Presentation on theme: "Knowledge Management Challenges in the Healthcare Delivery Market"— Presentation transcript:
1 Knowledge Management Challenges in the Healthcare Delivery Market Tonya Hongsermeier, MD, MBACorporate Manager, Clinical Decision Support and Knowledge Management,Clinical Informatics Research & DevelopmentPartners HeatlhCare System, Inc.
2 Agenda About Partners Healthcare Knowledge Management and Informatics Knowledge ApplicationKnowledge DiscoveryKnowledge Asset ManagementChallenges in Healthcare DeliveryWeak Organizational AlignmentWeak Investment in Asset ManagementImplications for Clinical R&DImplications for Personalized Medicine
3 Partners HealthCareMassachusetts General Hospital, Brigham and Women’s Hospital and several other hospitals in the networkLicensed BedsBirths ,478Admissions ,991Patient Days ,321Average LOSTotal Outpatient Visits 2,324,073
4 Partners Information Systems Much published on innovative use of informatics in healthcare (Bates, Teich, Glaser, Kuperman, Barnett, Chueh, and many others)800 applications520 active projects680 employees based in 19 locationsFY02 operating budget of $92.3MFY02 capital budget of $47MThese are relatively generous numbers as a percentage of operating expenses
5 Some Current Clinical Knowledge Assets Developed at Partners Medication Data Dictionary and DDIsInpatient alerts and interactive order rulesGerios and Nephros for proactive filtering of drug doses for elderly and/or renal insufficientRadiology Ordering decision supportPreventive health remindersOutpatient lab result decision supportOutpatient documentation templatesPiloting outpatient drug-lab, drug-disease interactive reminders
6 Current State Challenges Typical of Many Academic Healthcare Delivery Organizations 7 homegrown and 2 commercial CPOE systems, plan to evolve to “next generation CPOE” in next 5 yearsLimited implementation of structured (encoded) clinical documentationProprietary approaches to knowledge encodingNot re-usable or sharableMuch updating/maintenance is bottlenecked by resource constraintsResearch datawarehouse in place, but struggle to expand in face of fragmented clinical systems environment
7 Typical Committee and Project Structures Related to Medication Safety Illustrate Organizational Alignment ProblemInformation Technology ProjectsCommittees and DepartmentsPhysician Order Entry TeamClinical Data Repository TeamPharmacy System TeamClinical Documentation TeamElectronic Medication Adminstration TeamPharmacy and TherapeuticsPatient SafetyQuality or Performance ImprovementPolicies and ProceduresFormularyInfection ControlHealth care systems that succeed at organizational alignment are committed to measuring patient safety and have established clear goals, incentives, and accountability for achieving safety. Building such a culture is challenging in the context of today’s medicolegal environment and the legacy of fragmented approaches to care processes. In the medication use process, for example, it is sometimes unclear who is accountable for a safe outcome because each step of medication ordering, dispensing, and administration is managed separately by the physician, pharmacy, and nursing staff respectively. Further, it is not uncommon that systems for physician order entry, pharmacy dispensing, and medication administration are purchased and implemented somewhat independently of one another. The table above is a sampling of typical committees and project teams relevant to the systems and clinical knowledge that impact the medication use process . It is evident that as project teams implement their respective systems, they can be highly challenged to manage all the inherent interdependencies of achieving both workflow simplification for their respective customers and a safe medication use process. Many healthcare organizations are unprepared for the organizational investment necessary to continuously re-engineer their processes through clinical automation.Consider the topic of adverse drug event surveillance and the specific example of declining renal function in a patient on a nephrotoxic drug. It is sometimes the case that such surveillance may be performed by expert systems in more than one of the applications (Clinical Data Repository, Pharmacy System, Physician Order Entry System). Regardless of the potential number of inferencing mechanisms, if there is fragmented rather than interdisciplinary design approach, each application team will be challenged to determine whom and when to notify and how to escalate notification.
8 Medication Use Process Organization Medication Safety Steering CommitteeChief Medical Officer, Chief Nursing Officer, Chief Information Officer, Chief Quality OfficerInterdisciplinary Medication Use Process Advisory TeamPhysicians, Nurses, Pharmacists, Clinical Systems ArchitectsInformation Technology ProjectsCommittees and DepartmentsPhysician Order Entry TeamClinical Data Repository TeamPharmacy System TeamClinical Documentation TeamElectronic Medication Administration TeamPharmacy and TherapeuticsPatient SafetyQuality or Performance ImprovementPolicies and ProceduresFormularyInfection ControlThus, it is important to have an organizational superstructure (in this instance, an interdisciplinary medication use process advisory team), which is empowered to analyze and make design decisions that account for workflow, patient safety and technical feasibility. This advisory team needs to be empowered and sponsored by a steering committee of clinical and information systems leaders who set the strategy and priorities for medication safety. Resources must be dedicated to both the consensus development, design, and final knowledge encoding steps in order to leverage information technology investments in a timely fashion.
9 Knowledge Management: The Core Processes ApplicationKnowledgeAsset ManagementKnowledgeDiscoveryInfrastructure for knowledge management, particularly the knowledge that is tightly embedded in clinical applications, is a quality imperative. Winnie Schmeling, RN, PhD is Senior Vice President of Organizational Improvement at Tallahassee Memorial Healthcare. This site is among 12 sites that were selected by the Robert Woods Johnson Foundation for the Pursuing Perfection grant program ( She defines knowledge management as a “systematic process for making sure that everybody knows what the best of us knows” to deliver excellent care.Knowledge Management in practice is comprised of three key processes. These include knowledge application, knowledge discovery and knowledge asset management. They are organized in a circle to emphasize that the knowledge management process is one of continuous learning and knowledge dissemination
10 A Continuum of Clinical Decision Support and Knowledge Discovery* SurveillanceInteractiveProactiveLearningReference KnowledgeLinkingEvent MonitoringSafety NetAnticipationUnderstanding and PredictingPerformanceMonitoring patient data with passive decision supportIntercepting incorrect clinical decisionsMaking the right decisions the easiest decisionsPredictive ModelingCase-based ReasoningLearning Knowledge RepositoryHealthcare delivery is inherently complex and knowledge-dependent. To serve the needs of any clinical counter, relevant patient-centered knowledge must be accessible to the person supplying care at the “right time” in the workflow. Such knowledge can be delivered proactively before decisions are made, interactively as decisions are made, asynchronously such as via a pager, or passively as reference information that can be queried online. As biomedical research changes the standard of care at a rapid pace, the knowledge base required for safe care is growing exponentially. Knowledge Application is the science of leveraging knowledge at the right places in clinical workflow to support enhanced care-giver effectiveness, work-satisfaction, patient satisfaction and overall care quality. This view of Knowledge Application is broader and inclusive of the traditional definition of clinical decision support. Knowledge can be delivered through clinical applications in the form of templates for documentation, order sets for protocol-driven care management, dashboards and reports for feedback, and as alerts, reminders or suggestions. . In a high-performance learning organization, process and outcomes data is captured as a byproduct of workflow and rapidly analyzed to determine new models for process improvement. In essence, robust knowledge application evolves into and merges with knowledge discovery whereby the knowledge repository “learns” from the clinical outcomes in the patient database.*modified from the First Consulting Group Model of Clinical Decision Support
11 Medication Decision Support Categories at Partners REFERENCE INFORMATIONDrug-information knowledge linking via info button adjacent to drug namePartners handbook provides access to numerous drug information databasesPlanned drug-information knowledge linking via info button in electronic medication administration record in FY 04SURVEILLANCE AND MONITORINGDrug-induced abnormal lab result notification of physicianDrug-induced abnormal lab result notification of pharmacistRenal function decline in patient on renally excreted drug notification of physician and pharmacistINTERACTIVE DECISION SUPPORT FOR PHYSICIAN AND PHARMACIST:Drug-allergy checkingDrug-drug interaction checkingDrug-food interaction checkingDrug-herb interaction checkingDrug-disease interaction checkingINTERACTIVE DECISION SUPPORT FOR PHYSICIAN ORDER ENTRY ONLY:Drug-lab interaction checkingConsequent order recommendationsRelevant lab displayIndication-required ordersHeight, weight, allergy update required notificationDose calculation toolsIntravenous to oral conversion recommendation on renewal of intravenous order when patient receiving other oral medicationsFormulary substitution alertsAntibiotic restriction alertsPROACTIVE DECISION SUPPORTGerios for elderly patient medication dosingNephros for dosing in renal insufficiencyPreventive health remindersProblem-linked order setsAt Partners Healthcare Systems, a recent inventory of all medication-related decision support resulted in the above categorization. These are highly schematic representations. The taxonomy of a knowledge repository is highly informed not only by accepted standards for best practice, but by the local attributes of the healthcare organization, the features and functions of the applications that are used, and the clinical goals of the organization.
12 Reference information site at Partners used by clinicians across the system
13 Laboratory Notification with consequent order recommendations
14 Alternate Procedures, Redirects, Drug-Allergy, Drug-Drug, Drug-Lab etc.
15 Gerios: Dose-filters for age Nephros: Dose-filters for renal function At Partners, there are some examples of anticipatory decision support for safety in knowledge bases called Gerios for advising clinicians about appropriate drug dosing in elderly patients and Nephros for advising about drug dosing in patients with renal dysfunction. These types of prescribing errors are among the most common and costly (Bates, Yuen). As the clinician proposes antibiotic therapy, Gerios and Nephros surface dosing options in the order selection screen that are personalized to the patient taking into account his/her age and laboratory test results; highlighting the safest drugs and doses. The screen shot below of an order for a medication, temazapm illustrates that how information is presented by the Gerios system. This approach has been very well received by Brigham and Women’s clinicians because it is perceived as a non-task interfering design approach that makes the right care decisions the easiest care decisions.Yuen EJ, et al, “Sedative-hypnotic use and increased hospital stay and costs in older people”. J Am Geriatr Soc Nov; Vol. 44(11):Bates DW, Spell N, Cullen DJ et al. The costs of adverse drug events in hospitalized patients. Adverse Drug Events Prevention Study Group [see comments]. JAMA 1997; 277(4):Inappropriately sedated elderly inpatients on average incur $5600 excesscosts over expected for severity of illness
16 These reminders in our ambulatory health system remind physicians to perform preventive health interventions.Preventive Reminders
17 Problem-level anticipatory decision support Today, order sets and documentation templates are static which means that clinician must change them to personalize them to patientWe plan to use inferencing to dynamically generate problem-driven order sets and documentation templates that account for multiple co-morbiditiesMust be able to incorporate future onslaught of gene diagnostic and prognostic data
18 Clinical Standardization Knowledge Application must anticipate these dimensions of the clinical encounterClinical StandardizationStandards of Practice,Role/Venue RequirementsBilling/Regulatory RequirementsThe vendor community will be required to push development in the direction of anticipatory modes of decision support particularly in light of clinician concerns about speed and usability of information systems. This is illustrated by the recent events at Cedars-Sinai Medical Center in Los Angeles. The computerized physician order entry system was turned off in early 2003 after hundreds of physicians complained it slowed down the process of filling their orders (7).Further, with the expected arrival of gene-diagnostics as additional data to be considered in therapeutic decision making, exclusive reliance on the safety-net mode of decision support will become increasingly untenable.User PersonalizationEnd-user workflow preferencesLearning and User-definedImprovisationPatient Preferences
19 Notice in this screen shot that the expert system is accounting for numerous types of data, some of it inferred from other data, in order to make a patient assessment and formulate a treatment recommendation.Poly-hierarchical inferencing with actionable advice – surveillance, interactive, or proactive mode
20 Recommendation accounts for patient immune status and renal function This is an example from clinical decision support company called Theradoc
21 A Continuum of Clinical Decision Support and Knowledge Discovery SurveillanceInteractiveProactiveLearningReference KnowledgeLinkingEvent MonitoringSafety NetAnticipationUnderstanding and PredictingPerformanceMonitoring patient data with passive decision supportIntercepting incorrect clinical decisionsMaking the right decisions the easiest decisionsPredictive ModelingCase-based ReasoningLearning Knowledge RepositoryKnowledge Discovery is the clinical research process of analyzing data for the purpose of understanding performance, reporting, predicting, and harvesting new knowledge on how to improve. With each successive stage of decision support in Figure 3, healthcare performance becomes increasingly transparent. As processes for care-giver communication, results access, ordering, documentation, task completion, and patient monitoring become supported by expert systems, more data is collected about clinical rationale, process and outcomes. In a well designed analytic data repository, this data can be mined to identify new knowledge that can, in turn, be disseminated through knowledge-enriched clinical applications. Rapid-cycle learning is the goal of knowledge management.*modified from the First Consulting Group Model of Clinical Decision Support
22 Current Initiatives Quality data warehouse with Clinician Dashboards Early identification of patients at risk for case managementLonger term knowledge discovery goals to use performance data to enhance knowledge repositoryNeed to evolve towards non-human dependent modes of knowledge acquisitionCurrently, most hospitals have poor access to clinical data for measurement and rely, instead, on billing and administrative data. The architecture of a typical transaction-oriented Clinical Data Repository (CDR) is not optimized to support measurement. Further, the data that must be aggregated to enable deep analytics is often located in many databases across an integrated delivery system. In the absence of an analytic data repository, those engaged in the process of research and performance measurement must bear the cost and time delays of chart abstraction land data acquisition to aggregate clinical data, which consequently slows the translation of such insights into quality improvement.
23 Knowledge Asset Management Infrastructure: Analysis of clinical performance data to understand where knowledge deficits are to support performance goalsAuthoring and support of virtual, asynchronous collaborative authoring by knowledge editors and leaders of research, safety and quality improvement initiatives (reference knowledge specs for encodingKnowledge acquisition from commercial/etc knowledge basesValidation and audit trail maintenance (meta-knowledge)Inventory (knowledge librarian)Publishing and SharingReference information and knowledge modelKnowledge Asset Management is one of the more challenging aspects of knowledge management. In non-healthcare industries, the focus of asset management is typically on free-text document management. In healthcare, we must also focus on knowledge encoded in workflow applications to support clinical care, administrative functions, and clinical research. This is a list of processes that must be supporeted by that asset management system. These include…
24 What are the challenges today: Healthcare delivery organizations purchase systems but don’t invest in knowledge asset management, they install plumbingVendors sell knowledge editors, not knowledge management support infrastructureThere is no repository of “best clinical IT practices” at a national level, few among the vendorsNo knowledge encoding and representation standards to facilitate knowledge sharingRobust knowledge asset management systems must support the collaborative authoring, encoding, inventory, versioning, updating, sharing, and dissemination of encoded knowledge. As discussed above, many delivery systems adopting an EHR (whether single clinic, or large integrated delivery network) often have poorly developed or absent organizational structures in place for consensus development on how to convert reference knowledge into the specifications for encoding alerts, reminders, order sets, documentation forms and templates. The proprietary knowledge editors of most vendor systems support knowledge encoding but rarely the inventory in a repository or any of the collaborative processes required for agreeing upon how knowledge should be encoded. Commercial content management systems and groupware have emerged in recent years that support virtual, asynchronous collaborative authoring of documents whereby subject matter expert opinion can be elicited and shared through simple and list-serve interfaces. These capabilities will need to be merged with the capabilities of traditional commercial knowledge-editors to create a knowledge asset management system that supports end-to-end knowledge engineering.
25 Partners-Wide Knowledge Management Model KNOWLEDGE ASSET MANAGEMENTSignature Initiatives andSub-Committees setEnterprise-wide Strategy,Clinical Standards andPerformance MeasuresPerformance Feedback to Leaders, SMEs, Committees, and End-usersSubject Matter Expert (SME) PanelsAdvise on Entity, Venue, Role, Specialty,Primary Care, Disease Management, andSafety related requirements forapplication function and knowledge basesData WarehouseDecision Support Design Teams direct the design of cross-functional knowledge to be encodedPartners GeneticsComputing PlatformPERFORMANCEand OUTCOMES(KNOWLEDGE DISCOVERY)Applications for Virtual Collaborative Knowledge Authoring and MaintenanceThis is a schematic drawing of a knowledge engineering and management factory for Partners Healthcare. Key Features are:Signature Initiatives determine Partners-Wide clinical standards of care and performance measure benchmarks (ie. HgA1C maintained between 7 & 8)Design Teams representing Inpatient Care, Ambulatory, and Medication Decision Support work on the design of Partners-wide knowledge bases (ie if HgA1C>8, increase insulin). Focus on the 20% of Content that drives 80% of business performance. Remaining content is authored by end-users.Subject Matter Expert Panels (SMEs) representing Entity, Venue, Specialty Care, Primary Care, Safety, and Disease Management provide advice on optimal approach from a function and knowledge perspective. (ie Diabetes Council advises on HgA1C rules). Also direct subspecialization of LMR.Design Teams direct Common Services knowledge editors to build knowledge base. (ie rule is updated in LMR reminders)Clinical Portfolio Assembly supports Inter-application coordination to address legacy application challenges (ie EPO guideline must be re-encoded 7+ times)Knowledge is served up through workflow applications in the form of data, order sets, rules, alerts, dashboards, documentation templates, and the likeActivity data is collected by the Quality Data Warehouse for feedback and reporting back to end-users (dashboards), SI leaders, design teams.KnowledgeRepositoryClinical WorkflowApplications andServicesKnowledgeBuilding BlocksInformationModelCommon Services Knowledge EditorsDECISION SUPPORT(APPLIED KNOWLEDGE)
26 Knowledge Asset Management: Translating Goals into a Knowledge Repository TaxonomyCare Applications(Results, Observations, Orders, Tasks/Proc/Mar,Messaging, CDS,Measurement)and Knowledge BasesGoal Framework: Safety, Quality, Efficiency, ResearchData/Knowledge SeekingAssessmentDx/RxDecision MakingOrder Fulfillment,Communication and CoordinationBillingReportingTransfer/HandoffCORE CARE PROCESS AUTOMATION TAXONOMYMedical Management, Research, and ReportingClinical Knowledge for Personalized Medicine TaxonomyReference Information ModelRole and Venue Domain TaxonomyWhen a healthcare organization begins an EHR implementation, they must inventory and organize their enterprise-wide knowledge assets into a human-readable knowledge repository. Such an inventory might include existing paper order forms, documentation forms, policies, procedures, as well as any data on design of legacy systems. They must then define a taxonomy for their knowledge repository informed by their core processes, knowledge domains, venues, and roles. These become the basis for defining the requisite care applications, knowledge bases, and reference information model to support the goals. It is important to emphasize that knowledge-based sources constitute a resource not only for physician order entry, but for all processes (documentation, task completion, medication administration, etc) and for any role (physician, nurse, pharmacist) in the healthcare delivery system.RequirementsCare Applicationsand Knowledge Bases
27 MEDICATION USE PROCESS: Acetaminophen in a 2.5 Kg Premature Infant The medication use process illustrates the importance of a multi-dimensional approach to clinical systems design. It requires the integration of differing information needs and workflows to achieve a safe order fulfillment process. In the following acetaminophen order for a 2.5 Kg premature infant, notice how the information needs differ by role.
28 Sample High-level Example Taxonomy for Knowledge Assets
29 Center for Clinical Knowledge Engineering Welcome to the National Knowledge Engineering RepositoryGoContent searchHEDISAdvanced Search Filters (press Ctrl to select more than one):Clinical Discipline: SurgicalInformatics ModeCardiothoracic SurgeryInteractive RulesSearch File HierarchyInterventional CardiologyOrthopedicsEtc.Surveillance and NotificationsDocumentation TemplatesEtc.Knowledge Asset Management ToolkitLink to references, survey instruments, diagrams, descriptions, process flow diagrams, etc on Partners and VA approaches to asset managementClinical Discipline: Non-SurgicalAgeCardiovascularAdultSubmit Content to EditorEndocrinologyGastroenterologyEtc.PediatricsNeonateEtc.About UsClinical Discipline: SafetyRoleNosocomial Infection ControlNurseMedication SafetyDecubitus Ulcer PreventionEtc.PhysicianCase ManagerEtc.Clinical Discipline: Disease ManagementVenueDiabetes MellitusCCUCongestive Heart FailureMultiple SclerosisEtc.Ambulatory CareEmergency DepartmentEtc.
30 Knowledge Specifications For Encoded KnowledgeVsMeta-knowledge aboutThe knowledge
31 Future State KM Model Workflow Applications Portal CollaborativeKnowledgeAuthoringToolsWorkflow ApplicationsMeta-KnowledgeRepositoryKnowledge-based ServicesKnowledge RepositoriesInformation Model
32 Barriers to Success at the Intersection of Clinical Informatics and KM Leadership inadequately committedProducts inadequate to support processesBusiness case intangibleFear of exposure (technology increases transparency)Few roadmaps to success are proven in the healthcare arena
33 Market Drivers will Propel Progress Aging population: computer literate and population growth will outstrip service capacity, informatics must support self-managementBusiness community will aid transition from commodity to value based purchasing by employers and consumers, they know that the current inflation rate of the commodity is untenableLeapfrog and Government are beginning to purchase qualityGenomics: personalized medicine will require technologies for personalization, these same technologies will enable more user-friendly safety solutions