Title: COVID-19: a retrospective analysis comparing positive cases in the United States under the lens of age, location, and positive symptoms.
Authors: David Dommermuth, OMS1; Lindsey Petrelle, OMS1; Tianfu Shang, OMS1; Faculty: Dr. Anthony Stephas
Introduction
Emergency Departments (ED) have played a pivotal role in evaluating and characterizing the novel COVID-19 disease. Frontline data from select Washington state EDs were utilized to examine positivity rate as factored by age, location, and sentinel symptoms. The purpose of this research is to identify trends and correlations between these factors for the COVID-19 disease in 2020.
Methods
Data from 520,159 depersonalized patient encounters was obtained from Providence St. Joseph EDs within Washington State. This study calculated the positive test rate of COVID-19 in the categories of age group (0-14, 15-49, 50-64, 65+), location (n=17), and sentinel symptoms (headache, fever, cough, myalgia, nausea/vomiting, shortness of breath, sore throat, and wheezing). Significance was determined via p-tests and correlation tests.
Results
Through the lens of COVID-19 positivity rate, statistically significant variation was found between age groups (p=0.0015) and ED locations (p=8.7×10−7). The 0-14 age group had the lowest positive rate at 12.2%, while the 50-64 age group had the highest positive rate at 28.0%. In the context of ED locations, it was found Swedish Redmond had the lowest positive rate at 11.2%, while Kadlec Regional Medical Center had the highest positive rate at 47.6%. The average positive rate for all ages and all locations is 24.16%.
The positivity rate was investigated in correlation to reported symptoms for each age group. Overall, it was found that myalgia had the greatest correlation with positivity rate at 0.52. In the 0-14 age group, it was found that headache (0.78) and wheezing (0.66) demonstrated the strongest correlation. In the 65+ age group, all symptoms negatively correlated with the positivity rate.
Conclusions
Strong correlation between age group and positive test rate supports initial assumptions that risk of infection increases with age. The cause of significant variation in positivity rate between locations is an area for further investigation. Within the limitations of our methodology, our findings support that increasing test rate reduces the correlation between each symptom and positive test rate. Understanding the relationship between age, location, and symptoms in COVID-19 can assist front-line workers in maximally utilizing limited testing and treatment resources as COVID-19 continues to be a global risk factor.
A question from a judge: You seem to make a lot of the 6 x 8 table that showed one “statistically significant” correlation between age, symptom, and positivity. As I look at the table, you have 48 cells, and by chance I would expect TWO of these to have ” P < 0.05." Please consider adjusting your threshold P-value for multiple simultaneous comparisons. And what is "Unsymptomatic?" These patients probably had a specific symptom or another good reason to visit the ED, unless it was the only place to get a COVID-19 test (as happened in the beginning of the pandemic).
Hello Dr. Elliot,
Thank you for the inquiry. We utilized the Pearson Product Moment Correlation Coefficient Table to calculate our P-value. As we continue our research comparing 2020’s data to 2021’s data, we will keep your advice in mind as we consider other statistical methodologies for calculating P-value. To address the “Unsymptomatic” inquiry, this category in our analysis was patients who presented to the ED who did present with any of the eight sentinel symptoms (headache, fever, cough, myalgia, nausea/vomiting, shortness of breath, sore throat, and wheezing) but still tested positive for COVID-19.
Question: Your title implies that you looked at data from the entire US. However, your results are limited to Washington State. Any idea how these findings correlate to other states?
Hello Dr. Habecker,
Thank you for your question. Yes, I see the confusion. This is an error between where the project started to where the project ended. We first had more than a half-million data points as Providence St. Joseph stretches all over the United States; however, as we began to extrapolate the data, we found that it would be easier to narrow our analytics to just Washington State. I would love to give you an answer about the ability to utilize Washington States data thus implicating a representation of the entire United States, but alas, this is an error and should have been corrected when coping over from our presentation to the title box field.
Thank you – I was NOT assigned to judge your presentation, but I do have a question (and I agree that giving your project more accurate title to reflect the geographic focus would be wise). Do you have any other demographic information, aside from age group, that could explain the higher positive rates observed at Kadlec compared to Redmond? I’m sure you do, but did not hear it mentioned ;) Thanks again for your work.
Hello Dr. Briggs-Early,
Thank you for your inquiry! Unfortunately, due to the nature of the data we collected, we were not given information in the demographic category outside of that of age group, department visiting within the hospital, and location of the visit. However, in future work, we hope to include a wide range of demographic data like race and gender. We are currently waiting for IRB approval for 2021’s data to do a greater analysis between the years of 2020 and 2021. We hope that future work will give us the ability to compare urban locations to more rural locations like Kadlec to Redmond.
Thank you for presenting this work. As a judge I am hoping you can clarify the contribution to the science. Given the number of epidemiological studies that have assessed trends in COVID incidence by age, symptoms and state, what new insights does the present study add. For example, is there strength in focusing on local data versus national, or does the fact that the data were collected from the ED make a difference? Thanks!
Hello Dr. Fritz,
Thank you for your question! At the time of data collection and initial analysis, many of the now generally accepted conclusions were still novel or not widely known. I do concede that this data is not inducing any groundbreaking conclusion, but we hope to utilize this data as a starting point as we further explore 2021’s data in relation to 2020’s data. We hope that further extrapolation of data will bring novel conclusions that will continue to benefit scientific literature/understanding.
Thanks for presenting your analysis. As a judge I am wondering if you think, from this data analysis, that sentinel symptoms should be used at all to determine resource allocation (as in your first application point). With the low correlation to age groups and overall the high asymptomatic rate, how useful is using sentinel symptoms as a metric?
Hello Dr. Salido,
Thank you for your question. Symptoms have been a defining characteristic in the emergency department to characterize the possible presence of COVID-19. The prevalence of specific symptoms is quite low. Utilizing symptoms might be specific to age groups, but it is not sensitive at a population level. Sentinel symptoms are specific as well as sensitive at the individual level, yet this information cannot be blindly extrapolated to the general public.
In terms of public health education, this information could be used by individuals to combat the developed sigma that would bias others towards the individual. As to your point, it appears that using sentinel symptoms alone as a metric is not very specific for determining COVID-19 infection status. In this investigation, although some symptoms did indeed have a positive correlation, the significance varied widely from age group to age group, no single symptom has the ability to be a predictor of COVID-19 at the population level. However, further investigations could lead to age group-specific indicators for the presence or absence of COVID-19. Understanding the difference in COVID-19 geographic location, symptom, and age presentation can assist front-line workers in properly screening additional patients who may need further care. It is the hope of this investigation to not only utilize similar data collection methods to make predictions on how resources can best be allocated to combat future widespread infection but also to raise awareness for public compliance on further data collection efforts.