COVID-19 Forecast for El Paso County — Dec. 14
Plus, our resident microbiologist on how she makes her predictions
Good morning, and happy Monday.
We’d like to start today’s newsletter with some humility and a moment of transparency. We are nearing our 100th newsletter, and we pride ourselves on only ever issuing one correction. However, since publishing Wednesday’s newsletter, we have learned that some of the information we reported, as attributed to others, was inaccurate, and we are issuing some corrections.
CORRECTIONS:
How we quoted a source: “There are a gazebo and an empty pool full of dirt outside, and none of the doors had any locks.” How we corrected it: “There are a gazebo and an empty pool outside, and some of the doors did not have working locks.”
We quoted another source as saying, “I had a friend who literally, her ceiling fan was hanging off the ceiling, like by a wire.” We have since learned there are no fans in Bijou West, so we have deleted this quote.
❓How this happened: One of our first-year sources clarified for The CC COVID-19 Reporting Project that during her stay in Block 1, some grass in the courtyard was patchy, so some of the ground looked to her like dirt. The pool was empty, but it was not “full” of dirt. They also clarified that the individual rooms in Bijou West had locks on the doors, but we learned in our follow-up reporting that some of the locks in the building didn’t work. As for the ceiling fan, another one of our sources clarified they heard about the fan from a friend of a friend, and did not see it firsthand.
➡️Moving forward: When we make mistakes, we’ll own up to them. Today, we apologize for allowing these to slip through, and frankly, we’re embarrassed we have to write this. We have a rigorous fact-checking process, which you can read about here, and it will only grow more rigorous. We have learned from our mistakes of not thoroughly vetting information we receive from our sources, and are committed to doing a better job at this in the future.
Onward. In today’s newsletter, Phoebe Lostroh returns to give her weekly COVID-19 forecast for El Paso County and to explain how she creates her prediction models. Lostroh is a professor of molecular biology at Colorado College on scholarly leave who is serving as the program director in Genetic Mechanisms, Molecular and Cellular Biosciences at the National Science Foundation.
Phoebe’s Forecasts
NOTES: These forecasts represent her own opinions and not necessarily those of the National Science Foundation or Colorado College. She used the public El Paso County dashboard for all data. Lostroh prepared these forecasts on Dec. 12.
⚖️ How her predictions last week shaped up: Dec. 12 is the last day of Morbidity and Mortality Weekly Report week 50 in the national public health calendar. It is the 40th week since the first case was detected in El Paso County. Since March 13, 428 El Paso County residents have died of COVID-19. Last week, Lostroh predicted about 5,300-6,000 new cases in El Paso County during the week ending Dec. 10. Instead, there were many fewer, only 3,848.
“The estimation for last week was most likely too high because I did not take the possible impact of the safety precautions for ‘Level Orange’ sufficiently into account,” Lostroh said. “Because the other disease indicators were still very concerning, I thought that we would fall between the low and middle estimation curves.”
The COVID-19 outbreak in El Paso County has grown less predictable because of human behavior, according to Lostroh. Events that might increase spread include Halloween, which was about six weeks ago, and Thanksgiving, which was about two weeks ago. The implementation of county-wide safety precautions associated with “Level Red: Severe Risk” might decrease spread.
Predicted cumulative reported cases in El Paso County
🗝️ Key points: The calculated estimates are based on several assumptions. The first assumption is that the virus is spreading exponentially with a reproduction number greater than one, which would mean every known case infects more than one person on average. For the best-case scenario, Lostroh assumes the reproduction number is exactly one. A second assumption is that most people in the county still remain susceptible to the coronavirus. Given the high percent positivity, the inability of El Paso County to do sufficient contact-tracing for all cases, and El Paso County’s level in the Colorado dial framework, Lostroh estimates next week’s cumulative cases will fall somewhere between the best-case scenario and lowest exponential curve-fit. That would mean 4,330-4,525 new cases for the week ending Dec. 17.
Cumulative reported hospitalizations, deaths, and cases of COVID-19 in El Paso County
🗝️ Key points: Cumulative hospitalizations are in red squares, and deaths are plotted on the left-hand Y-axis in black triangles. Cumulative reported cases are represented by blue circles on the right-hand Y-axis. Horizontal boxes filled with a gradient from white to purple indicate the six weeks following an event that might reduce spread. Horizontal boxes filled with a gradient from white to pink indicate the six weeks following an event that might increase spread. Triangles show when incidence and contact-tracing crossed important thresholds, while boxes underneath the curve show the rolling 14-day percent positivity.
14-day cumulative rolling incidence, annotated with changes in local policy and activities
🗝️ Key points: The actual calculated incidences are in black Xs, while the other symbols provide estimates based on exponential curve-fitting. When percent positivity is higher than 5%, some cases in the community are going undetected, Lostoh said. The percent positivity in El Paso County is currently about 15%, as it has been for about three weeks.
Q-and-A with Lostroh: Our resident microbiologist on how she creates her weekly forecasts
This interview has been edited for length and clarity.
CC COVID-19 Reporting Project: Walk us through your process for putting forecasts together. Why do you use exponential curve-fits?
Lostroh: I take the cumulative cases and I plot them over time. Then I make exponential curves that fit the most recent 24, 14 or seven days of the data, and I calculate what would happen in one week, two weeks, three weeks, and so on, and write that down. The reason I use an exponential curve is that when cases are increasing, spread is exponential. So when the cases are consistently going up, what I see is that the seven-day curve is steeper than the 14-day curve, which in turn is steeper than the 21-day curve, because they’re just all increasing. Under those conditions, that’s when the forecast is the most accurate.
CCRP: Why has it been more difficult lately to issue accurate predictions for COVID-19 in El Paso County?
Lostroh: For the previous three weeks, sometimes the 21-day curve has been in between the 7- and 14-day curves. That indicates that human behavior is not constant, and it’s harder to predict that the virus will continue to spread exponentially. The virus will only spread exponentially if there’s more than one new case for every existing case. If there’s exactly one case per existing case, then the spread is just linear, and it’s not so bad — obviously, we want to eventually get to a situation where there’s less than one case for a new case. So the spread potentially not being exponential makes the forecast less reliable. Another thing that makes the forecast less reliable is events and interventions that ought to either increase or decrease spread have been overlapping in time. It takes between two and six weeks to see the impact of a policy change or a big event in the case record. When you make changes that overlap by two weeks, you really are not letting the first one have its impact before the next one starts. So what we would like to do is not make any changes for four weeks and see what happens before we make another change, but we haven’t been doing that. Instead, what happens is people get impatient, and we change public health measures with a shorter frame of time.
CCRP: How do you use current and past data to predict future COVID incidence rates?
Lostroh: For the incidence prediction, I use the real numbers to look at the incidence as it actually is, and then I use those predicted numbers to calculate the incidence if the worst-case, medium-case, or best-case scenario were to happen. Now this week, and in other weeks, when it appears that the case rate might be going flat or even going down, I just use a linear fit for the best-case scenario. That’s because if the spread is not exponential anymore, a linear curve will describe the situation better than an exponential curve. But we are in a situation where human behavior is not predictable at the moment. We just had Thanksgiving about two ago, which is probably going to turn out to be a spreading event, and we also changed the dial to “Level: Red” safety precautions a day after Thanksgiving. So, to what extent are those two things going to counteract one another? That can’t be solved by math; that’s a judgment call. So what I’m going to do for next week is predict that the most likely number of cases will fall between the best case, where it’s been a very slow linear rise for the last three weeks, and the lowest exponential fit curves. The reason I haven’t just predicted it will be the best case is that the percent positivity is still almost 15%, which says to me that we are not detecting all of the cases that there are. When the percent positivity is that high, it’s almost certain that spread is still exponential and not just one new case per existing case. So I’m hedging my bets and thinking it’ll be between the linear fit and the lowest exponential curve for next week.
About the CC COVID-19 Reporting Project
The CC COVID-19 Reporting Project is created by Colorado College student journalists Isabel Hicks, Esteban Candelaria, Lorea Zabaleta, and Miriam Brown in partnership with The Catalyst, Colorado College’s student newspaper. Work by Phoebe Lostroh, Associate Professor of Molecular Biology at CC and National Science Foundation Program Director in Genetic Mechanisms, Molecular and Cellular Biosciences, will appear from time to time, as will infographics by Colorado College students Rana Abdu, Aleesa Chua, Sara Dixon, Jia Mei, and Lindsey Smith.
The project seeks to provide frequent updates about CC and other higher education institutions during the pandemic by providing original reporting, analysis, interviews with campus leaders, and context about what state and national headlines mean for the CC community.
📬 Enter your email address to subscribe and get the newsletter in your inbox each time it comes out. You can reach us with questions, feedback, or news tips by emailing ccreportingproject@gmail.com.