Our study builds on a growing body of evidence that WS for IAV is highly concordant with the results of other disease incidence measures (13,28). During the 31-week period of retrospective analysis, influenza-related hospitalizations within Windsor-Essex and Michigan covaried with the concentration of IAV RNA measured in wastewater. However, the IAV signal was not a leading indicator of influenza incidence in either community when analyzed on an epidemiologic week basis; wastewater signal either lagged or was synchronous with hospitalization data. Observation of synchronous or delayed wastewater signal is not without precedent. Recent surveillance efforts have noted lagging wastewater signals (29). Other studies have cited the predictive ability of WS in the context of influenza monitoring (13,14,30). In our study, application of blockwise averaging to produce average concentrations of IAV RNA by epidemiologic week could have masked lead time within the data.

Viral load in influenza patients may peak 1–2 days after symptom onset on the basis of nose and throat swab testing results, and shedding may last 6–7 days (31). A meta-analysis of challenge studies examining respiratory tract shedding found shedding lasts an average of 4.8 days, and peak shedding rates occur 2 days after exposure (32). A clinical study reported that 41% of IAV-positive patients produced detectable levels of IAV RNA in their feces (11). Another study of hospitalized patients found that 47% of people infected with IAV shed IAV RNA in their feces (33). Because only some people shed IAV RNA in feces, incubation periods are short, and viral loads rapidly peak, WS loses its ability to predict influenza-associated hospitalizations when the temporal granularity of incidence data is limited. However, producing meaningful data through clinical testing takes longer than results from WS. Case data obtained through laboratory-based virology are released weeks after testing occurs and are often subject to revision because results may reflect data compiled from multiple laboratories. WS can provide more timely measures of incidence because sample processing, RNA quantification, data analysis, and reporting are often completed the same day as sample collection in Windsor-Essex. WS can be considered de facto lead-time because data may be disseminated to public health officials well in advance of case data.

The utility of WS is not restricted to predictive ability. WS is an independent and sensitive measure of disease prevalence (34), enabling it to be used as an additional metric for trend comparison across jurisdictional boundaries, and it may be helpful when testing conventions and public health policies differ. Unlike WS, influenza cases and hospitalizations likely represent only the most severe cases of influenza in which people sought medical testing and treatment, and they do not necessarily represent population-wide trends. WS has the potential to aid in accurately tracking infection dynamics when testing capacity is limited or few patients seek medical care.

In the case of Windsor-Essex and Detroit, the cross-border movement of persons and goods is vital to the region because of strong economy integration (35). Many people, such as healthcare and automotive workers, commute across the border daily to work in Michigan while living in Ontario (16). Almost 18,000 people crossed into Windsor-Essex from Detroit daily over the course of our study, which suggests the communities could have concurrent influenza seasons (Appendix Figure 1). Contrary to this expectation, we observed the peak and onset of the 2022–23 IAV wastewater signal in Detroit was delayed by ≈3 weeks when compared with Windsor-Essex. An explanation for this discrepancy could be the lingering effect of travel restrictions implemented during the COVID-19 pandemic, which limited travel between these interconnected cities (Appendix Figure 1). There was no restriction on trade or the commutes of essential workers, and testing requirements for cross-border travel ended in April 2022. The remaining COVID-19 related travel restrictions were lifted at the start of October 2022, before the onset of the influenza season. Despite the removal of restrictions, the number of people crossing into Windsor-Essex each day during the second half of 2022 was ≈25% less than the number of people crossing prior to the COVID-19 pandemic (17,867 vs. 24,260) (Appendix Figure 2). The residual effect of border restrictions, evidenced by the suppressed cross-border traffic in the leadup to the 2022–23 respiratory season, shows that border restrictions could have played a role in the observed discrepancies between Windsor-Essex and Detroit.

Another potential cause of the timing difference in the influenza seasons between Detroit and Windsor-Essex could be pandemic mitigation strategies, such as mask mandates and social distancing guidance. Ontario was slower to implement a mask mandate but kept the mandate in place much longer than Michigan did. Michigan lifted all masking requirements on June 22, 2021, whereas Ontario ended its mask mandate 272 days later, on March 21, 2022 (36,37). Michigan ending the mask mandate early could have enabled the circulation of influenza in Detroit before the 2022–23 respiratory season, increasing levels of natural immunity in Michigan. However, clinical data and wastewater surveillance in Windsor-Essex show that influenza was circulating in Windsor-Essex in spring 2022 (Appendix Figure 3).

Differences in influenza immunization campaigns might also explain the differences in influenza season onset between Detroit and Windsor-Essex. Influenza vaccines could have played a role in determining the effective reproduction number for IAV in the 2022–23 season. Preliminary research showed the vaccine effectiveness (VE) in Canada against IAV subtype H3N2 was 54% for people <65 years of age (17). H3N2 was the dominant IAV subtype, representing ≈95% of cases (17), in contrast with the limited sequencing results of a selection of amplicons (Appendix Figure 4). Similar estimates of VE for this cohort were produced for Wisconsin (60%) (38) and across the United States (51%) (39). Michigan and Ontario use similar vaccines, with a focus on administering quadrivalent-inactivated influenza vaccines to the population (40,41). Although inoculation with some vaccine types, such as live attenuated vaccines, is associated with viral shedding, it is unlikely to contribute to wastewater signals (42). Because VE and vaccine type were similar between Michigan and Ontario, the differentiating factor could be inoculation timing. Influenza vaccine distribution in Michigan typically begins earlier than Ontario, with inoculations happening as early as July. By November 2022, a total of 2,632,430 Michigan residents were vaccinated (≈25% of the population) (43). Inoculation efforts in Ontario began later, and vaccines were not made available to all residents until November (44). Mass influenza inoculation efforts in Ontario began after IAV RNA concentration started increasing in the wastewater, signaling the start of the influenza season. Vaccination campaigns were already well under way in Michigan when the IAV RNA concentration began to increase in wastewater. We speculate that people from Ontario were less likely to have vaccine-induced immunity than those from Michigan early in the season, potentially contributing to the earlier peak in Windsor-Essex wastewater IAV signal. However, it is unclear if the timing of vaccine campaigns contributed to the observed difference between the Michigan and Ontario influenza seasons. Additional factors, including socio-economic status (45), access to healthcare, racial demographics (46), population age structure (47), and virus–virus interactions (48) could have contributed to the differences. Population-level administration schedules for seasonal influenza vaccines merit further investigation; those schedules help determine when herd immunity is reached and if herd immunity is reached before the spread of illness within a community.

The first limitation of our study is that WRRF does not serve the Michigan counties where FluSurv-NET–participating hospitals are located. However, both the hospitalization data garnered from FluSurv-Net and wastewater data can be considered a proxy for statewide incidence. All influenza-associated hospitalizations were included in this dataset, for both Michigan and Windsor-Essex, even though the WS focused only on IAV. Our analysis could have been affected by focus on IAV, despite its dominance in the 2022–23 influenza season (17). Not all hospitalizations recorded in the Windsor-Essex data were laboratory confirmed cases of influenza. The temporal resolution of sample collection at the WRRF was limited, and weekly sampling might have failed to capture variability in the wastewater signal. Finally, the wastewater treatment plants monitored in both Windsor-Essex and Detroit, although serving representative populations, do not encompass all the residents and could have failed to capture variability in IAV wastewater signal.

In conclusion, our study demonstrates how wastewater surveillance can shed light on regional differences that may have otherwise gone unnoticed, or remain unvalidated, because of the inherent limitations of traditional metrics to capture populationwide trends. Future studies investigating influenza vaccine administration schedules should incorporate WS as an independent metric of disease incidence. WS can potentially provide more timely measures of incidence, rather than waiting for the release of laboratory testing. The utility of WS as a predictive metric, and as a metric for trend comparison across jurisdictional boundaries with different approaches to vaccination and collecting disease incidence metrics, demonstrates its usefulness when testing conventions and public health policies differ.