64 ISE Magazine | www.iise.org/ISEmagazine

with Michael L. Washington

Michael L. Washington, Ph.D., is an industrial

engineer and health scientist working with the

U.S. Centers for Disease Control and Prevention’s

Center for Emerging and Zoonotic Infectious

Diseases. He also works with the Bloomberg

Philanthropies Data to Policy Initiative through

the CDC Center for Global Health/Division of

Global Health Protection providing assistance

in developing countries. He and his team

have created two spreadsheet models, one

to help healthcare facilities plan for a surge

in COVID-19 cases during the pandemic and

another that illustrates the resources needed

to conduct contact tracing and monitoring of

COVID-19 cases and its potential impact. Both

can be downloaded at no cost at www.cdc.gov/

coronavirus. His full interview can be heard on

Problem Solved: The IISE Podcast at podcast.iise.

org. He is an IISE member.

What’s

Your

Story?

What is the general idea behind creating such models?

Models are an approximation of reality. There are people who think that models

are going to give you an exact answer. Models are just a simpliﬁed version of re-

ality and a tool to help manage possible future outcomes. It’s not going to tell you

exactly what you should do or exactly what should happen. That rarely occurs.

It’s just a tool to help you prepare for what could happen, what could not happen.

One of the main problems that’s happening now, especially with COVID, is

when people are saying “these models are way off” or “they’re incorrect.” The

fact that this is a new virus; we’re still learning about it and we’re including as

much as we can in our models that are in dealing with COVID to get as close

to reality as possible. But there’s still a lot of stuff that we don’t know, and that

makes it very difﬁcult to model. We just need to help people, especially decision-

makers, understand that and give a range of outcomes that could occur based

upon the models, and that should help them make more informed decisions.

What factors can affect the range of possible outcomes?

There are a lot of things that can occur that will give you a wide range. One of

them is the uncertainty and the variables that we have. And so if you’re using

a more statistical or mechanicalistic model in which the variables change or are

randomly selected, then you will get a range of certain outcomes.

One of the assumptions people are making now is that people are going to

obey wearing masks and staying at home, those types of things, and a lot of

models assume that. We can’t model what people are actually going to do, and

of course, the model is going to be wrong because of that assumption. ... With

humans, things may change if 10% of the population stays home, if 30% or

100% – there’s a wide range of possibilities. If the model is set on 30%, then the

population does something else, then yes, the model is going to be incorrect. If

we change the assumptions to what people actually did, chances are the model

is going to be correct.

What is the purpose of the COVID Surge model?

The main reason our team developed this model is that hospitals would run out of

resources if the outbreak became too bad. So we created a model that looks at the

number of ICU beds that are available. And a lot people are very concerned about

ventilators, so we included that in the model. We created a mechanicalistic model,

SEIR, (susceptible, exposed, infectious, recovered) and we used that model to

estimate the number of COVID cases that would occur, how many would need

hospitalization, how many would need to go into the ICU, how many would

need a ventilator. Then we matched that with the people using the model and

what resources they have. That could be the resources in a hospital, in a region

or even in a country. The model tells you when you are going to run out of re-

sources, how long you’re going to run out of resources and eventually when you’re

going to go back down to a level where you can manage the surge of patients.

Can this model be used by ofﬁcials in various locations?

That’s what the model is designed to do – it is designed to look at your speciﬁc

situation, both from the disease’s perspective and also your resources. One of the

great things is we work for the CDC, and if there’s anybody that knows about

COVID or any outbreak of diseases, it’s the CDC. We consult them on the type

of parameters we should use. We can also send people out into the ﬁeld to hospi-

tals. We can gather information for the model from our colleagues and it makes

it more usable almost immediately. ... The model is pretty robust to handle a

October 2020 | ISE Magazine 65

lot of situations, including new information and changes that

occur.

We have had people provide us feedback in ways to improve

the model, and that’s been very helpful. We’ve also heard from

people who contacted us to learn a little more about the model

as far as the details of how it works. And we’ve gotten feedback

from people within the CDC who said this is a great tool, they

appreciate it and they’ve been using it.

How does the COVID Tracer model work?

We thought there was a need to create a model to estimate how

much staff would be needed to do appropriate contact tracing –

how many hours it would take and what impact would it have

on the outbreak. It’s very similar to the COVID Surge model

in that it uses an SEIR model and we have a number of differ-

ent inputs that go into the model. One of the more interesting

things we added is that we keep track of patients who are in

certain stages, such as if they have just gotten sick, if they are

pre-symptomatic but shedding, if they are symptomatic and

shedding and if they recover. That’s important with contact

tracing because you want to ﬁnd people as fast as possible and

isolate them, preferably before they start shedding. And the fact

that there are a number of people who are shedding and are as-

ymptomatic makes it very important to ﬁnd those people early

and isolate them.

With Ebola, you really don’t start infecting people until you

start showing symptoms. That was one of the things that people

thought about COVID in the very beginning was that we didn’t

think people could transmit without showing symptoms. What

we did ﬁnd out was people who are asymptomatic can transmit.

And so it makes it very difﬁcult for the contact tracers to actually

appropriately isolate people who are sick but not showing any

symptoms. ... It’s all about making sure you have enough staff to

actually do contact tracing appropriately.

What are the strategies included in the model?

Those are complex strategies that we researched and found that

people have been using around the world. You can use those

in your situation if you think that strategy matches up to what

you’re doing. ... That’s one of the things we try to emphasize

with the model – we have to tailor it to your situation. This

is also something I teach in my job working with Bloomberg

Philanthropies where we teach people in lower and middle-

income countries that we have models you are able to use. But

just because a situation works in New York doesn’t mean it’s

going to work in a rural village in Zambia. You have to modify

the model to match your situation. That’s even true within the

United States, from state to state. Gainesville, Florida, is prob-

ably going to be very different from Los Angeles. You have to

modify the data that’s going in to match your situation in order

for the model to work appropriately.

With the ﬁrst one, it’s more ﬁnding those who are symp-

tomatic and isolating them. That may not work as well because

by the time they are symptomatic, they have already been out

in the wild transmitting. In the second situation, you are again

focusing on those who are symptomatic and their contacts, and

you’re asking the contact as soon as they show symptoms to iso-

late yourself, contact the people in the public health department

so they can start tracking you. Again, that is a slight problem in

that by the time they show symptoms, there are some who are

asymptomatic and they’re already spreading.

With the third situation, it is again isolating those that you

ﬁnd who are sick but then also isolating the contacts. You also

tell them to stay at home, because by the time they show symp-

toms, they’re already at home and they’re not spreading. You

contact the health department and they can start monitoring

you. Now as you can see, with each of those different strategies,

more and more resources are going to be required. So the tool

takes all of that into consideration. But it also shows the impact

of isolating people quickly.

As I said before, these are humans, and it’s very difﬁcult to

predict what humans are going to do. And so we add a variable

in there such as compliance, how many people are actually go-

ing to comply, with staying at home or staying in contact with

the contact tracer. And the less people who are compliant, the

less likely the strategy is going to work.

The COVID models were based on previous

efforts with Ebola. Describe that project.

Back in 2014, when the Ebola outbreak was occurring, we de-

cided to create a tool to help decision-makers make decisions,

and that’s basically what all of these models are. We created the

model called Ebola Response and we took whatever informa-

tion that we could about Ebola, and at that time it was very lim-

ited. When you have something like Ebola where thousands of

people are dying every day, you have to make decisions quickly.

With the Ebola model, we were able to use that to push inter-

national agencies into contributing to all the efforts in western

Africa to help stop the outbreak.

Describe your work with Bloomberg Philan-

thropies to help people in other countries.

It’s been one of my passions since 2002 when I took my interna-

tional trip to Ghana to help with polio eradication. I went there

to teach the people on the ground about surveillance, about

checking data to appropriately identify if there are polio cases in

certain regions. I taught them how to use geographic informa-

tion software to help map out where all these different outbreaks

are occurring. ...

That international bug, and work, has really bitten me. In the

last couple of years working with Bloomberg Philanthropies,

I’ve accumulated a lot of miles. I work in a lot of different coun-

tries, and I get a lot of joy from that.

– Interview by Michael Hughes, IISE