AI: innovation in healthcare Dr Ali Connell Senior Research - - PowerPoint PPT Presentation
AI: innovation in healthcare Dr Ali Connell Senior Research Scientist, Google Health UK Health systems aim to meet The Quadruple Aim.. . 1 Betuer clinical outcomes Don Berwick Ex-Head of Medicare and Medicaid 2 Reduced costs 3 Improved
Dr Ali Connell Senior Research Scientist, Google Health UK
Private and Confidential
Health systems aim to meet The Quadruple Aim...
Betuer clinical outcomes Reduced costs Improved patient experience
Don Berwick Ex-Head of Medicare and Medicaid
Improved experience of care provision
Private and Confidential
...but face signifjcant challenges
>50% of healthcare not evidence based Stafg burnout rates on the rise Care continues to be episodic vs integrated Intractable increases in healthcare costs Failure to deliver shared decision making for patients Unwarranted variation exists across healthcare delivery > 10% of patients experience harm in hospitals Focus on illness at the expense
AI is already working in tools used by billions of people
Private and Confidential
Google Health Cloud Brain Streams Apigee Google Search Google Fit Google Assistant
Groups across Google/Alphabet have been working in healthcare
MEDICAL RECORDS
Together we are applying AI to a range of clinical challenges
RADIOLOGY OPHTHALMOLOGY PATHOLOGY
We are commitued to the highest standards of peer-reviewed academic research
130k examples
No DR Mild DR Moderate DR Severe DR Proliferative DR Image Quality L/R eye Field of View
Inception model
Inception
Labeling tool 54 ophthalmologists 880k diagnoses
Training a model to read fundus photographs
1-Specificity, % Sensitivity, %
Generalists
Dec 2016
Specialists
Feb 2018
100 90 80 70 60 50 40 2 4 6 8 10
The model now pergorms on par with retina specialists
ARTICLE
Prediction of cardiovascular risk factors from retinal fundus photographs via deep learning Age Self reporued sex Cardiovascular risk
14
A collaboration with Veterans Health Afgairs
Test orders and results 521,000,000 labs Procedures 181,000,000 instances Microbiology 465,000 reporus Demographics and health factors 227,000,000 instances Medications 293,000,000 prescriptions Discrete vitals 356,000,000 records Clinic visits, note titles 123,000,000 instances Comorbidities 166,000,000 ICD-9 codes Admissions, discharges and transfers 1,500,000 admissions 825,000 transfers Unique patients ~700k
Continuous prediction of patient deterioration from EHR data
Hospital admission
48 hours
Model AKI
Sepsis
Billing code
Moruality
Hypoglycemia
Kidney function (GFR) Current AKI fmags
Model warns clinicians
Time
48 hours
Enabling anticipatory care in AKI
Predicting 7 lab tests: substantial increases correctly predicted in 88.5% of instances Network predicted 90.2% of all AKI that needed dialysis ROC AUC 92.1%
Enabling anticipatory care in AKI
“Technology projects fail because we don’t take suffjcient account
Prof Trish Greenhalgh
Greenhalgh T. et al. Beyond Adoption: A New Framework for Theorizing and Evaluating Nonadoption, Abandonment, and Challenges to the Scale-Up, Spread, and Sustainability of Health and Care Technologies. J Med Internet Res. 2017 Nov 1;19(11):e367.
Understanding complexity
Model metrics != clinical applicability
Model metrics != clinical applicability It is vital to know how the model, user and socio-technical environment interact
Multidisciplinary perspectives
Pathumthani, Thailand
Human Computer Interaction Collaborative Design Process Ethnographic Studies
Understanding complexity: design and deployment
Processes of care Patient outcomes Hospital metrics Qualitative review Economic analysis
Understanding complexity: impacts
@draconnell