Asymmetric post-pandemic recovery of influenza A and B since 2020 in Hong Kong: an interrupted time-series (ITS) analysis of weekly surveillance data with subtype and lineage characterization Authors: Ling Yang, Muhammad Riaz, Najm Ur Rahman Published: 2026 ABSTRACT BackgroundInfluenza circulation declined markedly worldwide after the emergence of COVID-19 in 2020. Although early suppression was widely documented, less information is available on how influenza activity returned during later years, particularly for influenza A and influenza B and their subtype or lineage distributions.MethodsWeekly laboratory surveillance data from Hong Kong covering 2014–2024 were analyzed using interrupted time-series regression. Weekly numbers of tested specimens were included as an offset. Epidemiological week 1 of 2020 was used as the primary interruption point. Influenza A subtype activity, A(H1) and A(H3), and influenza B lineage activity, Victoria and Yamagata, were also examined descriptively. Other respiratory viruses (ORV) were included for general comparison.ResultsBoth influenza A and influenza B declined sharply during early 2020, although the decline was greater for influenza B. Influenza A activity increased again during later years, with recurrent seasonal peaks and variation in the relative activity of A(H1) and A(H3). Influenza B remained at comparatively low levels through much of the post-2020 period. Among characterized influenza B viruses, most detections belonged to the B/Victoria lineage, whereas B/Yamagata was not identified in later surveillance data. ORV circulation showed greater variability after 2020 and did not follow the same pattern observed for influenza.ConclusionRecovery patterns differed between influenza A and influenza B after 2020 in Hong Kong. Influenza A circulation resumed gradually, whereas influenza B activity remained lower and lineage diversity was reduced. Continued surveillance at the level of virus type, subtype, and lineage may help identify future changes in influenza circulation during the post-pandemic period.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12879-026-13714-8. Introduction Seasonal influenza remains a major respiratory infection worldwide and is mainly caused by influenza A and influenza B viruses [1]. Although these viruses are commonly discussed together, they differ in several important ways, including host range, evolutionary behavior, and patterns of circulation. Influenza A has greater antigenic diversity and circulates in both humans and animal hosts. As a result, dominant subtypes can shift considerably between seasons, sometimes producing large differences in epidemic intensity. Influenza B, in contrast, circulates almost exclusively in humans and evolves more slowly. Its circulation is mainly linked to the Victoria and Yamagata lineages [2]. Because of these biological differences, large disruptions in population mixing and respiratory virus transmission may not affect influenza A and influenza B equally. Changes in subtype or lineage dominance can also alter the timing and magnitude of seasonal epidemics. For this reason, examining influenza activity only as a combined outcome may overlook important differences between virus groups. At the beginning of 2020, circulation of respiratory viruses changed markedly across many parts of the world following the emergence of COVID-19 and the implementation of extensive public health and behavioral interventions. Influenza activity declined sharply in multiple regions and remained unusually low throughout 2020 and 2021 [3, 4]. Similar patterns were observed in Hong Kong, where long-standing laboratory surveillance systems recorded extremely low influenza activity and disruption of the usual seasonal epidemic pattern [5, 6]. As public health measures gradually relaxed and population mobility increased, influenza activity began to re-emerge. However, it remained unclear whether influenza A and influenza B resumed circulation symmetrically and whether subtype or lineage composition changed following the prolonged suppression period [5, 7]. Recovery from pandemic related suppression is not necessarily expected to be uniform across influenza virus types, subtypes, or lineages. Differences in viral fitness, population immunity, and transmission ecology may result in divergent post-disruption trajectories for influenza A and influenza B, as well as selective re-emergence of influenza A subtypes or influenza B lineages [8, 9]. Early descriptive reports have suggested asynchronous rebounds of respiratory viruses following 2020, but most analyses have relied on short observation windows or aggregated influenza indicators [10, 11]. Such approaches limit the ability to distinguish pre-existing temporal trends from pandemic-related changes and obscure subtype or lineage-specific dynamics that may underlie aggregate recovery patterns. Understanding how influenza circulation changed after 2020 requires separating the immediate disruption from longer-term shifts in transmission patterns. Interrupted time-series (ITS) analysis is useful in this context because it allows post-2020 observations to be examined against historical trends rather than treated as isolated changes. Using surveillance data from more than a decade also makes it possible to account for recurring seasonal variation, gradual temporal trends, and fluctuations in testing volume over time. This approach helps distinguish the abrupt decline seen during the early pandemic period from the slower recovery that followed. Subtype and lineage information provides additional context. Instead of treating influenza circulation as a single combined outcome, it allows assessment of whether post-2020 recovery reflected broad restoration of transmission or was driven mainly by a smaller number of viral groups [12, 13]. Weekly influenza surveillance data from Hong Kong between 2014 and 2024 were used to examine how circulation patterns changed after the disruptions of 2020. ITS analysis was applied to evaluate changes in influenza activity while accounting for long-term temporal trends, seasonality, and variation in testing volume over time. In addition to overall influenza activity, subtype distributions for influenza A and lineage patterns for influenza B were examined to better understand post-2020 changes in circulation. Attention was given to differences in recovery between influenza A and influenza B after the prolonged suppression observed during 2020–2021. Influenza trends were also compared with those of other respiratory viruses (ORV) to determine whether the observed changes were influenza-specific or reflected broader shifts in respiratory virus activity. Most early pandemic studies mainly described the initial decline in influenza circulation during 2020. By extending surveillance through 2024 and incorporating a longer pre-pandemic baseline, the present study was able to evaluate both the immediate disruption and subsequent recovery patterns across multiple post-pandemic seasons. To our knowledge, few studies from Hong Kong have examined these long-term post-2020 trends together with subtype- and lineage level patterns. Methods Data sources and surveillance system Weekly laboratory-based respiratory virus surveillance data were obtained from DATA.GOV.HK, the public open-data platform of the Hong Kong SAR Government. The dataset includes weekly numbers of respiratory specimens tested and the corresponding numbers positive for specific viral pathogens. Data covering January 2014 to December 2024 were included in the analysis [14]. For each epidemiological week, the surveillance system provided counts of specimens positive for influenza A, influenza B, and several ORV, including rhinovirus/enterovirus, respiratory syncytial virus, adenovirus, and human metapneumovirus. Weekly positivity was calculated as the proportion of tested specimens positive for the virus of interest. For comparison with influenza trends, non-influenza respiratory viruses were grouped into a combined category. Individual virus groups were not analyzed separately because the primary objective was to compare overall respiratory virus activity with influenza circulation rather than describe the epidemiology of each pathogen independently. Influenza subtype and lineage data were obtained from a separate publicly available surveillance dataset and were available only for a subset of influenza-positive specimens. As these datasets differ in testing scope and reporting structure, analyses of overall influenza activity and subtype/lineage distributions were conducted separately, and counts are not directly comparable across datasets. The proportion of influenza-positive specimens undergoing subtype or lineage characterization varied over time, and may have been influenced by changes in testing priorities and laboratory capacity during the COVID-19 period. Study design We applied an interrupted time-series (ITS) design to evaluate changes in weekly virus positivity associated with the onset of the COVID-19 pandemic [15]. This approach is appropriate for assessing population-level disruptions when randomized designs are not feasible. The interruption point in the primary analysis was defined a priori as epidemiological week 1 of 2020. This breakpoint was selected to capture the earliest phase of pandemic-related behavioral adaptation and heightened public awareness preceding formal large-scale interventions in Hong Kong. Because behavioral responses and precautionary changes often begin before official policy implementation, an early predefined breakpoint was considered more appropriate than a retrospectively optimized interruption date. In addition, defining the interruption point a priori reduced the risk of post hoc breakpoint selection based on observed outcomes. Sensitivity analyses using an alternative interruption date of 1 March 2020 were also conducted to assess robustness of the findings. The ITS framework was used to estimate three parameters: the pre-pandemic secular trend, the immediate level change at the interruption point, and the change in trend during the post-interruption period. For clarity, the term “post-2020 period” refers to the period following the onset of COVID-19 related disruptions in early 2020. Statistical analysis Weekly influenza positivity was analyzed using generalized linear models with a quasi-Poisson distribution to account for overdispersion. Models were fitted to weekly counts of positive specimens, with the logarithm of the number tested included as an offset. Estimates from the ITS models are therefore interpreted as rate ratios (RRs) reflecting changes in weekly positivity rather than absolute counts. Separate models were fitted for influenza A, influenza B, and total influenza A + B. Each model included a continuous time variable, a post-2020 indicator term, and a post-interruption time term to estimate changes in trend after 2020. Seasonal variation was adjusted using epidemiological week indicators. Regression coefficients were exponentiated to obtain RRs and 95% confidence intervals. For graphical presentation, fitted values were converted to weekly positivity estimates using the number of specimens tested, and observed trends were plotted together with model-fitted values across the study period. Alternative model specifications and sensitivity analyses Several additional analyses were performed to examine the stability of the primary findings. Poisson models using the same linear predictors and seasonal adjustment were re-estimated with Newey–West standard errors (lag = 4). To address short-term temporal dependence, lag-adjusted models incorporating a one-week autoregressive term were also evaluated separately for influenza A and influenza B. Alternative approaches to seasonality were examined by replacing categorical epidemiological week indicators with harmonic sine and cosine terms using a 52-week cycle. Additional analyses excluded the first eight weeks of 2020 to evaluate the influence of the early transition period. Because major behavioral and public health changes intensified during late January and February 2020, sensitivity analyses were also repeated using 1 March 2020 as an alternative interruption point. Across these analyses, the estimated direction and overall magnitude of the main ITS parameters remained broadly similar to those observed in the primary quasi-Poisson models. Model diagnostics and visualization Model fit was assessed by examining residual patterns and comparing observed with fitted trends. Overdispersion was evaluated for the quasi-Poisson models [16, 17]. Residual autocorrelation was assessed using autocorrelation function (ACF) plots of Pearson residuals from the influenza A and influenza B models, which showed remaining short-term temporal dependence at early lags (Figure S3). To assess the robustness of the primary findings, additional sensitivity analyses incorporated lag-adjusted models and Newey–West standard errors [18]. These alternative specifications produced findings broadly similar to those from the primary analyses. Figures display observed weekly positivity together with fitted trends, with a vertical reference line indicating epidemiological week 1 of 2020 as the primary interruption point. Supplementary analyses also considered an alternative interruption date of 1 March 2020. Subtype and lineage characterization Where available, influenza A detections were disaggregated by subtype and influenza B detections by lineage. These analyses were descriptive and intended to contextualize type-specific recovery patterns rather than support causal inference. Weekly counts and within-type proportions were summarized separately for influenza A and influenza B. Proportions were calculated only for weeks with non-zero denominators to avoid unstable estimates, and rolling means were used to reduce short-term variability [19]. Statistical software All analyses were conducted in R version 4.2 (R Foundation for Statistical Computing). Interrupted time-series models were fitted using generalized linear models with quasi-Poisson or Poisson distributions. Data management and visualization were performed using packages from the tidyverse framework. Results Interrupted time-series analysis of influenza A, influenza B, and ORV Interrupted time-series analyses were used to examine changes in influenza and respiratory virus positivity after 2020. The primary interruption point was defined as epidemiological week 1 of 2020. This breakpoint was selected before analysis to represent the early phase of COVID-19-related behavioral and public health changes in Hong Kong. Weekly positivity was modeled using quasi-Poisson regression with the number of specimens tested included as an offset. The models estimated pre-2020 temporal trends, immediate changes after the interruption point, and post-2020 trends in circulation. Main estimates from the primary models are summarized in Table 1, with fitted trends shown in Figs.1–3. Prior to 2020, influenza A positivity showed a modest declining trend, with clear seasonal peaks across successive years (Fig. 1). At the interruption point (epidemiological week 1 of 2020), a marked and immediate reduction in positivity was observed, consistent with the ITS estimates reported in Table 1. Influenza A positivity showed a modest declining trend before 2020, with clear seasonal peaks across successive years (Fig. 1). At the interruption point (epidemiological week 1 of 2020), a marked and immediate reduction in influenza positivity was observed. Influenza A declined by approximately 78% (RR 0.22, 95% CI 0.13–0.36), while influenza B showed a substantially larger reduction of nearly 98% (RR 0.02, 95% CI 0.006–0.074), indicating near-complete suppression during the early pandemic period. In contrast, ORV did not show a statistically significant immediate decline (RR 1.12, 95% CI 0.96–1.32), suggesting a more heterogeneous response to pandemic-related disruptions. After the marked decline observed in early 2020, influenza activity began to increase again over subsequent years. Influenza A showed a significant positive post-2020 trend (RR 1.009 per week, 95% CI 1.007–1.012), with seasonal peaks becoming more apparent during later years, although activity generally remained below pre-2020 levels. Influenza B also demonstrated a positive post-2020 trend (RR 1.010, 95% CI 1.003–1.017), but circulation remained limited throughout much of the post-2020 period. Patterns for ORV differed from those observed for influenza. Although week-to-week variability remained high, the overall post-2020 trend showed a modest decline (RR 0.997 per week, 95% CI 0.996–0.998) rather than evidence of sustained recovery. Overall, influenza B appeared more strongly affected by the post-2020 disruption and recovered more slowly than influenza A. Findings were consistent across sensitivity analyses, including harmonic seasonal adjustment, lag-adjusted models, and models with Newey West standard errors (Supplementary Tables S1–S3; Figure S1). Table 1Interrupted time-series estimates for influenza A, influenza B, and ORV positivity, 2014–2024ParameterInfluenza A Rate Ratio (95% CI)Influenza B Rate Ratio (95% CI)ORV Rate Ratio (95% CI)Pre-2020 time trend (per week)0.998 (0.997–0.999)0.999 (0.998-1.000)0.999 (0.999-1.000)Immediate post-2020 level change0.22 (0.13–0.36)0.02 (0.006–0.074)1.12 (0.96–1.32)Post-2020 time trend (per week)1.009 (1.007–1.012)1.010 (1.003–1.017)0.997 (0.996–0.998)Rate ratios (RRs) represent multiplicative changes in weekly virus positivity. Models were estimated using quasi-Poisson regression with the logarithm of weekly specimens tested included as an offset. Seasonality in the primary models was adjusted using categorical epidemiological week indicators. Harmonic sine and cosine terms (52-week periodicity) were evaluated in sensitivity analyses. The post-2020 indicator corresponds to epidemiological week 1 of 2020, which was used as the breakpoint in all interrupted time-series models. All estimates are expressed on the scale of virus positivity (test-positive proportion) Table 1Interrupted time-series estimates for influenza A, influenza B, and ORV positivity, 2014–2024ParameterInfluenza A Rate Ratio (95% CI)Influenza B Rate Ratio (95% CI)ORV Rate Ratio (95% CI)Pre-2020 time trend (per week)0.998 (0.997–0.999)0.999 (0.998-1.000)0.999 (0.999-1.000)Immediate post-2020 level change0.22 (0.13–0.36)0.02 (0.006–0.074)1.12 (0.96–1.32)Post-2020 time trend (per week)1.009 (1.007–1.012)1.010 (1.003–1.017)0.997 (0.996–0.998)Rate ratios (RRs) represent multiplicative changes in weekly virus positivity. Models were estimated using quasi-Poisson regression with the logarithm of weekly specimens tested included as an offset. Seasonality in the primary models was adjusted using categorical epidemiological week indicators. Harmonic sine and cosine terms (52-week periodicity) were evaluated in sensitivity analyses. The post-2020 indicator corresponds to epidemiological week 1 of 2020, which was used as the breakpoint in all interrupted time-series models. All estimates are expressed on the scale of virus positivity (test-positive proportion) Interrupted time-series estimates for influenza A, influenza B, and ORV positivity, 2014–2024 Rate ratios (RRs) represent multiplicative changes in weekly virus positivity. Models were estimated using quasi-Poisson regression with the logarithm of weekly specimens tested included as an offset. Seasonality in the primary models was adjusted using categorical epidemiological week indicators. Harmonic sine and cosine terms (52-week periodicity) were evaluated in sensitivity analyses. The post-2020 indicator corresponds to epidemiological week 1 of 2020, which was used as the breakpoint in all interrupted time-series models. All estimates are expressed on the scale of virus positivity (test-positive proportion) Influenza A Before 2020, influenza A positivity showed a slight downward trend over time (RR 0.998 per week, 95% CI 0.997–0.999), although clear seasonal peaks remained evident across successive years (Fig. 1). At the beginning of 2020, influenza A activity declined sharply. Weekly positivity decreased by approximately 78% (RR 0.22, 95% CI 0.13–0.36) following the interruption point. Influenza A activity remained low during the early pandemic period but gradually increased again during later years. The post-2020 trend was positive and statistically significant (RR 1.009 per week, 95% CI 1.007–1.012). Seasonal peaks reappeared over time, although activity generally remained below pre-2020 levels during the earlier post-pandemic years (Fig. 1). Fig. 1Weekly influenza A positivity in Hong Kong from 2014 to 2024, showing observed values and fitted interrupted time-series trends. Solid lines show observed positivity, and dashed lines show model-fitted trends. The vertical dashed line marks epidemiological week 1 of 2020, used as the primary interruption point Weekly influenza A positivity in Hong Kong from 2014 to 2024, showing observed values and fitted interrupted time-series trends. Solid lines show observed positivity, and dashed lines show model-fitted trends. The vertical dashed line marks epidemiological week 1 of 2020, used as the primary interruption point Influenza B Before 2020, influenza B showed recurring seasonal peaks across successive years (Fig. 2). At the beginning of 2020, influenza B activity declined sharply following the interruption point. Weekly positivity decreased by approximately 98% (RR 0.02, 95% CI 0.006–0.074), and influenza B detections remained very low throughout most of 2020 and 2021. Although the post-2020 trend was statistically positive (RR 1.010 per week, 95% CI 1.003–1.017), activity increased from a near-zero baseline and remained limited during much of the post-2020 period. Only sporadic detections were observed, and the regular seasonal pattern seen before 2020 did not clearly reappear during the study period (Fig. 2). Compared with influenza A, recovery of influenza B circulation appeared slower and less complete. Fig. 2Observed and model-fitted weekly influenza B positivity in Hong Kong, 2014–2024. The dashed vertical line indicates the primary interruption point at epidemiological week 1 of 2020 Observed and model-fitted weekly influenza B positivity in Hong Kong, 2014–2024. The dashed vertical line indicates the primary interruption point at epidemiological week 1 of 2020 Other respiratory viruses (ORV) Before 2020, positivity for ORV showed a slight downward trend over time (RR 0.9994 per week, 95% CI 0.9991–0.9998). Compared with influenza, positivity levels were generally higher and seasonal fluctuations were broader across the study period (Fig. 3). At the interruption point in early 2020, no significant immediate level change was observed (RR 1.12, 95% CI 0.96–1.32). Following 2020, the overall post-2020 trend showed a modest decline (RR 0.997 per week, 95% CI 0.996–0.998). Patterns during this period were more variable than those seen before 2020, with irregular peaks and less consistent seasonality. Because this category included several respiratory viruses with different transmission patterns, these analyses were intended mainly as a general comparison with influenza trends rather than as a detailed evaluation of individual pathogens. Fig. 3Observed and model-fitted weekly positivity of ORV in Hong Kong, 2014–2024. The dashed vertical line indicates the primary interruption point at epidemiological week 1 of 2020 Observed and model-fitted weekly positivity of ORV in Hong Kong, 2014–2024. The dashed vertical line indicates the primary interruption point at epidemiological week 1 of 2020 Descriptive temporal patterns of influenza activity, 2014–2024 In addition to the interrupted time-series analyses, weekly surveillance data were examined descriptively to assess long-term patterns in influenza activity between 2014 and 2024. These analyses were based on observed weekly detections rather than model-estimated positivity. Seasonal fluctuations, inter-annual variation, and changes in the timing and intensity of influenza circulation were examined across the study period. The descriptive analyses also helped place the ITS findings within the broader temporal patterns observed before and after 2020. Long term weekly trends in influenza detections Between 2014 and early 2020, influenza activity in Hong Kong showed clear seasonal variation, with recurrent winter peaks and substantial differences in activity between years. Influenza A accounted for most detections during this period and contributed to the largest seasonal peaks, whereas influenza B generally remained lower in magnitude and less frequently dominated seasonal epidemics (Fig. 4). Influenza detections declined sharply in early 2020 following the introduction of COVID-19-related public health measures. Activity remained very low throughout most of 2020 and 2021, and the usual seasonal pattern was largely absent during this period (Figs. 4 and 5). Influenza circulation began to increase again from late 2022 onward. Seasonal activity became more evident during 2023 and 2024, although peak levels generally remained below those observed before 2020. Influenza A accounted for most detections during the post-2022 period. Subtype and lineage patterns also varied across seasons after 2020. The relative contributions of A(H1) and A(H3) changed over time without a consistently dominant subtype. Influenza B lineage activity remained intermittent, partly because overall influenza B circulation was limited during much of the post-2020 period (Fig. 5). Fig. 4Weekly composition of influenza detections in Hong Kong by virus subtype and lineage, 2014–2024. The stacked area plot represents absolute weekly detections (counts) of influenza A(H1), A(H3), B/Victoria, and B/Yamagata. The vertical dashed line indicates the interruption point (epidemiological week 1, 2020), and the shaded region denotes the main period of COVID-19 related suppression Weekly composition of influenza detections in Hong Kong by virus subtype and lineage, 2014–2024. The stacked area plot represents absolute weekly detections (counts) of influenza A(H1), A(H3), B/Victoria, and B/Yamagata. The vertical dashed line indicates the interruption point (epidemiological week 1, 2020), and the shaded region denotes the main period of COVID-19 related suppression Fig. 5Within-type proportions of influenza subtypes and lineages over time (not absolute counts), 2014–2024. Proportions were calculated within each virus type using a 3-week rolling mean to reduce short-term variability. Values were set to go missing when weekly denominators were zero. The vertical dashed line indicates the interruption point (epidemiological week 1, 2020) Within-type proportions of influenza subtypes and lineages over time (not absolute counts), 2014–2024. Proportions were calculated within each virus type using a 3-week rolling mean to reduce short-term variability. Values were set to go missing when weekly denominators were zero. The vertical dashed line indicates the interruption point (epidemiological week 1, 2020) Disruption and partial restoration of influenza seasonality Prior to 2020, influenza detections were strongly concentrated in winter months, with consistent clustering of peak activity across calendar years. The highest detection intensity typically occurred within a relatively narrow range of epidemiological weeks each year, reflecting a stable and well-defined seasonal structure. During the COVID-19 control period, this seasonal concentration was markedly disrupted. From 2020 to 2022, influenza detections were sparse across most weeks of the year, and the characteristic winter clustering was largely absent, indicating a prolonged interruption of established seasonal dynamics. Seasonal activity became more evident again after 2023, particularly during winter weeks. Peak activity, however, remained lower and less regular than before 2020 (Fig. 6). Fig. 6Heatmap of influenza detections in Hong Kong, 2014–2024, showing marked suppression during 2020–2022 and partial seasonal recovery afterward Heatmap of influenza detections in Hong Kong, 2014–2024, showing marked suppression during 2020–2022 and partial seasonal recovery afterward During 2023 and 2024, influenza activity again increased mainly during winter weeks. Seasonal peaks reappeared, although they remained smaller and less regular than those observed before 2020 (Fig. 6). Comparative post-2020 dynamics of influenza A and influenza B Differential recovery of influenza A and influenza B Before 2020, influenza A accounted for most laboratory-confirmed influenza detections, representing approximately 68–90% of annual cases between 2016 and 2019 (Table 2). Influenza B contributed the remaining proportion, although activity varied between seasons. Overall influenza activity also fluctuated considerably across years before the pandemic. Influenza detections declined sharply after early 2020 and remained at very low levels through most of 2022. Activity increased again during 2023 and 2024, largely because of increasing influenza A circulation. Subtype and lineage data were available only for a subset of characterized specimens and are presented mainly to describe relative composition rather than absolute burden. After 2022, influenza A accounted for most detected influenza cases, whereas influenza B activity remained limited. In 2023, approximately 97% of laboratory-confirmed detections were influenza A, with a similar pattern observed in 2024 despite partial-year data (Table 2). Weekly trends showed increasing influenza A activity during the later post-pandemic period, while influenza B detections remained sporadic and comparatively low (Fig. 7). Fig. 7Weekly influenza positivity and fitted interrupted time-series trends in Hong Kong, 2014–2024. Separate panels show total influenza, influenza A, and influenza B. The dashed vertical line indicates the primary interruption point at epidemiological week 1 of 2020 Weekly influenza positivity and fitted interrupted time-series trends in Hong Kong, 2014–2024. Separate panels show total influenza, influenza A, and influenza B. The dashed vertical line indicates the primary interruption point at epidemiological week 1 of 2020 Table 2Annual totals of laboratory-confirmed influenza detections in Hong Kong, stratified by virus type, influenza A subtype, and influenza B lineage. Percentages indicate the proportion of total annual detections attributable to influenza A and influenza BSeason yearWeeksTotal influenzaInfluenza A (n)Influenza B (n)Influenza A (%)Influenza B (%)A(H1)A(H3)B(Victoria)B(Yamagata)20145212,0117,7564,25564.635.43,6894,2215423,23620155222,77020,0922,67888.211.827814,860831,76520165320,53013,9386,59267.932.11,5483,06174285120175227,64524,9732,67290.39.76684,03629899920185222,7089,05513,65339.960.12,297844921,24020195224,21520,8403,37586.113.92,4291,8471,002112020525,7155,44527095.34.748313636020215286384844.255.80210202253118922678.022.08407020235217,42016,91950197.12.99352,0749602024†2613,71812,5651,15391.68.49906131860† 2024 represents a partial year, including epidemiological weeks 1–26 only. Percentages indicate the proportion of annual influenza detections attributable to influenza A and influenza B. Annual influenza A and B totals were derived from the primary surveillance dataset. Subtype and lineage counts were obtained from a separate subtyping dataset and are available only for a subset of influenza-positive specimens. Because these datasets differ in sampling frame and coverage, subtype and lineage counts are presented for descriptive purposes and are not expected to sum to total influenza A or B counts Table 2Annual totals of laboratory-confirmed influenza detections in Hong Kong, stratified by virus type, influenza A subtype, and influenza B lineage. Percentages indicate the proportion of total annual detections attributable to influenza A and influenza BSeason yearWeeksTotal influenzaInfluenza A (n)Influenza B (n)Influenza A (%)Influenza B (%)A(H1)A(H3)B(Victoria)B(Yamagata)20145212,0117,7564,25564.635.43,6894,2215423,23620155222,77020,0922,67888.211.827814,860831,76520165320,53013,9386,59267.932.11,5483,06174285120175227,64524,9732,67290.39.76684,03629899920185222,7089,05513,65339.960.12,297844921,24020195224,21520,8403,37586.113.92,4291,8471,002112020525,7155,44527095.34.748313636020215286384844.255.80210202253118922678.022.08407020235217,42016,91950197.12.99352,0749602024†2613,71812,5651,15391.68.49906131860† 2024 represents a partial year, including epidemiological weeks 1–26 only. Percentages indicate the proportion of annual influenza detections attributable to influenza A and influenza B. Annual influenza A and B totals were derived from the primary surveillance dataset. Subtype and lineage counts were obtained from a separate subtyping dataset and are available only for a subset of influenza-positive specimens. Because these datasets differ in sampling frame and coverage, subtype and lineage counts are presented for descriptive purposes and are not expected to sum to total influenza A or B counts Annual totals of laboratory-confirmed influenza detections in Hong Kong, stratified by virus type, influenza A subtype, and influenza B lineage. Percentages indicate the proportion of total annual detections attributable to influenza A and influenza B † 2024 represents a partial year, including epidemiological weeks 1–26 only. Percentages indicate the proportion of annual influenza detections attributable to influenza A and influenza B. Annual influenza A and B totals were derived from the primary surveillance dataset. Subtype and lineage counts were obtained from a separate subtyping dataset and are available only for a subset of influenza-positive specimens. Because these datasets differ in sampling frame and coverage, subtype and lineage counts are presented for descriptive purposes and are not expected to sum to total influenza A or B counts Subtype and lineage composition of circulating influenza viruses To further characterize changes in influenza circulation beyond virus type, we examined subtype- and lineage-specific surveillance data for influenza A and influenza B across the study period. These analyses describe temporal patterns in subtype and lineage composition and complement the virus-level findings reported above. Influenza A subtype dynamics Before 2020, influenza A activity showed alternating predominance of the A(H1) and A(H3) subtypes across seasons. In several pre-pandemic years, both subtypes contributed substantially to overall influenza A circulation, although the dominant subtype varied between seasons. Influenza A activity remained very low during 2020 and 2021. After circulation resumed, subtype composition varied across weeks and seasons. A(H3) was more common during the earlier phase of re-emergence, whereas A(H1) became more prominent during later periods. Relative contributions of the two subtypes continued to fluctuate through 2023 and 2024 without a consistently dominant subtype (Fig. 4 and Figure S2). Influenza B lineage dynamics Before 2020, influenza B circulation included both the B/Victoria and B/Yamagata lineages, with variation in predominance between seasons. During several pre-pandemic years, both lineages contributed substantially to overall influenza B activity, although their relative frequencies differed across years. Influenza B detections declined sharply after 2020, and lineage diversity became more limited during the post-pandemic period. Most characterized influenza B viruses identified after 2020 belonged to the B/Victoria lineage. No B/Yamagata detections were identified in the available subtype data after 2020, including during the increase in influenza activity observed in 2023 and 2024 (Table 2; Fig. 4). Subtype and lineage analyses were based on a subset of characterized specimens and may not fully represent all circulating viruses. Model fit and agreement between observed and fitted trends Model-fitted trends were broadly consistent with the observed surveillance data (Fig. 7). In particular, the models reflected the marked decline in influenza activity during early 2020 and the slower recovery of influenza B compared with influenza A. Discussion Analysis of weekly influenza surveillance data from Hong Kong showed clear differences in post-2020 circulation patterns between influenza A and influenza B. Both virus types declined sharply during early 2020, similar to reports from other regions during periods of COVID-19-related public health measures [7, 20]. However, recovery patterns differed during subsequent years. Influenza A activity increased again during the later post-pandemic period, with seasonal peaks reappearing after 2022. In contrast, influenza B circulation remained low for a longer period and recovered more slowly, with only limited detections during much of the study period. These differences became easier to observe as surveillance data accumulated through 2024, extending beyond the early pandemic period examined in many previous reports. Both influenza A and influenza B declined sharply during early 2020, although the reduction was greater for influenza B. Influenza B activity approached near absence during the early pandemic period, whereas influenza A remained detectable at low levels. This difference was also reflected in the ITS estimates (Fig. 7) and continued beyond the initial disruption period. One possible explanation is that influenza B circulation was more strongly affected by changes in contact patterns and mobility during the pandemic. Influenza B has a more limited host range and lower evolutionary diversity than influenza A, which may have reduced its ability to re-establish circulation after the disruption. Similar post-2020 patterns have also been reported in other settings [2, 4]. The prolonged period of very low influenza B activity suggests that suppression continued for several years rather than representing only a short-term interruption in transmission [21]. The differing recovery patterns observed after 2020 were likely influenced by several biological and epidemiological factors rather than a single explanation. Influenza A viruses have greater genetic diversity and circulate in multiple host species, whereas influenza B circulation is largely restricted to humans. These differences may partly explain why influenza B activity remained low for a longer period after the initial pandemic disruption. Reduced travel, changes in contact patterns, and prolonged periods of low circulation during 2020–2021 may also have limited opportunities for influenza B transmission and re-establishment. Similar explanations have been proposed in previous studies [2, 22, 23]. However, the present analysis was based on surveillance trends and cannot directly evaluate the underlying mechanisms responsible for these patterns. In contrast to influenza A, influenza B circulation remained much more limited after 2020. Because influenza B circulates almost entirely in humans and has lower genetic diversity, prolonged suppression during 2020 and 2021 may have reduced opportunities for transmission and re-establishment once public health restrictions were relaxed. Similar explanations have been proposed in previous studies [24, 25], although the present analysis cannot directly determine the mechanisms underlying these patterns. Subtype and lineage data also supported differences in post-2020 recovery between the two virus types. For influenza A, changes in circulation after 2022 were accompanied by variation in subtype predominance across seasons rather than uniform recovery of all strains. Periods with greater A(H3) activity were followed by increased contributions from A(H1), indicating continued fluctuation in subtype composition during the later post-pandemic period. Similar subtype-specific differences in recovery have been described following relaxation of COVID-19 control measures in other settings [7]. The pattern observed for influenza B differed from that of influenza A. After 2020, influenza B detections remained limited and most characterized viruses belonged to the B/Victoria lineage, whereas B/Yamagata was not identified during the study period. This differed from the pre-pandemic years, when both lineages contributed to seasonal influenza activity. The reasons for the absence of B/Yamagata cannot be determined from the available surveillance data. Reduced international circulation, limited reintroduction, or differences in lineage fitness may have contributed to this pattern [26, 27]. However, lineage characterization was available only for a subset of specimens, and overall influenza B activity remained low during much of the post-2020 period. Consequently, the absence of detected B/Yamagata viruses should be interpreted as non-detection within the surveillance system rather than confirmation of complete disappearance. Changes in seasonal timing also suggest that influenza circulation did not return to its pre-pandemic pattern immediately after 2020. Before the pandemic, influenza activity in Hong Kong generally showed consistent seasonal peaks within a relatively narrow period each year. During 2020 and 2021, this pattern was largely absent because influenza detections remained very low across most weeks. Seasonal activity became more evident again during later post-pandemic years, although peaks were broader and less sharply defined than those observed before 2020. These findings suggest that post-pandemic recovery involved changes not only in the level of influenza activity but also in the timing and distribution of circulation across seasons. Patterns for ORV were also examined to determine whether the changes observed for influenza reflected broader disruptions in respiratory virus activity. These viruses were analyzed as a combined group to provide general epidemiological context rather than detailed comparison of individual pathogens. Unlike influenza A and influenza B, ORV did not show the same marked decline during early 2020. Although circulation patterns changed during the pandemic period, detections continued throughout much of the study period and varied across weeks and seasons. These findings suggest that the prolonged suppression observed for influenza, particularly influenza B, was unlikely to be explained only by reduced testing or surveillance disruption. The reasons for these differences cannot be determined directly from the present analysis. However, influenza viruses may have responded differently to changes in travel, contact patterns, and population mixing during the pandemic period compared with ORV. The prolonged reduction in influenza B activity may also indicate greater sensitivity to disruptions in seasonal transmission and reintroduction. These findings have implications for post-pandemic influenza surveillance. Influenza A and influenza B showed different recovery patterns after 2020, suggesting that analysis of overall influenza activity alone may overlook important differences between virus types. The prolonged reduction in influenza B circulation may also have implications for future epidemic patterns. If population immunity declined during the years of minimal circulation, future influenza B outbreaks could differ from those observed before 2020 in terms of timing or magnitude [28]. Reduced subtype and lineage diversity during the post-pandemic period further suggests that influenza activity may continue to change over coming seasons. These findings highlight the importance of continued surveillance at the level of virus type, subtype, and lineage [29, 30]. Such monitoring may help identify future changes in influenza circulation and improve understanding of post-pandemic recovery patterns. The subtype and lineage changes observed after 2020 should also be interpreted within the broader context of the COVID-19 pandemic and associated public health measures. The present data cannot separate the individual effects of travel restrictions, changes in population mobility, non-pharmaceutical interventions, healthcare-seeking behavior, or testing practices. Prolonged disruption of influenza circulation during this period may nevertheless have influenced seasonal transmission patterns and the relative predominance of circulating strains. However, these observations do not establish a direct causal effect of SARS-CoV-2 transmission itself. Several limitations should be considered when interpreting these findings. First, the surveillance data were observational and may have been influenced by changes in healthcare-seeking behavior, testing practices, and laboratory capacity during the COVID-19 pandemic. Although analyses adjusted for testing volume, residual bias related to testing patterns cannot be excluded. Second, the interruption point at epidemiological week 1 of 2020 was used to represent the beginning of major pandemic-related disruption in Hong Kong. Behavioral changes, travel reductions, and public health interventions developed over several weeks during early 2020, and the exact timing of the immediate interruption effect therefore remains uncertain. However, sensitivity analyses using alternative interruption dates produced broadly similar findings. Third, subtype and lineage characterization was available only for a subset of influenza-positive specimens. Interpretation of subtype- and lineage-specific trends should therefore be cautious, particularly during periods of very low influenza circulation after 2020 when the number of characterized specimens was limited. The absence of detected B/Yamagata viruses in the present surveillance data should not be interpreted as definitive evidence of global disappearance, although similar patterns have been reported in international surveillance studies. In addition, the present analysis could not separate the effects of SARS-CoV-2 transmission from those of public health interventions, travel restrictions, changes in population mobility, or healthcare-seeking behavior during the pandemic period. The subtype and lineage patterns observed after 2020 should therefore be interpreted within the broader context of pandemic-related disruption. Fourth, ORV were analyzed as a combined category for general comparison with influenza trends. Because this group included viruses with differing transmission patterns and seasonality, these analyses were intended mainly as descriptive context rather than direct comparison between pathogens. Finally, the analyses were conducted using aggregate surveillance data and did not include age-specific or setting-specific information. Recovery patterns may also have been nonlinear during the post-pandemic period, whereas the primary ITS models assumed linear post-2020 trends. Conclusion Analysis of weekly influenza surveillance data from Hong Kong showed clear differences in post-2020 recovery patterns between influenza A and influenza B. Although both virus types declined sharply during early 2020, influenza A activity gradually increased again during later years, whereas influenza B circulation remained limited and did not clearly return to its previous seasonal pattern. These findings suggest that influenza circulation after the COVID-19 pandemic did not recover uniformly across virus types. Differences in subtype and lineage patterns also indicate that post-pandemic influenza activity differed from that observed before 2020. Continued surveillance at the level of virus type, subtype, and lineage will remain important for monitoring future changes in influenza circulation and seasonal activity. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1