Epidemiological impact of three major respiratory coronavirus epidemics on influenza transmission: a multicountry analysis using surveillance data and mathematical modelling

Abstract

Background

With the acceleration of globalisation and the increasing frequency of international exchanges, the risk of cross-border transmission of emerging respiratory infectious diseases (ERIDs) has significantly increased. Since the year 2002, epidemics of SARS, Middle East respiratory syndrome (MERS) and COVID-19 have exemplified this trend. These epidemics have impacted the prevalence and transmission of traditional respiratory infectious diseases (RIDs), such as influenza, which share similar transmission routes and control measures. To better explore the impact of ERIDs epidemics on influenza, our study quantitatively evaluates the epidemiological changes in influenza during three representative emerging respiratory coronavirus epidemics: SARS, MERS and COVID-19.

Methods

Using Global Influenza Surveillance and Response System data, we examined influenza trends across different periods and regions affected by the three coronavirus epidemics. The impact of the epidemic on influenza was revealed by comparing and analysing the reported positive cases (RPCs) of influenza during the pre-epidemic and epidemics, and during the three postpandemic periods. Based on the Susceptible-Exposed-Infected-Asymptomatic-Recovered (SEIAR) compartmental model, the time-varying effective reproduction number ($R_t)$) over time was calculated, and the Farrington surveillance algorithm was used to calculate the RPCs in the absence of an epidemic to analyse the characteristics of influenza transmissibility during the epidemics of the three respiratory coronavirus changes.

Results

There was a significant decline in the RPCs of influenza and transmissibility. The suppressive effect of the COVID-19 epidemic on influenza prevalence was the most pronounced. During the COVID-19 epidemic, the RPCs of the three major influenza subtypes showed the largest decrease compared with historical predictions, with reduction rates of −53.30% for A(H1N1), −57.50% for A(H3N2) and −48.56% for influenza B (p<0.01), with A(H3N2) being the most significantly affected, as most countries experienced reductions exceeding 50%. The impact of the SARS epidemic on influenza was secondary, with total RPCs of A(H1N1) and influenza B decreasing by approximately 84.39% and 45.31%, respectively (p>0.05). During the MERS epidemic, the RPCs of A(H1N1) and A(H3N2) decreased by 28.75% and 17.62%, respectively, although influenza B partially rebounded in the later stages, resulting in a relatively smaller overall impact.

Conclusions

The COVID-19 epidemic demonstrated the most pronounced suppressive effect on influenza prevalence. The impact of SARS was secondary, while MERS had the least effect. Among different influenza subtypes, A(H3N2) and influenza B exhibited greater declines compared with A(H1N1). The decrease in RPCs during coronavirus epidemics highlighted the importance of non-pharmaceutical interventions (NPIs), demonstrating the broad applicability and high efficacy of comprehensive control strategies for RIDs. Furthermore, when NPIs are lifted during the later stages of coronavirus epidemics, attention should be paid to the potential rebound of traditional respiratory diseases such as influenza.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• We employed the Farrington surveillance algorithm method to predict influenza transmission dynamics excluding interference from other coronavirus pandemics.

• It remains possible that influenza characteristics change during epidemics, a possibility our study cannot yet rule out.

Introduction

Influenza is a recognised global infectious disease that poses a continuous threat to humanity, with epidemics occurring approximately every 10–50 years.¹ The accelerated globalisation in recent years has accelerated the emergence and transmission of emerging respiratory infectious diseases (ERIDs). Typical among these are the various respiratory coronavirus pandemics since SARS in 2002. The transmission route of respiratory coronaviruses including respiratory, droplet and contact transmission,² which shared the same spread patterns with influenza, while the specific extent of the association between ERIDs, particularly pandemic-causing respiratory coronaviruses and influenza, remains unclear.

Since 2000, influenza has experienced multiple coepidemics with respiratory coronaviruses. The more typical representatives are the three respiratory coronaviruses, SARS, Middle East respiratory syndrome (MERS) and COVID-19, all of which are characterised by seasonality, periodicity, diversity of viral variation and inter-regional variability, the same as influenza. The sudden arrival of SARS and its widespread transmission raised awareness and prioritisation of respiratory infectious diseases (RIDs), including influenza, in public health planning and resource allocation.^{3 4} Based on the scale and impact of the SARS epidemic in 2003, it is possible that the epidemic had an influence on the transmission of influenza. Taiwan saw a significant reduction in excess mortality from influenza among the elderly.⁵ MERS is a respiratory coronavirus which transmits between humans and camels, but with limited transmission among the population resulting in localised outbreaks in certain regions.^{6 7} Current research efforts primarily focus on comparing the epidemiological and virological differences between MERS and other coronaviruses.⁸ There have also been many studies⁹,¹³ showing that the epidemic characteristics of influenza have been influenced by the COVID-19 epidemic, which had a suppressive effect on influenza epidemics in global countries, including China, Japan, USA and Brazil. This has led to a decrease in influenza incidence, a significant delay in peak influenza seasons, and may even have resulted in the global extinction of the B/Yamagata lineage.¹⁴ As of now, no studies have further explored and summarised the epidemic patterns of various subtypes of influenza during the three major respiratory coronavirus epidemics.

The three coronaviruses, SARS, MERS and COVID-19, represent typical ERIDs, sharing highly similar transmission routes and response measures with influenza. Analysis of the differences in the epidemic and transmission characteristics of influenza during epidemics of respiratory coronaviruses can aid in formulating more comprehensive and effective response strategies, enhancing the adaptive capacity of public health systems. Therefore, we selected countries and regions that have experienced one of the SARS, MERS and COVID-19 epidemics. We compared the reported positive cases (RPCs) in the periods before, during and after the three epidemics, used Farrington surveillance algorithms (FSAs) to predict the real state of influenza circulation in the absence of other major epidemics, and fit the time-varying effective reproduction number ($R_t$) through the compartmental model. We aimed to analyse the changes in the epidemic characteristics of the main subtypes of influenza A (H1N1), A (H3N2) and influenza B during the occurrence of three major respiratory coronavirus epidemics, and explore the social factors that may contribute to these changes.

Methods

Data collection

In our study, data on global influenza of RPCs provided by the Global Influenza Surveillance and Response System (GISRS) from 26 December 1994 to 31 December 2023, were analysed.¹⁵ This data set includes country name, country code, epidemic weeks, type of surveillance site, number of collected samples, number of processed samples, cases for each subtype, and positive and negative samples, covering 181 countries, containing a total of 117 352 entries, with 35 919 (30.6%) entries requiring multiple imputations, totalling more than 152 000 entries to study. The GISRS provides a standardised framework for data collection and reporting, ensuring comparability across different locations.^{16 17}

We included data from all types of surveillance sites to obtain the total number of RPCs for each subtype. Where A(H1N1) includes seasonal A(H1N1) before the year 2009 and A(H1N1) pdm09 after the 2009 epidemics, and influenza B is the sum of the number of RPCs for B/Yamagata, B/Victoria and unidentified subtypes.

For non-continuous data and instances of missing weekly data, we backtracked on the missing data and found that the missing data were correlated with the time, with more missing data in the early period of underdeveloped network, and only a few countries had missing data in the COVID-19 period, which was judged to be non-randomly missing. Therefore, ‘mice’ function in R V.4.2.1 to perform multiple imputation for missing data to obtain continuous time series data. Moreover, countries with fewer than 52 weeks of data in non-leap years and fewer than 53 weeks in leap years were excluded from the study, which better ensured the quality of the study data. Ultimately, our research selected nine countries/regions that experienced SARS outbreaks, South Korea for MERS, while 112 countries/regions for COVID-19 based on strict indices of intervention measures. The timeline of the epidemic of the three main ERIDs in the study regions is provided in online supplemental file 1.

Patient and public involvement

None.

Compartment model

The Susceptible (S) - Exposed (E) - Infectious (I) - Asymptomatic (A) - Recovered (R) (SEIAR) model has been used to simulate the spread of influenza in different epidemic periods. It is aimed to evaluate changes in the transmissibility of influenza across different epidemic periods. We collected and compiled a database of natural history parameters for the three main influenza subtypes: A(H3N2), A(H1N1) and influenza B, along with birth rate, death rate and population data for each region. The natural history parameters were sourced from historical references, while birth and death rates for each study region were obtained from (https://ourworldindata.org), and total population data were sourced from (https://data.worldbank.org.cn). For further details, please refer to online supplemental file 2.

The basic assumptions of the model are as follows:

1. Assuming the population’s birth rate and death rates are denoted by br and dr, respectively. All newborns are defaulted to susceptible individuals, thus at time t, the increase in population is brN, where N represents the population size of each area. The natural mortality in each compartment of the model is denoted as drS, drE, drI, drA and drR.

2. Assuming the transmission rate coefficient after effective contact between S and I is $\beta$, and considering that asymptomatic infections have a transmissibility $k$ times that of I, the new infections at time t are given by $\beta S\left(I+kA\right)$.

3. Assuming the proportion of asymptomatic infections is p, with an incubation period of 1/$w$ and latency period of 1/$w^{`}$. The number of people from E to A and I is $p{w}^{`}$E and (1- $p$)$w$E, respectively.

4. Assuming the interval from the onset of symptoms in I to the time of first diagnosis is 1/$\gamma$, the number of people from I to R at moment t is $\gamma$I.

5. Assuming that individuals can be reinfected by the influenza virus after recovering from infection, with a reinfection probability of $\sigma$, the number of reinfections at time t is $\sigma$R.

The differential equations are as follows:

$${\displaystyle {\displaystyle \begin{array}{c}\frac{dS}{dt}=brN-\beta S\left(I+kA\right)-drS+\sigma R\end{array}}}$$ (1)

$${\displaystyle {\displaystyle \begin{array}{c}\frac{dE}{dt}=\beta S\left(I+kA\right)-p{w}^{\prime }E-\left(1-p\right)wE-drE\end{array}}}$$ (2)

$${\displaystyle {\displaystyle \begin{array}{c}\frac{dI}{dt}=\left(1-p\right)wE-\gamma I-drI\end{array}}}$$ (3)

$${\displaystyle {\displaystyle \begin{array}{c}\frac{dA}{dt}=p{w}^{\prime }E-{\gamma }^{\prime }A-drA\end{array}}}$$ (4)

$${\displaystyle {\displaystyle \begin{array}{c}\frac{dR}{dt}=\gamma I+{\gamma }^{\prime }A-drR-\sigma R\end{array}}}$$ (5)

Estimation of the key parameter

Constructing the state space expression:

Initially, establish the state equations by discretising the system of ordinary differential equations mentioned above to obtain (6), and introduce noise to yield (7).

$${\displaystyle {\displaystyle \begin{array}{c}f{\left(x_t\right)}_{}=x_t+h_k\times \frac{dx}{dt}\end{array}}}$$ (6)

$${\displaystyle {\displaystyle \begin{array}{c}{x}_{t+1}=f{\left(x_t\right)}_{}+w_t\end{array}}}$$ (7)

Where $x_t$ represents the n-dimensional state vector at time t, here x = (S E I A R$\beta$)^T, n=6, the superscript T denotes transpose, $h_k$ denotes the fitting step size, $h_k$=1, and $w_t$ represents the state prediction error, following an n-dimensional normal distribution.

Next, establish the observation equation.

$${\displaystyle {\displaystyle \begin{array}{c}{y}_{t+1}=g{\left(x_t\right)}_{}+v_t\end{array}}}$$ (8)

Where $y_{t+1}$ represents the observation at time t+1, the observed data we obtain is the number of new cases of influenza. Therefore, $g_{{(x}_t)}=\omega \left(1-p\right)$, $v_t$ denotes the measurement error, following a normal distribution. Since the positive reported case count is a fraction of the actual cases, we introduce a coefficient λ into $g_{{(x}_t)}=\omega \left(1-p\right)$, where λ$\in$(0,1). Through testing, λ=0.3 yields the best filtering effect.

We use the Ensemble Kalman filter¹⁸ to infer the transmission rate coefficient $\beta \left(t\right)$, and subsequently calculate the $R_t$ using a defined method.¹⁹ $R_t$ was defined as the expected number of secondary cases produced by an infected individual over the course of their infectious period in a population affected by public health and social measures. The formula is as follows:

$${\displaystyle {\displaystyle \begin{array}{c}{R}_t=\frac{\beta S}{p{\omega }^{\prime }+\left(1-p\right)\omega }\cdot \left[\frac{\kappa p{\omega }^{\prime }}{{\gamma }^{\prime }}+\frac{\left(1-p\right)\omega }{\gamma }\right]\end{array}}}$$ (9)

Due to significant limitations in the actual reported influenza case counts during non-epidemic seasons, where reported counts are 0 and differ from reality, we assume that the $R_t$ of influenza is 0.2 during non-epidemic seasons. The implementation of the compartment model is completed with MATLAB V.2020.

Statistical analysis

We employed the FSA, using data from 5 years preceding the SARS, MERS and COVID-19 epidemics to predict influenza trends, cases and 95% CIs during epidemic periods. This approach aimed to exclude the influence of interventions taken during the major epidemic on influenza epidemiology. FSA uses a discrete Poisson generalised linear model and spline terms to estimate and predict the weekly trend of RPCs, which has been widely applied to the impacts of COVID-19 on zoonotic diseases, vectorborne infectious diseases and other RIDs.^{20 21}

Normality tests were conducted on the total RPCs and $R_t$ values for each country during the SARS, MERS and COVID-19 period by Shapiro–Wilk tests. The value of p>0.05 is defined as a normal distribution and vice versa for non-normal distribution. Based on these test results (online supplemental file 3), the Wilcoxon rank-sum test was applied to statistically compare the differences in total RPCs and $R_t$ values between each pair of periods. For the pre-epidemic period, the mean of the corresponding epidemic weeks was used to represent the average level before the epidemic. All statistical analyses were performed using R V.4.2.1.

Results

Temporal trends in major influenza subtypes

The epidemic curves for the 5 years preceding the SARS (figure 1), MERS and COVID-19 epidemic periods are illustrated in onlinesupplemental files 4 5. A(H1N1) showed significant peaks in some countries and years, such as the USA, Italy, China, Argentina, etc, but overall epidemic intensity was weaker or less variable in many other countries. A(H3N2) was the predominant subtype in most countries, with prominent epidemic peaks, distinct seasonality and peaks often concentrated in the first half of the epidemic season of weeks 5–20. Influenza B exhibited greater regional and annual differences, with high incidence in some countries and years, while being weaker or less variable in others. Overall, the three influenza subtypes showed significant differences in epidemic intensity and peak timing across different countries and years, reflecting the complexity and heterogeneity of global influenza epidemics.

Figure 1. Influenza epidemic curves corresponding to epidemic weeks in the 5 years before the SARS epidemics. RPCs, reported positive cases.

Total RPCs of influenza in pre-epidemics, epidemics and postepidemics

The comparison of total RPCs results showed that all three influenza subtypes showed a significant decrease during the epidemic period compared with the pre-epidemic period (p<0.01) (figure 2). During the SARS pre-epidemics vs epidemics (figure 2A), A(H1N1) decreased from 18 443 cases (10 913, 23 565) to 567 cases (29, 2707), A(H3N2) decreased from 16 184 cases (14 189, 18 050) to 1875 cases (1549, 4122) and influenza B decreased from 8051 cases (4819, 12 112) to 1434 cases (613, 2807). The postepidemic period remained at a lower level, with no significant difference compared with the epidemic period, but still significantly lower than the pre-epidemic period (p<0.01).

Figure 2. Comparison of the total number of RPCs for influenza A (H1N1), A (H3N2) and influenza B in the prepandemic, pandemic and postpandemic periods. (A) is the comparison of the three subtypes in each period during the SARS background period. (B) is the comparison of the three subtypes in each period during the COVID-19 background period. RPCs, reported positive cases.

During the MERS periods, RPCs of A(H1N1) decreased from 964 to 5, A(H3N2) decreased from 369 to 9 and influenza B decreased from 153 to 13. In the postepidemic period, there was a slight resurgence of the A(H3N2) epidemic, while the other two subtypes continued to decline.

During the COVID-19 epidemic (figure 2B), the total number of influenza cases for all three subtypes showed a marked decline. The RPCs of A(H1N1) decreased from 427 (181, 933) in the pre-epidemic period to 253 (45, 627) during the epidemic. A(H3N2) dropped from 396 (152, 1037) to 147.5 (20, 927), and influenza B cases declined from 266 (91, 753) to 139 (29, 629). In contrast to the SARS and MERS epidemics, the postepidemic period of COVID-19 exhibited a significant rebound (p<0.001), with the RPCs for A(H1N1), A(H3N2) and influenza B rising to 889 (407, 1368), 1380 (929, 1667) and 467.5 (210, 943), respectively. However, these values remained lower than those observed in the pre-epidemic period (p<0.05). The RPC curves for each country for the three periods are shown in onlinesupplemental files 68.

Analysis of the variability in influenza transmissibility during different periods

During SARS, the $R_t$ of all three influenza subtypes showed a significant downward trend (figure 3A). Before the epidemic, the $R_t$ for A(H1N1), A(H3N2) and influenza B was 0.43 (0.31, 0.47), 0.42 (0.35, 0.48) and 0.42 (0.35, 0.43), respectively. In the SARS epidemics, the $R_t$ decreased to 0.28 (0.28, 0.30), 0.29 (0.29, 0.33) and 0.28 (0.27, 0.34), respectively (p<0.001). In the postepidemic period, the $R_t$ values for the three subtypes rebounded slightly to 0.28 (0.28, 0.30), 0.30 (0.28, 0.34) and 0.29 (0.28, 0.33), but remained lower than the pre-epidemic levels (p<0.05).

Figure 3. Comparison of $R_t$ of influenza A (H1N1), A (H3N2) and influenza B in the prepandemic, pandemic and postpandemic periods. (A) is the comparison of the three subtypes in each period during the SARS background period. (B) is the comparison of the three subtypes in each period during the COVID-19 background period.

During MERS, the $R_t$ of the three influenza subtypes also exhibited varying degrees of change. Before the pandemic, the $R_t$ for A(H1N1), A(H3N2) and influenza B were 0.38, 0.51 and 0.39, respectively. During the epidemics, the $R_t$ were 0.29, 0.63 and 0.35, respectively. In the postepidemic period, the $R_t$ for the three subtypes was 0.28, 0.31 and 0.73, respectively. Overall, the $R_t$ for A(H1N1) and A(H3N2) during and after the MERS epidemics were lower than those before the epidemics, whereas the $R_t$ for influenza B increased during the pandemic and then decreased markedly in the postpandemic period.

During the COVID-19 epidemics, the $R_t$ of all three influenza subtypes also showed a significant decline (figure 3B). Before the epidemics, the $R_t$ for A(H1N1), A(H3N2) and influenza B were 0.36 (0.36, 0.37), 0.38 (0.36, 0.41) and 0.40 (0.37, 0.47), respectively. During the pandemic, these values dropped to 0.28 (0.27, 0.28), 0.29 (0.28, 0.32) and 0.29 (0.28, 0.31), respectively (p<0.001). In the postepidemic period, the $R_t$ for the three subtypes increased slightly to 0.28 (0.27, 0.28), 0.31 (0.30, 0.38) and 0.30 (0.28, 0.34), but overall remained below prepandemic levels (p<0.05). The $R_t$ curves for each subtype in each country for the three periods are shown in onlinesupplemental files 911.

Influenza epidemic reality in the absence of major epidemics

We used the FSA to calculate the predicted values in the context of the exclusion of the epidemic in each country as shown in onlinesupplemental files 68. In the SARS period (figure 4A), the RPCs of A(H1N1) and influenza B were lower than predicted. The rate of decrease is −84.39% and −45.31%, respectively (p>0.05). The RPCs of A(H3N2) were 438 cases (31, 668), which was close to the predicted median of 430 cases (218, 1324). Specifically, the RPCs of A(H1N1) in several countries such as Canada, Hong Kong, Italy, Sweden and England were lower than predicted, with change rates ranging from −52.17% to −99.52%. For A(H3N2), the change rates ranged from −28.09% to −91.36%, with Canada and the USA showing most changes of −91.36% and −86.73%, respectively. Influenza B showed change rates ranging from −4.77% to −99.76%, while other unmentioned countries and subtypes exhibited an increasing trend.

Figure 4. Comparison of predicted values of FSA excluding influenza A (H1N1), A (H3N2) and influenza B in the context of an epidemic and real RPCs of influenza. (A) is the comparison of the predicted and real RPCs of the three subtypes during the SARS period. (B) is the comparison of the predicted and true RPCs of the three subtypes during the COVID-19 period. FSA, Farrington surveillance algorithm; RPCs, reported positive cases.

During the MERS epidemic, A(H1N1) and A(H3N2) were lower than predicted, with change rates of −28.75% and −17.62%, respectively. Influenza B was higher than predicted with a change rate of 176.61%.

During the COVID-19 period (figure 4B), the total RPCs of the three major influenza subtypes were significantly lower than predicted (p<0.01), with reduction rates of −53.30%, −57.50% and −48.56%, respectively. In over 71% of countries, the RPCs of A(H1N1) were lower than predicted, with countries such as Australia, Bangladesh, Brazil, Canada, China, Spain, Germany, India, Japan, Mexico, Saudi Arabia, Republic of Korea, the UK and USA, showing reductions generally ranging from 4.52% to 100%. A(H3N2) and influenza B exhibited similar trends with A(H1N1) in many countries such as Australia, Bangladesh, Canada, China, Brazil, Germany, India, Japan, Mexico and the USA showing reductions in RPCs compared with the predicted, generally ranging from 0.72% to 100%.

Discussions

Our study systematically compared the global trends in influenza prevalence before, during and after the SARS, MERS and COVID-19 epidemics. The results indicate that during all three epidemics, there was a significant decline in the RPCs of influenza and transmissibility. The suppressive effect of the COVID-19 epidemic on influenza prevalence was the most pronounced. During the COVID-19 epidemic, the RPCs of the three major influenza subtypes showed the largest decrease compared with historical predictions, with reduction rates of −53.30% for A(H1N1), −57.50% for A(H3N2) and −48.56% for influenza B, with A(H3N2) being the most significantly affected, as most countries experienced reductions exceeding 50%. The impact of the SARS epidemic on influenza was secondary, with total RPCs of A(H1N1) and influenza B decreasing by approximately 84.39% and 45.31%, respectively. During the MERS epidemic, the RPCs of A(H1N1) and A(H3N2) decreased by 28.75% and 17.62%, respectively, although influenza B partially rebounded in the later stages, resulting in a relatively smaller overall impact. In summary, coronavirus epidemics have had a profound impact on global influenza prevalence. The COVID-19 epidemic had the strongest suppressive effect on influenza prevalence, followed by SARS, with MERS having the least impact. Among the subtypes, A(H3N2) was the most affected during these periods.

During coronavirus epidemics, there were notable differences between the total RPCs and those predicted for excluding epidemics, which may be partly related to interventions, disease perception, vaccination and other factors. The extent of transmission varies between epidemics and therefore the intensity of interventions varies, with non-pharmaceutical interventions (NPIs) playing a significant role in attenuating influenza transmission. Although the SARS epidemic was far-reaching, it was mainly concentrated in East Asia and some South-East Asian countries. SARS epidemics are characterised by a narrower transmission scope; primarily within healthcare settings and stringent control measures,^{22 23} the epidemic duration was relatively short, typically lasting 1–2 months in most study areas. During the SARS epidemic, two categories of NPIs were predominantly implemented.²⁴ One of the measures aimed at reducing contact between infections and susceptibility, including isolation, quarantine, travel restrictions and enhanced social distancing. Another intervention targeting the reduction of effective contact between cases and the environment encompasses hand hygiene, mask wearing, disinfection and ventilation practices. In the periods of MERS and COVID-19, in addition to the aforementioned NPIs, extensive interventions such as school closures, restrictions on international flights and mass nucleic acid testing were introduced to mitigate both international and domestic population movements.²⁵,²⁸ These measures were similarly deemed effective in influenza protection,^{29 30} effectively reducing opportunities for influenza virus and coronavirus release by infected individuals, diminishing virus survival rates in the environment and weakening the transmissibility of the influenza virus. Consequently, reductions in outpatient consultation rates for influenza, as well as decreases in both the influx and spread of influenza cases, were observed. Large-scale interventions such as school closures were estimated to reduce total influenza cases by 15% and peak case numbers by 40%.³¹ Although similar interventions were taken during the MERS and COVID-19 periods, COVID-19 was a global epidemic with unprecedentedly stringent preventive and control measures implemented in almost all countries, and MERS was an outbreak in Korea with a limited prevalence and a weak human-to-human transmission of MERS. Furthermore, heightened psychological awareness and fear of SARS, MERS and COVID-19 further bolstered the strength and efficacy of control measures, consequently reducing influenza transmission during other epidemics.^{32 33} Furthermore, unique pharmaceutical interventions and vaccination during the COVID-19 era contributed to influenza suppression. Antiviral medications³⁴,³⁶ such as ribavirin and azithromycin, proven effective in treating COVID-19, have demonstrated efficacy in treating patients with influenza, which can reduce illness duration and the incubation period to lower population infection rates. Influenza vaccination coverage in the USA and the UK saw significant increases during the COVID-19 era (2020–2021), with 28% of the global population expressing a notable willingness to receive vaccination.³⁷,³⁹ This trend suggests that public awareness and acceptance of prophylactic vaccination has increased after the COVID-19 epidemic and that this increased willingness to vaccinate has contributed to higher rates of influenza vaccination and thus lower rates of influenza incidence. Lastly, the existence of ecological competition between SARS-CoV-2 and influenza viruses within epithelial cells may confer COVID-19 dominance, thereby reducing the risk of influenza virus infection within the host.⁴⁰,⁴²

In the background of the same coronavirus epidemic, there are significant differences in the impact on the prevalence intensity of different influenza subtypes, particularly with A(H3N2) showing greater declines compared with A(H1N1) and influenza B. This disparity can be explained by several factors: the biological characteristics and transmission dynamics of the viruses differ. The antigenic drift of the A(H3N2) subtype occurs more rapidly than that of A(H1N1) and influenza B, resulting in a generally lower level of immunity in the population. Consequently, the prevalence intensity of A(H3N2) depends on the proportion of susceptible individuals and the frequency of social contacts.^{43 44} Under large-scale NPIs, the transmission chain of A(H3N2) is more easily disrupted, leading to a significant reduction in case numbers. Additionally, influenza B primarily affects children and adolescents, whose social activities are heavily dependent on school environments. Studies have shown that school closures have a particularly significant suppressive effect on the low incidence of influenza B.⁴⁴ Differences in immune background and previous exposure history also play a crucial role. Since the 2009 epidemic, the global population has generally maintained a higher level of immunity to the A(H1N1) subtype. Therefore, even under NPIs, the decline in the prevalence intensity of A(H1N1) is relatively limited. Approximately 500 million doses of influenza vaccines are produced globally each year, most of which are egg-grown inactivated subunit or split virus particle vaccines, offering moderate efficacy against A(H1N1) and influenza B, but slightly lower efficacy against A(H3N2).⁴⁵ As a result, the better immune foundation for A(H1N1) and influenza B in the population leads to their slightly lower sensitivity to NPI measures compared with A(H3N2).

Our study primarily explored the changes in influenza epidemiological characteristics during the SARS, MERS and COVID-19 periods, including variations in RPCs, changes in transmissibility and trend changes after removing epidemic interference. This provides a theoretical basis for understanding the epidemiological patterns of influenza during other coronavirus epidemics. Notably, the decline of influenza during the three coronavirus epidemics highlights the synergistic effects of NPIs. Large-scale public health interventions not only effectively curbed the spread of ERIDs but also significantly suppressed the prevalence of common respiratory viruses like influenza. This phenomenon suggests that comprehensive control strategies for RIDs are broadly applicable and highly effective. However, as epidemic control measures are gradually relaxed, there is a potential for a rebound in the prevalence of influenza and other viruses, possibly leading to larger outbreaks due to a decline in population immunity barriers. Therefore, future RID control policies should fully consider multipathogen collaborative control strategies and enhance monitoring and vaccination of immunologically naive populations to improve global capacity to respond to ERIDs and re-ERIDs.

Limitations

Our study has some limitations: First, previous research has indicated that influenza surveillance systems and diagnostic methods vary across countries, resulting in the influenza case data we obtained lacking a unified definition. However, the GISRS has grown to include 143 WHO-accredited National Influenza Centres, 6 WHO Collaborating Centres, 4 Basic Regulatory Laboratories and 13 H5 Reference Laboratories in the 65 years of its existence, which is the only global system with comprehensive influenza case reporting, and numerous studies have used data from this system for analysis.^{16 17} Second, due to data acquisition constraints, we were unable to obtain influenza data from Saudi Arabia before 2012, so only South Korea was included as a representative region to assess the impact of MERS on influenza. Saudi Arabia has been more heavily affected by MERS, better reflecting the impact of its pandemic on influenza, whereas the MERS outbreak in South Korea may have had a short-term effect, which may weaken the representativeness of the study regions. What’s more, particularly with the low number of reports for influenza B subtypes, some years and regions had only single-digit reports. This resulted in many units having unreported or extremely sparse data after further stratification, which would severely affect the stability of statistical analysis and the reliability of results. Therefore, in this study, we combined different subtypes of influenza B (Yamagata, Victoria and unknown subtypes) for analysis to ensure the representativeness and robustness of the analysis. Lastly, during other coronavirus epidemics, the focus and intensity of influenza surveillance may have decreased, and this study cannot yet rule out the possibility that changes in influenza characteristics during epidemics may be influenced by inherent deficiencies in the surveillance reporting system.

Conclusions

Our study revealed that during the three coronavirus epidemics, there was a significant decline in both influenza-positive case numbers and transmissibility, with varying degrees of impact. The COVID-19 epidemic demonstrated the most pronounced suppressive effect on influenza prevalence, showing reductions exceeding 50% in most countries globally. The impact of SARS was secondary, while MERS had the least effect. Among the different influenza subtypes, A(H3N2) exhibited greater declines compared with A(H1N1) and influenza B. The decrease in RPCs during coronavirus epidemics highlighted the importance of NPIs, demonstrating the broad applicability and high efficacy of comprehensive control strategies for RIDs. Furthermore, when NPIs are lifted during the later stages of coronavirus epidemics, attention should be paid to the potential rebound of traditional respiratory diseases such as influenza. These findings provide important insights into the mechanisms by which coronavirus epidemics affect influenza prevalence and offer valuable references for developing influenza control strategies in the context of multiple RIDs.

Data availability statement

Data are available in a public, open access repository.