News

When Colorado Gov. Jared Polis and the Colorado Department of Public Health and Environment (CDPHE) needed to respond to the rapidly worsening COVID-19 pandemic, they needed data and projections of the pandemic’s course based upon Colorado data. A team was quickly assembled by the Colorado School of Public Health (ColoradoSPH) to collaborate with CDPHE on epidemic modeling—the basic approach for generating the needed information for decision-making. In March, CDPHE asked ColoradoSPH Dean Jonathan Samet, MD, MS, four questions: when will infections peak in Colorado, how soon will we reach ICU bed capacity, how many hospital beds will be needed when we do peak, and how will social distancing policies impact those numbers? 

To answer these questions, experts in the ColoradoSPH and other schools across the University of Colorado put their heads together to model and project “the curve” of the virus’s course using epidemiological models. These models are mathematical representations that reflect how viruses affect populations: infecting those who are susceptible, making some ill and in need of hospital and critical care, and leading to death for some. The models are used to examine how different measures, like closing restaurants and bars, slow and diminish the epidemic. Remembering that a model is a simulation of the world, none “are correct” but they should be useful and allow people to make hypotheses about the trajectory of a disease and the impacts of different policies and behavior changes. 

Epidemiology Avengers Assemble 

In response to the CDPHE request, Dean Samet put together an interdisciplinary and interinstitutional team that brought complementary expertise for epidemiological modeling. Katie Colborn, PhD, MSPH, assistant professor in the Department of Surgery at the CU School of Medicine and in biostatistics and informatics at ColoradoSPH, is an expert in modeling vector-borne diseases like malaria. Beth Carlton, PhD, MPH, assistant professor of environmental and occupational health at ColoradoSPH, brings experience in infectious disease epidemiology and disease control. David Bortz, PhD, associate professor of applied mathematics at CU Boulder, is a mathematical biologist who has used his experience in HIV infection spread to hone the model. Other members of the team include Andrea Buchwald, PhD (an experienced modeler and postdoctoral fellow in environmental and occupational health at ColoradoSPH), Debashis Ghosh, PhD (professor and chair of the Department of Biostatistics and Informatics at ColoradoSPH), Richard Lindrooth, PhD (professor of health systems, management and policy at ColoradoSPH), Tatiane Santos, PhD (instructor in health systems, management and policy at ColoradoSPH), social scientist Jimi Adams, PhD (associate professor of health and behavioral sciences at CU Denver), and Jude Bayham (assistant professor of agricultural and resource economics at Colorado State University). 

The Evolution of a Model 

In the time since the team started its work, its approaches have become more and more refined. At the start, Colorado-specific data about COVID-19 was limited, so the team turned to experience in other countries. It soon put together a Colorado-specific SEIR model, so-named because it moves people through four categories: Susceptible to the disease, Exposed to the disease, Infected with the disease, and Recovered from the disease (therefore immune). For those infected, the model identifies a proportion who need hospitalization and intensive care, and estimates how many will die. The model incorporates the infectiousness of the virus and contact among people, and also reflects the incubation period and how long people are ill. 

As the outbreak has progressed, the team has been able to utilize data gathered in Colorado to make better estimates for each of the inputs into the Colorado model. More accurate inputs mean more accurate outputs, and greater confidence in projections of when cases will peak, the number of ICU beds that will be needed, and the number of potential deaths. “It’s okay that models change over time,” Katie Colborn said. “They should change over time. We’re learning every day.” 

Differing and changing model results complicate understanding of the trajectory of the COVID-19 pandemic. Different groups are using different approaches to predict outcomes, and even those using similar models may use different input data. For example, the Institute for Health Metrics and Evaluation (IHME) model, originally predicted that COVID-19 cases peaked in Colorado on April 4 and has since produced estimates ranging from early to mid-April. The Colorado team has carefully explained how its model differs from those developed by IHME and others, commenting that having multiple modeling groups take on the same challenge is valuable, particularly if there is some convergence in the findings. 

Modeling What Comes Next 

With Gov. Polis shifting the statewide stay-at-home order to “safer-at-home” social distancing on April 27, the modeling team is now trying to answer new questions such as what is the best way to reduce physical distancing requirements while preventing a second peak? how might other measures, such as better case detection, help control the epidemic? and who are the most vulnerable populations and why? The answers to many of these questions hinge on being able to implement robust testing and screening protocols, but in the meantime, Colborn says there are elements of the model that the team can explore and change. They’re interested in looking at the disease’s behavior in different groups—for example, people of retirement age and older may be more willing and able to distance themselves because they aren’t working, while some people may not have jobs that allow them to work from home, leading to lower levels of physical distancing. With any model, it’s important to be aware that the results are just estimates—scientifically-supported estimates, but estimates nonetheless. Predicting future behaviors and their consequences inherently carries uncertainty that scientists, policymakers, and the general population should be aware of. “Model simulations are often used to illustrate assumptions and hypotheses,” said Colborn. “The exact predictions will never be perfect. They’re meant to aid in decision making.” 

This is where the CDPHE comes in. According to Dr. Rachel Herlihy, Communicable Disease Branch Chief at the CDPHE and the state epidemiologist, the team’s modeling work is being used on a daily basis to help guide policymaking and decision making in the state. “As public health scientists, we are fortunate to find ourselves in a state with science-minded leadership,” she said. 

Right now, the biggest policy decision is how to balance disease control and economic priorities. As more data comes in about the effects of individual physical distancing policies, like the closure of ski resorts, restaurants, and bars, and then the full stay-at-home order, the team is working with policymakers to find the appropriate combination of physical distancing measures to control the epidemic while also allowing the economy to begin to recover. 

CDPHE is very aware that modeling work will be critical to striking that balance. “We can allow some amount of disease transmission to happen to build immunity in the population, but we need to do it in a controlled way that protects the most vulnerable and keeps hospitals from being overwhelmed,” Herlihy said. 

What the modelers can say with confidence is that the SARS-CoV-2 virus isn’t going to disappear in the next few weeks, and that some level of precautionary measures will likely need to be practiced for months. “All signs point to another outbreak as soon as you stop effective intervention strategies,” Colborn said. “There’s a long road ahead, and even when the stay-at-home order ends, we need to be diligent about protecting ourselves, even if we go back to working in person.” 

In the meantime, the team is doing its best to provide timely updates to CDPHE and the governor’s office. Gov. Polis has presented the team’s data at several press conferences as a way of explaining the importance of physical distancing policies. Team members have found themselves frequently in mainstream news. “It turns out that the only time the media pay attention to mathematical and statistical modeling is during a global pandemic. Or maybe during a U.S. presidential race,” Colborn observed. 

Mid April, some Colorado Model team members presented a ColoradoSPH webinar, the COVID-19 Pandemic and Colorado: Epidemiology in Action, to more than 1,400 viewers—a record number of attendees for a lecture or discussion at ColoradoSPH. Drs. Samet, Carlton, and Herlihy of CDPHE presented, talking about the modeling team’s results to date and how those results are being used to inform decisions at the state level. The team is planning to present more webinars with updated results and policy impacts in the future. For at least one team member, it’s this collaboration that has been a shining light during the epidemic. “There has been an amazing sense of collaboration within our modeling team, between our modeling team and the state, between other scientists we’ve reached out to for input on parameters and their papers,” Beth Carlton said in closing the webinar. “And that really does make me hopeful that we will find a way forward that is positive for health, both in terms of COVID and all the other health ramifications.”  

Written by Tori Fosheim

When Colorado Gov. Jared Polis and the Colorado Department of Public Health and Environment (CDPHE) needed to respond to the rapidly worsening COVID-19 pandemic, they needed data and projections of the pandemic’s course based upon Colorado data. A team was quickly assembled by the Colorado School of Public Health (ColoradoSPH) to collaborate with CDPHE on epidemic modeling—the basic approach for generating the needed information for decision-making. In March, CDPHE asked ColoradoSPH Dean Jonathan Samet, MD, MS, four questions: when will infections peak in Colorado, how soon will we reach ICU bed capacity, how many hospital beds will be needed when we do peak, and how will social distancing policies impact those numbers? 

To answer these questions, experts in the ColoradoSPH and other schools across the University of Colorado put their heads together to model and project “the curve” of the virus’s course using epidemiological models. These models are mathematical representations that reflect how viruses affect populations: infecting those who are susceptible, making some ill and in need of hospital and critical care, and leading to death for some. The models are used to examine how different measures, like closing restaurants and bars, slow and diminish the epidemic. Remembering that a model is a simulation of the world, none “are correct” but they should be useful and allow people to make hypotheses about the trajectory of a disease and the impacts of different policies and behavior changes. 

Epidemiology Avengers Assemble 

In response to the CDPHE request, Dean Samet put together an interdisciplinary and interinstitutional team that brought complementary expertise for epidemiological modeling. Katie Colborn, PhD, MSPH, assistant professor in the Department of Surgery at the CU School of Medicine and in biostatistics and informatics at ColoradoSPH, is an expert in modeling vector-borne diseases like malaria. Beth Carlton, PhD, MPH, assistant professor of environmental and occupational health at ColoradoSPH, brings experience in infectious disease epidemiology and disease control. David Bortz, PhD, associate professor of applied mathematics at CU Boulder, is a mathematical biologist who has used his experience in HIV infection spread to hone the model. Other members of the team include Andrea Buchwald, PhD (an experienced modeler and postdoctoral fellow in environmental and occupational health at ColoradoSPH), Debashis Ghosh, PhD (professor and chair of the Department of Biostatistics and Informatics at ColoradoSPH), Richard Lindrooth, PhD (professor of health systems, management and policy at ColoradoSPH), Tatiane Santos, PhD (instructor in health systems, management and policy at ColoradoSPH), social scientist Jimi Adams, PhD (associate professor of health and behavioral sciences at CU Denver), and Jude Bayham (assistant professor of agricultural and resource economics at Colorado State University). 

The Evolution of a Model 

In the time since the team started its work, its approaches have become more and more refined. At the start, Colorado-specific data about COVID-19 was limited, so the team turned to experience in other countries. It soon put together a Colorado-specific SEIR model, so-named because it moves people through four categories: Susceptible to the disease, Exposed to the disease, Infected with the disease, and Recovered from the disease (therefore immune). For those infected, the model identifies a proportion who need hospitalization and intensive care, and estimates how many will die. The model incorporates the infectiousness of the virus and contact among people, and also reflects the incubation period and how long people are ill. 

As the outbreak has progressed, the team has been able to utilize data gathered in Colorado to make better estimates for each of the inputs into the Colorado model. More accurate inputs mean more accurate outputs, and greater confidence in projections of when cases will peak, the number of ICU beds that will be needed, and the number of potential deaths. “It’s okay that models change over time,” Katie Colborn said. “They should change over time. We’re learning every day.” 

Differing and changing model results complicate understanding of the trajectory of the COVID-19 pandemic. Different groups are using different approaches to predict outcomes, and even those using similar models may use different input data. For example, the Institute for Health Metrics and Evaluation (IHME) model, originally predicted that COVID-19 cases peaked in Colorado on April 4 and has since produced estimates ranging from early to mid-April. The Colorado team has carefully explained how its model differs from those developed by IHME and others, commenting that having multiple modeling groups take on the same challenge is valuable, particularly if there is some convergence in the findings. 

Modeling What Comes Next 

With Gov. Polis shifting the statewide stay-at-home order to “safer-at-home” social distancing on April 27, the modeling team is now trying to answer new questions such as what is the best way to reduce physical distancing requirements while preventing a second peak? how might other measures, such as better case detection, help control the epidemic? and who are the most vulnerable populations and why? The answers to many of these questions hinge on being able to implement robust testing and screening protocols, but in the meantime, Colborn says there are elements of the model that the team can explore and change. They’re interested in looking at the disease’s behavior in different groups—for example, people of retirement age and older may be more willing and able to distance themselves because they aren’t working, while some people may not have jobs that allow them to work from home, leading to lower levels of physical distancing. With any model, it’s important to be aware that the results are just estimates—scientifically-supported estimates, but estimates nonetheless. Predicting future behaviors and their consequences inherently carries uncertainty that scientists, policymakers, and the general population should be aware of. “Model simulations are often used to illustrate assumptions and hypotheses,” said Colborn. “The exact predictions will never be perfect. They’re meant to aid in decision making.” 

This is where the CDPHE comes in. According to Dr. Rachel Herlihy, Communicable Disease Branch Chief at the CDPHE and the state epidemiologist, the team’s modeling work is being used on a daily basis to help guide policymaking and decision making in the state. “As public health scientists, we are fortunate to find ourselves in a state with science-minded leadership,” she said. 

Right now, the biggest policy decision is how to balance disease control and economic priorities. As more data comes in about the effects of individual physical distancing policies, like the closure of ski resorts, restaurants, and bars, and then the full stay-at-home order, the team is working with policymakers to find the appropriate combination of physical distancing measures to control the epidemic while also allowing the economy to begin to recover. 

CDPHE is very aware that modeling work will be critical to striking that balance. “We can allow some amount of disease transmission to happen to build immunity in the population, but we need to do it in a controlled way that protects the most vulnerable and keeps hospitals from being overwhelmed,” Herlihy said. 

What the modelers can say with confidence is that the SARS-CoV-2 virus isn’t going to disappear in the next few weeks, and that some level of precautionary measures will likely need to be practiced for months. “All signs point to another outbreak as soon as you stop effective intervention strategies,” Colborn said. “There’s a long road ahead, and even when the stay-at-home order ends, we need to be diligent about protecting ourselves, even if we go back to working in person.” 

In the meantime, the team is doing its best to provide timely updates to CDPHE and the governor’s office. Gov. Polis has presented the team’s data at several press conferences as a way of explaining the importance of physical distancing policies. Team members have found themselves frequently in mainstream news. “It turns out that the only time the media pay attention to mathematical and statistical modeling is during a global pandemic. Or maybe during a U.S. presidential race,” Colborn observed. 

Mid April, some Colorado Model team members presented a ColoradoSPH webinar, the COVID-19 Pandemic and Colorado: Epidemiology in Action, to more than 1,400 viewers—a record number of attendees for a lecture or discussion at ColoradoSPH. Drs. Samet, Carlton, and Herlihy of CDPHE presented, talking about the modeling team’s results to date and how those results are being used to inform decisions at the state level. The team is planning to present more webinars with updated results and policy impacts in the future. For at least one team member, it’s this collaboration that has been a shining light during the epidemic. “There has been an amazing sense of collaboration within our modeling team, between our modeling team and the state, between other scientists we’ve reached out to for input on parameters and their papers,” Beth Carlton said in closing the webinar. “And that really does make me hopeful that we will find a way forward that is positive for health, both in terms of COVID and all the other health ramifications.”  

Written by Tori Fosheim

When Colorado Gov. Jared Polis and the Colorado Department of Public Health and Environment (CDPHE) needed to respond to the rapidly worsening COVID-19 pandemic, they needed data and projections of the pandemic’s course based upon Colorado data. A team was quickly assembled by the Colorado School of Public Health (ColoradoSPH) to collaborate with CDPHE on epidemic modeling—the basic approach for generating the needed information for decision-making. In March, CDPHE asked ColoradoSPH Dean Jonathan Samet, MD, MS, four questions: when will infections peak in Colorado, how soon will we reach ICU bed capacity, how many hospital beds will be needed when we do peak, and how will social distancing policies impact those numbers? 

To answer these questions, experts in the ColoradoSPH and other schools across the University of Colorado put their heads together to model and project “the curve” of the virus’s course using epidemiological models. These models are mathematical representations that reflect how viruses affect populations: infecting those who are susceptible, making some ill and in need of hospital and critical care, and leading to death for some. The models are used to examine how different measures, like closing restaurants and bars, slow and diminish the epidemic. Remembering that a model is a simulation of the world, none “are correct” but they should be useful and allow people to make hypotheses about the trajectory of a disease and the impacts of different policies and behavior changes. 

Epidemiology Avengers Assemble 

In response to the CDPHE request, Dean Samet put together an interdisciplinary and interinstitutional team that brought complementary expertise for epidemiological modeling. Katie Colborn, PhD, MSPH, assistant professor in the Department of Surgery at the CU School of Medicine and in biostatistics and informatics at ColoradoSPH, is an expert in modeling vector-borne diseases like malaria. Beth Carlton, PhD, MPH, assistant professor of environmental and occupational health at ColoradoSPH, brings experience in infectious disease epidemiology and disease control. David Bortz, PhD, associate professor of applied mathematics at CU Boulder, is a mathematical biologist who has used his experience in HIV infection spread to hone the model. Other members of the team include Andrea Buchwald, PhD (an experienced modeler and postdoctoral fellow in environmental and occupational health at ColoradoSPH), Debashis Ghosh, PhD (professor and chair of the Department of Biostatistics and Informatics at ColoradoSPH), Richard Lindrooth, PhD (professor of health systems, management and policy at ColoradoSPH), Tatiane Santos, PhD (instructor in health systems, management and policy at ColoradoSPH), social scientist Jimi Adams, PhD (associate professor of health and behavioral sciences at CU Denver), and Jude Bayham (assistant professor of agricultural and resource economics at Colorado State University). 

The Evolution of a Model 

In the time since the team started its work, its approaches have become more and more refined. At the start, Colorado-specific data about COVID-19 was limited, so the team turned to experience in other countries. It soon put together a Colorado-specific SEIR model, so-named because it moves people through four categories: Susceptible to the disease, Exposed to the disease, Infected with the disease, and Recovered from the disease (therefore immune). For those infected, the model identifies a proportion who need hospitalization and intensive care, and estimates how many will die. The model incorporates the infectiousness of the virus and contact among people, and also reflects the incubation period and how long people are ill. 

As the outbreak has progressed, the team has been able to utilize data gathered in Colorado to make better estimates for each of the inputs into the Colorado model. More accurate inputs mean more accurate outputs, and greater confidence in projections of when cases will peak, the number of ICU beds that will be needed, and the number of potential deaths. “It’s okay that models change over time,” Katie Colborn said. “They should change over time. We’re learning every day.” 

Differing and changing model results complicate understanding of the trajectory of the COVID-19 pandemic. Different groups are using different approaches to predict outcomes, and even those using similar models may use different input data. For example, the Institute for Health Metrics and Evaluation (IHME) model, originally predicted that COVID-19 cases peaked in Colorado on April 4 and has since produced estimates ranging from early to mid-April. The Colorado team has carefully explained how its model differs from those developed by IHME and others, commenting that having multiple modeling groups take on the same challenge is valuable, particularly if there is some convergence in the findings. 

Modeling What Comes Next 

With Gov. Polis shifting the statewide stay-at-home order to “safer-at-home” social distancing on April 27, the modeling team is now trying to answer new questions such as what is the best way to reduce physical distancing requirements while preventing a second peak? how might other measures, such as better case detection, help control the epidemic? and who are the most vulnerable populations and why? The answers to many of these questions hinge on being able to implement robust testing and screening protocols, but in the meantime, Colborn says there are elements of the model that the team can explore and change. They’re interested in looking at the disease’s behavior in different groups—for example, people of retirement age and older may be more willing and able to distance themselves because they aren’t working, while some people may not have jobs that allow them to work from home, leading to lower levels of physical distancing. With any model, it’s important to be aware that the results are just estimates—scientifically-supported estimates, but estimates nonetheless. Predicting future behaviors and their consequences inherently carries uncertainty that scientists, policymakers, and the general population should be aware of. “Model simulations are often used to illustrate assumptions and hypotheses,” said Colborn. “The exact predictions will never be perfect. They’re meant to aid in decision making.” 

This is where the CDPHE comes in. According to Dr. Rachel Herlihy, Communicable Disease Branch Chief at the CDPHE and the state epidemiologist, the team’s modeling work is being used on a daily basis to help guide policymaking and decision making in the state. “As public health scientists, we are fortunate to find ourselves in a state with science-minded leadership,” she said. 

Right now, the biggest policy decision is how to balance disease control and economic priorities. As more data comes in about the effects of individual physical distancing policies, like the closure of ski resorts, restaurants, and bars, and then the full stay-at-home order, the team is working with policymakers to find the appropriate combination of physical distancing measures to control the epidemic while also allowing the economy to begin to recover. 

CDPHE is very aware that modeling work will be critical to striking that balance. “We can allow some amount of disease transmission to happen to build immunity in the population, but we need to do it in a controlled way that protects the most vulnerable and keeps hospitals from being overwhelmed,” Herlihy said. 

What the modelers can say with confidence is that the SARS-CoV-2 virus isn’t going to disappear in the next few weeks, and that some level of precautionary measures will likely need to be practiced for months. “All signs point to another outbreak as soon as you stop effective intervention strategies,” Colborn said. “There’s a long road ahead, and even when the stay-at-home order ends, we need to be diligent about protecting ourselves, even if we go back to working in person.” 

In the meantime, the team is doing its best to provide timely updates to CDPHE and the governor’s office. Gov. Polis has presented the team’s data at several press conferences as a way of explaining the importance of physical distancing policies. Team members have found themselves frequently in mainstream news. “It turns out that the only time the media pay attention to mathematical and statistical modeling is during a global pandemic. Or maybe during a U.S. presidential race,” Colborn observed. 

Mid April, some Colorado Model team members presented a ColoradoSPH webinar, the COVID-19 Pandemic and Colorado: Epidemiology in Action, to more than 1,400 viewers—a record number of attendees for a lecture or discussion at ColoradoSPH. Drs. Samet, Carlton, and Herlihy of CDPHE presented, talking about the modeling team’s results to date and how those results are being used to inform decisions at the state level. The team is planning to present more webinars with updated results and policy impacts in the future. For at least one team member, it’s this collaboration that has been a shining light during the epidemic. “There has been an amazing sense of collaboration within our modeling team, between our modeling team and the state, between other scientists we’ve reached out to for input on parameters and their papers,” Beth Carlton said in closing the webinar. “And that really does make me hopeful that we will find a way forward that is positive for health, both in terms of COVID and all the other health ramifications.”  

Written by Tori Fosheim