Long Term Assessment of SARS-CoV-2 in Wastewater and the Transition to Evaluate Additional Viral Targets

Abstract

The COVID-19 pandemic caused by the SARS-CoV-2 virus dramatically impacted society over five years ago and continues to have an impact today. Since the beginning of the pandemic there have responses and strategies implemented to maintain the public safety of communities affected by SARS-CoV-2. This study is a unique opportunity to analyze nearly four years of SARS-CoV-2 wastewater-based surveillance (WBS) data, obtained from five different laboratories, combined with four years of human health data from three adjacent regions of a large urban community (Miami-Dade County). The objective of this study was to analyze that data and evaluate longitudinal and geographic trends of SARS-CoV-2 levels in wastewater (WW) during the extensive time frame of this study. Additionally, WBS data were analyzed for multiple targets (influenza A/B, norovirus, RSV, HMPV, PMMoV) other than SARS-CoV-2 to assess the potential for expanding WBS to a wider range of targets. We found that SARS-CoV-2 levels correlated strongest with clinical positivity rates across all three geographic regions (Spearman r=0.81 for the entire period of record), with the most geographically restricted region showing higher correlations (South, r=0.86) than the region with populations with higher geographic mobility (North, r=0.69). Stronger correlations (0.80<r<0.97) were observed when correlations were established by variant waves rather than single or multiple year time frames (0.73<r<0.88). When analyzing the data for targets beyond SARS-CoV-2, results show promise as two laboratories detected norovirus, influenza A/B, RSV, and HMPV at statistically not different frequencies (Chi-squared≥0.6). Overall, results suggest that the clinical metrics used (e.g., positivity), geography, and the time frames of data analyses influence the ability of WBS to predict disease prevalence in a community. The consistency among the laboratories supports that the measurement of a wider range of viral targets can be disaggregated among different laboratories providing flexibility for building national-level WBS programs.

1. Introduction

In 2020, COVID-19 was declared a global pandemic by the World Health Organization (WHO, 2020). Although the number of deaths attributed to COVID-19 has decreased considerably, it continues to impact public health as 500 to 1000 deaths per week in the US are currently attributed to COVID-19 (CDC 2025a). Within that time, wastewater-based surveillance (WBS) has been used to track COVID-19 worldwide (Ahmed et al., 2020; Haramoto et al., 2020; La Rosa et al., 2020; Lu et al., 2020; Medema et al., 2020a; Sherchan et al., 2020; Fitzgerald et al., 2021; Giraud-Billoud et al., 2021; Kumar et al., 2021; Zhan et al., 2023; Radvák et al., 2024; Sanguino-Jorquera et al., 2024). WBS has also been used in more localized approaches to document and predict community level (Miyani et al., 2021; Zhu et al., 2021), neighborhood level (Starke et al., 2024), and building-level (Zambrana et al. 2022; Amirali et al. 2024; Solo-Gabriele et al. 2025) health trends. Accurate tracking of the disease within wastewater (WW) is possible because SARS-CoV-2 sheds in the feces, and other bodily fluids, of both symptomatic and asymptomatic individuals (Jones et al., 2020; Medema et al., 2020b; Sharkey et al. 2021). As a result, WBS can track pathogen presence without requiring individuals to report their illness, thereby avoiding some of the biases associated with clinical surveillance (Chiolero et al., 2013; Kitajima et al., 2020; Haak et al., 2021).

Most studies involving WBS of COVID-19 performed their analysis on data spanning less than a year (Hata et al., 2020; La Rosa et al., 2020; Medema et al., 2020a; Nemudryi et al., 2020; Peccia et al., 2020; Randazzo et al., 2020; Greenwald et al., 2021; Weidhaas et al., 2021), fewer analyzed one to two years (Acosta et al., 2022; Herrera-Uribe et al., 2022; Amirali et al. 2024; Duvallet et al. 2022; Sovová et al., 2024; Kaplan et al. 2022; Yale 2025), and even fewer have published data spanning more than two years (Boehm 2023, 2024; Zhan et al. 2023; Schenk et al. 2024; Zulli et al. 2025). There is a shortage of research analyzing the relationship between clinical human health (HH) metrics and WW SARS-CoV-2 levels over long periods of time. Longer time frame analysis can provide evidence for how the relationships between WW SARS-CoV-2 levels and clinical HH metrics (e.g., positive case counts, testing numbers, and positivity rates) are affected by changes in outside factors like variant dominance, testing availability, social perceptions, and vaccination uptake.

The primary objective of this study was to analyze WW SARS-CoV-2 data and the corresponding COVID-19 HH data of an extended period to evaluate longitudinal and geographic trends associated with COVID-19 across Miami-Dade County (MDC). Analysis of matching WW SARS-CoV-2 data from the three (North, Central, and South) regional wastewater treatment plants (WWTP) servicing MDC over a 46-month record (March 2020 to December 2023) allowed for an evaluation of the geospatial relationship between WW SARS-CoV-2 data and reported clinical HH data. WW data for the 46 months corresponded to the total of five laboratories many with many months of data collection that overlapped in time.

The secondary objective of this study was to compare WW data for multiple molecular targets from different laboratories to assess how WBS can be expanded in the future for monitoring a broader range of infectious diseases. Each laboratory was permitted to process samples as per laboratory-specific internal standard operating procedures which provided the opportunity to evaluate consistency among laboratories each implementing their own quality control criteria. Each of the five labs processed some combination of molecular targets including but not limited to SARS-CoV-2, Influenza A/B, Norovirus, RSV, HMPV, and PMMoV over some unique time frame that fell within the 46-month study period. We compared WW data for common targets across labs to investigate the applicability of WBS to a broader range of targets.

2. Methods

2.1. Data Sets

This study analyzed two major data sets, a WW data set and a HH data set. The WW data set was comprised of calculated levels of seven viral targets (6 pathogens, 1 human waste indicator) from among the five laboratories for which data were available. The maximum period of record provided from any single lab spanned nearly four years (March 2020 to January 2024). Not all labs provided data spanning the entire four years and on all viral targets (Table 1). Two laboratories (Laboratory 1 and 2) provided data spanning over 40 months. Laboratory 3 provided data for 20 months, and Laboratories 4 and 5 each provided data for 12 months. WW samples were obtained from among the three WWTPs in Miami Dade County. Each plant services a different population size. The South District (SD) plant services a population of 920,000, the Central District (CD) plant services 830,000, and the North District (ND) plant services a population of 780,000.

Table 1.

SARS-CoV-2, PMMoV, and additional disease targets by laboratory, inclusive of periods of record.

	Laboratory 1	Laboratory 2a	Laboratory 3b	Laboratory 4c	Laboratory 5d
Period of Record	Mar. 2020 to Nov. 2023	Jan. 2021 to Jan. 2024	Jan. 2021 to Aug. 2022	Jan. 2023 to Jan. 2024	Jan. 2023 to Jan. 2024
WWTP sampled	SD, CD, ND	CD	CD	CD	SD, CD, ND
SARS-CoV-2	X	X	Xa	X	Xb
PMMoV	X	X			X
Additional Disease Targets Measured by Multiple Laboratories
Influenza A		X		X	X
Influenza B		X		X	X
RSV A/B		X		X	X
Human metapneumovirus				X	X
Norovirus GII				X	X

^aLaboratory 2 also analyzed for poliovirus. Poliovirus was chosen by laboratory 2 due to a case of paralytic polio in New York during July 2022. All samples analyzed for poliovirus were negative. Mpox and C. auris were also analyzed by laboratory 2 during the summer of 2022. See Sharkey et al. 2023 and Babler et al. 2023a, respectively, for details.

^bLaboratory 3 also analyzed for emerging variants of SARS-CoV-2 through targeted sequencing. See Tierney et al. 2024 for details.

^cLaboratory 4 also analyzed for adenovirus, Mpox, and the GI strain of norovirus.

^dLaboratory 5 also analyzed for specific variants of SARS-CoV-2 as primers were developed. In addition, laboratory 5 analyzed for rotavirus, adenovirus, Mpox, EVD68, and hepatitis A. (Boehm et al, 2024).

Laboratories 1 and 5 received WW samples from all three plants, whereas the remaining laboratories (2, 3, and 4) received samples from only the CD plant. Wastewater treatment at the plants includes a grit chamber followed by pure oxygen aeration and secondary clarification. All samples were raw untreated wastewater collected from the head of the grit chamber. Each sample was collected as 24-hour (mid-night to mid-night) flow-weighted composite sample using a refrigerated autosampler (HACH AS950 fitted with an IO9000 for flow proportional sampling). Samples for laboratories 1 (hybrid commercial and research lab) and 5 (hybrid commercial and research lab) consisted of raw wastewater shipped on wet ice by overnight mail directly from the WWTP’s. Samples for Laboratories 2, 3, and 4 were collected in-person, transported on ice, then split, concentrated, and preserved in lysis solution. All sample splits designated for Labs 2, 3, and 4 went through initial processing (concentrated, preserved in lysis, and stored at −80 degrees Celsius) at a centralized laboratory at the University of Miami the same day as sample collection, as per standard operating procedures (Babler et al. 2025). Concentrates designated for Lab 4 were shipped in batches to Lab 4 for extraction and ddPCR processing. Concentrates for Lab 2 and Lab 3 were extracted and run by qPCR at the University of Miami.

Pathogenic target data included in this study included only targets that had at least two laboratories analyze for that target over a consistent period of at least 11 months. Among the six pathogenic targets, SARS-CoV-2 was the only target analyzed for samples from all three WWTPs and by all five laboratories. The human waste indictor, Pepper Mild Mottle Virus (PMMoV), was also a common target measured by 3 of the 5 laboratories. PMMoV is commonly found in wastewater due to the human dietary intake of peppers and is therefore used as a human waste control and, in some studies, used to normalize pathogen target data to account for dilution of human waste in wastewater (Maal-Bared et al. 2022; Zhan et al. 2022, Hsu et al. 2022). The five pathogenic viral targets analyzed beyond SARS-CoV-2 included four respiratory targets influenza A, influenza B, respiratory syncytial virus A/B (referred herein as RSV), and human metapneumovirus (HMPV)) which were reportedly circulating among the population at the time of this study. The one additional viral target that met the inclusion criteria, norovirus GII (referred to herein as norovirus), is known to transmit via fecal oral routes. Norovirus is one of the most common causes of vomiting and diarrhea in the United States (CDC 2025b). Although, as it is a generally self-limiting disease, it is typically not treated and therefore not as extensively tracked by clinical methods as the respiratory diseases.

2.2. Laboratory Methods for WW

All laboratories followed similar steps when processing the collected wastewater for molecular analysis which included concentration of the wastewater, nucleic acid extraction, and finally either digital droplet polymerase chain reaction (ddPCR) [Lab 4 and Lab 5] or quantitative PCR (qPCR) [Lab 1, 2, 3] analysis. Raw wastewater samples were concentrated through either precipitation or magnetic beads (Lab 1), filtration or magnetic beads (Lab 3), filtration only (Lab 2, and 4), or sedimentation/centrifugation (Lab 5). Raw wastewater sample volumes processed varied from 30 to 150 mL. Nucleic acids were extracted from the concentrates using traditional methods involving TRIzol and chloroform or using commercially available kits. The purified nucleic acids were then subject to either droplet digital PCR or quantitative PCR (qPCR) for detection using target-specific primers and fluorescent reporter probes. PCR quantification was performed through either standard curve analysis or through the most probable number/direct quantification approach employed in digital droplet PCR technologies. All wastewater data measured by PCR were reported in gene copies (gc) per liter, and the data were not normalized. Additional details are available in Table 2 and within the cited references listed in the last row of this table. Lab 2 was considered the home laboratory for this study from which sample splits were sent to Labs 3, and 4. The checklist for Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE, Bustin et al. 2009) for Lab 2 is provided in the Supplemental Materials.

Table 2.

Summary of WW laboratory methods.

Laboratory Processing Step	Laboratory 1	Laboratory 2	Laboratory 3	Laboratory 4	Laboratory 5
Pasteurization	Yes	No	No	No	No
Large solids removal	Yes, 0.2 μm	Yes, 2 mm	Yes, 2 mm	Yes, 2 mm	No
Concentration	PEGb and centrifugal ultrafiltrationc	Electronegative Filtration	Electronegative Filtration and Magnetic Beadsa	Electronegative Filtration	Sedimentation and Centrifugation
Lysis and preservation	TRIzol and AVL buffer	DNA/RNA Shield (Zymo)	DNA/RNA Shield (Zymo)	DNA/RNA Shield (Zymo) and Applied biosystems MagMAX Microbiome Lysis Solution	DNA/RNA Shield (Zymo) plus bead beating (Geno/Grinder)
Nucleic Acid Extraction	Phenol/chloroform/ethanol and Qiagen RNeasy MiniKit	Zymo Quick-RNA Viral Kit	MagMAX Viral/Pathogen II Nucleic Acid Isolation Kit IFU	Phenol/chloroform/isoamyl alcohol, MagMAX DNA/RNA binding beads	Chemagic™ ViralDNA/RNA 300 Kit H96
Extraction Platform	Manual	Manual	Manual and Thermo Fisher Scientific KingFisher Apex	Thermo Fisher Scientific KingFisher Apex	Perkin Elmer Chemagic 360
Nucleic Acid Preparation of RNA Viruses	Reverse Transcriptase	Volcano Second Generation	Reverse Transcriptase	Reverse Transcriptase	Reverse Transcriptase
PCR Amplification Platform	Bio-Rad, CFX	BioRad CFX	Applied Biosystems	ddPCR and Taqman-based probe technology	ddPCR, Bio-Rad AutoDG Droplet Digital PCR system with AutoDG Automated Droplet Generator and QX200 droplet reader
SARS-CoV-2 gene	N1, N2 genes	N3 gene	N1, ORF1ab genes	N1, N2 genes	N gene
Wastewater Control	PMMoV	PMMoV, B2Md		Total nucleic acids (subset includes RP gene and PMMoV)	PMMoV
Process Control		OC43 (added to raw sewage prior to concentration)	OC43 (added to raw sewage prior to concentration)	None	Bovine Coronavirus (added to lysis buffer)
Inhibition Control	Review raw qPCR curves	HIV	Ran 3 dilutions	Dilution to 200 ng/μL total nucleic acid concentration	Used Bovine Coronavirus Process Control with Zymo OneStep PCR Inhibitor Removal Kit
PCR Positive Control	Synthetic SARS-CoV-2 RNA	PMMoV, B2M, OC43, HIV	MS2 coliphage	Synthetic gene blocks for SARS-CoV-2 RNA, Influenza A/B, RSV A/B, HMPV, Norovirus (GI/GII)	SARS-CoV-2 gRNA, BCoV, and PMMoV gene block
PCR Negative Control	Yes, no template	Yes, no template	Yes, no template	Yes, no template	Yes, no template
References for Method Details	Duvallet et al. 2022	Sharkey et al. 2021 Babler et al. 2022, 2023 Zhan et al. 2022, 2023	Sharkey et al. 2021 Babler et al. 2022, Zhan et al. 2022	Peccia et al. 2020    Zulli et al. 2025	Topol et al. 2021a,b,c Boehm et al. 2024

^aElectronegative filtration utilized for concentration through March 2021. From March 2021 to November 2021 concentration utilized a manual approach based upon Ceres magnetic beads. After November 2021, concentrates were generated using Ceres Magnetic Beads using a KingFisher Apex robotic platform.

^bPEG=Polyethylene glycol precipitation withTRIzol, and phenol/chloroform extraction with ethanol precipitation used from March to June 2020.

^cCentrifugal ultrafiltration (Amicon Ultra-15) coupled with AVL buffer (Qiagen 19073) and Qiagen Rneasy Mini Kit used from June 2020 through November 2023.

^dB2M= beta-2 microglobulin. For details about the analysis of B2M. See Babler et al. (2023b,c, 2025) for details.

2.3. Clinical Data

The clinical health data set, provided by the Florida Department of Health (Delegation of FDOH IRB to UM IRB #20210164), included all daily COVID-19 PCR tests and recorded positive PCR tests for each zip code within the service area of each WWTP. The zip codes serviced by each WWTP were provided by the local wastewater operator (Miami-Dade Water and Sewer Authority, MD-WASD). For zip codes that were split among two WWTPs we estimated the percent of the zip code area that fell into each WWTP service area from geographical maps and split the corresponding HH data by the proportion.

The clinical data were then aggregated by the zip codes serviced by each regional WWTP, providing clinical data specific to a region that coincides spatially with the community contributing to a specific plant. From the aggregated data, we then calculated positivity rates as the number of positives within the region divided by the number of tests for the region. The HH data set spaned four years (March 2020 to January 2024) and coincided with the four-year length of the viral data set. The clinical data reported in this study was limited to COVID-19. Additional disease end points were not available on a zip code basis from the FDOH. Therefore, no comparisons were made between HH and WW for targets beyond SARS-CoV-2.

To create a uniform time scale between the WW and HH data sets all values in both data sets were processed into average weekly values. The WW data were subject to considerable sample-to-sample variability. To smooth this variability the weekly wastewater data was converted to 3-week moving averages. Similarly, the aggregated weekly HH data was converted to 3-week moving averages. Each value in a week was averaged with the value in the week prior to it and the week after it. This allowed for each week’s health metrics (positivity rate, positive cases, and testing numbers) to be compared against an averaged viral level in WW for that same week. All correlations between health metrics and WW used 3-week moving average values.

2.4. Statistical Analysis

Results from the Shapiro-Wilks normality test indicated that the data was not normally nor log-normally distributed. As a result, non-parametric tests were used to evaluate correlations (Spearman) and to evaluate statistical differences between groups of data (Mann-Whitney U tests). Spearman correlations (r) were performed using SPSS, between different combinations of viral WW 3-week moving average values and 3-week moving average HH metric values. SARS-CoV-2 values from each of the three WWTP’s, as well as their combined area, were correlated against each health metric for different time periods to establish geographic and longitudinal trends. Additionally, viral WW data for numerous targets from each laboratory were correlated against each other to establish relationships between labs. Correlations were considered significant for p-values less than 0.05.

Due to inconsistency in viral WW data available for certain targets we used three different statistical tests to establish the association, or lack thereof, between the results provided by each laboratory. For targets that were commonly below detection limits (influenza A, influenza B, RSV, and HMPV) we converted the data to represent percent positive detections. We then analyzed those targets using the Chi-squared test to establish if there was an increased likelihood of a sample being either positive (detected) or negative (not detected) due to the specific lab that processed it. For targets that were usually above detection limits (SARS-CoV-2, PMMoV, and norovirus) we performed the Mann-Whitney U test to establish if there was a significant difference in the median values obtained between labs for the same target. We also ran Spearman correlations on six targets (influenza A, influenza B, HMPV, RSV, norovirus, and PMMoV) to compare results between laboratories.

Data were analyzed by varying time frames spanning from one year to four years (e.g., the entirety of 2020 to 2021, 2020 to 2022, 2020 to 2023, etc.). Data showed distinct waves of WW and HH values over time, which seemed to follow the dominant SARS-CoV-2 variant-of-concern present within the community at the time. To account for, what we are dubbing “the variant waves”, the data sets were analyzed by variant wave according to the following dates (Table 3).

Table 3.

Date ranges for SARS-CoV-2 variant waves.

Variant Period	Date Range of Occurrence
Pre Delta 1 wave	May 26th, 2020 – September 29th, 2020
Pre-Delta 2 wave	September 29th, 2020– June 1st, 2021
Delta wave	June 1st, 2021– Nov 2nd, 2021
Omicron wave	Nov 2nd, 2021 – March 1st, 2022
Omicron 2 wave	March 1st, 2022 – Oct 4th, 2022
Omicron 3 wave	Oct 4th, 2022 – May 23rd, 2023
Omicron 4 wave	May 23rd, 2023 – Oct 17th, 2023

Results

3.1. Comparison of HH parameters across regions

To establish context for the correlations values between the WW and clinical data for each region, the results in this section focus on describing the clinical data. Comparisons of HH data between each region of MDC (South, Central, and North, Figure 1 Panels a through c) as well as their combined area (Figure 1, Panel d) shows similar patterns across each region. Between the three regions the most uniform parameter was positive cases (Figure 1, Panel c). Testing numbers (total number of tests performed) and positivity rates showed broader ranges of values between regions (Figure 1, Panel b and Panel a). At points during the Omicron wave, the Central region had 35,000 more tests per week than the Northern region and about 45,000 more than the Southern region. Overall, the Central region performed about two million more tests than both the Northern and Southern regions from 2020 to 2024.

Figure 1.

HH positivity (positive case counts divided by testing numbers) (Panel a), testing numbers (Panel b), positive case counts (Panel c) for the three WWTP (South, Central and North). Panel d shows results for positivity, testing numbers, and positive cases for all three WWTPs combined. Wastewater data shown for Laboratory 1 for its period of record (March 2020 to November 2023) (Panel e) and for Laboratory 5 for its period of record (January 2023 to January 2024) (Panel f). Panel g provides a table of p values for Mann Whitney U Test with values bolded and showing statistically different means (p<0.05). All plots show the data in 3-week moving averages.

Normalizing each region by its population showed that the Southern, Central, and Northern areas performed over 340 tests, 585 tests, and 381 tests for every 100 people, respectively. Using the Mann Whitney U test, certain regions were found to have statistically different (Figure 1, Panel g) testing numbers when compared to each other.

When positive cases were compared between regions, the Central region had about 45,000 more positive cases than the Southern region and about 120,000 more positive cases than the Northern region. However, when normalizing by their populations, the Southern, Central, and Northern regions had 49, 60, and 49 positive cases out of every 100 people, respectively. Comparing the number of positive cases between regions showed statistical differences between the Central region versus the Northern region (p=0.016), but no statistical difference between the Central and Southern regions (p=0.371) and the Southern and Northern regions (p=0.125). In other words, the number of positive cases for the Central region was only statistically different than the number of positive cases for the Northern region despite there being a statistically significant difference in the number of tests performed between the Central region and both the Southern and Northern regions.

Positivity rates showed that the Central region had the lowest positivity rates while having the highest testing numbers (Figure 1, Panel A). The average positivity rate for the entire study period for the Southern, Central and Northern regions were 13.3%, 10.6%, and 11.3%, respectively. Mann Whitney U testing between regions showed that positivity rates were only statistically different between the Southern and Northern regions, and Southern and Central regions. (Figure 1, Panel g).

Testing numbers and positive cases for all three regions saw a steep decrease in magnitude in 2023, a decline that was not present in positivity rates. Average testing numbers dropped by 86%, 90%, and 86% for the South, Central, and North regions in 2023, respectively. Average positive case count dropped by 85%, 88%, and 86% for the South, Central, and North regions in 2023, respectively. However, average positivity rates increased by 21%, 22%, and 6% in 2023 for the South, Central, and North regions, respectively.

3.2. Comparison between WW and HH data

Time series graphs of HH and WW data (Figure 2, Panels a, b, c, and d) show that peaks in testing numbers and peaks in positive cases coincide with each other, and both generally coincide with peaks in the WW values during each of the variant waves. Although peaks and valleys in the WW SARS-CoV-2 data mostly align with peaks in the HH data, the WW levels measured were different between laboratories. WW SARS-CoV-2 levels varied from 10⁴ gc/L to 10⁷ gc/L as determined from Lab 1. For Lab 5, the trends were similar but reduced by two log units. This could be, in part, because each lab has a different limit of detection and/or capacity to isolate SARS-CoV-2 with the internal methods employed for analysis.

Figure 2.

Time series plots of HH parameters (testing numbers (green bars), positive case counts (black bars), positivity (red line)) and WW SARS-CoV-2 measurements (gc/L) from two laboratories (Lab 1 (orange line) and 5 (blue line)) for the Southern (Panel a), Central (Panel b), Northern (Panel c), and the Combined (Panel d) regions. HH and WW parameters are displayed in 3-week moving averages. Bottom panels provide Spearman r values between HH parameters and WW RNA levels for each of the regions as measured by laboratory 1 (Panel e) and as measured by laboratory 5 (Panel f). Darker green color corresponds to higher correlation coefficients whereas yellow signifies intermediate correlation coefficients and red signifies weak correlations. All correlation coefficients were statistically significant (p<0.05) except for those shown with a white background.

However, the patterns between WW SARS-CoV-2 and HH data show changes over the entire time frame. During the omicron wave there was a significant increase in the test counts and positive case counts in every service area (South, Central, and North). That increase was mirrored in the WW SARS-CoV-2 levels and positivity rates. Additionally, positive case counts and testing numbers saw a steady decline starting near the end of 2022 throughout 2023. That decline was not reflected in the WW SARS-CoV-2 data nor the positivity data. As testing counts and positive case counts decreased steadily throughout 2023, the positivity and WW SARS-CoV-2 levels remained near early pandemic levels.

Focusing on correlations, in general, correlations were highest for the SD WWTP (Figure 2, Panel e). Also, correlations tended to be highest when using positivity as the HH metric. For example, the correlations between positivity and WW SARS-CoV-2 were commonly high across all three WWTP for data clustered by both years and waves. Correlations between positivity and WW SARS-CoV-2 were not strongly impacted by the time frame of analysis. Spearman r values were commonly greater than 0.7 for positivity (especially for the SD and combined data) whether the data were analyzed by year or by wave.

Although, when the HH metric was positive cases and testing numbers, time frames used for WW SARS-CoV-2 correlations did have an impact on Spearman r values. The correlations between WW SARS-CoV-2 and positive cases or testing numbers followed a general trend of decreasing as the time frame of analysis increased. For example, for the CD plant, the r values for positive cases ranged from 0.09 for a three-year period (2021–2023) to an value of 0.95 for a 7-month period (Omicron 2 wave). A similar trend was observed for testing numbers which ranged from −0.50 for a two-year period (2021–2022) to an r value of 0.90 for the seven-month span of the Omicron 3 wave.

3.3. SARS-CoV-2 within the Central District WWTP only

When comparing WW SARS-CoV-2 data between laboratories for the Central district, certain trends were observed (Figure 3). All laboratories showed peaks in WW SARS-CoV-2 during the Omicron and Omicron 2 waves. Positivity rates generally mirrored the fluctuations seen in the WW SARS-CoV-2 values. Positive case counts and testing numbers showed less overall consistency than WW SARS-CoV-2 measurements, coupled with a continuous decrease in testing numbers after 2022. Spearman r values for correlations between WW SARS-CoV-2 levels from different labs (1–5) and the three HH data parameters (positivity rates, positive cases, and testing numbers) were generally highest for Lab 1. However, the range of the Spearman r values between labs was relatively small. For example, the r values for correlations between WW SARS-CoV-2 and positivity for the Omicron 4 wave for labs 1, 2, 4, and 5 were 0.86, 0.77, 0.69 and 0.85, respectively. Similarly, the r values between WW SARS-Cov-2 and positivity for the Pre-Delta, Delta, and Omicron waves were higher than 0.6 for 8 of the 9 correlations computed.

Figure 3.

Time series plots of HH parameters (positive case counts, testing numbers, and positivity) and WW SARS-CoV-2 values for all 5 laboratories for samples collected at the CD WWTP (Panel a). Spearman r values between WW SARS-CoV-2 results for different laboratories for different time frames and waves for laboratories with data sets prior to 2023 (Panel b). Spearman r values between WW SARS-CoV-2 results for different laboratories for laboratories with data set available during 2023 (Panel c). Green color corresponds to high, yellow to intermediate, and red to weak correlation coefficients. All correlation coefficients were statistically significant (p<0.05) except for those shown with a white background. p values for Mann Whitney U tests between SARS-CoV-2 levels from different laboratories (Panel d).

When evaluating statistical differences between the means of SARS-CoV-2 reported by different laboratories, Mann-Whitney U testing showed that all combinations of lab results, except for one combination, had a p value less than 0.001 indicating statistically significant differences between the levels of SARS-CoV-2 detected among the laboratories. The only combination of lab results found to not be statistically different were from Lab 4 and Lab 5 which had a p value of 0.279.

3.4. Comparisons of additional disease targets between laboratories

When comparing alternative targets other than SARS-CoV-2 the results show some comparability between labs. All the data illustrated in Figure 4 comes from the CD-WWTP and aside from the PMMoV target all other targets span only 2023. Two targets (PMMoV and norovirus) had no samples reported as below detection and were therefore analyzed by Mann-Whitney U test. Panel h in Figure 4 shows the comparison of results from the four labs (Lab 1, Lab 2, Lab 4, Lab 5) that processed PMMoV and norovirus. Every combination of labs for those targets had a p value less than 0.001 indicating statistically significant differences between the median levels of PMMoV and norovirus.

Figure 4.

Time series plots of human waste marker PMMoV (Panel a), Norovirus (Panel b) Influenza A (Panel c), Influenza B (Panel d), RSV (Panel e), and HMPV (Panel f), between laboratories. Table of Chi-squared results (Panel g), table of Mann Whitney U test results (Panel h), and Spearman correlations between labs (Panel i). Correlations with a “*” listed in Panel i are not statistically significant.

The last of the additional targets (influenza A, influenza B, RSV, and HMPV) were more frequently below detection limits than other targets and as such were analyzed for positive percent detection using Chi-squared tests. Analysis showed that Lab 2 results did not align with results from Lab 4 and Lab 5. As Panel g in Figure 4 shows, a p value below 0.001 was consistently obtained whenever analysis included Lab 2, indicating that analysis by Lab 2 influenced the proportion of positive or negative results. Whereas analysis by Lab 4 and by Lab 5 (all p > 0.05) did not differ among the proportion of positive or negative results. From the plots the specific days that Lab 4 and 5 recorded positive results differed, but the proportions of positives among all the analyses were not statistically different.

Analysis also showed a generally positive Spearman correlation between labs for five targets (influenza A, influenza B, RSV, HMPV, and PMMoV). PMMoV showed a relatively weak positive correlations between Lab 1 and Lab 2 (r=0.36) and Lab 2 and Lab 5 (r=0.38). The correlation between Lab 1 and Lab 2 involved data covering roughly three years while the correlations between Lab 2 and Lab 5 covered only one year. RSV had two relatively weak positive correlations and one strong positive correlation. The weaker correlations were between Lab 2 and 5 (r=0.37) and Lab 2 and 4 (r=0.40). The strong correlation was between Lab 4 and 5 (r=0.83). Norovirus resulted in a weak negative correlation (r=−0.27) between Lab 4 and 5. HMPV had a weak positive correlation (r=0.13) between Lab 4 and 5. Both influenza A and B had weak positive correlations, but those correlations were found to not be statistically significant.

4. Discussion

4.1. Geographic associations

The consistency in correlations between SARS-CoV-2 data and HH parameters for all three WWTP’s as well as the combined data reflects that surveillance trends in SARS-CoV-2 tend to persist through geographic space. However, comparison of nonadjacent areas should be performed to establish how surveillance trends fluctuate through extended geographic space. The consistently higher correlations observed for the SD-WWTP may be due to its geography.

The service area of all three plants is contained on the eastern coast of Florida by the Atlantic Ocean and towards the west by the Florida Everglades. However, the area of service for the SD-WWTP is also contained on its southern boundary as it is located at the bottom of the Florida peninsula with only the relatively small populations of the Florida Keys further south. Unlike the Southern region, the Central and Northern regions have high population densities below and above them. This results in the individuals living in North and Central regions being more interconnected with the regions that surround them. An individual could live and perform their COVID-19 testing in one region but work and contribute to SARS-CoV-2 WW levels of another region. There is more movement across county lines towards the north of Miami-Dade County than towards the south. The reduced likelihood of population drifting in the Southern region is one possible contributor to the higher correlations observed in that area between the HH and the WW SARS-CoV-2 data sets.

4.2. Stability in positivity

WW SARS-CoV-2 levels correlated strongly with positivity rates both during shorter (one year or less) and longer (more than one year) time frames. Positive case counts on the other hand showed strong correlations with WW SARS-CoV-2 levels only when shorter time frames were considered and not for longer time periods. Testing numbers showed the weakest correlations when analyzed over longer time periods. The weaker correlations observed for positive case counts and testing numbers over longer periods of analysis is likely due in part to the sharp decline in both parameters beginning in 2022. Towards the end of 2022, the total tests performed, and the number of recorded positive case counts began to decline, and this trend continued throughout 2023. However, no decline occurred in the WW SARS-CoV-2 levels. Positivity also did not decline and instead, like the WW SARS-CoV-2 levels, maintained a stable range throughout 2023.

When testing centers were free and readily available throughout the city when social concerns of COVID-19 were high, testing throughout the county was high. However, as testing centers become less abundant, vaccination rates increased, social pressures to provide proof of a negative COVID-19 status lessened, and symptoms of variant COVID-19 strains weakened, coupled with the increase in home testing which was not reported, the impetus diminished to regularly test for COVID-19. This resulted in a significant and continuing drop in COVID-19 test counts which also led to a decline in recorded positive case counts.

However, since positivity is used here as a normalized parameter, and not a direct count, it was able to maintain stable levels more easily despite the economic and social factors that affected positive case counts and testing counts. Positivity was able to better mirror WW SARS-CoV-2 levels throughout the entire study period, and as such, served as the better indicator of COVID-19 infections in the community. This is supported by the stronger correlations obtained between SARS-CoV-2 WW levels and positivity for all three WWTP’s as well as their combined area. The significance of positivity observed in this study in its association with WW SARS-CoV-2 levels is consistent with other studies that have found significant associations with positivity but not with case counts (Bivins et al. 2021).

4.3. Variant wave symptoms

The general trend of stronger correlation obtained for variant waves as versus complete years might be due to the distinct symptom manifestation associated with each variant of SARS-CoV-2. Since each variant of COVID-19 can have its own unique combination of symptoms it is reasonable to say that each variant also has differing levels of discomfort and noticeability. A variant with multiple symptoms that causes discomfort will likely lead to more individuals getting tested. Whereas a variant with milder symptoms might result in less testing and more positive cases going unreported. Analyzing data that span multiple variants together does not allow for the individual discrepancies between variants to be accounted for and related back to the HH data. When correlations were performed across multiple years, and even single years, multiple variants were analyzed together. The discrepancies between variants could be a factor causing the weaker correlations seen for yearly versus single wave analysis.

4.4. Trends across labs

The comparison of data from different WWTP’s shows that the trends between SARS-CoV-2 WW levels and HH data parameters hold up across geographic distance. Whereas the comparison of data between labs shows that the trends are present across different processing protocols and even across target genes. Each lab that processed samples in this study had its own unique equipment and protocols. Beyond that, some labs targeted different genes (N, N1, N2, N3, ORF1ab) when analyzing SARS-CoV-2 levels in the wastewater. Even with differing procedures and different target genes the same trends were observed across all three WWTP’s and across the five different labs involved in our study. All five labs showed that clinical positivity functioned better as indicators of SARS-CoV-2 WW levels in a community than positive case counts and testing numbers.

4.5. Correlation strength and increased WW sampling

Labs 1 (r=0.81 for all data) and 5 (r=0.82), in general, had the strongest SARS-CoV-2 correlations across every time frame of analysis in the study period. Both labs frequently had more than one sample collected per week. Labs 2, 3, and 4 only had single samples collected and analyzed each week. Lab 1 had an average of 1.45 samples per week and Lab 5 had an average of 2.91 samples per week. Lab 1 samples were not as uniformly spaced as Lab 5. Also, due to the longer period of record, the Lab 1 sampling frequency varied from single sample weeks to 7-day sampling weeks. Sampling frequency for Lab 5 was more uniform by comparison, with almost all weeks having three consecutive sample days. The results support the conclusion that sampling multiple times weekly can increase correlations between WW SARS-CoV-2 levels and HH data. More analysis would be needed to evaluate exactly how sampling variations (e.g., frequency, time of day, weekly distribution) affect correlations.

4.6. Other targets across the labs, possible expansion of surveillance efforts

All the results we have discussed so far support that WW surveillance can be effectively correlated to clinical data. WW surveillance provides insight into the prevalence of disease in a community and even the potential progression of a disease. This information can be useful in determining the optimal response and plan of action for reducing impacts to public health from outbreaks. COVID-19 had a drastic impact on the entire world and as a result efforts were made to acquire consistent and accurate clinical data associated with the disease. Although our analysis of other WW viral targets was not as extensive as what was done for SARS-CoV-2, results support the potential for expanding WW surveillance targets as part of public health monitoring for a community.

Our data showed positive detection of viral targets other than SARS-CoV-2 in the WW. However, the uniformity in results across different labs for those alternative targets was less consistent than seen for SARS-CoV-2. As a result, we used a multitude of statistical tests to evaluate the results of different targets and establish the relationship between results of different labs.

Our analysis using the Chi-squared test revealed that Lab 4 and Lab 5 had the most consistent results. This indicated that sample processing from either Lab 4 or Lab 5 did not increase the probability of obtaining either a positive or negative sample, but processing by Lab 2 did increase the probability of yielding a negative sample. Our Mann-Whitney U analysis of SARS-CoV-2 results further supported the relationship between Lab 4 and Lab 5. A possible explanation for the lack of statistical difference between Lab 4 and Lab 5 could be that those two labs were the only labs to perform ddPCR as opposed to qPCR. The increased sensitivity of ddPCR could account for the higher instances of detection for those two labs.

Our Spearman correlation analysis of the alternative targets showed differing levels of correlation between all labs. While correlation is a more generalized comparison than Chi-squared and Mann-Whitney U, it is still a promising sign that statistically significant correlations (p<0.05) can be seen between four distinct laboratories and across four different WW viral targets. If analysis methods for alternative viral targets can be optimized to allow more consistent detection and matching with clinical data, then WW surveillance can be expanded.

4.7. Limitations

This study would have benefited from clinical data for additional disease targets evaluated in this study. The FDOH does measure for influenza-like illnesses but obtaining this information on a regional basis was not possible. For WW, the study would have benefited from increased frequency of sample collection at the CD-WWTP. This could have eliminated the need to conduct moving averages such that the average of multiple measurements on different days within the week could have been used to smooth the variability.

5. Conclusions

Through the analysis of a four-year record of WW SARS-CoV-2 data and its corresponding clinical data, we found geographic and longitudinal trends between WW data and testing numbers, positive cases, and positivity rates in the regions serviced by Miami-Dade county’s three major WWTPs. The scientific contribution from this study were that higher correlations were observed in areas with geographic limitations in population mobility. Among the health metrics, positivity (number of clinical cases divided by the number of tests) was the parameter that provided the highest correlations with levels observed in wastewater. In addition, more significant correlations were observed between WW SARS-CoV-2 data and HH metrics when time frames were focused on variant waves as opposed to standard yearlong time frames. These results imply that wastewater based tracking of disease should utilize positivity when possible to track illness in a community, and that the correlations between wastewater and positivity may change as different variants of the virus emerge.

When evaluating WW SARS-CoV-2 results between laboratories for the Central district, all laboratories consistently aligned with HH data trends. For targets beyond SARS-CoV-2, results showed that while there is variation in the results between labs there is also uniformity and consistency. Our Chi-squared and Mann-Whitney U analysis showed that processing by either Lab 4 or Lab 5 did not influence whether a sample would turn out negative or positive. Even further it showed that Lab 4 and Lab 5 RSV values were not statistically different from each other. Additionally, our correlations showed that there was a statistically significant relationship between the results of four different labs. The consistency between the laboratories supports the possibility of widespread implementation of wastewater-based surveillance of targets beyond SARS-CoV-2. Overall, results support that WW surveillance for a multitude of viral targets can be disaggregated among different laboratories facilitating national-level capacity for the surveillance of circulating viruses.