© 2017 Health Catalyst
Proprietary and Confidential
July 12, 2017
Demystifying Text Analytics
and NLP in Healthcare
Dr. Carolyn Simpkins
Mike Dow
Agenda • What is text analytics and NLP?
• Clinical scenarios
• Hands-on NLP algorithm creation
© 2017 Health Catalyst
Proprietary and Confidential
“The process of deriving high quality information
from text, by applying natural language processing
(NLP) to transform text into data for analysis.”
What is text analytics?
3
© 2017 Health Catalyst
Proprietary and Confidential
Why text analytics matters in healthcare
Procedure and op notes
Progress notes
Chief complaint
History of present illness
Physical exam
Assessment and Plan
Cardiology reports: echo,
stress test, EKG
Radiology reports
Pathology reports
Discharge summaries
Consults
Complications,
incidental findings
during procedures
Working
diagnoses
Character of chest pain
Lung exam: Wheezing?
Crackles?
Key findings on daily
post-op wound checks
Duration of
presenting
symptoms
Ejection fraction, wall
mobility
Baseline EKG: ST
segment depression, in
which leads? Old Q
waves?
Social hx notes for
discharge planning Prior end of life
discussions
© 2016 Health Catalyst
Proprietary and Confidential
One clinical scenario:
managing heart failure
© 2017 Health Catalyst
Proprietary and Confidential
Imagine you’re a cardiologist and a patient with chronic congestive
heart failure walks into your office for an appointment. What would you
ideally find in a quick review of the chart as you’re walking into the
room?
• Current med list, current weight, current BP
• Key events from most recent hospitalization eg new cardiac events,
discharge weight, new echo report with EF and wall motion, reason for
decompensation
• Latest ejection fraction and perhaps a graph of trend in EF over time
• Current symptoms or complaints: weight gain, shortness of breath,
peripheral edema
Hidden data: Clinician view of Heart Failure
6
© 2017 Health Catalyst
Proprietary and Confidential
Imagine you’re a cardiologist and a patient with chronic congestive
heart failure walks into your office for an appointment. What would you
ideally find in a quick review of the chart as you’re walking into the
room?
• Current med list, current weight, current BP
• Key events from most recent hospitalization eg new cardiac events,
discharge weight, new echo report with EF and wall motion, reason for
decompensation
• Latest ejection fraction and perhaps a graph of trend in EF over time
• Current symptoms or complaints: weight gain, shortness of breath,
peripheral edema
Discrete data vs Buried in text notes vs Mix
7
© 2017 Health Catalyst
Proprietary and Confidential
Now imagine you’re the clinical director of a heart failure program and you’re
looking to assess performance in managing heart failure across your
practice or health systems population and identify opportunities to target
improvement efforts to a subpopulation
How are subgroups of patients with heart failure doing?
• ejection fraction
• ACC heart failure stage
• hospital days
• mortality
compared to practice patterns
• medications, procedures, rehab
• practices, individual providers
Hidden data: Clinician view of Heart Failure
8
© 2017 Health Catalyst
Proprietary and Confidential
Now imagine you’re the clinical director of a heart failure program and you’re
looking to assess performance in managing heart failure across your
practice or health systems population and identify opportunities to target
improvement efforts to a subpopulation
How are subgroups of patients with heart failure doing?
• ejection fraction
• ACC heart failure stage
• hospital days
• mortality
compared to practice patterns
• medications, procedures, rehab
• practices, individual providers
Discrete data vs Buried in text notes vs Mix
9
© 2016 Health Catalyst
Proprietary and Confidential
Unearthing buried data
with text analytics
© 2017 Health Catalyst
Proprietary and Confidential
Ejection fraction – heart failure
Live walk-through
with EF…
© 2017 Health Catalyst
Proprietary and Confidential
What should a Text Analytics solution for healthcare look like?
13
Search engine
Discovery & Synonyms
Context
Extraction
Validation
NIDDM ~ diabetes
! Patient denies history of diabetes
! Family history of diabetes
Ejection fraction of 60 – 70%
© 2017 Health Catalyst
Proprietary and Confidential
Using the data
15
© 2016 Health Catalyst
Proprietary and Confidential
Wide clinical application
© 2017 Health Catalyst
Proprietary and Confidential
Problem:
• Must tailor oncology treatment plan to specifics of patient’s
condition
Challenges:
• High stakes complex decision making
• Requires numerous pieces of information not coded in
discrete data fields
• Constantly evolving treatment algorithms require tracking outcomes
against matrix of patient and treatment decisions
Use case: oncology treatment program
17
© 2017 Health Catalyst
Proprietary and Confidential
Brain tumor program intake form
18
Extract key MRI findings from text report
Search for key Sx (seizures, headache) & neuro
exam findings (deficits) in text notes
Search text notes for path diagnosis, stage
© 2017 Health Catalyst
Proprietary and Confidential
Problem:
• Must assess the risk of cardiac events perioperatively for
noncardiac surgery
Challenges:
• Risk assessment algorithm is complex
• Requires numerous pieces of information not coded in
discrete data fields
• Overestimate of risk results in excessive application of expensive
and exposure-laden nuclear stress imaging, as well as
unwarranted delay of the surgery
Use case: pre-op cardiovascular risk
19
© 2016 Health Catalyst
Proprietary and Confidential20
Current cardiac chest pain (Sx), pathological Q waves (ECG)
Pulmonary edema (CXR), rales or S3 (exam)
History of TIA symptoms
© 2017 Health Catalyst
Proprietary and Confidential
Using the data
• Compare outcomes of patients with similar MRI findings
undergoing different treatments
• Analyze accuracy of pathological staging vs clinical staging vs
survival
• Identify patients with specific symptom profiles or neurological
exam findings newly recognized to track with better success in new
treatment regimen
• Rapidly find complete cardiac risk elements from numerous text
sources to accelerate pre-operative evaluation and adjust peri-
operative treatment
© 2017 Health Catalyst
Proprietary and Confidential
What should a Text Analytics solution for healthcare look like?
22
Search engine
Discovery & Synonyms
Context
Extraction
Validation
NIDDM ~ diabetes
! Patient denies history of diabetes
! Family history of diabetes
Ejection fraction of 60 – 70%
Thank you
© 2017 Health Catalyst
Proprietary and Confidential
Healthcare Analytics Summit 17
Summit highlights
Industry Leading Keynote Speakers
We’ll hear from well-known healthcare visionaries. We’ll also
hear from two C-level executives leading large healthcare
organizations.
CME Accreditation For Clinicians
HAS 17 will again qualify as a continuing medical education
(CME) activity.
30 Educational, Case Study, and Technical
Sessions
We have the most comprehensive set of breakout sessions of
any analytics summit. Our primary breakout session focus is
giving you detailed, practical “how to” learning examples
combined with question and opportunities.
The Analytics Walkabout
Back by popular demand, the Analytics Walkabout will feature
24 new projects highlighting a variety of additional clinical,
financial, operational, and workflow analytics and outcomes
improvement successes.
Analytics-driven, Hands-on Engagement for
Teams and Individuals
Analytics will continue to flow through the three-day summit
touching every aspect of the agenda.
Networking and Fun
We’ll provide some new innovative analytics-driven
opportunities to network while keeping our popular fun run and
walk opportunities and dinner on the down.
Sept. 12-14, 2017
Grand America Hotel
Salt Lake City, UT

Demystifying Text Analytics and NLP in Healthcare

  • 1.
    © 2017 HealthCatalyst Proprietary and Confidential July 12, 2017 Demystifying Text Analytics and NLP in Healthcare Dr. Carolyn Simpkins Mike Dow
  • 2.
    Agenda • Whatis text analytics and NLP? • Clinical scenarios • Hands-on NLP algorithm creation
  • 3.
    © 2017 HealthCatalyst Proprietary and Confidential “The process of deriving high quality information from text, by applying natural language processing (NLP) to transform text into data for analysis.” What is text analytics? 3
  • 4.
    © 2017 HealthCatalyst Proprietary and Confidential Why text analytics matters in healthcare Procedure and op notes Progress notes Chief complaint History of present illness Physical exam Assessment and Plan Cardiology reports: echo, stress test, EKG Radiology reports Pathology reports Discharge summaries Consults Complications, incidental findings during procedures Working diagnoses Character of chest pain Lung exam: Wheezing? Crackles? Key findings on daily post-op wound checks Duration of presenting symptoms Ejection fraction, wall mobility Baseline EKG: ST segment depression, in which leads? Old Q waves? Social hx notes for discharge planning Prior end of life discussions
  • 5.
    © 2016 HealthCatalyst Proprietary and Confidential One clinical scenario: managing heart failure
  • 6.
    © 2017 HealthCatalyst Proprietary and Confidential Imagine you’re a cardiologist and a patient with chronic congestive heart failure walks into your office for an appointment. What would you ideally find in a quick review of the chart as you’re walking into the room? • Current med list, current weight, current BP • Key events from most recent hospitalization eg new cardiac events, discharge weight, new echo report with EF and wall motion, reason for decompensation • Latest ejection fraction and perhaps a graph of trend in EF over time • Current symptoms or complaints: weight gain, shortness of breath, peripheral edema Hidden data: Clinician view of Heart Failure 6
  • 7.
    © 2017 HealthCatalyst Proprietary and Confidential Imagine you’re a cardiologist and a patient with chronic congestive heart failure walks into your office for an appointment. What would you ideally find in a quick review of the chart as you’re walking into the room? • Current med list, current weight, current BP • Key events from most recent hospitalization eg new cardiac events, discharge weight, new echo report with EF and wall motion, reason for decompensation • Latest ejection fraction and perhaps a graph of trend in EF over time • Current symptoms or complaints: weight gain, shortness of breath, peripheral edema Discrete data vs Buried in text notes vs Mix 7
  • 8.
    © 2017 HealthCatalyst Proprietary and Confidential Now imagine you’re the clinical director of a heart failure program and you’re looking to assess performance in managing heart failure across your practice or health systems population and identify opportunities to target improvement efforts to a subpopulation How are subgroups of patients with heart failure doing? • ejection fraction • ACC heart failure stage • hospital days • mortality compared to practice patterns • medications, procedures, rehab • practices, individual providers Hidden data: Clinician view of Heart Failure 8
  • 9.
    © 2017 HealthCatalyst Proprietary and Confidential Now imagine you’re the clinical director of a heart failure program and you’re looking to assess performance in managing heart failure across your practice or health systems population and identify opportunities to target improvement efforts to a subpopulation How are subgroups of patients with heart failure doing? • ejection fraction • ACC heart failure stage • hospital days • mortality compared to practice patterns • medications, procedures, rehab • practices, individual providers Discrete data vs Buried in text notes vs Mix 9
  • 10.
    © 2016 HealthCatalyst Proprietary and Confidential Unearthing buried data with text analytics
  • 11.
    © 2017 HealthCatalyst Proprietary and Confidential Ejection fraction – heart failure
  • 12.
  • 13.
    © 2017 HealthCatalyst Proprietary and Confidential What should a Text Analytics solution for healthcare look like? 13 Search engine Discovery & Synonyms Context Extraction Validation NIDDM ~ diabetes ! Patient denies history of diabetes ! Family history of diabetes Ejection fraction of 60 – 70%
  • 15.
    © 2017 HealthCatalyst Proprietary and Confidential Using the data 15
  • 16.
    © 2016 HealthCatalyst Proprietary and Confidential Wide clinical application
  • 17.
    © 2017 HealthCatalyst Proprietary and Confidential Problem: • Must tailor oncology treatment plan to specifics of patient’s condition Challenges: • High stakes complex decision making • Requires numerous pieces of information not coded in discrete data fields • Constantly evolving treatment algorithms require tracking outcomes against matrix of patient and treatment decisions Use case: oncology treatment program 17
  • 18.
    © 2017 HealthCatalyst Proprietary and Confidential Brain tumor program intake form 18 Extract key MRI findings from text report Search for key Sx (seizures, headache) & neuro exam findings (deficits) in text notes Search text notes for path diagnosis, stage
  • 19.
    © 2017 HealthCatalyst Proprietary and Confidential Problem: • Must assess the risk of cardiac events perioperatively for noncardiac surgery Challenges: • Risk assessment algorithm is complex • Requires numerous pieces of information not coded in discrete data fields • Overestimate of risk results in excessive application of expensive and exposure-laden nuclear stress imaging, as well as unwarranted delay of the surgery Use case: pre-op cardiovascular risk 19
  • 20.
    © 2016 HealthCatalyst Proprietary and Confidential20 Current cardiac chest pain (Sx), pathological Q waves (ECG) Pulmonary edema (CXR), rales or S3 (exam) History of TIA symptoms
  • 21.
    © 2017 HealthCatalyst Proprietary and Confidential Using the data • Compare outcomes of patients with similar MRI findings undergoing different treatments • Analyze accuracy of pathological staging vs clinical staging vs survival • Identify patients with specific symptom profiles or neurological exam findings newly recognized to track with better success in new treatment regimen • Rapidly find complete cardiac risk elements from numerous text sources to accelerate pre-operative evaluation and adjust peri- operative treatment
  • 22.
    © 2017 HealthCatalyst Proprietary and Confidential What should a Text Analytics solution for healthcare look like? 22 Search engine Discovery & Synonyms Context Extraction Validation NIDDM ~ diabetes ! Patient denies history of diabetes ! Family history of diabetes Ejection fraction of 60 – 70%
  • 23.
  • 24.
    © 2017 HealthCatalyst Proprietary and Confidential Healthcare Analytics Summit 17 Summit highlights Industry Leading Keynote Speakers We’ll hear from well-known healthcare visionaries. We’ll also hear from two C-level executives leading large healthcare organizations. CME Accreditation For Clinicians HAS 17 will again qualify as a continuing medical education (CME) activity. 30 Educational, Case Study, and Technical Sessions We have the most comprehensive set of breakout sessions of any analytics summit. Our primary breakout session focus is giving you detailed, practical “how to” learning examples combined with question and opportunities. The Analytics Walkabout Back by popular demand, the Analytics Walkabout will feature 24 new projects highlighting a variety of additional clinical, financial, operational, and workflow analytics and outcomes improvement successes. Analytics-driven, Hands-on Engagement for Teams and Individuals Analytics will continue to flow through the three-day summit touching every aspect of the agenda. Networking and Fun We’ll provide some new innovative analytics-driven opportunities to network while keeping our popular fun run and walk opportunities and dinner on the down. Sept. 12-14, 2017 Grand America Hotel Salt Lake City, UT