BE PART OF THE REVOLUTION TRANSFORMING HEALTHCARE WITH AI - - PowerPoint PPT Presentation

Insert Logo, text, URLs, etc here 1000 ATTENDEES 80 SPEAKERS 10 WORKSHOPS 2 SOCIAL EVENTS 1 AIMed19

www.aimed.events/northamerica-2019/ #AIMed19

BE PART OF THE REVOLUTION

TRANSFORMING HEALTHCARE WITH AI

CALIFORNIA — THE RITZ-CARLTON, LAGUNA NIGUEL 11–14 DECEMBER 2019

Edward H. Shortliffe, MD, PhD; Professor (Adjunct, Retired) Columbia University Arizona State University Weill Cornell Medical College Senior Executive Consultant, IBM Watson Health AIMed NORTH AMERICA, CALIFORNIA 11–14 DECEMBER 2019 www.aimed.events/northamerica-2019

Insert Photo Here

tedshortliffe @tshortliffe

Look Looking g Ah Ahea ead with Both Ca Caution and Eager r An Anticipa pation

• n

JAMA 2018;320(21):2199-2200 CDSS = Clinical Decision-Support System

Varying Models of Decision Support: Medical-Device Data Interpretation

Decision-specific clinical data (e.g., from measurement device) Software produces interpretive report e.g.

• Electrocardiogram
• Electroencephologram
• Radiologic images

Well accepted: May be reviewed first by expert; clinician decides how to use result Deliver to clinician Clinician verifies interpretation and incorporates it into decision making process

AI & Machine Learning for Data Analytics

• JAMA. 2017;318(22):2211-2223

Varying Models of Decision Support: Event Monitoring and Alerts

Repository for clinical information

• n a patient (e.g.,

EHR) Interface to clinician for chart review and

• rder entry

Event monitoring software Knowledge base of what to watch for (e.g. rules) If: patient is receiving digoxin and serum potassium is low Then: warn physician that patient may require potassium replacement Decision support:

• Warnings
• Alerts
• Guidance

Two Major Consulting Questions for Medical AI to Address What does my patient have? (a question about what is true right now) What should I do for my patient? (a question about action-oriented decision making)

Varying Models of Decision Support: Direct Consultation with Clinical User

Clinical advisory tool (typically for diagnosis or treatment) Repository for clinical information

• n a patient (e.g.,

EHR) Analytical methods (e.g. rules, scoring function, statistical) Interface to clinician that integrates decision support with chart review and order entry Issues:

• Transparency/explanation
• Time requirements
• Usability and complexity
• Relevance and insight
• Respect for user
• Scientific foundation

“Greek Oracle” model Preferred model

www.aimed.events/northamerica-2019/

AIMed NORTH AMERICA, CALIFORNIA 11–14 DECEMBER 2019

Cr Criter eria for

• r CD

CDSS Accept cceptance nce and nd In Integ egra ration

• n into
• Wo

Workf kflow

• Black boxes are unacceptable:
• A Clinical Decision-Support System (CDSS) requires

transparency so that users can understand the basis for any advice or recommendations that are offered.

www.aimed.events/northamerica-2019/

AIMed NORTH AMERICA, CALIFORNIA 11–14 DECEMBER 2019

Cr Criter eria for

• r CD

CDSS Accept cceptance nce and nd In Integ egra ration

• n into
• Work
• rkflow
• w
• Time is a scarce resource:
• A CDSS should be efficient in terms of time requirements

and must blend into the workflow of the busy clinical environment.

www.aimed.events/northamerica-2019/

AIMed NORTH AMERICA, CALIFORNIA 11–14 DECEMBER 2019

Cr Criter eria for

• r CD

CDSS Accept cceptance nce and nd In Integ egra ration

• n into
• Work
• rkflow
• w
• Complexity and lack of usability thwart use:
• A CDSS should be intuitive and simple to learn and use

so that major training is not required and it is easy to

• btain advice or analytic results.

Users vote with their feet….

www.aimed.events/northamerica-2019/

AIMed NORTH AMERICA, CALIFORNIA 11–14 DECEMBER 2019

Cr Criter eria for

• r CD

CDSS Accept cceptance nce and nd In Integ egra ration

• n into
• Work
• rkflow
• w
• Relevance and insight are essential:
• A CDSS should reflect an understanding of the pertinent

domain and the kinds of questions with which clinicians are likely to want assistance.

Rapid pulse, sweating, shallow breathing. According to the computer, you’ve got gallstones.

www.aimed.events/northamerica-2019/

AIMed NORTH AMERICA, CALIFORNIA 11–14 DECEMBER 2019

Cr Criter eria for

• r CD

CDSS Accept cceptance nce and nd In Integ egra ration

• n into
• Work
• rkflow
• w
• Delivery of knowledge and information must be

respectful

www.aimed.events/northamerica-2019/

AIMed NORTH AMERICA, CALIFORNIA 11–14 DECEMBER 2019

Cr Criter eria for

• r CD

CDSS Accept cceptance nce and nd In Integ egra ration

• n into
• Work
• rkflow
• w
• Delivery of knowledge and information must be

respectful:

• A CDSS should offer advice in a way that recognizes the

expertise of the user, making it clear that it is designed to inform and assist but not to replace a clinician.

X

www.aimed.events/northamerica-2019/

AIMed NORTH AMERICA, CALIFORNIA 11–14 DECEMBER 2019

Cr Criter eria for

• r CD

CDSS Accept cceptance nce and nd In Integ egra ration

• n into
• Work
• rkflow
• w
• Scientific foundation must be strong:
• A CDSS should have rigorous, peer-reviewed scientific

evidence establishing its safety, validity, reproducibility, usability, and reliability.

www.aimed.events/northamerica-2019/

AIMed NORTH AMERICA, CALIFORNIA 11–14 DECEMBER 2019

De Decisi sion Ma Making is s Not the Only Topic fo for E Eval valuati ations ns

• Time requirements and impact on workflow
• Usability and acceptability
• Impact of system use on clinician’s subsequent

decisions

• Ultimate impact on patient outcomes
• Cost-effectiveness of the introduction and use of the

technology

www.aimed.events/northamerica-2019/

AIMed NORTH AMERICA, CALIFORNIA 11–14 DECEMBER 2019

So Some Im Implic licatio tions

• Commercialization of clinical decision support systems

requires a time-consuming staged evaluation plan that demonstrates more than quality of advice

• Early adopters naturally become partners in evaluation

studies, since many questions cannot be answered in the laboratory and must be assessed in actual clinical use environments

www.aimed.events/northamerica-2019/

AIMed NORTH AMERICA, CALIFORNIA 11–14 DECEMBER 2019

Th The F Futu ture

• Decision support evaluation partnerships similar to

those that have long existed in pharmaceutical industry

• Interactive medical AI systems and their

promulgation will be highly dependent on success in standards setting, interoperability, and rigorous evaluations

Th Thank k You

ted@shortliffe.net