Ai: Solving Healthcare Challenges
1 like230 views
The document discusses the application of AI in addressing challenges faced by healthcare organizations, focusing on making patient data actionable. It highlights the importance of data consolidation, standardization, and enrichment, as well as the potential benefits of AI in enhancing productivity and improving clinical outcomes. Key opportunities for AI integration include reducing readmission risks, enabling differential diagnoses, and supporting clinical decision-making.
![Leverage cNLP
to Make Your
Data Actionable
Admission Date: 2015-12-25
New Patient
The patient is a 59-year-old female, who was
experiencing mild chest pain on the left side. The
patient states that she slipped on a newly waxed
floor and fell on her tailbone and low back region.
She would like to lose some weight through
dieting. Complains of dry mouth even though
drinking plenty of water.
The pt is allergic to Morphine Sulfate.
No fh of colorectal CA. Last colonoscopy in Kansas
City, May 2012.
Past Medical History:
• DM2
• Asthma
Shx:
Pt has a remote 10 pack-year smoking history.
Occasional ETOH drink.
Height 173 cm, weight 78.8 kg, blood pressure
114/80
Lab data:
2014-6-10 03:20PM GlycoHgB A1c - 7.0*
2014-6-10 03:20PM Creatinine, 24 hour urine
g/24 h-1.2
Patient Problems:
29857009 (R07.9) – Chest Pain
Severity: Mild
Laterality: Left
Temporal: Present
Subject: Patient
87715008 – Dry Mouth
Temporal: Present
Subject: Patient
44054006 (E11.9) - Type 2 diabetes mellitus
Temporal: Past
Subject: Patient
195967001 (J45.909) – Asthma
Temporal: Past
Subject: Patient
Family History:
363406005 (C18.9) - Malignant tumor of colon
Subject: Family member
Negated: Yes
Social History:
77176002 (F17.200) – Smoker
Subject: Patient
219006 (Z72.89) – Drinks alcohol
Subject: Patient
Labs:
41995-2 – Hemoglobin A1c [Mass/volume] in Blood
Value: 7.0
Date: 2014-06-10
14684-5 - Creatinine [Moles/time] in 24 hour Urine
Value: 1.2
Date: 2014-06-10
Demographic:
397669002 - Age
Value: 59
263495000 – Gender
Value: Female
Vitals:
50373000 - Height
Value: 173 cm
726527001 - Weight
Value: 78.8kg
75367002 - Blood pressure
Value: 114/80
Allergy:
RX 30236 - Morphine Sulfate
Procedure:
73761001 – Colonoscopy
Date: 5/2012
Admission Date: 2015-12-25
New Patient
The patient is a 59-year-old female, who was
experiencing mild chest pain on the left side. The
patient states that she slipped on a newly waxed
floor and fell on her tailbone and low back region.
She would like to lose some weight through
dieting. Complains of dry mouth even though
drinking plenty of water.
The pt is allergic to Morphine Sulfate.
No fh of colorectal CA. Last colonoscopy in Kansas
City, May 2012.
Past Medical History:
• DM2
• Asthma
Shx:
Pt has a remote 10 pack-year smoking history.
Occasional ETOH drink.
Height 173 cm, weight 78.8 kg, blood pressure
114/80
Lab data:
2014-6-10 03:20PM GlycoHgB A1c - 7.0*
2014-6-10 03:20PM Creatinine, 24 hour urine
g/24 h-1.2](https://image.slidesharecdn.com/aisolveshealthcarechallenges-190809132344/85/Ai-Solving-Healthcare-Challenges-16-320.jpg)
![Reviewing Patient Risk
UNSTRUCTURED TEXT
Patient Problems:
44054006 (E11.9) - Type 2 diabetes mellitus
Temporal: Past
Subject: Patient
[HCC19] Diabetes without Complication
DATA ELEMENTS
Admission Date: 2015-12-25
New Patient
The patient is a 59-year-old female, who was
experiencing mild chest pain on the left side. The
patient states that she slipped on a newly waxed
floor and fell on her tailbone and low back region.
She would like to lose some weight through dieting.
Complains of dry mouth even though drinking plenty
of water.
The pt is allergic to Morphine Sulfate.
No fh of colorectal CA. Last colonoscopy in Kansas
City, May 2012.
Past Medical History:
• DM2
• Asthma
Shx:
Pt has a remote 10 pack-year smoking history.
Occasional ETOH drink.
Height 173 cm, weight 78.8 kg, blood pressure
114/80
Lab data:
2014-6-10 03:20PM GlycoHgB A1c - 7.0*
2014-6-10 03:20PM Creatinine, 24 hour urine g/24 h-
1.2
Patient Problems:
87715008 – Dry Mouth
Temporal: Present
Subject: Patient
(E11.638) - Type 2 diabetes mellitus with other
oral complications
[HCC18] Diabetes with Chronic Complications](https://image.slidesharecdn.com/aisolveshealthcarechallenges-190809132344/85/Ai-Solving-Healthcare-Challenges-17-320.jpg)
Ai: Solving Healthcare Challenges
- 1. Leveraging AI to Solve Common Healthcare Challenges: Hear from the Experts SARAH BRYAN | Wolters Kluwer CHRIS FUNK | Wolters Kluwer JOHN LANGTON | Wolters Kluwer
- 2. Today’s Speakers 2 SARAH BRYAN Director of Product Management, Health Language Wolters Kluwer JOHN LANGTON, Ph.D Director of Applied Sciences Wolters Kluwer CHRIS FUNK, Ph.D Sr. Medical Informaticist, Health Language Wolters Kluwer
- 3. 3 Agenda 1. Challenges healthcare organizations face in making patient data actionable 2. How AI can address these challenges by optimizing staff productivity and efficiency 3. Looking to the future: examples of the value AI can bring to the healthcare industry and how to get started
- 5. 5 BILLING DATA ICD-10, CPT®, HCPCS, DRGs EMERGING DATA Telehealth, Genomics, Lifestyle, Social Determinants of Health CLINICAL DATA EHR Data, Labs, Drugs, Imaging, Behavioral Health, Pathology, Clinical Notes MAXIMIZE VALUE BY LEVERAGING ALL DATA SOURCES Healthcare (Big) Data
- 6. Healthcare Data is Disparate 6 Cerner EHR Radiology Information Systems (RIS) Pharmacy Information System (PIS) Lab Information Systems (LIS) Epic EHR Ambulatory EHR eClinical Works EHR HEALTH PROVIDERS CLINICS HOMES Telehealth
- 7. Types of Healthcare Data 7 STRUCTURED CLAIMS DATA CPT®, ICD-10, MS-DRG, HCPCS UNSTRUCTURED FREE TEXT Clinician Notes, Clinical Documents, Pathology Reports, Patient Charts DISCRETE CLINICAL DATA EHR Data, Local Labs, Medications SEMI-STRUCTURED OTHER “?” Social Determinants of Health, Genomics, Patient-generated Data, Telehealth
- 8. Step 1: Consolidate, Standardize & Enrich Your Data 8 STANDARDIZE Normalize your data to interoperability standards such as SNOMED, RxNorm, LOINC, and HL7 CONSOLIDATE Consolidate your patient data from across the various data sources incorporating all data types ENRICH Enrich your data by extracting and normalizing clinical concepts locked in unstructured notes and categorizing data into clinical concepts
- 9. Survey Question #1 9 What areas do you feel most excited about when thinking about adopting AI technology within your organization? Select all that apply A. Automating chart review for quality measures, medical necessity review, etc. B. Categorizing patient risk for appropriate reimbursement in capitated payment models C. Enhancing diagnostics, enabling differential diagnosis D. Discovering correlations with predictive analytics E. Automating administrative functions, such as scheduling, follow-up care, etc. F. Other
- 10. 10 Agenda 1. Challenges healthcare organizations face in making patient data actionable 2. How AI can address these challenges by optimizing staff productivity and efficiency 3. Looking to the future: examples of the value AI can bring to the healthcare industry and how to get started
- 11. Artificial Intelligence Continuum 11 WEAK STRONG Supports human decisions Examples: Natural language processing, computer vision Automates simple tasks Example: Lane correction systems in cars Makes autonomous decisions Example: Self-driving cars Mimics human behavior Examples: HAL 9000 (2001: A Space Odyssey) Data (Star Trek: TNG) COGNITIVE INTELLIGENCEAUTONOMOUS INTELLIGENCEAUGMENTED INTELLIGENCEASSISTED INTELLIGENCE HEALTHCARE IS PRIMARILY HERE
- 12. Artificial Intelligence, Machine Learning, and Natural Language Processing 12 ARTIFICIAL INTELLIGENCE EXPERT SYSTEMS PLANNING ROBOTICS Speech to Text Text to Speech SPEECH Image Recognition Machine Vision VISIONNATURAL LANGUAGE PROCESSING Text Generation Text Analytics Question Answering Context Extraction Classification Machine Translation MACHINE LEARNING Deep Learning Unsupervised Supervised
- 13. Artificial Intelligence Applied Across Healthcare 13 CLINICIAN WORKFLOW INTEROPERABILITY FINANCIAL MANAGEMENT REPORTING & ANALYTICSKNOWLEDGE MANAGEMENT CLINICAL DECISION SUPPORT
- 14. 80% of Patient Data is Locked in Unstructured Text 14 NOTE Admission Date: 2015-12-25 New Patient The patient is a 59-year-old female, who was experiencing mild chest pain on the left side. The patient states that she slipped on a newly waxed floor and fell on her tailbone and low back region. She would like to lose some weight through dieting. Complains of dry mouth even though drinking plenty of water. The pt is allergic to Morphine Sulfate. No fh of colorectal CA. Last colonoscopy in Kansas City, May 2012. Past Medical History: • DM2 • Asthma Shx: Pt has a remote 10 pack-year smoking history. Occasional ETOH drink. Height 173 cm, weight 78.8 kg, blood pressure 114/80 Lab data: 2014-6-10 03:20PM GlycoHgB A1c - 7.0* 2014-6-10 03:20PM Creatinine, 24 hour urine g/24 h-1.2
- 15. 15 Current Process for Extracting Valuable Patient Information Time Consuming • Manually browsing multiple documents • Searching for key words • Pasting key findings into separate forms Expensive • Review process is commonly done by subject matter experts Risk of Inaccuracy • Hard for humans to consistently maintain high level of accuracy due to volume and demand of review
- 16. Leverage cNLP to Make Your Data Actionable Admission Date: 2015-12-25 New Patient The patient is a 59-year-old female, who was experiencing mild chest pain on the left side. The patient states that she slipped on a newly waxed floor and fell on her tailbone and low back region. She would like to lose some weight through dieting. Complains of dry mouth even though drinking plenty of water. The pt is allergic to Morphine Sulfate. No fh of colorectal CA. Last colonoscopy in Kansas City, May 2012. Past Medical History: • DM2 • Asthma Shx: Pt has a remote 10 pack-year smoking history. Occasional ETOH drink. Height 173 cm, weight 78.8 kg, blood pressure 114/80 Lab data: 2014-6-10 03:20PM GlycoHgB A1c - 7.0* 2014-6-10 03:20PM Creatinine, 24 hour urine g/24 h-1.2 Patient Problems: 29857009 (R07.9) – Chest Pain Severity: Mild Laterality: Left Temporal: Present Subject: Patient 87715008 – Dry Mouth Temporal: Present Subject: Patient 44054006 (E11.9) - Type 2 diabetes mellitus Temporal: Past Subject: Patient 195967001 (J45.909) – Asthma Temporal: Past Subject: Patient Family History: 363406005 (C18.9) - Malignant tumor of colon Subject: Family member Negated: Yes Social History: 77176002 (F17.200) – Smoker Subject: Patient 219006 (Z72.89) – Drinks alcohol Subject: Patient Labs: 41995-2 – Hemoglobin A1c [Mass/volume] in Blood Value: 7.0 Date: 2014-06-10 14684-5 - Creatinine [Moles/time] in 24 hour Urine Value: 1.2 Date: 2014-06-10 Demographic: 397669002 - Age Value: 59 263495000 – Gender Value: Female Vitals: 50373000 - Height Value: 173 cm 726527001 - Weight Value: 78.8kg 75367002 - Blood pressure Value: 114/80 Allergy: RX 30236 - Morphine Sulfate Procedure: 73761001 – Colonoscopy Date: 5/2012 Admission Date: 2015-12-25 New Patient The patient is a 59-year-old female, who was experiencing mild chest pain on the left side. The patient states that she slipped on a newly waxed floor and fell on her tailbone and low back region. She would like to lose some weight through dieting. Complains of dry mouth even though drinking plenty of water. The pt is allergic to Morphine Sulfate. No fh of colorectal CA. Last colonoscopy in Kansas City, May 2012. Past Medical History: • DM2 • Asthma Shx: Pt has a remote 10 pack-year smoking history. Occasional ETOH drink. Height 173 cm, weight 78.8 kg, blood pressure 114/80 Lab data: 2014-6-10 03:20PM GlycoHgB A1c - 7.0* 2014-6-10 03:20PM Creatinine, 24 hour urine g/24 h-1.2
- 17. Reviewing Patient Risk UNSTRUCTURED TEXT Patient Problems: 44054006 (E11.9) - Type 2 diabetes mellitus Temporal: Past Subject: Patient [HCC19] Diabetes without Complication DATA ELEMENTS Admission Date: 2015-12-25 New Patient The patient is a 59-year-old female, who was experiencing mild chest pain on the left side. The patient states that she slipped on a newly waxed floor and fell on her tailbone and low back region. She would like to lose some weight through dieting. Complains of dry mouth even though drinking plenty of water. The pt is allergic to Morphine Sulfate. No fh of colorectal CA. Last colonoscopy in Kansas City, May 2012. Past Medical History: • DM2 • Asthma Shx: Pt has a remote 10 pack-year smoking history. Occasional ETOH drink. Height 173 cm, weight 78.8 kg, blood pressure 114/80 Lab data: 2014-6-10 03:20PM GlycoHgB A1c - 7.0* 2014-6-10 03:20PM Creatinine, 24 hour urine g/24 h- 1.2 Patient Problems: 87715008 – Dry Mouth Temporal: Present Subject: Patient (E11.638) - Type 2 diabetes mellitus with other oral complications [HCC18] Diabetes with Chronic Complications
- 18. Quality Measure Reporting UNSTRUCTURED TEXT Family History: 363406005 (C18.9) - Malignant tumor of colon Subject: Family member Negated: Yes DATA ELEMENTS Procedure: 73761001 – Colonoscopy Date: 5/2012 Admission Date: 2015-12-25 New Patient The patient is a 59-year-old female, who was experiencing mild chest pain on the left side. The patient states that she slipped on a newly waxed floor and fell on her tailbone and low back region. She would like to lose some weight through dieting. Complains of dry mouth even though drinking plenty of water. The pt is allergic to Morphine Sulfate. No fh of colorectal CA. Last colonoscopy in Kansas City, May 2012. Past Medical History: • DM2 • Asthma Shx: Pt has a remote 10 pack-year smoking history. Occasional ETOH drink. Height 173 cm, weight 78.8 kg, blood pressure 114/80 Lab data: 2014-6-10 03:20PM GlycoHgB A1c - 7.0* 2014-6-10 03:20PM Creatinine, 24 hour urine g/24 h- 1.2
- 19. 19 Additional Areas Where cNLP Can Be Applied JOURNAL ARTICLES PATHOLOGY REPORTS CONSUMER PAMPHLETS
- 20. 20 Benefits of Using cNLP REDUCE REVIEW TIME INCREASE STAFF EFFICIENCY REDUCE ADMINISTRATION COST INCREASE ACCURACY OPTIMIZE ANALYTICS IMPROVE KNOWLEDGE MANAGEMENT
- 21. Survey Question #2 21 Where are you in adoption of AI technology within your organization? A. Exploration, data-gathering stage B. Beginning to use AI and NLP to gain insights C. Implemented but facing challenges D. Implemented and deployed successfully
- 22. 22 Agenda 1. Challenges healthcare organizations face trying to make patient data actionable 2. How AI can address these challenges by optimizing staff productivity and efficiency 3. Looking to the future: examples of the value AI can bring to the healthcare industry and how to get started
- 23. 23 Opportunities for AI in Healthcare Building intelligent content libraries Example: Article tagging and pharmacovigilance use cases Reducing readmission risk Example: Use of post-discharge care management data with EMR data to assess risk and intervention opportunities Enabling differential diagnosis Example: Use of patient-reported symptoms at intake to discover potential diagnosis Providing clinical decision support Example: Extract family history to customize clinical decision support Predicting onset of disease Example: Improve speed and accuracy of C-Diff and Sepsis detection Targeting population health initiatives Example: Opioid abuse and intervention services and support
- 24. 24 Emerging Opportunities to Leverage Advancements in Data and Technology in Healthcare • Telehealth • Chatbots • Wearables (patient-generated data) • Genomics • Social Determinants of Health • Digital assistants / smart devices
- 25. Survey Question #3 25 What do you feel are your organization’s biggest barriers or challenges in adopting AI technology? Select all that apply A. Cost prohibitive B. Difficulty in getting leadership buy-in C. Resistance to change; lack of change management D. Competing priorities E. Not sure where to start
- 26. 26 Action Plan for Adopting AI Technology Identify high-value areas that could benefit the most from improved data quality and optimized workflows 1 Leverage NLP, AI, and non-AI technologies to accelerate data extraction and enrichment 2 Empower your machine learning models and other AI investments with clean, enriched data 3
- 27. Health Language Solutions 27 WE BELIEVE IN THE POWER OF ACCURATE DATA REFERENCE DATA MANAGEMENT Centralize and manage your clinical, claims, administrative data, and clinical concepts, to streamline data governance and improve operational efficiencies. CNLP FOR UNSTRUCTURED DATA Extract amd codify clinical concepts locked in unstructured text by leveraging clinical synonyms and acronyms to identify gaps-in-care and improve quality of care. INTEROPERABILITY / DATA NORMALIZATION Achieve semantic interoperability by mapping non-standard/standard data to standard terminologies for accurate analytics.
- 28. 28 Questions? SARAH BRYAN Director of Product Management, Health Language Wolters Kluwer Sarah.Bryan@WoltersKluwer.com CHRIS FUNK, Ph.D Sr. Medical Informaticist, Health Language Wolters Kluwer Chris.Funk@WoltersKluwer.com JOHN LANGTON, Ph.D Director of Applied Sciences Wolters Kluwer John.Langton@WoltersKluwer.com