done completed. The CDA R2 Implementation Guide for Consent Directives will be == posted to the CBCC, Security and perhaps the Structured Documents work lists.
SHIPPS Project Scope Statement (PSS)
- Submission of PSS to Steering Committee this week
- Mary Ann to forward PSS to Austin and cc coChairs for Steering Committee approval
- Submission will be in queue for TSC/PMO approval
- PSS already in Project Insight
- If questions from Steering Committee CBCC will need to be responsive
- Publishing of the Draft for Comment submission will need to be decided (e.g. pdf vs doc format or use of section headers, sections numbers, and line numbers).
Action: CBCC to discuss format of SHIPPS publication by February 2011
Interested Parties: Security WG, NQF Walter Suarez (KP hat), MITRE Corporation (in progress) Co Sponsors: Structured Document WG, EHR working group (in progress)
Outreach to interested parties in Behavioral Health
- In scope: EHR and the subgroup Records Management and Evidentiary Support (RMES) e.g. tremendous overlap
- Out of scope: PHER and RCRIM at this point in time
Action: Richard and Serafina to present to EHR agenda working group Action: Serafina to approach MITRE Corporation
Review of documents NQF measures
We looked at the descriptions of the NQF measure, the numerator, and denominator in the measure in order to determine the information objects, attributes and associations. We need to approach Floyd Eisenberg to see if there are the quality data sets for all the concepts that are encoded. We are relying on the NQF version of these measures which originate from various other organizations. We looked at the types of information that the measure is focused on.
The EHR system maintains encoded information and from all this information within the EHR we would be interested in several different types of objects that would be required to compute a specific measure. We tried to break down what those types of information would be and what specific attributes would be relevant. Deferring the value set discussion until we identify whether the QDS is set up for these measures.
Issue: How do we deal with an NQF having a CPT code as well as non-coded concepts within the measure?
QDS addresses the reference to existing standards they’ve included the HL7 v2 item ID for all the non-coded data elements and they provide 2 or 3 different options for a value set for all the coded data.
The purpose of the analysis is to give the implementer, of a system that would compute these quality measures automatically, an idea of what information they would have to extract from the EHRS. Not all the information about a patient is relevant, but the fact that they had a visit, at least one visit e.g. 1..* visits during the measure period is important and any patient who did not have a visit is not part of this population – we are not interested in them. They have to have a DOB as we need to know if they are adults.
We further elaborate what the visit information and DOB information includes. We know that diagnosis is assigned during the visit so we further identify the time of the visit e.g. during the reporting year. The NQF measures often time will provide explicit list of all types of encounters.
NQF enumerates encounter types. We include this in our analysis because we know that they expect encounter types to be enumerated. The important thing is that we identify during that year a diagnosis, if this is a reoccurring episode of MDD then there would be an optional date of onset associated with that diagnosis. We also expect as part of the visit the patient was administered an assessment. The score for this assessment is not important but the fact that the person was administered an assessment.
The list of measures that SAMHSA requested be included in the analysis; NQF 004, NQF 105, NQF 103, NQF 104, NQF 112 and PHQ 9. The two smoking cessation codes could also be included
Rob: If the NQF has been retooled a lot of issues have been worked through that with the right guidance will answer the questions you are trying to deal with. Ask NQF if measure has been retooled – if so they can provide definitive information.
If the measure was not retooled the approach is still to find the right resource to provide guidance. You would go through the measure that was retooled and get a sense of what they did by looking at the content and elements they identified for the retooled measures, which would give you an idea of what they learned, which can be applied to other measures.
Ioana: If the NQF measure is not retooled, the QDS even with incomplete value set should be better than the description of the numerator and denominator. If we have the old value set those would be useful in determining what the data elements they are looking for and then we can assemble them.
Rob: What diagnostic evaluation was done in order to diagnosis the patient with MDD and when it was done? The CPT code says I did an assessment, for MDD and it was documented, now what the specific thing that led to diagnosis in order to get MDD you have to have 5 out of the 9 symptoms that occur over a particular time. The fact that you have captured that is what you get paid for when you put use a CPT code.
Ioana: There is an assessment that evaluated the 9 specific symptoms this is the basis for the MDD diagnosis
Rob: In order to qualify as an MDD you have to have 5 or more of the symptoms for 2 weeks – it may not be the only way of documenting that they’ve done work and therefore they should be paid for the diagnosing the patient. The measure is attempting to assess and this is the problem. If all you have to look for is that the CPT code exists this may be a problem as it may not be the only way the clinician will document his work and therefore should be paid for the diagnosis.
Look at the results of the retooled measures.
Action item: touch base with Floyd requesting which one of NQF on our list have been retooled
Rob: Many of the quality measures are moderately good at charge capturing and not the documenting. The measures look for the billing code and for something else and it probably means you have to do some of the detail analysis and if you look at some of the retooling logic you will see the kind of complexity that tends to be expected and that shows the kind of elements needed for the tooling to work, which is what you are looking to document. There are probably not too many that are different, the value sets may be different but the kind of elements are probably covered in the retooled measures.
Richard: To what extent are we adding value to what NQF is doing?
Ioana: NQF provides a QDS spreadsheet for each measure. What we are doing is looking at common concepts e.g. Alcohol and drug dependent (X 4 in spreadsheet), which has many vocabulary bindings e.g. ICD-9-CM, ICD-10-CM and SNOMED-CT. That’s how they describe there is a problem of alcohol dependent as an active diagnosis. There is a diagnosis or problem which has 3 different terminology bindings with specific value sets for each coding system listed. What we are saying there is a single concept called diagnosis that is relevant for the measure and for different situations e.g. if you are using clinical information you would favor that binding for the SNOMED-CT value sets. But there is a single concept of diagnosis. 20 rows in this measure that contain the combination of individual attributes, group value sets and what we are doing is to combine them to create a very simple model across the board to look at what all these measures are trying to look at. If this information / measures were being computed in real time e.g. every night the data is pulled from the EHR and this measure is computed. So for this measure period every 3 months you could find out the percentage of people receiving the assessment whenever they are getting an MDD diagnosis. You would do this by extracting the data from your EHR system and look for the specific attributes to see if you have a diagnosis of MDD. So if all these criteria occur you found your population, the number of people who meet these criteria. Then you find all the patients who have MDD and have been administered the assessment and you compute the percentage.
Ioana: Our contribution is a very simple analysis model shows you for all these patient centric measures you need to know some things about the patient , very few things, you need to know the patients have had visit(s) and out of these people those who have specific diagnosis that receive specific services are of interest.
For all the measure you may have 5 or 6 different classes that are related to each other and for different measures you have different value sets that are applicable – e.g. when you are computing a measure that has to do with diabetes you plug in different value sets for diabetes
This is the data you need to extract from your EHR to compute these measures real time.
Richard: More comprehensible way of how to draw out information from the EHR
Rob: QDS spreadsheet
- D = standard_concept
- E = standard_category
- F = QDS_data_type
- H = standard_taxonomy
- J= standard_code_list
I expect you will focus on columns D, E, and F because D is a combination of understanding E, F and applies to the specific values that are in J the Standard code list. So you would see the name alcohol or drug dependence as what’s the desired piece of information’. The desired piece of information to look at the grouping ICD-10.CM or SNOMED and that would be found as a type of standard category and diagnosis that is active that binds the patient that have alcohol dependence. While it is of value to look at H,J … it is D, F that is what you are building your classes on and by linking this to HL7 is of value.
Ioana: Also linking concept to each other e.g. in the spreadsheet e.g. the diagnosis has to occur in the context of the encounter. Having that clear which is an element of logic that retooling does is not clear on a spreadsheet. So what we are discovering is the relationship of objects that is buried in the numerator and denominator calculations and we want to make them overt.
When people extract information …let’s say they don’t show a relationship between the diagnosis and encounter they have a problem as they will not be able to compute it. If we show them the relationship between objects they can compute it, which can then be done automatically.
Rob: Looking at the NQF process is not enough you also have to look at the retooling in order to understand this and be clear about the data model you see in the spreadsheet
Ioana: We can also us the approach of documenting the problem and issues and ask specific questions. If they have any non-retooled QDF spreadsheets (e.g. older versions) for this data we should be ok. The point of this exercise is that people will extract data from their EHR system; it’s not going to look like a set of spreadsheets. We will make it more visible for people doing extracts from their database.
NQF will respond to CMS pressure, Ken could help with this. Floyd will probably not give us retooled measures as they are owned by CMS.
Meeting was adjourned at 3:05 PM Eastern