Clinical Decision Support
Technical Expert Panel Meeting

August 19, 2009

3:00 – 5:00 PM ET

Facilitator: Eta Berner

---

Agenda
____________________________

• Welcome

• Review of June 26 TEP Meeting at AHRQ

• Discussion of Contractors’ Status Reports

Progress and accomplishments; challenges, with a focus on technical challenges; questions for the TEP; and next steps

I. GLIDES

II. CDSC

• Recap

---

Review of June 26th TEP Meeting at AHRQ

---

GLIDES UPDATE
August 19, 2009

---

Summary

• GLIDES CDS now live in multiple locations in CT, DE, FL

• We are on schedule and within budget for first 18 months of contract

• Continuing to focus on

  • Roll-out of Phase II CDS deliverables at Yale and Nemours

  • Evaluation of Phase I Asthma CDS at Yale Specialty Clinic

  • Phase III design and workflow analysis of primary care at Yale and in the Delaware Valley primary care practices

• Balance of our work through February 2010 will focus on increasing and evaluating utilization of the CDS applications

• Promoting voluntary use appears to be the most challenging aspect of this project

---

Accomplishments

 

Nemours Asthma Implementation 

• Roll-out activities for the Nemours Asthma CDS SmartForms are now complete

• Forms are functioning and performing correctly in production, but we are seeing low utilization

  • Missing clinical champions “on the ground” in each clinical location (following recent Nemours organizational changes)

  • Uses SmartForm in EpicCare – more complex and time-consuming to use

• We are working with Nemours to recapture clinical attention and boost adoption after the summer vacation period.

 

Nemours Obesity Implementation

• Roll-out activities for the Obesity CDS are now complete

• We are beginning to see usage but plan to work to improve adoption in September

• Uses SmartText in EpicCare - more intuitive/simpler to select and launch

• Beginning to receive requests for enhancements – evidence of usage and commitment to make the CDS work well for the clinical community 

---

Accomplishments

 

Yale Obesity Implementation

• Obesity CDS is now in use at the Primary Care clinic

• Every clinician seeing a child age 2 and older for a health maintenance visit has the nutrition and activity decision support automatically inserted into the documentation stack

 

Evaluation

• Qualitative evaluation and thematic categorization is underway

• Preparing a detailed discrepancy analysis which analyses differences between clinician decisions and the CDS recommendations

• Detailed review and comparison of the both the Yale and Nemours Obesity CDS designs in September

 

Phase III Planning

• Asthma CDS at Nemours Primary Care clinics in Delaware Valley will require minor changes to the CDS design prepared for Phase II (pending assessment of Phase use)

• Asthma CDS at Primary Care clinics at Yale (New Haven) will require design changes and more significant changes to workflow

---

Technical Challenges - GEM

• Multi-Platform Operation

  • MAC – PC transfer

  • Consistent handling of Unicode characters

• Embedded Graphics

• Guideline Versions

• Guideline Quality

  • Implementability

  • Missing details

• Use of controlled vocabulary

---

Extractor Overview

• Provides users with the ability to view XML formatted guideline knowledge using structured, extensible stylesheet transformations (XSLT)

• XSLT, which uses XPATH as a query language, enables

  • Flexible access to the content of elements of a GEM document

  • Flexible reporting and presentation of guideline knowledge components

• Extractor is Java web application built using the Struts 2.0 Web Framework

• GEM Documents are uploaded to an Apache Tomcat web server via a web interface that allows users to select the format of the generated report

• The uploaded file is transformed to HTML using a Xalan XSLT processor and formatted via an XSL stylesheet to meet view criteria specified in the request

• Various views of the data are available

  • Recommendations

  • Imperative and conditional statements (IF...THEN …ELSE rules)

  • Knowledge Components

  • Decision Variables

  • Actions and Directives

  • GEM-COGS, GEM-Arden

---

Technical Challenges - Centricity

• User Interface

(UI) Design

• Access to Data

• Time Stamping

---

Technical Challenges - EpicCare

• Asthma vs. Obesity CDS Usability

• SmartForms vs. SmartText

• Locating Database Fields That Store CDS Actions

• Prioritizing Enhancements To The CDS

---

Questions For The TEP

• How to design clinical workflow so that CDS is triggered for specific patient encounters, especially for multi-problem patients?

• How is guidance best prioritized for patients with multiple problems?

• Under what circumstances should mandatory requirements be imposed on clinicians?

• How are regulatory edicts (e.g., recording of pain scores for all patients, including 1 year old children presenting for health maintenance, TB risk factor screening on all patients) most effectively integrated (or resisted)?

• What practices are most effective in improving use of the EHR at the point of care for clinicians who believe that the computer interferes with the clinician-patient relationship?

---

Next Steps

Phase II Implementation

• Continue to work to improve utilization at all sites

Phase III Implementation

• Commence Phase II design and delivery (Asthma at Nemours Delaware Primary Care, Asthma at Yale Primary Care).

Evaluation

• Continue formal Phase I Evaluation data gathering and quantitative/qualitative evaluation activity

• Complete compilation of “lessons learned” and draft CDS design and implementation-related papers

• Commence Phase II Evaluation

---

CDSC UPDATE
August 19, 2009

---

Accomplishments Summary

• Focused on development and testing of our service in preparation for implementation in the Partners LMR (scheduled for November 2009)

• Conducted first end-to-end test of 75% of the service components in our own development environment and moved them into QA

• Worked on our option year proposals and began meetings with Regenstrief (second demonstration site)

---

Accomplishments, cont.

• Conducted site visit to Zynx from 7/28-7/30

• Recorded first demo of CDS best practices from our qualitative assessment featuring Regenstrief

• Completed install of Documentum components with customizations on production servers

• Developed assessment tool for health IT vendors on their KM capabilities

• Finalized clinic selection for demonstration at PHS

• Developed eRoom database to store comments on editorial policies for submitted content

• Completed analysis of actionable DM rules from across the consortium sites

---

Challenges: KMLA, KTS, CCHIT, CGC

• Vendors are “pushing-back” on answering questions regarding their content and CDS capabilities

• Team lead for KTS left Harvard and moved but will continue to work on the project

• Difficult to create easy-to-use editing tools and manage GELLO’s complex syntax

• Difficult to plan for CCHIT recommendations since we are waiting for ONC’s decision on its future role

• Challenging to make editorial policy decisions within the Content Governance Committee since these must go “up the chain” at the consortium sites

---

Challenges: KM Portal

• Design and implement metadata constraints associated with all four content specification levels

• Define document relatedness for the evolution of content through multiple specification levels

• Difficult to synchronize enhancements across both the portal and repository, which are independent applications

• The content management system at Partners underwent a major version upgrade, which required redesign of some CDSC capabilities

• Difficult to assess the portal usage patterns without requiring users to log on

---

Challenges: Service/Demonstration

• An operations resource is still required to support the decision support service pilot

• Delays in finalizing content selection delays specification development and rule authoring

• Coordinating multiple components required for the decision support solutions presents logistical challenges, given the tight timelines

• Coordinating with all required data sources and mapping local terms to standard codes

---

Services: Decisions, Decisions . . .

• Actually making a decision is fast and easy

• Preparing the decision

  • Ruleset identification

  • Data inventory

  • Data assembly, translation

  • Inference model assembly

  • Inference model refinement (i.e., higher-order classifications)

• Returning something meaningful

• Informatics + high-performance software engineering

---

High Level Sequence Diagram for CDSC Pilot in PHS LMR App

---

Calling ECRS

---

Patient Information Model

---

Action Model

---

Demo: Co-existing

• Clinical governance of existing content vs. new ‘stateless’ content based on research clinics vs. ‘operational’ clinics to avoid overlaps.

• Application layer to capture and maintain mapping data from externalized service for trigger rules.

• Not all content will replicate on GoLive thus strategy on parallel DSS for the short term.

• Dependencies between rules and actions limit how many rules are replicable in an externalized service in the short term.

• Trigger Logic: Pre-generate vs. real-time

---

Synchronization

• Middleware (called the ‘Orchestrator’) communication schema is generalized; thus coordinating application layer services is critical

• User defined actions and iterations of snoozing logic

• Error handling: who is capturing what, when and how and communication protocols following it.

• Testing

• Performance

• Reporting: Transaction events distributed across application, middleware & rules engine.

---

Process Flow Diagram
CDSC LMR Integration

---

Reminders in LMR

---

Questions for TEP

• Advice for dealing with push-back from vendors when we ask about their decision support capabilities

• How to plan for CCHIT recommendations while waiting for ONC’s decision on future of CCHIT and other certification organizations

• How to make editorial policy decisions with the participants (decisions need to go “up the chain”)

• Suggestions for Option years 3 to 5

---

Next Steps

• Visit another clinical content vendor (First Data Bank)

• Continue evaluation of model using GLIA

• Develop and test portal function

• Conduct interviews with health IT vendors on their KM capabilities

• Continue to write and test AHRQ/CDSC rules in QA

• Continue analysis of LMR Reminder data, and how to get it into QDW each month

• Review CAD and HTN rules from consortium sites

---

Recap