Respiratory viruses are a major public health concern, contributing significantly to global morbidity and mortality, particularly among vulnerable populations such as young children, the elderly, and individuals with underlying health conditions. These infections place a substantial burden on healthcare systems, leading to seasonal surges in hospitalizations, severe complications, and, in some cases, mortality. Infection by one virus could enhance or reduce infection and replication of a second virus, resulting in positive (additive or synergistic) or negative (antagonistic) interactions [1]. Mechanisms for these interactions range from cellular-level (e.g., blocking of cell surface receptors by influenza A virus [IAV], competition for resources like sialic acid by respiratory syncytial virus [RSV]), to host-level (impaired immune response caused by influenza leading to increased bacterial colonization), to population-level interactions (behavioral changes like self-isolation reducing transmission of other pathogens) [2,3]. Additionally, the initial viral infection can induce long-term specific immunity against the same virus or transient nonspecific immunity that offers broader protection against different, but related, viruses, particularly among taxonomically similar viruses [4].

Ecological factors also play an important role in shaping the circulation patterns of respiratory viruses. These include environmental conditions such as temperature and humidity, which can affect virus stability and transmission efficiency [5]. Seasonal behaviors, such as increased indoor crowding during colder months, can facilitate spread of certain viruses. Variations in host susceptibility due to prior exposure or vaccination campaigns further influence which viruses predominate in a given season [6]. For instance, influenza viruses often dominate during the winter, while adenoviruses and parainfluenza viruses circulate year-round but peak under specific conditions [7]. Positive virus–virus interactions describe a scenario where one virus creates conditions that aid another virus’s replication, transmission, or overall impact. Negative virus–virus interactions occur when a primary viral infection hinders the replication, transmission, or overall impact of a secondary viral infection [1].

Understanding both positive and negative interactions among viruses is useful for public health, as they can influence disease severity, transmission patterns, and effectiveness of interventions [1,8]. The increased availability of population-level etiological surveillance data—facilitated by advances in multiplex diagnostic testing and expanded testing during the COVID-19 pandemic—provides valuable insights into these interactions. However, limitations exist. Routine diagnostic practices often focus on one or two pathogens based on presenting symptoms and circulating trends, leading to biased incidence rates if the tested subpopulations differ significantly between viruses. Multiplex testing on representative clinical samples can mitigate this issue, but uncovering interactions requires large datasets collected over extended periods, often beyond the capacity of many studies [9]. Robust statistical methods are necessary to identify true virus–virus interactions while accounting for confounding factors such as seasonal and annual climatic variations, age-dependent contact patterns, and fluctuations in testing frequencies [10]. For instance, RSV peaks often coincide with colder weather, complicating the isolation of independent effects of temperature and co-circulation of other viruses on its transmission [11]. Similarly, co-circulation of human metapneumoviruses (HMPV) with influenza can result in more severe illness. Advanced analytic techniques, such as Bayesian hierarchical models and wavelet coherence analysis, can separate biological interactions from seasonal trends by explicitly modeling and adjusting for seasonality, long-term trends, and other confounders. These approaches isolate residual variation attributable to virus–virus interactions after accounting for known seasonal and environmental patterns.

To better understand the drivers and implications of virus–virus interactions, we use a novel Bayesian framework building on prior work for analyzing virus–virus interactions in public health surveillance data [8,10,12]. This approach is designed to evaluate population-level patterns of viral interactions while accounting for seasonal and long-term trends. However, it does not establish causality or distinguish biological mechanisms from shared environmental drivers. Here, we leverage this framework to evaluate positive and negative interactions among seven acute respiratory viruses: IAV, influenza B (IBV), RSV, human parainfluenza viruses 1 and 3 (HPIV-1, HPIV-3), human adenovirus (HAdV), and HMPV in Puerto Rico over an 11-year period.

Our findings suggest potential interactions among respiratory viruses in Puerto Rico over 11 years, which may influence disease severity, transmission, immune response, and vaccine effectiveness [1,3,8]. Identifying such correlations can inform public health strategies. For example, a positive correlation of 0.52 between HMPV and HPIV-1 suggests that when HMPV prevalence deviates from its expected seasonal trend, HPIV-1 prevalence tends to deviate in the same direction and by a moderate degree. Recognizing such correlations could allow hospitals to anticipate potential increases in certain infections, improving resource allocation for managing complications like croup and bronchiolitis. However, the observed interactions appear to be context-dependent rather than consistent drivers of virus transmission.

The impact of the COVID-19 pandemic on respiratory virus circulation was substantial, altering observed interaction patterns. From March 2020 to April 2021, SARS-CoV-2 became the dominant respiratory virus among SEDSS participants, reflecting trends in the continental U.S. [20]. This coincided with a sharp decline in IAV, IBV, RSV, HAdV, and other respiratory viruses following the implementation of stay-at-home orders and other nonpharmaceutical interventions [21,22]. These reductions persisted beyond the initial mitigation measures, likely due to changes in social interactions, testing behaviors, and potential viral interference [21]. Our analysis of the prepandemic period found that most virus–virus correlations identified in the full dataset did not remain statistically credible, suggesting that many observed interactions were influenced by pandemic-related disruptions. The only correlation that remained significant across both analyses was the negative association between IAV and HAdV, while other pairs, including IBV-HMPV and RSV-HPIV-3, were no longer significant in the prepandemic period. This finding suggests that some covariations may have been transient, potentially influenced by external factors such as testing practices, public health measures, or shifts in viral transmission dynamics. On the other hand, it could also be that the prepandemic data is underpowered to detect some interactions, especially the ones for which the magnitude and sign of the correlations remained consistent with those based on the whole study period, e.g., the pair of IBV and HMPV (Figure 2). These results underscore the need for caution in interpreting population-level correlations as definitive biological interactions.

Several studies from other countries have documented similar shifts in respiratory virus circulation due to viral interference with SARS-CoV-2. Nenna et al. [23] conducted a multicenter study in Italy and found that RSV hospitalizations peaked earlier than usual during the 2021–2022 season, coinciding with the relaxation of COVID-19 mitigation measures. Their findings suggest that SARS-CoV-2 may exert competitive pressure on RSV circulation, aligning with our observation of disrupted viral synchrony during the COVID-19 pandemic in Puerto Rico. Matera et al. [24] further reviewed global evidence of viral interference, highlighting how the emergence of SARS-CoV-2 altered epidemic patterns of influenza, RSV, and other respiratory viruses. Their review emphasizes the importance of considering virus–virus interactions when interpreting epidemiological trends, which is consistent with our findings using Bayesian hierarchical modeling. These studies reinforce the need for robust surveillance and modeling approaches to better understand how respiratory virus interactions shape epidemic patterns across different settings.

Consistent with previous research [10,25], positive correlations emerged between RSV/HPIV-3 and HMPV/HPIV-1, viruses commonly associated with lower respiratory tract infections in young children. HMPV/HPIV-1 also showed synchronous co-variation at approximately 1-year cycles in wavelet coherence analyses, suggesting their seasonal fluctuations might be driven by similar factors. Increased susceptibility in young children could result from frequent close contact in daycare and school settings, and limited prior exposure to these pathogens [26]. HPIVs, a leading cause of childhood lower respiratory tract infections, are slow-replicating viruses with extended infection times and a gradual decline in viral load [4]. This prolonged presence could increase the window of opportunity for encountering other respiratory viruses. The consistently high viral load observed for HPIV-3, regardless of coinfection status [27], further supports the notion of its prolonged presence within the host.

Our analysis confirms seasonal influenza patterns in Puerto Rico, with influenza A typically peaking from November to March and influenza B from April to June. While these peaks align with temperate regions, local tropical factors may cause slight variations in timing. Influenza A, which poses a higher risk of severe illness [28], peaks earlier in the season. This finding underscores the need for early vaccination efforts, as delaying vaccination could leave individuals vulnerable to influenza A, the more dangerous strain. Given that influenza B tends to extend later into the year, physicians should emphasize the importance of getting vaccinated as early as possible, ideally by the end of October, to provide protection during the critical influenza A season [29].

The negative correlation between IAV and HAdV in both partially adjusted and Bayesian analyses may reflect viral interference, potentially influenced by differences in replication cycles, tissue tropism, and seasonal variations, which can create a competitive environment unfavorable for simultaneous infections. Biological mechanisms, such as blocking cell surface receptors or competition for cellular resources, could underlie this interaction [1]. For example, neuraminidase expression in influenza A-infected cells removes sialic acid, preventing subsequent infections, and RSV replication is inhibited during co-infection with IAV due to competition for viral protein synthesis [30]. Ecological factors, including fluctuations in susceptible populations, age distribution, and testing practices, may also contribute to this negative correlation [25]. These findings likely reflect a combination of biological, ecological, and population-level dynamics rather than definitively confirming underlying biological mechanisms [8]. Host immune responses, particularly interferon induction, may mediate some interactions by providing transient protection against subsequent infections with different viruses [1]. Further research is needed to disentangle these drivers and better understand the observed dynamics.

Our study identified consistent positive correlations between IBV and HMPV across both Pearson correlation and Bayesian modeling, as well as between IBV and HAdV in Bayesian analyses. Although the mechanisms underlying these correlations require further investigation, IBV infection can damage respiratory epithelium and suppress the host’s immune response by downregulating antiviral proteins or interfering with antigen presentation, increasing susceptibility to secondary infections [31]. However, we acknowledge that our Bayesian approach, while adjusting for seasonality, testing patterns, and demographic factors, does not fully eliminate the possibility that shared environmental or epidemiological factors influence the observed associations. Differences in temperature, humidity, and human behavior (e.g., school attendance, holiday travel) could drive synchronized viral circulation independent of biological interactions [25].

Although the low individual-level co-infection rate among acute respiratory viruses (0.5%) in this study suggests single virus infections are most common, co-infections were observed and warrant further study. Studies have shown co-infections with respiratory viruses can lead to more severe illness, longer hospital stays, and worse clinical outcomes [32,33]. HAdV was involved in the most co-infections, consistent with other research [34,35].

Our study was subject to several limitations. First, it focused on seven respiratory viruses, excluding other pathogens like rhinovirus and bocavirus, which may influence co-infection patterns. Second, our data originated from hospital surveillance, capturing only symptomatic individuals seeking medical care. Respiratory viruses can present with mild or no symptoms, and these asymptomatic infections would not be captured through our healthcare-facility-based surveillance system, potentially affecting the observed correlations. Third, vaccination records were unavailable, which could affect infection and co-infection patterns. Fourth, we did not assess higher-order interactions between three or more viruses. Fifth, Bayesian modeling did not consider sensitivity and specificity of testing methods. Sixth, while our Bayesian hierarchical model accounts for many confounding factors, it does not fully isolate biological mechanisms from shared environmental drivers of viral circulation. Seasonal behaviors and climate factors may contribute to the observed correlations. Seventh, we could not determine the sequence of infections in individuals, and most virus pairs had too few co-infections (e.g., only 2 cases of RSV/IBV) to draw meaningful conclusions about within-host interactions. Future studies tracking repeated infections in the same individuals would be needed to distinguish biological interference from coincidental co-circulation. Notwithstanding these limitations, our study had a high testing rate (nearly 100%) for all seven viruses. The unique island context of our study potentially mitigates cross-contamination from external regions, strengthening the generalizability of the observed correlations between respiratory viruses.

The high testing rate for multiple viruses highlights the value of leveraging multiplex diagnostic data from hospitals and public health authorities to inform vaccine campaigns, resource allocation, and surveillance efforts. However, routine testing for viruses such as HPIV and HMPV is uncommon outside of high-resource hospitals with multiplex capabilities, as most facilities focus on influenza, COVID-19, and RSV. For hospitals with multiplex testing, using existing data to monitor trends in test positivity rates could help predict shifts in respiratory virus circulation—for example, a rise in HMPV cases may signal an impending increase in HPIV-1 prevalence.

This study in Puerto Rico identified potential correlations between respiratory viruses, providing preliminary insights for public health strategies. However, given the variability in observed associations across different periods, these findings should be interpreted with caution. While some interactions persisted across both full-period and prepandemic analyses, others appeared to be influenced by external factors such as the COVID-19 pandemic and shifts in testing and transmission dynamics. This highlights the need for continued surveillance to better understand the drivers of virus–virus co-circulation and their implications for public health.

Future research should continue using platforms like SEDSS to track respiratory virus interactions as new vaccines, such as the high-dose influenza vaccines for older adults and recently introduced RSV vaccines and monoclonals, enter the scene. The impact of widespread antiviral treatments on virus–virus dynamics also warrants investigation, as aggressive treatment for one virus (e.g., COVID-19 or influenza) may have positive or unanticipated effects on the prevalence and severity of other respiratory viruses. Exploring these effects will help refine prevention strategies and improve our understanding of respiratory pathogen interactions.

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.ijid.2025.107878.