Can Facebook’s Two Billion Users Slow the Pandemic? Dr. Farzad Mostashari on the COVID-19 Symptom Data Challenge
This week, hosts Mark Masselli and Margaret Flinter speak with Dr. Farzad Mostashari, Founder and CEO of Aledade, former National Coordinator for Health IT, and Chair of the COVID-19 Symptom Data Challenge, a partnership with Facebook Data for Good, Carnegie Melon, Duke, University of Maryland, Resolve to Save Lives and organized by Catalyst @Health 2.0. The challenge is encouraging developers to create tools to mine data submitted by tens of millions of Facebook users, tracking real time COVID-19 symptoms to identify potential pandemic hotspots, for better epidemiological forecasting.
D r. Farzad Mostashari
[Music]
Female : Welcome to Conversations on Health Care with Mark Masselli
and Margaret Flinter, a show where we speak to the top
thought leaders in health innovation, health policy, care
delivery, global health, and the great minds who are shaping
the hea lthcare of the future.
This week, Mark and Margaret would speak with Dr. Farzad
Mostashari, Founder and CEO of Aledade, former National
Coordinator for Health IT and share of the COVID - 19 Symptom
Data Challenge, a partnership with Facebook Data for Good
an d other organizations, Resolve to Save Lives, Catalyst
@Health 2.0 as well as Duke, Carnegie Mellon in the University
of Maryland. They ’ re tapping into Facebook ’ s two billion users
to assist in creating a real time predictive tool for forecasting
when COVI D - 19 outbreaks will occur.
Lori Robertson also checks in, the Managing Editor of
FactCheck.org, looks at misstatements spoken about health
policy in the public domain, separating the fake from the facts.
We end with a bright idea that ’ s improving health an d well
being in everyday lives. If you have comments, please email us
at [email protected] or find us on Facebook, Twitter, or
wherever you listen to podcast. You can also hear us by asking
Alexa to play the progra m Conversations on Health Care. Now
stay tuned for our interview with Dr. Farzad Mostashari here
on Conversations on Health Care.
M ark Masselli: We ’ re speaking today with Dr. Farzad Mostashari, Chair of the
COVID - 19 Symptom Data Challenge, a partnership be tween
Facebook Data for Good, Carnegie Mellon , Duke University of
Maryland, Catalyst @Health 2.0 and Resolve to Save Lives.
They ’ re seeking a novel solution from developers using data
from early detection of COVID - 19 outbreaks.
Margaret Flinter: Dr. Mostas hari was the National Coordinator for Health IT
under President Obama. He was a fellow at the Brookings
Institution, the Engelberg Center for Public Health. He served
the New York City Department of Health under Tom Frieden,
and prior to that, at the Cente rs for Disease Control
Epidemiological Intelligence Service . Dr. Mostashari, we
welcome you back to Conversations on Health Care.
Dr. Farzad Mostashari: Thank you.
Mark Masselli: Yeah, Farzad, welcome back , and really the outbreak of
COVID - 19 caught the wo rld off guard from a regional epidemic
to a global pandemic , and really getting reliable data on
D r. Farzad Mostashari
outbreaks prov e to be a real challenge for public health
officials around the world . Y ou chair their COVID - 19 Symptom
Data Challenge and it ’ s a multi - stakehold er collaborative
seeking really a new approach leveraging the power of social
media to better pr edict the spread of c oronavirus. I think our
listeners like to know, maybe three things . O ne, tell us about
the partners in the challenge. Then also the global
participation already underway using the Facebook Data for
Change portal , and finally, what are the breakthroughs you ’ re
looking to find in this new real time pandemic surveillance?
Dr. Farzad Mostashari: Yeah, I think it ’ s worth reviewing what are our cur rent data
sources for being able to understand what ’ s happening with
the outbreak ? T hat ’ s obviously key . We have to tailor our
approach to the level of activity in every state and locality.
Ideally , there would be a national strategy, that would be
clear, and there would be data that ’ s clear to everybody . In
the absence of that we ’ re all having to make our own
decisions. Do we open business? Do we go out? Do we send
our kids to school or not? For that, we need data. What are
the data that we are using?
If y ou go to any New York Times or any other dashboard to see
how do we k now what ’ s going on with COVID? Well, one thing
you can look at are deaths, we just passed the 200,000 death
mark. The problem with deaths as an indicator of activity in
the community is they ’ re late. It takes weeks, and now up to
45 to 60 days before an increase in COVID will show fully in
terms of the death. Deaths are a hard outcome, but they ’ re
much delayed.
The second is obviously looking at case counts. The problem
with case counts i s it depends on testing , and if you have
different intensity of testing, if you don ’ t have a design, in your
testing program, if you don ’ t have sentinel surveillance, where
you ’ re testing the same kinds of people repeatedly then you
don ’ t get reliable info rmation from testing . We ’ ve seen as you
know, all sorts of debates about whether the increased case
count is a function of more testing or less testi ng or testing
different people.
T hen the third data source is a data source that I and others
innovated abo ut 20 years ago when I was in New York City
Health Department, which is well let ’ s just look at what ’ s
happening with hospitalizations or emergency room visits , and
those have been quite useful but , again, limited when people
change their behaviors . When t hey stopped going to the
emergency or going to the doctor , going to the -- because
D r. Farzad Mostashari
they ’ re afraid of getting tested that can c hange our receptors
as it were.
The idea h ere was and this was something that Carnegie
Mellon proposed to Facebook back in March , and they stood it
up beginning in April, is why don ’ t we just ask people on a
massive scale , have you had se t any of these symptoms in the
past 24 hours ? T hey have collected 30 million survey results.
Every week, there are surveys being answered by folks i n
every state and territory and in 70 plus countries across the
world in 55 different languages, asking , answering the
question, have you had symptoms ? I think it ’ s an incredibly
promising data source. But it really hasn ’ t been used as part of
the one of t he main pillars of us determining whether what ’ s
happening with COVID in the communities, and I think it ’ s
be cause it hasn ’ t been validated.
T his symptom challenge is basically putting the call out to
citizen scientists around the world saying, here ’ s the data ,
here is publicly available, aggregated counts by county week
or by state , by day , by country, see what you can do with it,
match it up against your understanding of what ’ s going on , the
best gold standards that we have today . S ee whether this is
wort h , giving us advanced notice compared to other sources?
Margaret Flinter: Well, there ’ s a lot to look back on and critique about the way
testing was rolled out in this country, the uncertainty that it
led to about actual infection rates . I do think America ns are
paying more attention to data and things like infection rates
tha n I ha ve ever seen the pay attention to about anything
before . P eople can tell us what county of our state , where the
rates are high, where the rates are low. T hen we began as you
know , measuring the pandemic in terms of deaths now
crossing the 200,000 marke d. Y ou ’ ve said counting death is a
lagging indicator of disease. Share with us a little bit about the
COVID - 19 map and dashboard. Is that part of this Facebook
initiative or separate from this? What better indicators of
disease do we find there? What are people saying when they
respond to those survey questions?
Dr. Farzad Mostashari: Yeah, we think that it seems to be a leading indicator for sure
of hospitalizations and deaths, and t racking super closely with
cases and case positivity. There ’ s an interesting proof point
that just recently looking at the wildfires in the Pacific
Northwest, and clearly seeing a big spike in people responding
to the surveys in those states saying, y es, I ’ m having a higher
incidence of cough, for example. When you look at Florida, in
the months leading up to their big increase, where people are
D r. Farzad Mostashari
saying, l ook, it ’ s a Florida m iracle , we know we can reopen,
and there ’ s not going to be an increase in cases. It was quite
clear on the symptom surveys that t here was an increase
happening , s o I think it ’ s promising.
The other thing you asked is, well, what else can you find out
from this? One of the other things you can do with symptom
surveys is you can ask other questions like, are you a
healthcare worker? Are you trying to social distance as much
as possible? Did you get tested ? Are you able to get tested ?
One of the just small observations from that was the vast
majority of people who have symptoms that are quit e
consistent with COVID including loss of sense of smell or taste,
the vast majority of them never get tested . I think that ’ s a
particular concern if we think that, for example, contact
tracing is going to be one of our tools in the toolkit to s top the
cha ins of transmission.
Mark Masselli: You ’ve round ed up really a great crew of partners here
Catalyst @Health 2.0, your old colleague in New York City, Dr.
Tom Fried en’s Resolve to Save Lives as well as the very
prominent universities, Duke , University of Ma ryland, and the
Carnegie Mellon . I ’ m wondering what types of data they ’ ve
already extracted or insights they ’ ve gained from the data
collected so far, or are you still at early stage? I know you were
just finishing up phase one going into phase two of recr uiting
candidates to start looking at the data , but has there been a ny
preliminary insights so far?
Dr. Farzad Mostashari: Yeah, we published some of the illustrative analyses, some of
the preliminary analyses done by data scientists at Facebook,
and Carne gie Mellon in particular who ’ ve done a really
wonderful first cut of this, the basic findings I described,
looking at the timing of increases , say , in Florida compared to
others. Other indicators, the geographically looking to see that
the tight correlatio n between states that have high levels of
activity giving those traditional metrics compared to high
levels of symptom survey.
The example I gave also of the smoke , the wildfires and smoke
in the Pacific Northwest. T here ’ s a set of analyses of folks go to
the website, www.symptomchallenge.org under illustrative
analyses, you can read those and click on it . But the data is
also publicly available, you can just download a CSV file, and
you can do it in Excel, r ight, and take a look at your
jurisdiction if you want to look at it by county week, or if you
want to look at it by state by day . A ll those -- as you pointed
out, all those folks who are more engaged than ever in data,
D r. Farzad Mostashari
this is an incredibly rich data sou rce for people to play with.
We are accepting still accept -- we haven’t closed phase one
yet in terms of receiving applications . We ’ re actually -- I ’ ll
break it here, we are going to extend th e period for people to
be able to submit applications so there ’ ll be a little bit more
extra time for peo ple to get their analyses done.
Mark Masselli: $5,000 for those that are selected, or –
Dr. Farzad Mostashari: The first five, yeah, the first five semi finalists will get $5,000
and then we ’ ll present publicly . T h en we will select the -- have
an additional period where we ’ ll work with them on
visualizations and so forth. T hen the final grand prize winner is
$50,000 and second prize would be 25.
Mark Masselli: Wonderful.
Margaret Flinter: Well, that ’ s great. Thanks for breaking that news here on
C onversations. On a very serious note , there ’ s a lot of tough
news these days, but certainly for all of us in healthcare or
public health , one difficult thing has been to hear about what
seems to be a politicization of CDC, a nd really calling it t o
question whether this beacon of the gold standard for most of
us whether we can rely on in the way we have based on what
we ’ re reading and hearing . Y ou worked in disease surveillance
at the CDC, what ’ s alarming you about what ’ s goin g on?
What ’ s at stake ? What should the public know about how
important it is that the CDC scientist voice be untarnished?
Dr. Farzad Mostashari: Yeah . Look, let me start with the good news. The good news is
that there are thousands of, as you said, scienti sts with each
with decades of experience who are the best in the world .
S ometimes I ’ ll go on Twitter, and I ’ ll interpret some
surveillance finding or play epidemiologist . I always feel like
I ’ m not the one who should be doing this. There are people
who wer e built in a lab to be doing this work, and yet we ’ re
not hearing from them. T he problem isn ’ t that we don ’ t have
expertise at CDC. The problem is that that their voices aren ’ t
being heard. I think it ’ s something that can be fixed, and I
hope will be fixed .
Mark Masselli: We’re speaking today with Dr. Farzad Mostashari, Chair of the
COVID - 19 Symptom Data Challenge. Farzad, I want to pull the
thread on the conversation about trust , and clearly Facebook
has two billion users, willing users, but there are many who
feel the data is being used for the benefit of other ’ s profit . I ’ m
wondering if you could talk about the privacy risk, and what
Facebook is doing to ensure privacy is secured , and just sort of
D r. Farzad Mostashari
the message to our citizen scientist about why this is so
important for people to be engaged in and the work you ’ re
doing on the back end to ensure the safety of the ir
information?
Dr. Farzad Mostashari: Yeah. Look , I can ’ t comment on Facebook as an actor in so
many other ways. But I can say that in terms of t his Facebook
Data for Good project, I think it is doing data for good and I do
think they ’ re doing it in a responsible way. Let me tell you a
few of the things that are relevant for the listeners. This data
that we ’ re talking about here, the symptom data is n ot from
scraping people ’ s social media or anything like that. It ’ s
literally a banner ad that says d o you want to take part in this
survey from Carnegie Mellon? If they click on it, opt in,
obviously voluntarily and they ’ re taken to a screen where
there ’ s consent basically to take part in the survey, and they
leave the Facebook ecosystem, leave the platform and go to
Q ualtrics where the surveys are done.
It’s consent , it’s voluntary , t here ’ s no information goes back or
forth other than the consent ed individ uals plus the survey
sample plates that are get pushed to Carnegie Mellon
University’s Delphi Survey Research Group, which is a highly
credible academic group. T hat ’ s in terms of where the survey
comes from, it ’ s people willingly answering questions ,
conse nted, and with a, I think , a high security bar in terms of
the Carnegie Mellon side .
The second part of it is, well, who has access to that
information ? T he information that we ’ re putting out of making
publicly available is aggregated information. M inimum cell
size s rolled up to the county level or the state level, there ’ s no
identifiability there . If you ’ re a university academic, you can
apply to access the micro data and I understand that there are
over a dozen university folks who have done that. But aga in,
it ’ s all done under what you would consider traditional IRB a nd
academic research provision s .
Mark Masselli: Will it pop up on everyone ’ s screen ? Is it on everybody ’ s
Facebook f eed?
Dr. Farzad Mostashari : They select a sample of people and say, hey do you want to
participate in the survey? It’s an ongoing offer to do it .
Mark Masselli: Great.
Margaret Flinter: Great. Well, Dr. Mostashari , one of the things that keeps us
going through these challenging times is there ’ s always some
unanticipated advances, right, that come with these great
D r. Farzad Mostashari
challenges . T he acceleration of use of telehealth in primary
care, behavioral health has certainly been one of those. But
another one is really this possible game changer for primary
care, and the ability to maybe aggrega te real time data to
really do a much better job of watching over our large cohorts
of patients who have chronic illness, or we have need s for
additional p revention and h ealth p romotion. A lot of practices
have had trouble integrating data like that into t heir workflow.
Do you want to opine a bit on how platforms like the Facebook
Symptom Database could be modified to assist frontline
providers such as primary care providers . You ’ ve done a lot of
innovation in this area . We’d really be curious about some of
the things you ’ re excited about.
Dr. Farzad Mostashari : Yeah, look I -- moving away , stepping away from the Facebook
Symptom s urvey, I think the point is, how do we get
situational awareness of the greater world into clinical
workflows? That ’ s what you ’ re really talking about.
Margaret Flinter: Yeah, well said.
Dr. Farzad Mostashari : For many years, decades, my hope was that electronic health
records were going to be those tools for population health,
that they were going to provide that visibility , 360 de gree
visibility , into the care of the patient, but also not just the
people who are here today, where you ’ re going to see and
you ’ re going to do a visit . B ut of your entire panel of patients
who should you be worried about what ’ s happening to them
out ther e? When someone goes to the hospital , do you get
notified? When someone leaves the emergency room , do you
get a ping in your practice? Do you have workflows for then
doing something with that information? If your patient doesn ’ t
fill a prescription , do you ever know about it? Do you ever
have an opportunity to do something about it? That is the
heart of populat ion health.
Margaret Flinter: Right.
Dr. Farzad Mostashari : T hen I don ’ t think electronic health records, despite five years
of our efforts , are goin g to be the tools to do that . W hat we ’ ve
done at Aledade we support 550 practices across 30 states.
They ’ re on 90 different electronic health record and practice
management systems. But they ’ re using our tool to do
population health work , and it ’ s a single c loud based piece of
software that does care management, panel management,
work lists, and point of care and now telehealth as well , and
just part of how we ’ re able to get success at reducing
hospitalizations . W e save 10,000 hospitalizations last year .
$ 1 80 millio n in savings for Medicare alone, and that’s the key.
D r. Farzad Mostashari
The key part of it is providing that information in context for
population health purposes. We do provide as part of a small
part of that , right, like COVID information for the individuals as
we ll as for the communities, but it ’ s just the tip of the iceberg.
Mark Masselli: Well, you ’ re doing great work there and we ’ ve been following
you for a long time. Remember that your work in New York
City with electronic health records then the National
Coor dinator for Health IT . As the nation was really trying to
shift over to electronic health records you oversaw the early
adoption of meaningful use of technology, and you were a
champion of the Health Datap alooza gallery in these tech
enthusiast and develo per s ---
Dr. Farzad Mostashari : Open d ata.
Mark Masselli: Pardon?
Dr. Farzad Mostashari : Open d ata.
Mark Masselli: Open d ata, yeah, liberate data.
Margaret Flinter: Liberate data.
Mark Masselli: My pardon, right, Todd Park. I just wonder as you sort of loo k
at sort of this inflection point we ’ re at. If there ’ s any silver
lining people have been saying about the pandemic, I don ’ t
really believe there ’ s been any , but there is ways that we ’ re
rethinking, re - imagining the delivery of healthcare, and
leveraging technology, but probably also doing the things that
you ’ re doing now leveraging AI. How do you see -- what do
you see the next five years loo king like in the landscape with, I
think , this robust adoption by people of telehealth, but also all
of the folks w ho are coming in and trying to make real inroads
on s ocial d eterminants of h ealth in other factors that are
playing a role in the transformation that ’ s happening now ?
Dr. Farzad Mostashari : 2020 has been a pretty horrible year in many ways. But let ’ s
look for the bright spots . I ’ ll pick three. One bright spot is just
recognizing that healthcare can move darn fast, when it has to
and regulators can move darn fast. I mean, what CMS did on
an almost daily basis was extraordinary to help keep practices
alive an d afloat and to modify regulations with OIG, with OCR,
with the payment side to enable telehealth to happen and
then the response from healthcare providers to -- literally in
12 days, we went from 100 telehealth visits to 10,000
telehealth visits. W e stood up 150 practices on telehealth over
a weekend . Y ou would have said that’s impossible , right ?
T hat ’ s the one bright spot is kind of recognizing we ca n still
sprint when we have to.
D r. Farzad Mostashari
The second bright spot I think was, as you mentioned this
recognition that fee for service is actually not reliable and that
we do need to , painful as it is, right w e need to think about
ways of moving off of -- you only get paid if someone walks in
the door. I hope that in the next five years, we will see an
embrace of value bas ed care. That ’ s even great er than the
trends were before.
The third bittersweet recognition and just hard truth of 2020,
in the aftermath of George Floyd was that we , all of us can ’ t be
bystanders . T he fact that we ’ re looking at racial disparities and
deat hs from COVID , w e in our company are looking not only at
how are we doing on the various things we ’ re working on our
core initiatives, but also what is the degree of racial disparities
in those ? W e are committing to move to action to remove
those and elim i nate those racial disparities. T hese have been
hard times for many of us and for many of our patients, our
providers, o u r communities, but I do think that we have an
opportuni ty to learn and grow from them.
Margaret Flinter: Well, thank you for calling out those three very important
points. We ’ ve been speaking today with Dr. Farzad
Mostashari , the Founder and CEO of Adelaide and Chair of the
COVID - 19 Symptom Data Challenge . You can learn more about
this exciting work by going to www.symptomchallenge.org or
you can follow Dr. Mostashari on Twitter @Farzad_MD.
Farzad, we want to thank you for your decade ’ s long
dedication to innovation and public health and for joining us
again today on Conversations on Health C are.
Dr. Farzad Mostashari : Thank you.
[ Music ]
Mark Masselli: At Conversations on Health Care we want our audience to be
truly in the know when it comes to the facts about healthcare
reform and policy. Lori Robertson is an award winning
journalist and Mana ging Editor of FactCheck.org, a
nonpartisan, nonprofit consumer advocate for voters that aim
to reduce the level of deception in U.S. politics. Lori, what
have you got for us this week?
Lori Robertson: President Donald Trump and Democratic Presidential Nom inee
Joe Biden have made competing claims about Biden ’ s earl y
statements on the c oronavirus. Following the disclosure of
comments he made to journalist Bob Woodward in March
about downplaying the c oronavirus , Trump has tried to turn
the tables on Biden, cl aiming it was Biden who maintained
through late February that the c oronavirus was , “ N ot even
D r. Farzad Mostashari
going to be a problem . ” That ’ s not accurate. The former vice
president warned early about the potential danger posed by
the virus and about the need for a thoughtf ul response by the
federal government.
At a February 28 campaign rally in South Carolina, Biden said
the Coronavirus is , “ A serious public health challenge. ”
Conversely, Biden stretched the facts at a CNN Town Hall on
September 17, claiming that in January , he wrote an op a d ,
“Saying we ’ ve got a pandemic, we ’ ve got a real problem. ” The
op a d did not go that far. In it he warned about the possibility
of a pandemic. He said Trump was not prepared to lead the
country through a , “ G lobal health challenge . ” W hich Biden did
predict would , “ G et worse before it gets better. ” We reviewed
all of Biden ’ s public comments that we could find in early 2020
about the c oronavirus. For more see our website
www.factcheck.org . I ’ m Lori R obertson, Managing Editor of
FactCheck .org.
Margaret Flinter: FactCheck .org is committed to factual accuracy from the
country ’ s major political players and is a project of the
Annenberg Public Policy Center at the University of
Pennsylvania. If you have a fact that you ’ d like check, email us
at www.chcradio.com . We ’ ll have FactCheck .org’s Lori
Robertson check it out for you here on Conversations on
Health Care .
[ Music ]
Margaret Flinter: Each week Conversation highlig hts a bright idea about how to
make wellness a part of our communities and everyday lives. If
music soothes the savage beast, the question I want to answer
at the Sync Project is how exactly ? There are lots of anecdotal
studies supporting music ’ s ability t o trigger memory or boost
endurance or focus. But virtually nothing is known about how
music truly impacts our physiological and neurological state.
This is the question that intrigued scientist Ketki Karanam, a
Systems Biology PhD from Harvard, who wonder ed how could
music be scientifically harnessed as a powerful precision
medicine tool? They formed the Sync Project with a cross -
section of neuroscientist, biologist , audio engineers , even
some rock stars like Peter Gabriel, and started by using
a rtificial i ntelligence systems to analyze existing playlists that
purport or promote relaxation induce sleep enhance focus or
athletic performance.
Ketki Karanam: Once we have this set of songs that a m achine l earning
algorithms predict to be effective for a specifi c activity , we can
then go on studies using these devices like your heart rate
D r. Farzad Mostashari
monitor s, your smart watch es, activity trackers and actually
look at how effective indeed it is that song for that purpose.
Margaret Flinter: Karanam and her colleagues note tha t most of us self -
medicate with music already , s o why not harness this
ubiquitous tool that ’ s available to all of us and develop
strategies and systems that might replace pharmacological
i nterventions with musical ones.
Ketki Karanam: We’re literally walki ng around with 14 million songs in our
pocket every single day. We saw a great opportunity on really
being able to understand how music was affecting us to
measure how different types of music affect both our
psychological health as well as the physiolog y.
Margaret Flinter: Karanam and her team see n vast potential for reducing
reliance on drugs by crafting personalized music interventions
and the management of a variety of complex conditions, such
as pain management, PTSD, even Parkinson ’ s disease.
Ketki Ka ranam: In Parkinson ’ s disease, patients have t rouble coordinating
movements. B y playing them the right kind of music, it can be
an external auditory support they have that ’ s going to help
them walk more smoothly.
Margaret Flinter: The Sync Project combinin g computer technology and
neuroscience , physiology and musicology to harness the
healing powers inherent in music. Now that is a bright idea.
[ Music ]
Mark Masselli: You've been listening to Conversations on Health Care. I'm
Mark Masselli.
Margaret Flinter: And I'm Margaret Flinter.
Mark Masselli: Peace and h ealth .
[ Music ]
Female : Conversations on Health Care is recorded at WESU at
Wesleyan University, streaming live at www.chcradio.com ,
iTunes, or wherever you listen to podcasts. If you have
comments, please email us at [email protected] or
find us on Facebook or Twitter. We love hearing from you. Th e
show is brought to you by the Community Health Center.