Bringing Clinical Guidelines to the Point of Care with HIT

www.CenterForUrbanHealth.org

MN HSR ConferenceMarch 3, 2009

Bringing Clinical Guidelines to the Point of Care with HIT Intelligent Designers & Adaptive Agents

ComparedYiscah Bracha, MS

Minneapolis Medical Research Foundation

Gail Brottman, MDHennepin County Medical Center

Kevin Larsen, MDHennepin County Medical Center

Robert GrundmeierThe Children’s Hospital of Philadelphia

Angeline Carlson, PhDData Intelligence, Inc.


The issue:

• Medical care delivered ≠ medical care recommended in evidence-based guidelines

• Is this a “problem”? Health policy: It is a problem. Docs not

following guidelines, pts don’t get best care.

Docs: Construction as “problem” depends on reason for differences


Reasons identified empirically:

1. Information overload Too many guidelines Unaware of specific recommendation Need info in the moment of delivering care

2. Limited resources for implementation. Docs don’t have time or staff No reimbursement

3. Recommendations not useful or relevant


Reason 1: Info overload

• Assumptions: Docs want to use guidelines, but don’t know what they are

• Source of the problem: Limits to human cognitive capacity

• Solution to the problem: Use information technology to enhance human cognitive capacity


Reason 2. Limited resources

• Assumptions: Docs want to use guidelines, know what they are, but cannot implement them with existing resources (e.g. time, staff)

• Source of the problem: Inadequate material resources

• Solution to the problem: Change reimbursement systems.


Reason 3. Not useful or relevant

• Assumptions: Contested!• Epistemic legitimacy: Biomedicine vs.

epidemiology• Source of authority: Front-line

clinicians vs. university-based researchers

• Credible “evidence”: Clinical practice vs. controlled experiments

• Problem? Contested!


Using HIT as soln to info overload:

HIT Tool: Experience:

PDF of guideline on screen

Clinicians don’t access it

Pop-up reminders and alerts

“Alert fatigue”.

Electronic clinical decision support tools for:

Diagnostic tests to use Tools exist. Varied effect on doc behavior. Pt outcomes unknownInitiating therapy

Modifying therapy over time

“The final frontier”


A Vision of the Future

• Individual practice sites will have up-to-date electronic decision support tools.

• Tools based on guidelines’ recommendations

• Recommendations based on evidence• Tools integrated into EHR systems.• Tool updates disseminated

electronically


(assumptions behind the vision)

• Material resources are adequate• Contests over legitimacy & authority

resolved.


Current HIT work leading to vision:

• Intelligent Designers Practiced by subgroup in academic medical

informatics community Supported by govt grants & contracts to

universities Current efforts guided by vision of future

• Adaptive Agents Practiced by vendors & their healthcare

customers Supported by market forces Current efforts guided by immediate needs


Intelligent Designers

• Situated at: Source of guidelines• Looking towards: Universe of practitioners• Concept: Convert guidelines into executable

code; disseminate code to practice sites.• Implementation: Standards (to help local

implementation) adopted by: Guideline developers EHR systems Guideline coders.

• Information channels: Academic conferences & peer-reviewed journals


Adaptive Agents

Situated at: Practitioner sites Tool Developers

Looking towards:

Universe of potential tools

Universe of potential sites

Development concept:

Find tools that meet local needs

Develop tools that meet local

needs

Implementation concept:

Use whatever is available

Relationships w. EHR vendors

Communication

mechanisms:Healthcare product marketplace


Properties of developed tools:

Properties: Single

tool

Current Efforts By:

Intelligent Designers Adaptive agents

ContentLanguage – can

represent guidelines as executable code

Support - clinical decision making &

administrative documentation.

ScopeBroad – all guidelines thru entire lifecycle

Narrow – single clinical condition or issue

Development effort

Extensive. Modest

Local Install Effort

ExtensiveDepends on local

environment

Local Use Effort

Intended to be minor Depends on installation

HIT Asthma Project:An Adaptive Agent Example

Project supported by the Agency for Health Research and Quality.

Contract No. HHSA290200600020Task Order No. 5

The findings and conclusions are the responsibility of the authors, not the AHRQ.


HIT Asthma Tool Supports:

• Assessment & documentation: Asthma severity for untreated patients Asthma control for treated patients

• Selection of age-specific therapy: Initial therapy for untreated patients Modified therapy for treated patients

• Production of: Asthma progress note for patient’s chart List of selected meds & instructions for use Patient-friendly Asthma Action Plan.



HIT Asthma Tool: Properties

• Coding language: Java applet• Installation mechanism: Hyperlink

opens applet on delivery site’s Intranet.• System tool data exchange:

Encrypted data in URL of hyperlink• Tool system data exchange:

Individual patient record: Doc opens applet in read-only mode to get asthma-specific hx.

Aggregate records: Merge data extracted from EHR & data generated by applet


EHR Data Repository

Information forPopulations

Local EHR

Information forIndividual Patients

HIT Asthma Data Model for EHR


EHR Data Repository


Local EHR


ASTHMA APPLET

Asthma info for Individual patients

ASTHMA REGISTRY

Asthma info for Populations

Asthma Summary

Read-only invocation

tool

Patient & user context

HIT Asthma Data Exchange Model


EHR Data Repository


Local EHR


ASTHMA APPLET


HIV APPLET

HIV info for Individual Patients

ASTHMA REGISTRY


HIV REGISTRY

HIV info for Populations

Asthma Summary


tool


tool

HIVSummary



Evolutionary Emergence:


EHR Data Repository


Local EHR


ASTHMA CIG*


HIV CIG*

HIV info for Individual Patients

ASTHMA REGISTRY


HIV REGISTRY

HIV info for Populations

Intelligent Design

interface engine

interface engine

* Computer-Interpretable Guideline


Both designers & adaptive agents:

• Face challenges: Converting natural language narrative

into computer-executable code Implementing uniform code into widely

disparate local systems

• Respond to challenges: Designers: Design global solutions Adaptive agents: Locally adapt


Proposed Solutions vs. Adaptations:

Intelligent Designers Adaptive Agents

PerspectiveGuideline developers & researchers

Healthcare delivery sites & vendors meeting their needs

ViewDisparate local delivery sites

Sites products; Product developers sites

Goal

Convert narratives to CIGs, disseminate to local sites electronically

Meet immediate needs

Implementation challenge

Proposed Solution: Standards

Adaptation: Whatever works

Conversion challenge

Proposed Solution: Standards

Adaptation: Contest or accept

epistemic assumptions.


Conclusions: Role of HIT

• HIT can encourage guideline-based care Addresses information overload Does not address:

Material resources Contested assumptions


Predicted response to challenges:

• Dispersed, creative, adaptive efforts will continue: If regulatory & reimbursement systems create needs Successful adaptations will spread

• Global design efforts will continue If grant funding continues

Convergence towards standard representation language No convergence towards standard implementation

• Standards: May assist locally adaptive agents• Vision will manifest (if at all) by evolution, not

design Loosely coupled data exchange No one in control Messy, but will bring more guidelines to more points of

care.


Thank YouThank You

For more information:Yiscah Bracha.

[email protected]


Live data transfer & registry

Registry & user entry

3. Peak flow, triggers, weight.

Live data transfer from EHR

Lookup in registry

2. Identify patient

Live data transfer from EHR

Log in by user & system check.

1. Authenticate user

2. Verify Level Of

Severity

Choose next step.

2. Validate meds in record

3. Identify current asthma meds

1. Assess Control

Age 0-4 Age 5-11 Age 12+

1. Classify Severity

Age 0-4 Age 5-11 Age 12+

4a. Current meds map to

recognized plan

4b. Current meds do not map to

recognized plan

4. Choose next step.

Existing treatment recordNo existing

treatment record

Not validated

Validated.Asthma not

well controlled

5. User chooses what to do next.

1. Choose Treatment Plan

1. Next Visit Info

2. View/Print Asthma Action Plan.

Selected medsVisit Summary

2. Choose daily controllers

3. Choose quick relievers

4. Choose burst meds (for exacerbations)

Validated.Asthma

well controlled

New values of variables and PDF of AAP to registry

New values of variables and PDF of AAP to registry

Saves.

4a. Existing record of severity

(From EHR or registry, if exists)(From registry, if exists)

4. Classify severity or assess control.

4b. No existing record of severity

Return user to EHR system

Return user to operating system

Clo

sure

HIT Asthma.Model workflow for an outpatient asthma visit

Consistent with recommendations in 2007 NEAPP Guidelines.

EHR-compatible versionDesktop version

Bringing Clinical Guidelines to the Point of Care with HIT

Health & Medicine

clinical guidelines

source of guidelines

asthma tool

evidence tools

evidencebased guidelines

medical care

asthma project

time initiating therapy