healthcare use of workflow engine technology with emphasis on data analysis and decision support

Healthcare use of workflow engine technology with emphasis on data analysis and decision support

Vojtech Huser MD PhDLaboratory for Informatics Development

National Institutes of Health, Clinical CenterBethesda, MD, USA

Vojtech Huser, MD, PhDVojtech Huser, MD, PhD2

Objectives 1. Describe the abstract notion of a workflow engine and

workflow technology in general 2. Understand the relationship of flowcharts (common in

medical guidelines) to executable models of processes used by workflow engines

3. Understand current use of workflow engines in healthcare in production environment and in research context (phenotype modeling, data analysis, clinical decision support, process mining and discovery)

Includes description of some of my research projects 4. List the evidence for benefits and challenges of using

workflow engines in healthcare

Introduction Workflow technology: focus on my research

for the last 8.5 years

This talks is a mixture of Workflow technology tutorial Larger context of WT My prior and current work


HealthcareWorkflow.wordpress.com

Workflow Technology (WT)

Vojtech Huser, MD, PhD 4


Workflow Technology (WT) Business process management system (BPMS)

Workflow = The automation of a business process, in whole or part, during which documents, information or tasks are passed from one participant to another for action, according to a set of procedural rules.

Workflow Process = A set of one or more linked procedures or activities which collectively realize a business objective or policy goal, normally within the context of an organizational structure defining functional roles and relationships (definition vs. instance)

Worklist Handler = A software component that manages the interaction between the user and the worklist maintained by a workflow engine (task display, completion, acceptance, referral)

WfMC: Terminology & Glossary, Document Number WFMC-TC-1011, Feb 99 http://www.wfmc.org/index.php?option=com_docman&task=doc_download&gid=93&Itemid=74


http://www.wfmc.org/index.php?option=com_docman&task=doc_download&gid=93&Itemid=74

WT history History

1960 Carl Adam Petri 1999 SDO, standard terminology 2000s

2002: XPDL 1.0, 2008 XPDL 2.1

Since 2000s offered by large IT companies

(MS, Oracle, SAP, IBM, Fujitsu) Healthcare

2008 Gartner report



Humans & machines working together


Executable flowchart


Minimum componentslanguage editor

execution engine


Workflow engine


Engine


Use of WT Pioneer domains

Banking Manufacturing Shipping Government authorities Pharmaceutical

Healthcare Gartner 2008 report: <5% use

More in the third section

Flowcharts


Flowcharts Present in clinical guidelines Can summarize several pages of text Take-home message Algorithmic nature (step by step) Usually very well defined terms, thresholds

and treatment options Used even more often in local guidelines


Clinical Guidelines Quality Improvement measures

Process definition (workflow editor)

http://www.ncqa.org/


Examples of WT use in healthcareBed management

Infections control (MRSA)

J. Emanuele and L. Koetter, "Workflow Opportunities and Challenges in Healthcare," in 2007 BPM & Workflow Handbook, 2007.L. Koetter, "MRSA infection control with workflow technology," Spring AMIA Conference, Orlando, FL, 2007.R. Hess, "The Chester County Hospital: Case Study," in 2007 Excellence in Practice: Moving the Goalposts., 2007.


Stroke guideline (WfMS)


Soarian


Current Soarian Example

Main Line Health (Harm Sherpbier)Soarian: Workflow engine (Tibco), MLM (Arden Syntax)


Rheumatology decision support


My research in workflow technology


1.

2.



Medinfo 2007


Implemented processes RetroGuide (v2) (2004-2008) (published)

Quality improvement Osteoporosis Cholesterol management Blood pressure control in

diabetics Decision support

Computerized glucose management protocol

Adverse Drug Events (naloxon, sleep apnea)

Research Hodgkin's lymphoma and

pregnancy Hepatitis C

+ several other experimental

RetroGuide (v3) 20+ scenarios

US Prev. Services TF Health plan rules hsCRP Genetic counseling Thyroid disease, ACEi

monitoring Osteoporosis

FlowGuide Surgical patient discharge

(feasibility) Cholesterol management Rheumatoid arthritis


Huser (2010) J Biomed Inform


Problems which HealthFlow tries to address? EHR systems need user-friendly customization Representation must be able to represent

complex logic and understandable Execution must be integrated into an routine-

care EHR system (standalone engine vs. integrated)

Same representation/execution must be deployed at multiple institutions

Switchable representation (retrospective experiment, prospective observation)

Re-uses existing cross-industry technology



HealthFlow System

Actions

Knowledge Base(workflow process

definitions)

Data Inquiry

Workflow Execution

Engine(Shark)

EHR System layer:

Workflow editor(JaWE)

Data Visualization

Additional Components:

Event Data Model

code.google.com/p/HealthFlow


HealthFlow: RetroGuide, FlowGuide

EHR systemEvent

Listener

Notifications

Actions


definitions)

Data Inquiry

Workflow Execution

Engine(Shark)

EHR System layer:

Prospective mode (FlowGuide):

Retrospective mode(RetroGuide):

Reports

Data Warehouse


Data Visualization


Event Data Model

Additional External Services(Terminology services, NLP services,

Statistical Analysis)

code.google.com/p/HealthFlow


HealthFlow: RetroGuide + FlowGuide


EHR systemEvent

Listener

Notifications

Actions


definitions)

Data Inquiry

Workflow Execution

Engine(Shark)

EHR System layer:



Reports

Data Warehouse

Workflow mining(ProM)


Data Visualization


Event Data Model

Ontology Model

Host EHR Passive Interaction Model

Host EHR Active Interaction Model



Workflow log analysis


HealthFlow : key concepts

Flowchart layer + code layer (applications) Set of HF external applications (EAs) Single patient execution model (DSS)

Works with time ordered chart Resembles manual chart review Concept of current position in EHR Use of variables

No AI component No autonomous DM algorithm, advanced ML methods Simple execution of human-specified steps (prior knowledge), no machine input

Data querying, answering questions, data exploration

Vojtech Huser, MD, PhDVojtech Huser, MD, PhD40self service tools (i2b2)




EHR systemEvent

Listener

Notifications

Actions


definitions)

Data Inquiry

Workflow Execution

Engine(Shark)

EHR System layer:



Reports

Data Warehouse



Data Visualization


Event Data Model

Ontology Model







Code Layer (node; right click properties)





EHR systemEvent

Listener

Notifications

Actions


definitions)

Data Inquiry

Workflow Execution

Engine(Shark)

EHR System layer:



Reports

Data Warehouse



Data Visualization


Event Data Model

Ontology Model







Together Workflow Engine (Community edition)





EHR systemEvent

Listener

Notifications

Actions


definitions)

Data Inquiry

Workflow Execution

Engine(Shark)

EHR System layer:



Reports

Data Warehouse



Data Visualization


Event Data Model

Ontology Model







Example of retrospective execution

0.9% of patients with high

risk patients with

subsequent MI


Example of execution

Process mining Arrive at process definition by examining

retrospective data Compare processes from different settings Even younger field then workflow technology

Examples Prior work on Chronic Kidney Disease Current work on Clinical Research Protocols

www.processmining.org

http://www.processmining.org/

Chronic kidney disease (mining)

Benefits and challenges of using workflow engines in healthcare


Decision support (wish list) Understandable format

Graphical - task network paradigm – SAGE,GLIF, ProForma Textual (Arden Syntax), GELLO

Several implementations (compare to databases, BI, prog. IDE) Choice of editor Choice of execution engine

Handles well logic and context Logic

>66yo, history of fracture -> recommend bone mineral density scan Context

Display when and to whom, Resource modeling (nurse role, current nurse, hospitalist; in-house machine vs. referral) Modeling the opportunities for decision support (SAGE)

Explanation function Authoring, Reviewing

Not-developed-here syndrome, owner of logic vs. happy user vs. victim Knowledge exchange (business model)

Benefits of WT Ability to customize processes within an EHR Graphical representation

author from scratch, modify slightly the logic, modify to fit local context Programmer vs. analyst (hard coded processes) Understandability (clinician, explanation function) (flowcharts)

Auditing / change management EHR with APIs/links (add task, track tasks, see roles, provide event

triggers) Observation mode/ Intervention mode

Ability to share workflows Sharing on a graphical level only (no “code”)

Cross industry technology

Challenges of WT Process administration

Owner of a process Medical home, infections control nurse, consumer (PHR)

Physician autonomy Cook book medicine

Workflow technology enabled EHR systems Add-on integration (TheraDoc, iList) Direct integration (Siemens Soarian, % of hardcoded steps)

Graphical modeling Several standards (BPMN, YAWL) Workflow patterns Handling exceptions (model in a flowchart), swim-lanes (roles) Very complex processes (code + HTML docs, just flowchart, simplified

flowchart)

Conclusion 4 objectives

Workflow engine, flowcharts (guidelines), current use in healthcare, benefits and challenges

WT is an emerging technology Compared with databases, started in 1999 Not everyone means executable process definitions (within a workflow

engine) when they say “workflow” Cross industry standard to watch/use

vs. healthcare specific guidelines engines

Thank you for attention. [email protected]

http://healthcareworkflow.wordpress.com

mailto:[email protected]

http://healthcareworkflow.wordpress.com/

Extra slides


Evaluation 18 users, 2 parts, qualitative 9 task questions (T1-T9)

T5: Find all patients who had at least 2 creatinine lab results flagged too high (but they must be at least 180 days apart).

5 choice questions (C1-C5) Solved problem 1 -> extended problem 2

+ choices A), B), C) Results

Statistically significant difference in scores (paired t-test, 2-sided)

RG: 11.1± 1.8 vs. SQL: 6.3 ± 2.1 (p<<0.0001)


Evaluation (SQL and RG) Which technology do you prefer (SQL or RG)?

94% of participants preferred RG

and why do you prefer it? 1. easy to learn/use/understand 2. temporal modeling capabilities 3. more intuitive/natural/logical

Disadvantages of SQL? 1. must know exact syntax/be expert 2. difficult to use 3. insufficient support for temporal criteria

Disadvantages of RG? 1. need to know function of various apps and new terminology 2. none 3. hard to understand what data user gets back 4. can be slow for queries involving a larger population

Hancock (2004), Lacey (2004)


Evaluation General important features: (general question)

Highest ranking features Intuitive modeling paradigm Short training time

Technology acceptance: (RG only) Unified Theory of Acceptance and Use of Technology (UTAUT) Performance expectancy, Effort Expectancy, Behavioral

intention Favorable scores predicting high actual use

CDS architecture evaluation (Wright, 2009)

Venkatesh (2003), Hulse (2006), Hu (1999), Chismar (2002), Wilson (2004), Hoof (2003)

healthcare use of workflow engine technology with emphasis on data analysis and decision support

Health & Medicine

workflow technology

vojtech huser

phd healthcareworkflow

set

wordpress

healthcare