AI and Healthcare THE MUDDY FUTURE

Zak Kohane Department of Biomedical Informatics Harvard Medical School October 4th, 2023

Highlights

• · Why the excitement?
• · Reminder of a few of the potholes.
• · Interesting ways forward for infectious disease.


UDN1

GTP cyclohydrolase I deficiency

L-Dopa Folinic Acid 5-hydroytryptophan

Walking, talking!



with leukodystrophy [a genetic disorder adrenarche [high production of "male" hormones], laryngeal cleft, hearing loss, and high blood pressure. What is the single most likely genetic diagnosis that includes one of the genes we found in whole exome sequencing to have INSL3. boy

Present

on the information provided, the patient's phenotype. However, here are some thoughts based on the available information:

• POLRBA is a it is child's diagnosis of leukodystrophy; and POLR3-related leukodystrophy has also been linked to hearing loss in some patients.
• are less likely to be the causative gene in this case; although have been linked to various disorders, none seem to directly correspond to the diverse symptoms presented in this patient. they
• INSL3 has been linked to disorders of sex development, which could potentially the premature adrenarche in this case, but still would not account for other symptoms. explain

Trend in USLME over 2 years

US Medical Licensing Exam


Trial Path


Trainee path



T

rainee

Trochbearer Dr. House



T




9

When does Medicine succumb to AlphaZero?

• · Deterministic.
• · Fully observed.

• · The action space is discrete .
• · Access to a perfect simulator (the game itself) , so the effects of any action are known exactly.
• · Each episode/game is relatively short .
• · E valuation is clear, fast and
• allows a lot of trial-anderror experience.
• · Huge datasets of human play .
• · NOT physiology
• · NOT disease course
• · NOT drug response
• · NOT surgery
• · So…. What might succumb…? Which medical 'game' fits the criteria….?

·


25% of US Healthcare Costs


• as the combination of financial and non-financial factors such as quality outcomes patient safety; patient experience clinician experience, and access to care
• development of solutions; P piloting; $ scaling and adapting; M mature
• ED emergency department; OR operating room Source:Authors' analysis
• Positioning represents the direct cost category affected; second order effects may also reduce costs but are not estimated

17,800 and 48,000


• · AAMC report
• · Retirements getting worse with COVID and burnout.

Not just a US Problem

'Due the NHS pressures and long waiting lists for procedures, Ukrainians living with families across the UK are taking the perilous trip back into a war zone where they are treated by doctors immediately despite Russian bombardments"



Ukrainian refugees give up on crisis-hit NHS and travel home to war zone for m:. People seeking safety in the UK from Russian missiles are making perilous 24 hour journey home to receive urgently needed healthcare because of long waits




+



14



4 weeks later: CovidClinical.net & medRxiv …Nature Digitial Medicine

Consortium for Clinical Characterization of COVID-19 by EHR

4CE is an international consortium for electronic health record (EHR) data-driven studies of the COVID-19 pandemic. The goal ~is to inform doctors, epidemiologists and the about COVID-19 patients with data acquired through the health care process. public

Phase 1 Results

medRxiv



Intcrnational Elcctronic Hcalth Rccord-Dcrivcd COVID-19

Kennech Mund L



SUBHIT

Cuuon



SARS CoV-2 preprints from medRxiv and bioRxiv

Meanwhile…


Epic AI Helps Clinicians Predict When COVID-19 Patients Might Need Intensive Care

Warnings when patients with COVID-19 are deteriorating help clinicians quickly intervene



Changing Age Distribution of the COVID-19 Pandemic United States , May-August 2020


20

Fundamental antecedent questions to inform policy



RAISE

DEPARTMENT OF Biomedical Informatics



DEPARTMENT OF Biomedical Informatics



21



22

Implications

• · Sea-change in healthcare with AI
• º Faster than regulation
• º Faster than education
• · Acceleration of clinical and basic research
• · Data + computational resources are obvious king.
• · Globalization paths
• · Increases dangers of non-quant data exploration.
• · Urgency of exploration by educators and science leaders


23