Evolution of the Knowledge Management Program at Partners HealthCare - - PowerPoint PPT Presentation

Evolution of the Knowledge Management Program at Partners HealthCare (PHS)

Roberto A. Rocha, MD, PhD, FACMI

Clinical Informa9cs Director, Partners eCare, Partners Healthcare System Assistant Professor of Medicine Division of General Internal Medicine and Primary Care, Department of Medicine, Brigham and Women’s Hospital, Harvard Medical School Learning Health System (LHS) Seminar

• October 7, 2016 -

Overview

• History

– Goals, types of assets, ini9al challenges, tac9cs

• Current state

– Program, principles, types of assets – New challenges, revised tac9cs

• Implementa9on framework

– Levels of support, areas of focus

• Future direc9ons
• Conclusions

HISTORY

Knowledge Management

• Established in 2003 as a formal func9on within

the Clinical Informa9cs R&D group – clinical team within Informa9on Systems

• Responsibility for designing, developing, and

suppor9ng processes, tools, and assets – e.g. governance, lifecycle, authoring, deployment

• Enterprise content areas – e.g. terminologies,

catalogues, rules, reference sources, etc.

Original Strategic Goals

• Reduce the cost and increase the speed of knowledge acquisi9on and

maintenance for decision support

• Speed transla2on of clinical innova9on and evidence into clinical prac9ce
• Proac2ve, an2cipatory decision support architecture to set founda9on for

personalized medicine – avoid “interrup9ve” decision support

• Improve Partners’ organiza2onal effec2veness as a learning organiza9on

through organiza9onal alignment and data-driven performance improvement

• Align knowledge assets with business, regulatory, safety and quality

requirements

• Only build what we cannot buy
• Partners has created some of the best decision support in produc9on in

the world, the goal here is to keep the knowledge up to date

Slide created by Tonya Hongsermeier, MD, MBA (Mar 2006)

Content Examples at PHS

• Medica2on Data Dic9onary with default doses, weight-based doses, dose

strings, and drug-drug interac9on checking – mul9ple applica9ons and popula9ons via Common Medica9on Services

• Gerios and Nephros for proac9ve dosing for elderly and/or renal

insufficient

• Drug-lab monitoring, duplicate drug checking, drug-group checking,

drug-disease checking, drug-pregnancy checking

• Primary and secondary preven9ve health reminders
• Results Manager (abnormal test results)
• Inpa9ent and outpa9ent order sets
• Inpa9ent interac2ve rules (applica9on specific, hard-coded)
• Concept dic2onary and problem list
• Pa9ent monographs
• Radiology ordering decision support
• Outpa9ent documenta2on templates

Slide created by Tonya Hongsermeier, MD, MBA (Mar 2006)

Content Life-Cycle Challenges

Slide created by Tonya Hongsermeier, MD, MBA (Mar 2006)

Clinical Content Committee

Slide created by Tonya Hongsermeier, MD, MBA (Mar 2006)

Deployed Tactics (2008)

• Transparency and Governance (2004)

– Document library portal of decision support knowledge specifica9ons in produc9on

§ 600 documents represen9ng 10s of 1000s of rules

– New governance and subject ma`er expert panels

• Collabora2on and Content Life-cycle Tools (2005)

– Collabora9on Portals aligned with governance and business goals of Partners

§ 60 spaces to date, Documentum’s eRoom

– Content Management infrastructure for sharing, versioning, audi9ng, scheduling, tracking

§ Documentum’s Content Management Services

• Content Edi2ng Re-architecture (2006)

– Once decision is made for knowledge to change, change will be implemented rapidly at all touch-points – Primary focus of KM development going forward

Slide created by Tonya Hongsermeier, MD, MBA (2008)

List Ingredient(s)? Get Request Research Med (if applicable) Choose Ingredient/ Ingredient Set Select Type

• f Entry

Commercially Available (default) Select Route Grouping Assign Allowable Route(s) Preview GCNSEQNO return list( (optional) Select or free text Strength(s) (if applicable) Select or free text Dose Form(s) (if applicable) Select or free text Indication(s) (if applicable) Assign/create Internal Name Assign/create External Name “Empty” list of Brand names ID Drug / Compound Choose Ingredients) Yes No Yes View list of Brand names Create new Synonym(s) Select none, 1, or more; or create new Synonym(s) Assign other defining characteristics? No ID Drug / Compound Assess DDIs Similar to pre- existing concept? Labs data applicable? Incorporate into DDI KB Yes Flag for DDI review No Yes DDIs applicable? ID Drug with Ingredient(s)? Compond with Ingredient(s)? ID Drug without Ingredient(s)? KM Comments needed? No systemic absorption Commercially Available Drug Enter Labs data Yes Enter Comments Yes No Select Population Accept or

• verride

Frequency list No Accept or

• verride

Dose+Unit(s) list Chemo drug? Dispensing Advice needed? Prescribing Advice needed? Enter text Yes Enter text Yes No No Enter Rounding Increment Enter Max Dose (Daily) Select Max Dose (Daily) calculation Yes Enter Max Dose (Weekly) Select Max Dose (Weeky) calculation Enter Max Dose (Lifetime) Select Max Dose (Lifetime) calculation Administration Advice needed? Patient Advice needed? Enter text Yes Enter text Yes No No No Review Summary/ Preview screen SAVE as Enterprise Version Input site- specific data Select Rounding Increment units Other Population(s)? Yes

ENTERPRISE

No FDIs applicable? Assess FDIs Flag for FDI review Similar to pre- existing concept

• r FDB has data?

Yes Nephros applicable? Assess Nephros data Flag for Nephros review Similar to pre- existing concept

• r FDB has data?

Yes Gerios applicable? Assess Gerios data Flag for Gerios review Similar to pre- existing concept

• r FDB has data?

Yes No No Linked to “external” editors Create Generic Name Use as chemo? Yes No

Medication Dictionary: KM Workflow

KM software tools

10-15 loosely connected tools used just to maintain Meds

CURRENT STATE

CKM Program

• Systema2c and sustainable acquisi9on, adapta9on

(localiza9on), and management of knowledge assets for different “modali9es” of CDS

• Includes the adapta2on of “reference” knowledge to

reflect local and ins9tu9onal requirements, resources, and priori9es

• Follows a well-defined lifecycle, including specific

stages for documenta9on, tes9ng, and monitoring – supported by integrated set of tools and resources

Rocha RA, Maviglia SM, Sordo M, Rocha BH. Clinical knowledge management program. In: Greenes RA, editor. Clinical Decision Support - The Road to Broad Adop9on (Second Edi9on). Burlington: Academic Press; 2014. p. 773-817

Program guiding principles @ PHS

• Objec9vely improves safety, quality, and efficiency
• Supported by evidence, clinical best prac2ces, and

sound clinical thinking

• Aligns with and promotes clinical and business goals
• Acceptable to prac9cing end users (workflow

integra9on)

• Adheres to informa9cs and knowledge management

best prac2ces

• Best u9lizes talent, resources, and capital of Partners
• Supports research and teaching missions of Partners

Inventory of Knowledge Assets Managed Centrally at Partners (1/2)

Rocha RA, Maviglia SM, Sordo M, Rocha BH. Clinical knowledge management program. In: Greenes RA, editor. Clinical Decision Support - The Road to Broad Adop9on (Second Edi9on). Burlington: Academic Press; 2014. p. 773-817

Inventory of Knowledge Assets Managed Centrally at Partners (2/2)

Rocha RA, Maviglia SM, Sordo M, Rocha BH. Clinical knowledge management program. In: Greenes RA, editor. Clinical Decision Support - The Road to Broad Adop9on (Second Edi9on). Burlington: Academic Press; 2014. p. 773-817

CKM Lifecycle @ PHS

Request (new or update) Authorize & Priori9ze Design Implement Test & Deploy Monitor Evaluate

Ongoing activities @ PHS

• Transi2on!

– Partners eCare: implementa9on of Epic

§ System is live at BWH, MGH, NWH, MEEI, and PCPO

– Migra9on (and preserva9on) of legacy assets – Evolving understanding of what Epic can/cannot do

• Implemen9ng analy9cs plakorm for Clinical KM

– Monitoring and evalua9on of CDS interven9ons – Op9miza9on of KM ac9vi9es

• Completed new KM solware plakorm (CKMS)

– Repository + Portal + Authoring + SME Collabora9on – System live since February 2015

Partners eCare

Program milestones @ PHS

ü Establish governance structure with clear guiding principles ü Define priori2es considering ongoing programs & ini9a9ves ü Catalog features & content available in legacy systems ü Assimilate features & content available in new EHR system ü Resolve or mi9gate iden9fied gaps (features & content) ü Define work plan aligned with EHR implementa9on 9meline ü Implement KM lifecycle (available tools) ü Implement monitoring process (CDS interven9ons)

• Replace isolated tools with integrated infrastructure
• Expose assets and processes to users and stakeholders
• Expand monitoring process (KM lifecycle & CDS evalua2on)
• Engage and collaborate with other organiza9ons

CDS Monitoring Portal

CDS Monitoring: Example

Revision (Nov) Released ac2ve (Dec) Released silent (Sep)

Reminder to document a principal problem 09/15/15 - Released silent for monitoring; firing was excessive 11/17/15 - Revised to fire only on admi`ed pa9ents (exclude ED pa9ents) 12/22/15 - Ac9vated; ~200 pa9ent-alerts/day

CKM software platform (CKMS)

• Integra2on

– Complete lifecycle: authoring, collabora9on, and publica9on – (Replace and consolidate legacy editors and repositories)

• Extensibility

– Core set of knowledge types and lifecycles with rich metadata – Configurable extensions to support local and reference assets

• Integrity

– Versioning and dependency management across knowledge types – Configurable valida9on pa`erns and rules (global or type-specific)

• Interoperability

– Extensive set of web services – Import & Export assets and type defini9ons (simple XML format)

• Seman2c “intelligence”

– Decision support for knowledge curators – Prevent errors and suggest op9ons: sustainable long-term maintenance

Content Sources

(e.g., EHR content, Open source content, Licensed content, etc.)

IMPORT

(XML format compa2ble with available standards – e.g., CTS2)

Content Consumers

(e.g., Clinical applica2ons and services, EHR systems, etc.)

Publication Authoring Linking Validation Review & Vetting

ETL

CKMS

EXPORT

(same XML format used for Import)

ETL

IMPLEMENTATION FRAMEWORK

Levels of support

• Reac9ve Support

– Troubleshoo9ng and maintenance of exis9ng assets – Preserva9on of original scope: quan9ty and coverage

• Periodic Improvement

– Sporadic improvements beyond iden9fied problems – Limited expansion of original scope

• Con9nuous Improvement

– An9cipatory iden9fica9on of problems and needs – Expanded (dynamic) scope, including new data types – Scalable and cost-effec9ve processes and interven9ons

Why Continuous Improvement?

• New EHR system includes a significantly larger collec2on of data &

knowledge assets, es9mated at 4 to 5 9mes the number of assets previously managed/curated

• Level of integra9on of the new EHR system greatly increases the number

& complexity of interdependencies between assets, aggravated by important limita9ons of “internal” configura9on tools

• Sites are demanding site-specific customiza2on & filtering of assets, given

disparate needs, target pa9ent popula9ons, resources, and processes

• Need for targeted & con2nuous CDS interven2ons increases as more

emphasis is made to manage high-risk popula9ons and episodes, taking into account different payer contracts and unique pa2ent characteris2cs

• Need for consistent data defini2ons aligned with standards is even more

cri9cal given increased demand for Interoperability, CDS, and Analy9cs

Complexity challenge

1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 500 1000 1500 2000 2500

Interdependencies vs. Number of Assets

Assuming that a given asset is typically connected to 3 other assets

CDS framework

• Clinical Decision Support (CDS)

– Ac2vi2es: analysis, design, specifica9on, build, tes9ng, monitoring, and troubleshoo9ng; evaluate and report the effec9veness of interven9ons; integra9on with user workflows and repor9ng; dependencies management – Assets: alerts, reminders, med warnings, duplica9on warnings, therapeu9c alterna9ves, infobu`ons, order sets, smart lists, etc. – mul9ple applica9ons – Support enterprise, site-specific, and research/innova2on efforts

Reac9ve Support

• Alerts, reminders, and

Infobu`ons

• +800 interven9ons for

Enterprise needs

Periodic Improvement

• Alerts, reminders,

Infobu`ons, warnings

• +1,200 interven9ons

for Enterprise and Site-specific needs

Con9nuous improvement

• All CDS, including

pa9ents and devices

• +2,000 interven9ons

for Enterprise, Site- specific, Research & Innova2on needs

Why Clinical Decision Support?

• Opportunity

– Complete CDS interven9ons – advanced protocols, risk stra2fica2on, management of comorbid condi9ons – Extend to pa2ents and allied healthcare professionals – Devices, genomic, pa2ent generated data, preferences – Performance data to con9nuously improve interven9ons

• Challenges

– Numerous interven9ons – some stakeholders not represented – Op9mal implementa9on – limited EHR features, maintenance – Poor design rejected by users – nega9ve percep2on about EHR reliability and u9lity – Interdependencies – isolated tools, resource intensive process, fragmenta2on, inconsistencies

Data Definitions framework

• Data Defini2ons

– Ac2vi2es: selec9on, analysis, mapping; management of local extensions; defini9on of reference models for high priority clinical topics; integra9on with user workflows, CDS, and repor9ng; dependencies management – Assets: data elements, local extensions, and standard reference models – Support enterprise, site-specific, and research/innova2on efforts

Reac9ve Support

• Data defini9ons, local

extensions, and reference models

• +20 clinical topics for

Enterprise needs

Periodic Improvement

• Data defini9ons, local

extensions, and reference models

• +40 clinical topics for

Enterprise and Site- specific needs

Con9nuous improvement

• Data defini9ons, local

extensions, and reference models

• +100 clinical topics for

Enterprise, Site- specific, Research & Innova2on needs

Why Data Definitions?

• Opportunity

– Data defined with standard reference models – consistency, completeness, and interoperability – Sustainable process – new domains, single electronic record for all sedngs and disciplines

• Challenges

– Data defini9ons not shared across EHR modules or serngs – similar data stored and encoded differently – Large libraries of defini9ons – promote inconsistency, dis9nc9ons not documented, overlapping domains/topics – Manual verifica9on – constantly evolving data collec9on tools (e.g. forms, flowsheets, templates, macros, etc.) – Poor data quality – affec9ng CDS, reports, registries, etc.

Additional frameworks

• Data classifica9on
• Terminologies (dic9onaries)
• Solware infrastructure
• Process analy9cs

FUTURE DIRECTIONS

KM opportunities (1/2)

• Individualized interven9ons

– Integra9on and effec9ve use of "new" data sources – e.g. devices, genomic, pa9ent entered, preferences

• Minimize fa2gue and maximize u2lity

– Detailed representa9on of context and state – e.g. interven9on targe9ng and con9nuous learning

• Prevent malfunc2ons

– Decision support to assist knowledge engineers – e.g. automate configura9on, customize valida9on rules

KM opportunities (2/2)

• Proac9ve maintenance

– Automated detec9on of malfunc9ons – e.g. con9nuously monitor ac9ve interven9ons, detec9on algorithms

• Localized knowledge transla2ons

– Automated detec9on of new evidence (from data) – e.g. evidence augmented by pa`erns of use, concurrent interven9ons with prospec9ve evalua9ons

• Mi9gate complexity

– Adopt new decision support methods and tools – e.g. combine inference methods, “dynamic” knowledge assets

CONCLUSIONS

Challenges – implementation

• Data availability and quality
• e.g. data not coded, coded inconsistently, not enough detail (codes)
• Rudimentary tools (edi9ng, versioning, linking)
• e.g. incorrect logic, missing values, related elements, automated

valida9on; content management tools are not a solu9on

• Labor intensive tes2ng
• e.g. posi9ve and nega9ve tests, regression tes9ng, automated tes9ng;

crea9on and maintenance of useful test data/pa9ents

• Prolifera9on of CDS vendors
• e.g. single vendor will not fulfill all your needs, overlapping

interven9ons (reconcilia9on) not iden9fied as a significant problem

• Internal (EHR) versus external service
• e.g. immature standards, complex configura9on, peculiar features

Challenges – large scale

• Labor-intensive acquisi2on process (SMEs)
• Inability to achieve proper domain coverage
• Rudimentary tools & resources
• Costly and inefficient maintenance
• Well-trained personnel using efficient processes
• Distributed and collabora9ve ac9vi9es – i.e. coopera2on

across teams and ins2tu2ons

• Common tools, processes, and standards

CDS has to follow the patient

• Clinical systems might have very similar CDS

features, but are frequently not configured the same way

– CDS triggered in one serng may not be confirmed

• r re-enacted in subsequent serngs
• Without con9nuity and consistency across

serngs and ins9tu9ons, interven9ons have decreased effec2veness for dissemina9ng evidence and reducing unwarranted variability

Acknowledgements

Saverio Maviglia Eileen Yoshida Charles Lagor Margarita Sordo Christopher Vitale Priyaranjan (Raj) Tokachichu All members of the Clinical Informa2cs Team at Partners Beatriz Rocha Dan Bogaty Tonya Hongsermeier Blackford Middleton

Thank you!

Roberto A. Rocha, MD, PhD

rarocha@partners.org http://scholar.harvard.edu/rarocha

• This work by Roberto A. Rocha is licensed under a

Crea9ve Commons A`ribu9on-NonCommercial-ShareAlike 4.0 Interna9onal License