Query Health Concept-to-Codes (C2C) SWG

Meeting #10

February 14, 2012

1

Today’s Agenda

Topic Time Allotted

• Presentation by Dr. Peter Hendler on Convergent Medical Terminology (Kaiser Permanente) 2:30 - 3:15 pm

• Quick Review of the Current and Future Tasks• Quick Review of Constraints and Criteria• Review of Summaries and Extracting Key Themes

3:15 – 4:00 pm

• Note - Next week’s meeting will be cancelled due to HIMSS Conference

2

Kaiser’s CMT and SNOMEDPeter Hendler MD

and CMT team Kaiser Permanente

How Kaiser’s Convergent Medical Terminology (CMT) is used to manage standardization of clinical information

How Kaiser leverages Systematized Nomenclature of Medicine (SNOMED)

How use of SNOMED Description Logic (DL) allows automated evaluation of clinical information

3

Kaiser donates CMT to IHTSDO and to NLM

Kaiser has agreed to donate CMT to the IHTSDO organization as well as NLM in the USA

The donation will include Kaiser mappings between various terminologies

It will include our modifications to the IHTSDO work bench

This will greatly reduce the burden of adding SNOMED to an enterprise EHR

4

Domains covered by CMT (1)

• Diagnosis–Support Problem List, Encounter Diagnosis, and Medical History–Mapped to SNOMED and ICD9, and in the process of mapping to ICD10• Procedures (Medical, lab, radiology, etc.)–Supports Order Entry–Supports documentation of Performed Procedures–Mapped to SNOMED and CPT4 or HCPCS • Lab Results–Supports Results Review–Mapped to LOINC• Reasons for visits/Chief Complaints–Supports scheduling/intake–Mapped to SNOMED• Immunizations–Supports Immunization/Vaccine Administration–Mapped to CVX codes (CDC’s Vaccine codes)

5

Domains covered by CMT (2)

• Organisms–Supports Results Review–Mapped to SNOMED codes• Clinical Observations–Supports clinical assessment documentation–Mapped to SNOMED and Clinical LOINC (in process)• Nursing Documentation–Supports nursing documentation of nursing interventions and nursing assessments–Mapped to SNOMED and Clinical LOINC (in process)• Specimen Type and Body Site–Supports Order Entry for cultures–Mapped to SNOMED• Allergens–Supports Allergy entry–Mapped to Medispan, FDB, and RxNorm (in process)

•

6

Why “Enterprise Terminology”?

• There is no one terminology that meets all needs.• Standard terminologies were created to meet certain specific needs.• We have a need use all or most of the different terminologies.• We have a need to integrate or “converge” these disparate terminologies in a

central model that leverages the efficiencies of each of these terminologies, provides interoperability, and meets other business requirements.

• None of the standard terminologies are ever “complete”. Therefore, there is a need to create new concepts or enterprise specific concepts.

• Turnaround time from reference terminologies is slow (i.e. SNOMED is 2x/year).

• Having a central terminology system eliminates duplication of effort and provides a “common definition” of concepts/terms across the enterprise.

•

7

What is Kaiser’s CMT?

• CMT is KP’s Enterprise Terminology System that includes several components:

–End user terminology–Standard terminology–Administrative codes–Query and Decision support–Request process

8

Kaiser CMT

You use at least 4 different kinds of vocabularies to run a healthcare enterprise

Physician interface (locally created)

Billing (CPT and ICD)

Patient facing (locally created)

Reference (SNOMED LOINC others)

9

Example of mapping of 4 terms

• Amaurosis fugax <Physician interface term>

• Amaurosis fugax (disorder) <SNOMED FSN>

• AMAUROSIS FUGAX (ONE SIDED TEMPORARY VISION LOSS) <Patient Display Name>

• Transient arterial occlusion <ICD9 Name>

10

End User Terminology component (1):

• Terminology used in KPHC (KP EMR)• End User terms are mapped to the standard

terminologies and have attributes the application needs

• End Users use/see the terms that are familiar to them, and the application uses the codes and attributes it needs

11

End User Terminology component (2):

• Protects end users from changes in Standard terminology or coding schemes. Examples:• Majority of the diagnosis terms currently being used does not have

to be changed or deleted because of transition from ICD9 to ICD10.

• When LOINC names change, it doesn’t have to change the Result display name that end users are seeing.

• When SNOMED descriptions change, it does not change the display names end users see.

• Neither ICD9 or SNOMED’s focus is end user usability. KP CMT is a bi-product of years of actual user experience, and continues to improve.

12

Standard Terminology Component:

• CMT is mapped/integrated to Standard Terminology, such as SNOMED and LOINC.

• CMT can be mapped to other terminology as needed.

• Supports requirements for standard terminology for Meaningful Use and Health Information Exchange.

13

Administrative Code Component:

• CMT supports Revenue Cycle and Charge Capture.• Diagnosis terms are mapped to ICD9 (and ICD10 in

process). Clinicians can pick a problem list and can use the same term as encounter diagnosis.

• Procedure terms are mapped to CPT4 or HCPCS codes. When a lab order is resulted/completed, CPT4 codes mapped to it can be sent to the Billing System.

14

A view of a single concept:

15

ICD9

Course grainedDiagnosis and FindingsReasons for visits or proceduresNot a DLSingle Hierarchy only (Diabetic Retinopathy is either a Diabetes or a Retinopathy, can’t be both)Difficult to find patient cohorts (must be lexical or single Hierarchy searches)

16

CPT

Only for ProceduresRequired for BillingNot DL basedOnly Name/Code pairsOnly concerned with “how much work was done? How much time was spent?”

17

SNOMED

SNOMEDSystematized Nomenclature Of MEDicineCan describe anythingDiagnosisProcedureEntityMorphologyEtiology (To name just a few)

18

SNOMED

SNOMEDBased on Formal Description LogicConcerned with clinical meaning, not billingFine grained enough to be clinically meaningfulCan be used for Outcomes measurementsCan be used by machines to make inferences

19

Inferences possible with SNOMED

Strep throat is caused by streptococcusPneumococcal pneumonia is caused by pneumococcusStrepococcus and pneumococcus are both sub types of gram positive cocciTherefore both pneumococcal pneumonia and strep throat are gram possitive cocci infections.

20

Query and Decision Support Component (1):

• CMT is based on SNOMED.• Can leverage SNOMED’s structure, including poly-hierarchy and

description logic (formal definitional attributes).• Can query different ways to identify subsets terminology:

• for supporting decision support modules in KPHC

• for identifying patient cohorts for Population Care

• for identifying KPHC terminology for reporting criteria, etc.

21

Query and Decision Support Component (3):

• Easily identify patient cohorts for certain conditions for Population Care.

• Identify subsets for use as “input criteria” for KPHC decision support modules, such as Best Practice Alerts, Reminders, etc.

• Do precise queries, such as “find all conditions where causative organism is “Aspergillus”, etc.

• Do large aggregate queries, such as “find all patients with cardiovascular system disorders”, etc.

22

Query and Decision Support Component (2):

•Example of SNOMED Description Logic:

Concept: Anthrax pneumonia

23

24

25

26

27

28

Web Ontology Language (Coming Soon)

OWL Has a large community independent of Health Care Domain

The Description Logic of OWL is more expressive than EL+ (SNOMED)

Allows for Negation and Disjunction

All Non Pulmonary conditons caused by Mycobacteria (example)

29

SNOMED to OWL Project

We will modify the IHTSDO toolbench so that it can use OWL

Will add Negation and Disjunction

Will have a new reasoner that can classify all of SNOMED/OWL in 2 minutes

30

In My Mind it's all Triplets

There are probably a dozen ways to think about this but in my mind it all boils down to triplets

I'll explain later as we go along

31

How Kaiser uses the IHTSDO Toolbench

We call our modification of the toolbench the Enterprise Terminology Tool or ETT

In order to talk about and understand what it does, it is very helpful to think in terms of Triplets

32

Triplets

Triplets basically say “this is related to that”. Everything that the ETT does can be explained

in terms of triplets The “this” part can be either a whole concept or

a code or a description The That part can also be any of these three

things

33

Triplets

The relationships can be these kinds Defining relationships Non defining relationships Core model relationships Relationships represented only in Refsets

34

ETT and triplets

When you think in terms of triplets, you only have to think to yourself, what is “this” what is “that” and how are they related.

In technical terms “this” is called the “subject”, “that” is called the “object” and the relationship is called the “predicate”

35

ETT and Triplets

So you can think “this is related to that” and you can say “subject predicate object” if you talk to a semantic web person

For one defining triplet example consider “Pneumonia has finding site Lung”

36

ETT and Triplets

This defining relationship or triplet would be modeled as part of the core model.

A non defining triplet might be this. “CHF <is the KP physician friendly name for>

Congestive heart failure (disorder)” Or “CHF <is part of> the NLM cardiovascular

disorders subset”

37

Kaiser ETT

The ETT has two ways of modeling triplets You can add relationships (triplets) to the core

model You can add relationships (triplets) using the

“Refset” mechanism.

38

Kaiser ETT Refsets

Kaiser adds descriptions, codes and defining relationships to the core model

We use Refsets for 3 different non core types of relationships (triplets.

They are:

39

Refsets used for three things

Generating the Epic specific load files For example, specifying that CHF is the “dot

2” for the Epic EDG master file.

Specifying sub sets of concepts like specialty diagnosis groups, or NLM donations groups

Linking our concepts to administrative billing codes like ICD9 and 10 and CPT

40

Load up toolbench Screen Shots?

41

KP ETT Screen Shots

IHTSDO Toolbench modified for KP CMT

42

43

44

45

46

47

Timeline

TODAY

Coordinate offline activities to summarize approaches and develop draft deliverable from presentations

Meeting 1 – Dec 6

Meeting 2 – Dec 13

Meeting 3 – Dec 20

Meeting 4 – Jan 03

Meeting 5 – Jan 10

Meeting 6 – Jan 17

Meeting 7 –Jan 24

Meeting 8 – Jan 31

Meeting 9 –Feb 7

Meeting 10 – Feb 14

Meeting 11 – Feb 28

Presentation •NQF –Value Set presentation

•Begin Overview of Next steps

•Overview of Constrains and Criteria

Presentation•I2b2 (Cont.)•Intermountain Health•DOQS (Data Warehousing / Mapping)

Tasks•Introductions •Scope•Proposed Approach•Identify SME and presentation timeline for next few meetings

Meeting times extended from 2:30-4:00pm

Presentation•hQuery •i2b2

Presentation•DOQS (Data Warehousing / Mapping) Cont.•PopMedNet•NLM

Presentation•Ibeza•CDISC SHARE

Presentation•Convergent Medical Terminology (CMT) Kaiser

•Discussion of presentation summaries and extraction of key themes

Presentation•RELMA (LOINC)•3M•NY Presbyterian Hospital Vocab Team

Presentation•AHIMA•LexEVS and CTS2•Jacob Reider - ONC

Presentation•S&I Repository

48

Tasks •Finalize discussion of presentation summaries and extraction of key themes

Next Steps*

Current Focus and Next Steps

Current SWG Focus

Standards

Tools

Distributed Query Networks

C2C Output Summary of Various Approaches taken by Organizations Identification of Key Themes and Industry Best Practices

List of Constraints to analyze Best Practices for QH within the Technical framework

Identify and assign Value Sets for a core set of data elements within the Harmonized QH CEDD as part of the Cross Walk

Align Proposed Technical Expression with Existing Value Set(s) and Vocabulary Task Force Recommendations

Identify standardized approach to store and access Value Set(s)

Dec

201

1– F

ebru

ary

2012

Feb

2012

- TB

D

Conduct Environmental Scan

Develop Technical Expression of C2C Technical Expression of C2C Approach as it Aligns with the Reference Implementation

Identified Value Set Representations for core set of Data Element in the CEDD

Selection of Existing Value Set in Alignment with the QH CEDD

Reference implementation Guidance for QH

Harmonized CEDD and Selected Value Set

Technical Expression of C2C Approach

Value Set Representation

C2C Output / Recommendation

Suggested Inputs Suggested Outputs

Clinical CEDD

Technical

Technical

C2C / Technical

Task Team

49

Value Sets

* Steps are not in sequential order

Criteria and Constraints

1. The approach must be easily implemented as part of the Technical Framework. – The Reference Implementation has to easily be able to use the mappings and the

Value Sets .

2. Utilize NQF as starter Value Sets per Jacob Reider’s recommendation– Additional Value Sets can be identified and included as needed

3. Value set representation should utilize NQF and the IHE SVS– Integrating the Healthcare Enterprise -Sharing Value Sets (IHE SVS) Profile can be

thought of as a Value Set Repository that houses Value Sets– IHE SVS provides a standardized, easy to use, RESTful interface to the value set

4. Potential mechanisms to import Value Sets as part of the Reference Implementation should be identified (ex. - Excel or another format?)

5. Each participating organization within the query network should consider operational best practices for ongoing updates and maintenance of Value Set

6. Value Set Owners are expected to perform ongoing maintenance of Value Sets

50

DISCUSSION OF KEY THEMES

51

Initial Draft of Key Themes

52

1. Mappings should be Purpose and Goal Specific

2. Hierarchy within Mappings

3. Centralized Data Dictionary or Central Terminology System

4. Challenges identified around maintenance of mappings as Standards change or are modified

5. Identifying Best Match and Alternate/Default Mappings

6. Importance of Context of the Queries

7. Resource Intensive - Dedicated and skilled resources (Clinicians/ Informaticists) is crucial for ongoing maintenance and updating of mappings

8. Open Source –Tools are publicly available or accessible for use

9. Maintain data in its original form

Key Themes Categorization By Presentation (DRAFT)

53

hQuery i2B2 PMN

Intermountain DOQS

UMLS / RxNorm Ibeza

CDISC Share 3M

NYP Terminology Service

RELMA

AHIMA

LEXEVS CTS2

S&I Repository NQF

1Mappings should be Purpose and Goal Specific x x x x

2Hierarchy within Mappings x x x x

3

Centralized Data Dictionary or Central Terminology System

Utilize NCBO UMLs HDD MED

4

Challenges identified around maintenance of mappings as Standards change or are modified*

x x x x x x x x x x x x x x x x

5

Identifying Best Match and Alternate/Default Mappings

x x x

6 Importance of Context of the Queries x x x x

7

Resource Intensive - Dedicated and skilled resources (clinicians/informaticists) is crucial for ongoing maintenance and updating of mappings

x x x x x x x x x x x x

8

Open Source – Tools are publicly available or accessed for use

x x x

9Maintain data in its original form x x x x x x x x

Notes

* - Assumption is that all would be impacted.