Analysis of Respiratory viruses in Chicago from 2019 to 2026

Introduction

This public dataset contains an weekly respiratory virus laboratory data used by the Chicago Department of Public Health to monitor the activity of respiratory viruses in Chicago. The respiratory viruses include influenza, RSV, SARS-CoV-2, parainfluenza, rhinovirus/enterovirus, adenovirus, human metapneumovirus, and seasonal coronaviruses. The data is reported on a weekly basis and doesn’t contain patient information. The data is reported from viral PCR tests conducted by hospital laboratories and two commercial laboratories in Chicago. This dataset is accessed from here. The description of the columns in the dataset can be found here.

This dataset defines season as a reporting period that starts in MMWR week 40 of one year and ends in week 39 of the following year. The season doesn’t follow the traditional Jan-Dec year but follows the epidemiological year used by the CDC. For example, Week 1 starts in Jan and Week 40 starts in early October. A respiratory season would start on Week 40 (early Oct) and end at Week 39 (late Sept of next year).

My research questions are:

How does cumulative percent positivity evolve over the season for each pathogen from 2019 to 2026?
What is the trend of cumulative percent positivity of the top four viruses in each seasonal period (Early, Mid, Late) during 2019-2020?

Methods

I did most of my data cleaning and wrangling in the source script dr_clean.R. Since I didn’t need the entire data, I created two smaller datasets from the main one: seasonal_burden, and seasonal_groups. In seasonal_burden, I calculated the cumulative percent positive from total positives and total tested. I used this dataset to answer the first RQ. In seasonal_groups, I categorized weeks into three different seasonal periods (Early,Mid,Late) during 2019 to 2020. I used seasonal_groups to answer the second RQ.

Results

Seasonal Burden Across 2019 to 2026

Table 1: Cumulative percent positivity from 2019-2026
season	mean	median	min	max
2019-2020	7.86	5.22	1.69	19.82
2020-2021	4.49	2.44	0.02	14.77
2021-2022	5.64	2.95	1.93	17.80
2022-2023	5.74	4.22	2.21	16.37
2023-2024	5.06	4.06	1.90	13.39
2024-2025	4.76	2.90	1.65	13.62
2025-2026	4.69	2.85	0.14	14.50

Table 1 shows a brief summary of cumulative percent positivity of pathogens over the years. According to the table, 2019-2020 season had the highest mean of 7.86. Years 2021 to 2024 are closely similar.

seasonal_burden |> 
    ggplot(aes(x = total_tested, y = total_positives)) +
    geom_point() +
    geom_smooth() +
    labs(
        title = "Fig. 1: Total number of tests compared to total positive tests"
    )

`geom_smooth()` using method = 'loess' and formula = 'y ~ x'

Fig.1 shows a positive correlation between total positives and the total number of tests over the years. The outliers with the highest tests and total positives are attributed to SARS-CoV-2 which only shows how much of an impact COVID-19 had on respiratory testing and detection during this time.

Table 2:Summary statistics by intensity
intensity	n	mean_positives	mean_tests	mean_percent_positives
Low	21	307	11,343	3.36
Medium	26	1,970	56,040	4.97
High	17	12,736	145,562	10.52

Table 2 shows a summary statistics of intensity of the positive tests during this time. Intensity is based on the number of total positives. High intensity is total positives > 5000, Medium intensity is total positives between 1000 and 5000 and low intensity is total positives < 1000. It is interesting to see that the mean percent positive for both low and medium intensity are similar despite having a vastly different number of mean tests.

Fig.1 shows a stacked bar plot of cumulative positive specimens of the pathogen by season. I excluded the seasons from 2010 to 2019 because it was only Influenza during those time periods. In 2019-2020, Influenza and SARS-CoV-2 have the most cumulative positives but starting 2020, Rhinovirus became the dominant pathogen. According to the figure, Rhinovirus shows a larger cumulative positive segment than SARS-CoV-2 in later seasons, possibly due to fewer amount of testing compared to SARS-CoV-2.

To check this theory, I plotted a percentage of positive tests and total cumulative tests of both Rhinovirus and SARS-CoV-2 in 2019 to 2026 for side by side comparison.

Fig. 3 shows that SARS_CoV-2 tests significantly exceeded tests for Rhinovirus as shown on the right. Even though Rhinovirus tests were much fewer, the test percent postivity rate was much higher than SARS-CoV-2. This confirms my theory that a higher proportion of Rhinovirus tests were positive relative to the number of tests being performed.

Pathogen Activity During Seasonal Period

Table 3. Cumulative positive specimens by week period
week_period	n	mean	median	sd	max
Early Season	89	3,215	935	4,845	24,436
Mid Season	99	5,653	1,557	9,933	55,618
Late Season	61	678	66	1,414	7,494

The data shows that positive specimens had the highest average of 5653 during the Mid season (weeks 21-40). This suggests that mid season is the most active time for pathogen circulation as it dominates all the metrics compared to other periods. Late season (week>40) seems to be the least active time with a mean of 678 . The minimum shows that there were some weeks when there were no cumulative positive specimens, reflecting low activity during some periods.

The most frequent pathogens are identified to be SARS-CoV-2, Influenza, RSV, Seasonal Coronaviruses, and a mix of other pathogens during each seasonal period. Fig.4 shows that the cumulative number of positive tests was dominated by SARS-CoV-2 across all seasonal periods. During mid season, it seems like the cumulative positives of SARS-CoV-2 is almost half of the total cumulative positives. During Late season, the number of all cumulative positives from all pathogens declined sharply. It is also interesting to note that seasonal coronavirus is only present during mid season.

Discussion

Over the years, it seems like seasonal burden from respiratory pathogens was the highest during 2019 to 2020. The rest of the years following after had a similar trend in cumulative percent positive tests but the years 2021 to 2023 had almost the same seasonal burden (Figure 2). However, this figure doesn’t really capture which pathogen had the most impact in a given year because testing volume differed across pathogens. This can be seen in the case of Rhinovirus and Covid-19. In Fig.2, it seemed like Rhinovirus contributed the most to the seasonal burden. However, this was because it had a high proportion of positive tests despite being a lot fewer tests than Covid-19 (Fig.3).

When examining the seasons more closely divided by early, mid and late season in 2019 to 2020, we can see that Covid-19 had a larger increase in positive specimens compared to other pathogens. This suggests that examining trends at a finer scale may be more helpful in understanding how specific pathogens drive disease burden within a season.