How do you manage a COVID-19 surge? When intensive care units are beyond capacity, health care workers are stretched thin, and life-saving equipment is in short supply, there are a lot of tough choices to be made, and quickly. Joe Agor, assistant professor of industrial engineering, is gathering the data on how hospitals prepared and responded to the pandemic, as well as how their patients fared, with the goal of helping to better manage surge capacity for future pandemics or disasters.
Joe Agor, assistant professor of industrial engineering, reviews data with graduate student Farzaneh Mansourifard.
[MUSIC: “Orbit” by Corbyn Kites, part of the YouTube Audio Library. Licensed under a creative commons license]
[AUDIO: “California Hospitals Face Staff Shortages As Hospitalizations Soar,” National Public Radio]
ERIC WESTERVELT: Southern California and the state's Central Valley are reeling. Less than 2% of ICU beds there are available, and many hospitals are overwhelmed.
[AUDIO: “Montana Doctor Says Hospital Is ‘Strapped Thin’,” National Public Radio]
NOEL KING: The mountain states are getting hit hard now, including Montana, where many hospitals are overwhelmed with COVID-19 patients.
[AUDIO: “India's COVID-19 Cases Break World Records 4 Days In A Row,” National Public Radio]
LULU GARCIA-NAVARRO: A new record in India. Today, the country confirmed nearly 350,000 new coronavirus cases. And with so many COVID-19 patients, India's health system is collapsing.
PERRY: It’s become an all-too-common story throughout the pandemic. Even with social distancing and lockdowns, many medical systems still faced huge surges of patients. Some have managed, but some have become overwhelmed. But even in communities that have been spared the worst, those hospitals still need to be prepared.
How should they do that? How do they know what works or doesn’t work? And what can we learn from their experiences now to be better prepared for the next wave of cases? Or natural disasters? Or even the next pandemic?
I’m Owen Perry, and today we’ll be talking about research here at Oregon State that’s working to document what hospital are doing to manage and adjust to surges in real-time and what impact those changes are having on patients.
[MUSIC: The Ether Bunny, by Eyes Closed Audio, used with permission under a Creative Commons Attribution License]
PERRY: From the College of Engineering at Oregon State University, this is “Engineering Out Loud.”
In hospitals, countless decisions are being made every day. Which rooms need to be isolated? How much PPE needs to be ordered? Who needs to go to the ICU and who has to wait?
When we look back at the pandemic, we can view each decision as a single point of data. And if you gather that data, you might just start to see patterns. What’s working? What isn’t? What can we learn that will let us better respond to the next pandemic?
Joe Agor, an assistant professor of industrial engineering, has been doing just that, collecting a record of the decisions that have been made in two hospital systems during the pandemic.
AGOR: My area of expertise is in optimization methods and data mining. More specifically, I do work or have done work and using bi-level optimization for feature selection.
PERRY: In other words, he sifts through large amounts of data about systems that are embedded like nesting dolls within larger systems – like hospitals – and tries to figure how to use that data to make improvements. Specifically, he analyzes large scale electronic health records to identify patterns in patient health outcomes.
For most of the last year, he’s been working to document the decision-making processes of hospital systems before, during, and after the pandemic. The goal is to provide tools to help hospitals better manage medical surge capacity for future pandemics or natural disasters.
AGOR: This work was funded through a program called NSF RAPID program through the National Science Foundation – NSF. That program’s main goal is gathering information that may otherwise be lost. So, in emergency situations or disaster situations, but when things happen quickly, information wants to be gathered so that we don't lose what we either learn from it or what we did.
PERRY: The goal of this project was to gather information related to how hospitals prepared for, reacted to, and managed surges of patients during the pandemic.
Joe hopes to gain insights into what impact the real-time adjustments hospitals made had on both patients with and without COVID.
[MUSIC: “Divine Life Society” by Jesse Gallagher part of the YouTube Audio Library. Licensed under a creative commons license]
He and researchers at North Carolina State University have partnered with two hospital systems – MedStar Health System in the Washington, D.C., metro region and Samaritan Health Services here in Corvallis.
AGOR: We want it to get together in a way to start documenting these procedures that we're going to be, that we're inevitably going to be put into place to manage this surge, right, to manage the surge to the hospitals. And so that we could look back on it or like we're doing right now to see, okay, what worked, what didn’t work.
PERRY: With access to real-time information as well as historical, pre-pandemic health records from the hospitals, the group was able to document changes in things like intensive care units and hospital bed capacity, staffing, equipment availability (including personal protective equipment and testing kits), and more. Every decision to change a policy, reassign a doctor, or purchase more N95 masks was captured.
GOPAL: In essence, it's still very important to get an accurate documentation of what actually happened and how we responded and boots on the ground, what we were doing day by day. Things like that are very difficult to recollect in hindsight without inaccuracy of data, without accurate modeling. And I think that's what Joe's study's really going to provide us.
PERRY: That’s Dr. Bharat Gopal, a physician at Samaritan Health Services, who has been working closely with Joe and the team on this project. He’s talked to the researchers about what it's like from a clinician's perspective and what information is most important to capture.
GOPAL: My role in this is primarily to look at the fields that are being collected in the data set and determine whether this is something that we should, could consider to include or not include. Is this extraneous data, or is this going to be something that we want to use or that it will be utilized. To me, data is only important as so much as it's being used.
[MUSIC: “DeepInThought” by Will Havnaer]
PERRY: In the current pandemic doctors are having to respond without historical data to support their decisions.
GOPAL : Imagine in 1918, we had an accurate depiction data-wise of what had happened during that pandemic. I think it will, we will have a chronicle that we can use to accurately model or try to think through the layers of policy and practice that we used in making decisions and what exactly happened every time we did that. We make decision a and then we have consequence X and make decision B, and we have consequence Y and every time we did that, we lent, we ended up with something else happening, Z or something else happening that really was difficult to anticipate at the time, but now that we have this great data set, we can start to anticipate a little bit better. If we have another pandemic, which I would say not if, when we have another pandemic, um, a better understanding of how we can respond.
PERRY: As you can imagine, there’s a lot of data to work with. Ultimately, the team categorized it into eight buckets. I’ll let Joe rattle those off …
AGOR: Category one policies around staff responsibilities, shift and work incentives. Category two staff scheduling changes to manage demand, so scheduling of staff. Category three patient care policy changes. Four is COVID hospital testing policies. Five: interaction with external organizations. This could be government organizations or private organizations, nonprofits, you know, those types of policies revolving around those. Internal organization and system management. Seven was non-human resource. Resource allocation and management, not human resource, being a vehicle, transportation units, PPE, ventilators. Those are non-human resources. And then the eighth category was policies involving exposure mitigation.
PERRY: Like I said before, it’s a lot of data. And even within those categories, there are subcategories. Just thinking about staffing for example: You have doctors and nurses, yes, but here's policies and actions that were for the kitchen personnel, for janitors, for the people who change over rooms. There’s a lot of people who work at a hospital.
AGOR: We went through each of those, picked out certain actions, certain days and things that things that really had dates so that we could understand when those policies are, when those actions were taken by the hospital system.
PERRY: Once the team had a list of all these policies changes and when they went into effect, they could compare those to the hospitals health records and see what patterns emerged in regard to how well patient did.
So, now they have all this data. what did they find out? That’s kind of a tricky question to answer.
AGOR: When we're trying to establish what was successful, what was working with managing medical surge, it's not just counting policies and seeing what actions took place and making a correlation, because there are different variables like compliance or like movement of people, and many others that can inform causation, you can inform, different outcomes. So we're trying to capture now an excess amount of information and present that information alongside it.
PERRY: Still, there are tangible patterns that stand out. One example was the use of telehealth, where doctors meet with patients virtually.
AGOR: Something that we can say is that the telehealth initiative that was, that was instated here at SHS, was very successful in managing the surge of patients to the hospital system. That is something we see in the demand patterns to each of the different five hospitals within Samaritan Health Services. Telehealth was a thing for Samaritan Health Services before, but in 2019, there were maybe one or two throughout the year that was a telehealth.
Whereas in April of 2020, when we started transitioning, when they made that major transition, more than 50% of their patients moved to telehealth. Physicians needed to be trained. And again, this was in an effort to, to manage the amount of patients that were flowing and flowing in and out from the hospital, which again, that part was successful, that part was successful.
PERRY: There’s another part of the health care system that we haven’t really touched on yet. The patients. And I’m not talking just about people sick with COVID. What about the ones who should be seeing their doctor for another reason but aren’t?
AGOR: The thing that I think we want to get the message across is that surge is not just related to the pandemic at hand, whenever the influenza pandemic happened, they were not just only worried about the surge of influenza patients. We're not just only worried about the surge of COVID patients. And it's something that our physician we're working with here at some Samaritan Health Services keeps stressing is that there are still very deadly chronic conditions that were pre-pandemic that, you know, we need to make sure that we keep an eye on, we, we understand that we have to manage COVID absolutely cause that's killing people. But we have to make sure that we don't forget about the current chronic conditions and that surge of non-COVID is I think the thing that we want to present information on.
PERRY: Again here’s Bharat Gopal, the doctor from earlier.
GOPAL: When it comes to COVID, people have been scared, and with, with fear and being scared, they've avoided coming into health care institutions when they've been ill. The question is how much and for how long and what, what has happened to the severity of their chronic conditions or their acute conditions, if they've avoided the health care system.
If a patient comes in having waited three days or four days with chest pain, because they didn't want to, come in to the hospital and their heart attack has now progressed to heart failure, that can be a significant detriment to them and also the recovery, but also have a significant on the health care system. And I think that what Joe's study is really about is looking at what has happened, not only to the patient, but also to the health care system.
[MUSIC: “Spring Field” by Godmode, part of the YouTube Audio Library. Licensed under a creative commons license]
PERRY: The next phase of this project will go deeper into looking at non-COVID patient have fared. What happens with their hospitals stays? Were they significantly longer? Did they go to the ICU more often? Was there a higher mortality rate?
AGOR: Maybe that'll lead to people saying, well, let's investigate that a little bit further, right, and let's dive in, spend some money and identifying causation of these particular patient populations that are not COVID, but that somehow reflected in the data had some, some negative outcomes, negative impacts, if that actually is the result that comes out.
PERRY: That’s the thing about this kind of research, especially as the pandemic continues and our response to the virus shift and adapt: There’s always more data to look at, more decisions to document, causations to discover.
AGOR: Our group is having ongoing discussions about different ways that we can use this data we collected to push, to go in a different direction as we move forward. And it's going to be a function of what is needed as we move forward in this COVID pandemic. Is there another surge coming up? Now vaccination is a, main focus, right? And so that's, uh, I think my last tidbit is, uh, I'm hoping that we can take this and continue to use it in some force of good.
[MUSIC: “CloudStepping” by Will Havnaer]
PERRY: That’s a wrap on what we all hope is our one and only season about COVID.
This episode was hosted by me, Owen Perry. Will Havnaer mixed and edited the sound.
A special thank you to everyone on the Engineering Out Loud team who has worked to keep putting these episodes out despite all the challenges of the past year. Those folks are executive producer Rachel Robertson, hosts Steve Frandzel, Keith Hautula, and Chris Palmer. Johanna Carson takes care of all things visual and digital. We’ll be back soon with new episodes. Stay tuned.
Our intro music is “The Ether Bunny” by Eyes Closed Audio. You can find it on SoundCloud. We used with permission of a Creative Commons attribution license. The music and effects in this episode were also used with appropriate licenses. For more episodes, visit engineeringoutloud.oregonstate.edu, and subscribe by searching “Engineering Out Loud” on your favorite podcast app.