Global Perspectives for AI and Data Analytics in Healthcare Prashant Natarajan Principal – Analytics & AI, Deloitte Consulting | Co-Faculty Instructor, Stanford University
T echnology Innovation, AI, and Humans of Healthcare Innovation, progress, and human existence • Going beyond the status quo • Much to celebrate: saving more lives, living longer, and cheering more • Shared challenges remain – population needs, access, funding, time, • burnout, and the big data deluge Soldiering on – ‘cause WE CARE • Technology - the insufficient funds paradox • Ready to cash this cheque – put data to work with analytics, AI, and • insights-driven workflows/interactions Source: Natarajan, Prashant et al., “Demystifying Big Data & Machine Learning for Healthcare” (Taylor & Francis, 2017 )
What can AI do for You? Make us more knowledgeable via Separate signal from noise new discoveries and insights Increase the joy of working and Enable the right care to the right caring person at the right time A in AI is not merely artificial – it’s Process, predict, and engage augmented, assistive, and outside traditional care settings amplified Improve personalisation, Allow us to be more human empathy, and collaboration Source: Natarajan, Prashant et al., “Demystifying Big Data & Machine Learning for Healthcare” (Taylor & Francis, 2017 )
Perspectives on AI Thinking Acting Rationally Humanly (CIAs) (Cognitive Modelling) Thinking Acting Rationally Humanly (Logic) (Turing Test) Source: Norvig, Peter, AI: A Modern Approach
AI and Machine/Deep Learning “…field of study that Training Data gives computers th the e abil ab ility to to le learn arn without being What is mach achin ine le lear arning? explicitly Train the Machine programmed” Learning Algorithm ”… sea earc rchin ing a a ve very ry la larg rge spac ace of of pos ossible Feedback Loop hypo ypothese ses to determine one that best fits the Model observed data and any prior knowledge…” Input Data Run in Production Prediction Source: Natarajan, Prashant et al., “Demystifying Big Data & Machine Learning for Healthcare” (Taylor & Francis, 2017 )
Why Machine Learning? Machine learning enables new use cases by: Ameliorating the effects of certain human limitations • Enabling new knowledge creation or data reduction • Generating computational markers • Processing repetitive data management tasks • Serving as the foundation for workflows and comprehensive secondary use that includes: • – predictive and prescriptive analytics – intelligent search – speech to text conversion – image processing – NLP/NLU/NLG Source: Natarajan, Prashant et al., “Demystifying Big Data & Machine Learning for Healthcare” (Taylor & Francis, 2017 )
6
Applications of Machine and Deep learning Classifiers Memory-based Anomaly Forecasting learning Detection Recommender Probability Clustering NLP systems estimation Source: Natarajan, Prashant et al., “Demystifying Big Data & Machine Learning for Healthcare” (Taylor & Francis, 2017 )
Art of the Possible Disease/Population Management* • How do we predict/classify/manage care & predict the needs of What questions can we answer using AI? populations using time-series forecasting ? • How do we estimate the probability of outcomes and adverse events? Value Based Care • How do we forecast and proactively optimize care management to avoid 30-day readmissions? Operations • How can we use historical medical adherence • How do we help hospitals identify and manage using memory-based learning? clinical coding and claims using NLP ? • How do we proactively predict scheduling and rostering and classify based on capacity and skills? Digital Twins Patient/Member Engagement • How do we help proactively manage facilities • How do we classify or cluster real-time data and handle and patient flow using classification, anomalies from home health to deliver a better forecasting, and image/video experience? understanding ? • How do we engage with patients and caregivers in a timely and effective way using recommender systems ? Precision Medicine • How do we interpret phenotypic and genomic imaging using computer vision to create individualized patient outcomes? • How do we predict the appropriateness and outcomes for a participant in an oncology drug trial?
BI, AI, and NLP: the Connections Artificial Intelligence Machine Narrow AI Deep Learning Learning Data/ Cognitive Cognitive Business Integration / Analytics Insights Engagement Intelligence Exploration Analytical exploration of Organising and Sourcing, cleaning and Designing, planning, testing and deploying predictive and data applying statistical summarising data from a unifying multiple data other models to explore relationships within or between modelling and probability source to monitor how sources into a consistent multiple data sets and algorithms. to generate insights different variables are structure for more performing against pre- sophisticated reporting. defined benchmarks. NLP
IDO Maturity Curve How effective is your organisation at making insight driven decisions? Stage 5 Insight Driven Organisation Stage 4 Analytical Companies Analytics tra rans nsfor form m Stage 3 process and streamline decision making across all Analytical Stage 2 business functions Aspirations Industr ustrialis ialising ing analytics, Stage 1 enabling efficient creation Localised of trusted insights and Analytics Analytically driving innovation Impaired Expandin ing ad-hoc analytical capabilities beyond silos and into Adoptin ing analytics, mainstream business building capability and Awar ware of analytics, but functions articulating an analytics little to no infrastructure strategy in silos and poorly defined analytics strategy
Being Insight Driven is More than Data and T echnology Asking the right questions Doing the right analysis Taking the right actions Changing Vision alignment Purple people the mindset Value Information Digital first generation management Organising Ethics, compliance Improving for success & regulation outcomes Iterative & agile approach
The Walrus & the Carpenter Review AI in Healthcare (with apologies to Lewis Carroll) But wait a bit,' the Oysters cried, The Walrus and the Carpenter Before we have our chat; Were walking close in the sand; For some of us can’t share electronic data, They giggled like anything to see But all of us need stats!' Such quantities of data at hand: No worries!' said the Carpenter. If this were only put to work,' They thanked him much for that. They said, it would be grand!' Big and little data,' the Walrus said, The time has come,' the Walrus said, Is what we chiefly need: To talk of many things: NLP, sharing, and governance too Of data — and algos — and best-practices — Are very good indeed — Of people — and other healthcare things Now if you're ready, Oysters dear, And why policy should be boiling hot — AI can begin to feed.' And whether our dreams have wings.' Source: Natarajan, Prashant et al., “Demystifying Big Data & Machine Learning for Healthcare” (Taylor & Francis, 2017 )
Best Practices Ensure the support of Get access to the right Start simple and target the leadership and Treat your data with talent and AI experts “low hanging fruit” to that AI is embedded suspicion (business and technology) deliver quick wins in the strategy . Invest in the right Communicate and “ build” or “buy ” Monitor ongoing develop an AI culture and performance and keep new ways of working Try many algorithms choices and integrate & set up a feedback into process and track of model changes loop technology landscape 13 Source: Natarajan, Prashant et al., “Demystifying Big Data & Machine Learning for Healthcare” (Taylor & Francis, 2017 )
Glo lobal bal Review view
THANK YOU! 15
About the Speaker: Prashant Natarajan Prashant is a specialist leader in data, analytics, and AI with an award- winning track record of conceptualising and delivering innovative solutions for global customers. - Before joining Deloitte, Prashant was Senior Director for AI Applications at H2O.ai. - From 2008-2018, he was Global Director of Strategy and Product Management at Oracle USA, where he conceptualised and led a global portfolio of products & cloud services for health & life sciences. - Prashant is a lead author or contributor to 4 books on data science and machine learning, business intelligence, and precision medicine. - He is a Co-Faculty Instructor at Stanford University, a Distinguished Fellow at the Council for Affordable Health Demystifying Big Data and Machine Coverage, & an advisor to US Congress, Govt of California, Learning for Healthcare Author: Prashant Natarajan, Detlev H. and Pistoia Alliance. He can be contacted at Smaltz, John C. Frenzel www.LinkedIn.com/in/natarpr
Recommend
More recommend
