SLIDE 1
PAGE 1 23-9-2010
Introduction
This talk:
- Applicability of Process Mining in the healthcare
domain
- Challenges
- > Results from applying Process Mining in the AMC
Process Mining in Healthcare Ronny Mans Introduction This talk: - - PowerPoint PPT Presentation
Process Mining in Healthcare Ronny Mans Introduction This talk: - - PowerPoint PPT Presentation
Process Mining in Healthcare Ronny Mans Introduction This talk: Applicability of Process Mining in the healthcare domain Challenges -> Results from applying Process Mining in the AMC 23-9-2010 PAGE 1 Overview Introduction
SLIDE 2
SLIDE 3
PAGE 2 23-9-2010
Overview
- Introduction
- AMC
- AMC case study
- DBCs gynecological oncology
- DBCs GO + radiotherapy + chemotherapy
- Future work
- Conclusion
- Questions / Discussion
SLIDE 4
PAGE 3 23-9-2010
Academic Medical Center (AMC)
- University hospital, Amsterdam
- 1002 beds
- 25.000 patients admitted
- 35.000 day admissions
- 350.000 outpatient clinic visits
- Patients
- Own region
- Outside region
SLIDE 5
PAGE 4 23-9-2010
AMC
AND
- Healthcare processes are highly variable
- > Not known what happens in a healthcare process
- Process improvement projects
- Time consuming to collect data
SLIDE 6
PAGE 5 23-9-2010
Data?
- Each department has their system
- Integration often difficult
- Payment system contains all events
- DBC: Diagnosis Treatment Combination
- Episode of care
- All the care steps (inpatient, outpatient, day, and after
care) that may be delivered for a specific medical problem or condition.
- Each service delivered to a patient is linked to one DBC
Introduction
SLIDE 7
PAGE 6 23-9-2010
Data?
DBC code: Cervical cancer Patient: Sue
- Visit outpatient clinic
- Lab test
- X-ray
- CT
- MRI
DBC code: Ovarian cancer Patient: Rose
- Visit outpatient clinic
- Pathology
- Lab test
- MRI
SLIDE 8
PAGE 7 23-9-2010
Overview
- Introduction
- AMC
- AMC case study
- DBCs gynecological oncology
- DBCs GO + radiotherapy + chemotherapy
- Conclusion
- Future work
- Questions / Discussion
SLIDE 9
PAGE 8 23-9-2010
Case – AMC log
- 627 Gynaecological oncology patients
- 376 Events
- 24331
Audit trail entries
- Gynaecology, Nursing wards, Radiology, …
- All care steps for GO patients
Case
SLIDE 10
PAGE 9 23-9-2010
Data
Only day timestamps
SLIDE 11
PAGE 10 23-9-2010
Data
Single lab tests
SLIDE 12
PAGE 11 23-9-2010
Data Visit to one department
SLIDE 13
PAGE 12 23-9-2010
Log Preprocessing
Filtering: Getting the right abstraction
- Remapping
- Aggregation
SLIDE 14
PAGE 13 23-9-2010
Filtering
- Representative R
- Keep R
- Remove others
Remap Element Log
SLIDE 15
PAGE 14 23-9-2010
Example
Acceptance Lab Test Na Ka Xray Hg Acceptance Lab Test Na O2-sat Xray Hg Remap Element Log
SLIDE 16
PAGE 15 23-9-2010
Filtering
lab representative lab representative lab representative lab representative lab representative lab representative lab representative Repetition to Activity Remap Element Log
- No Representative R
- Define R
- Remap all to R
- Aggregate
repetitions in trace
SLIDE 17
PAGE 16 23-9-2010
Example
Na Ka Xray Hg Na O2-sat Xray Hg lab R lab R lab R lab R lab R lab R Remap Element Log Repetition to Activity
SLIDE 18
PAGE 17 23-9-2010
Result after filtering
Case
SLIDE 19
PAGE 18 23-9-2010
Log Preprocessing (2)
Clustering: Grouping similar behaviour
- Trace clustering plug-in
Case
SLIDE 20
PAGE 19 23-9-2010
Case – clusters
Case
SLIDE 21
PAGE 20 23-9-2010
Case – biggest cluster
Case
SLIDE 22
PAGE 21 23-9-2010
Case – Social Network
Case
Interaction with dietics department
SLIDE 23
PAGE 22 23-9-2010
Case – Basic Performance Analysis Plug-in
Case
SLIDE 24
PAGE 23 23-9-2010
Results so far
- Complex hospital logs can be mined
- Log pre-processing can be used to derive
understandable models
- Filtering – for getting the right abstraction
- Clustering – for analysing common behaviour
Results
SLIDE 25
PAGE 24 23-9-2010
Overview
- Motivation
- AMC case study
- DBCs gynecological oncology
- DBCs GO + radiotherapy + chemotherapy
- Future work
- Conclusion
- Questions
SLIDE 26
PAGE 25 23-9-2010
AMC data
- DBC: Diagnosis Treatment Combination
- Episode of care
- All the care steps (inpatient, outpatient, day, and after care)
that may be delivered for a specific medical problem or condition.
- Each service delivered to a patient is linked to one DBC
- Whole care path of gynecological oncology
- Gynecological oncology
- Radiotherapy
- Internal medicine
SLIDE 27
PAGE 26 23-9-2010
DBCs
DBC code: GO Cervical cancer Patient: Sue
- Visit outpatient clinic
- Lab test
- X-ray
- CT
- MRI
DBC code: GO Ovarian cancer Patient: Rose
- Visit outpatient clinic
- Pathology
- Lab test
- MRI
DBC code: IM Cervical cancer Patient: Sue
- Visit outpatient clinic
- Lab test
- Chemo therapy
- Lab test
- Chemo therapy
DBC code: gynaecological tumors Patient: Sue
- Visit outpatient clinic
- Radiotherapy
- MRI
- Radiotherapy
- MRI
SLIDE 28
PAGE 27 23-9-2010
AMC data
- Resulting log:
- 682
Gynaecological oncology patients
- 43615
Audit trail entries
- Gynaecology, Nursing wards, Radiology, Radiotherapy,
…
SLIDE 29
PAGE 28 23-9-2010
Diagnostic + therapeutic process
- Visual insights
- Process Mining
SLIDE 30
PAGE 29 23-9-2010
Visual insights
Visit to the
- utpatient clinic
Radiology Lab
SLIDE 31
PAGE 30 23-9-2010
Visual insights
SLIDE 32
PAGE 31 23-9-2010
Visual insights
31
First surgery
SLIDE 33
PAGE 32 23-9-2010
Process Mining
- Split log in two parts
- Diagnostic part
- Therapeutic part
SLIDE 34
PAGE 33 23-9-2010
Diagnostic process
SLIDE 35
PAGE 34 23-9-2010
Diagnostic process (HM)
ArtificialStartTask (complete) 100 Lab (complete) 215 0,978 46 OC Gyn Onc (complete) 148 0,974 54 Pathology (complete) 60 0,944 32 0,983 69 Pharmacy Lab (complete) 76 0,987 76 Radiology (complete) 161 0,947 64 Nursing Ward H5Z (complete) 263 0,809 80 0,944 47 0,917 95 0,947 43 0,984 167 Operating Rooms (complete) 100 0,989 96 ArtificialEndTask (complete) 100 0,99 100 0,982 67
- Focus on most
important events
- Fitness: 0,7
- Performance
related data
SLIDE 36
PAGE 35 23-9-2010
Overview
- Introduction
- AMC
- AMC case study
- Future work
- Conclusion
- Questions / Discussion
SLIDE 37
PAGE 36 23-9-2010
Research
Goal: Obtain understandable results for the process analyst and medical specialist (end-user)
- 1. Capturing flexible processes
- 2. Presentation of process related information
SLIDE 38
PAGE 37 23-9-2010
Flexible processes (Declarative PLs)
SLIDE 39
PAGE 38 23-9-2010
Flexible processes (Declarative PLs)
SLIDE 40
PAGE 39 23-9-2010
Flexible processes
Develop mining techniques for less procedural and declarative languages
SLIDE 41
PAGE 40 23-9-2010
Process related information
SLIDE 42
PAGE 41 23-9-2010
Process related information
- Good maps?
- Navigation by
PowerPoints?
- Traffic information?
- Where is the next fuel
station?
- Who is in charge?
- Seamless zoom?
- Customizable views?
- When will the
destination be reached?
SLIDE 43
PAGE 42 23-9-2010
Overview
- Introduction
- AMC
- AMC case study
- Future work
- Conclusion
- Questions / Discussion
SLIDE 44
PAGE 43 23-9-2010
Conclusions
- Insights into healthcare processes
- Remaining challenges
- Capturing flexible processes
- Presentation of process related information
- Only day timestamps
Results
SLIDE 45
PAGE 44 23-9-2010