Background
Each year, seasonal influenza is responsible for a substantial burden of disease [
1‐
3]. With a total population in excess of 1.4 billion people, mainland China experiences a large influenza morbidity and mortality burden [
4]. Each year, it is estimated there are approximately 2.5 influenza-associated influenza-like-illness (ILI) consultations per 1000 people and between 84,200 and 92,000 influenza-associated excess respiratory deaths [
5,
6]. Despite this, influenza vaccines are not currently covered by the government-funded National Immunisation Program (NIP) for most of the population in mainland China. Hence, the cost is usually borne by the consumer, resulting in low uptake [
7]. In recent years, some wealthier local governments have offered free influenza vaccination to older adults, leading to increased coverage [
8‐
10].
Influenza seasonality varies substantially both between and within countries [
11]. In non-pandemic periods, seasonal influenza epidemics are characterised by strong winter peaks in temperate high and low-latitude regions. Conversely, sub-tropical and tropical regions broadly experience semi-annual influenza peaks or year-round activity [
11‐
13]. Mainland China spans a vast geographic area and experiences large variations in influenza seasonality. High-latitude provinces have been shown to exhibit influenza seasons similar to those of other temperate regions in the northern hemisphere, whilst mid and low-latitude provinces experience a less pronounced seasonal burden, more reflective of tropical and subtropical regions [
14‐
16].
Whilst many studies have estimated the burden of influenza-related health outcomes across mainland China, results are often reported at the annual level, neglecting the within-year variations [
5,
6]. Meanwhile, studies that examine the within-year seasonal variation of influenza are usually limited to an individual region [
17‐
20], or only report on one specific outcome e.g. influenza test positivity rates [
14]. Some studies only report on a single influenza season, in a single setting, making it difficult to contextualise both spatially or temporally [
21,
22]. Whilst all these studies provide useful elements individually, we aim to synthesise all the available information from both the English and Chinese literature to update and improve the current characterisation of the within-year seasonal variation of multiple influenza health outcomes across mainland China, at multiple spatial scales. In turn, this will provide evidence for the planning and optimisation of influenza immunisation programmes and public health strategies across mainland China.
Discussion
To our knowledge, this is the first systematic review to simultaneously examine the within-year variation of multiple influenza-related health outcomes across mainland China. Utilising a robust search strategy and a range of spatio-temporal analytical methods, we synthesised all the available data to summarise the varied seasonal characteristics experienced across the region. We provide provincial and sub-provincial estimates of monthly average influenza activity, as well as highlighting the likely onset, duration, peak, and intensity of epidemics. In line with other studies, we found seasonal patterns to vary particularly in relation to latitude and geographic location [
14‐
16,
28]. For overall influenza activity, as measured by influenza test positivity among ILI outpatients, we show high-latitude provinces are characterised by having short and intense annual winter epidemics, whilst most mid-latitude and low-latitude provinces experience semi-annual epidemics or year-round activity, respectively. By utilising the most up to date and detailed strain and subtype-specific data available from the literature, we were able to identify varied compositions and timings of influenza epidemics across mainland China, which has important implications for optimising healthcare strategies and immunisation programmes. In addition, by including the use of prefecture and county-level seasonality data in our analysis we were able to show that smaller local areas broadly experience the same general influenza seasonality patterns as their wider province.
Our most notable results are those in relation to the differences observed between seasonal characteristics of specific influenza subtypes across mainland China. In high-latitude provinces, all influenza subtype activity remains largely concentrated in the winter and spring months. This is also true for most mid and low-latitude provinces, with the exception of A/H3N2 epidemics, which likely occur during the summer months. Yu et al., (2013) [
14] previously highlighted the semi-annual nature of influenza A epidemics in mid-latitude provinces, however, without stratifying by A/H3N2 and A/H1N1pdm09 subtypes, the specific viral composition behind this seasonal characteristic was not clear. By utilising more recent and strain-specific data, we provide additional information on this phenomenon which has since been observed to broadly occur in southern China as a whole [
28], in specific provinces and cities in the region [
18], and other subtropical areas outside mainland China [
29,
30]. However, evidence from Shu et al., (2010) [
15] suggests this viral composition of summer epidemics in southern China may not always be stable and may be subject to changes in the long term, as between 2006 and 2008 summer epidemics were largely driven by A/H1N1 and B/Yamagata activity.
The geographical distinctions we make between the influenza epidemics in high, mid and low-latitude regions are broad categorisations with outliers and a gradient of seasonal patterns in between. For instance, we observe A/H3N2 activity in Beijing and Tianjin to increase slightly during the summer months, aligning with the summer epidemic in mid and low-latitude provinces, and contrasting provinces in their direct proximity. This may in part be due to the high level of travel and immigration these provinces receive, both domestically and internationally [
31,
32], resulting in travellers from higher incidence areas (e.g. Southern China and the Southern Hemisphere) potentially seeding chains of transmission outside the traditional main winter influenza epidemics. The reverse relationship can also be observed in the winter months where high positivity rates of A/H3N2 in northern provinces may be leading to a slight increase in infections in Shanghai during December and January. However, the variation in seasonality experienced by these major provinces could be also attributed to several other reasons, such as differences in population age-structure and population density [
26,
33].
The disparity in the timing of A/H3N2 epidemics across mainland China may have substantial implications for influenza immunisation programmes. In general, vaccinations in the northern hemisphere are administered throughout the autumn, prior to the winter epidemic; however, this style of roll-out in parts of mainland China may be less effective at reducing incidence in A/H3N2 infections throughout summer epidemics due to waning within-season vaccine effectiveness [
34,
35]. Addressing this regional epidemiological nuance may be challenging as repeat revaccination has been shown to be ineffective at enhancing immune response [
36,
37]. We further provide evidence that the onset of influenza epidemics tends to be staggered across mainland China. Epidemics occur first in parts of the North of the county and then several months later in the Southeast. Thus, regional-based approaches to vaccination timing may optimise levels of vaccine effectiveness, however, this should be balanced with the practicalities of vaccinating large amounts of the population and the risk of an early influenza season. With the length and intensity of epidemics also varying across the county, regions that experience short epidemics may need to be more adaptable for potential surge capacity compared to regions with a year-round burden of disease.
This study suffers from a number of limitations. Firstly, to enable direct comparisons between the influenza-related seasonal profiles of different administrative regions, we calculated the MMR based on data from multiple seasons. Whilst this method successfully captures the mean seasonal trend across many years, and likely highlights months with regularly increased activity, it may, however, obscure seasonality patterns in settings where activity is variable from year to year. This could result in a situation where the MMR is diluted across more months, potentially giving the impression that a region experiences more prolonged epidemics than it does in reality. However, by using the seasonal decomposition analysis and epidemic intensity to interpret the MMRs, we are able to identify provinces with greater variability in their seasonality patterns. In addition, by separating the MMR aggregation of pre and post-pandemic seasons, and excluding the pandemic period itself (Additional file
1: Fig. S9), we were able to compare if there were any marked differences in the seasonal characteristics. Overall, we observed very similar patterns, suggesting that all-strain influenza test positivity rates among ILI outpatients were not substantially variable between seasons, or, not changing in any particular direction over time. The main difference between pre and post-pandemic seasons we could find was the higher test positivity rates at the peak of epidemics post-pandemic. However, this was not likely a result of any substantial changes in the epidemiology and intensity of influenza epidemics, but rather, the result of improved testing and better ascertainment of cases with the vast expansion of the influenza surveillance network and more widespread use of real-time reverse transcription PCR tests to identify subtypes or lineages [
5,
38].
Secondly, as we conducted all of our analysis at the monthly level, our results may suffer from a lack of temporal granularity, as shorter-term changes in activity may be concealed. To standardise the data between studies we aggregated weekly data up to the monthly level, rather than disaggregating down, as this would require making various assumptions about data interpolation. Moreover, a high level of temporal resolution may not be required for the intended purposes of this review. We aimed to provide insights to better inform the planning and optimization of future vaccination programmes and healthcare provision. Therefore, any additional short term fluctuations detected from conducting the analysis at a smaller temporal resolution would likely be the result of noise from the multi-year aggregation of seasonal data, rather than any meaningful seasonal characteristic not captured at the monthly level.
Thirdly, the majority of our analysis relies upon utilising influenza test-positive data among ILI outpatients. This is because the current surveillance systems which record this indicator are well established and widely available across mainland China, and in turn, this data is frequently utilised for epidemiological research. Although this extensive surveillance can provide great insight into the general seasonal characteristics of influenza activity across the region, it does not directly reveal anything about the within-year seasonal variation of disease severity or absolute levels of burden, as it is a function of testing intensity. Whilst we have shown influenza test-positive rates among ILI outpatients to be positively associated with test positivity among SARI inpatients and influenza-associated excess mortality rates in certain settings, this relationship may not be uniform across all of mainland China, as it is only based on a small sample size where both outcomes were available in the same region. Previous studies have estimated the variations in influenza-related mortality across China [
5]; however, results are reported at an annual level and do not examine the within-year variation. This information may be particularly useful for planning healthcare provision and anticipating surge capacity at the peak of epidemics. Greater surveillance of these measures is therefore required to provide further insight into the within-year seasonal disparities experienced across China.
Fourth, due to many studies included in this review not disaggregating influenza B outcomes by B/Yamagata and B/Victoria lineages, we decided to conduct our seasonal analysis for influenza B as a whole. This reduced our ability to distinguish any seasonal differences between these lineages. However, doing so would be based on sparse data and would not be geographically representative of mainland China. Previous studies suggest B/Yamagata and B/Victoria annually alternate in dominance of total influenza B activity [
18]. This observation may have important implications for the planning of vaccination strategies and the decision to roll out quadrivalent or trivalent vaccines to prevent mismatches with circulation [
39]. This observation may have important implications for the planning of vaccination strategies and the decision to roll out quadrivalent or trivalent vaccines to prevent mismatches with circulation. However, since the emergence of SARS-CoV-2 in late 2019, a broad range of interventions have been enacted by governments worldwide to reduce its transmission. As a result, this has led to a substantial reduction in the transmission of many other respiratory viruses, with very few cases of influenza B/Yamagata being reported globally in 2020 [
40]. Current evidence suggests the lineage may have been eliminated, as there are no known sustained animal reservoirs, or evidence of animal to human transmission [
41]. Further surveillance over the coming years is required to better understand this phenomenon, as it may have considerable implications in deciding future components of influenza vaccines.
Finally, our use of digitised data in the analysis may be subject to some slight inaccuracies. However, we believe that acquiring a much larger and richer dataset through the use of this method than otherwise would have been possible by just relying on traditional data requests, or just utalising the raw data provided by some studies, substantially outweighs the potential of slight inaccuracies in the data. With this extraction method, we were able to include data encompassing a much larger proportion of mainland China, at multiple spatial scales, for multiple influenza-associated health outcomes, over a greater time horizon, allowing for a more complete synthesis and review of regional-based within-year variations of influenza activity across mainland China.
Future work should utilise the data collected and reviewed here to inform influenza-associated burden of disease estimates at multiple spatial and temporal scales across mainland China. Further, vaccine impact analyses should account for these spatio-temporal variations to inform future vaccination policy decisions.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.