Machine Learning for Heart Failure Management
Machine Learning for Heart Failure Management
Ray Liao of MIT presents Machine Learning for Heart Failure Management at IdeaStream 2022.
Content
6.8 -> Good morning everyone. My name is
Ray, and this is Dr. Steven Hong.
12.56 -> Dr. Hong is an emergency room physician at Beth
Israel Deaconess Medical Center. Three years ago,
19.52 -> Dr. Hong reached out to our lab at MIT CSAIL,
and told us how difficult and frustrating it is
26.16 -> to manage heart failure. 20 percent. 20 percent
of heart failure patients get readmitted to the
34.16 -> hospital within 30 days of discharge. As a
leading cause of hospitalization in the US,
42.16 -> heart failure hospitalizations cost the US
healthcare system 12 billion dollars every year.
49.28 -> Despite the money we spend, patients are dying.
Half of all heart failure patients die within 4
56.24 -> years of diagnosis. The fundamental challenge
is a variability of this disease. A treatment
64.64 -> regimen that works for patient John may not work
for patient Jane, and when a patient is out of the
70.8 -> hospital the therapy has to be constantly adjusted
for the changing and dynamic physiological state.
79.28 -> Right now, the data is siloed and the intervention
is delayed. What if we can learn from this?
88.32 -> From millions of patients retrospective
records, we can predict exacerbations and enable
96.96 -> timely interventions. Empallo empowers caregivers
from the hospital to the home. Specifically, it
107.2 -> helps caregivers figure out when the intervention
is needed and what interventions are needed.
114.64 -> Our initial solution integrates data from the
electronic health records and imaging systems,
121.44 -> identifies patients who are at high risk
of getting readmitted to the hospital,
127.12 -> and provides actionable insights to providers
to help them adjust and titrate medications.
136.16 -> In the longer term, we'll integrate data from
wearable and implantable devices. We will provide
142.64 -> service to medical device manufacturers to help
them identify patients who may benefit the most
150.64 -> from their product and improve their clinical
trials. Eventually, we're going to enable patients
158.64 -> to get access to care in a timely fashion. My name
is Ray. I finished PhD in Computer Science at MIT
166.56 -> last summer. My co-founder Claire has a business
background in healthcare consulting and finance.
172.56 -> Our team has two machine learning
engineers and a scientific advisory board
177.68 -> from University of Colorado Hospital, MGH, Mayo
Clinic, MIT, BIDMC, and the VA healthcare system.
187.68 -> We are seeing strong traction.
Our initial prototype was built on
192.24 -> 500,000 patient records from Beth Israel
Deaconess Medical Center. We recently signed
198.4 -> agreements with the VA national healthcare system
and the University of Colorado Hospital system.
204.8 -> This will give us additional 2 million patient
records to enhance and test our solutions.
214.08 -> We're in an exciting time. In 2008, the adoption
rate of electronic health record systems in the US
222 -> was only eight percent. In 2010, right before
the Affordable Care Act was signed into law,
228.96 -> the adoption rate of EHR was 10 percent. In
2015, this rate has grown by nine times.
238.72 -> The last decade has left us this massive treasure
trove of digital clinical data, and a window
246.64 -> of opportunity. Our initial customers will
be hospitals because when a heart failure
253.68 -> patient gets readmitted to the hospital within
30 days that rehospitalization is not reimbursed.
260.96 -> Furthermore, Medicare/Medicaid financially
penalized hospitals for high readmissions of
266.88 -> heart failure by deducting the total reimbursement
from CMS by up to three percent. Our long-term
275.2 -> target customers will be insured because we
save hospitalization costs for their patients
281.2 -> as will be medical device manufacturers because we
help them identify patients who may benefit from
288.08 -> their devices, improve their clinical trials, and
enhance their data analytics capabilities. We're
296.4 -> addressing a big market. Our initial beachhead
market that targets hospitalization cost savings
302.56 -> will be 3 billion dollars in the United States. As
we grow and expand the home monitoring and other
309.68 -> adjacent disease applications, we're addressing
a total market size of about 60 billion dollars.
316.64 -> Very fortunately at MIT, we benefit
from various programs including Sandbox,
321.76 -> Deshpande, Delta V. We are also fortunate
to get grants from the state government
327.28 -> as well as NSF. We're at the stage where
additional capital will help us speed up growth.
334.8 -> We're using 2.4 million dollars to support
our multi-center retrospective study,
340.8 -> and perspective piloting with our partners, as
well as preparation for FDA pre-submission meeting.
347.84 -> If you're interested to learn more, please come
talk to us and visit empallo.com. Thank you.
Source: https://www.youtube.com/watch?v=n5Js1Wu9bmM