Incidence, circulation, and spatiotemporal analysis of seasonal influenza in Shandong, China, 2008–2019: A retrospective study

Abstract Background Understanding the influenza‐like illness (ILI) incidence, circulation pattern of virus strains and spatiotemporal pattern of influenza transmission are important for designing control interventions. Based on the 10 years' surveillance data, we aimed to provide a baseline characterization and the epidemiology and dynamics of influenza virus in Shandong. Methods We extracted surveillance and laboratory testing data. We estimated the ILI incidence and analyzed the predominant virus. A wavelet power analysis was used to illustrate the periodicity. In addition, we applied a linear regression model to characterize the correlation of influenza seasonality with longitude. Results The average ILI incidence was estimated to be 3744.79 per 1 million (95% confidence interval [CI]: 2558.09–4931.45) during 2009–2018. Influenza A/H1N1 and A/H3N2 strains predominated in the most influenza seasons in Shandong. The annual amplitude of influenza epidemics decreased with longitude (P < 0.05). In contrast, the epidemic peak of influenza emerged earlier in the western region and increased with longitude in influenza A (P < 0.05). The annual peak of the influenza B epidemic lagged a median of 4.2 weeks compared with that of influenza A. Conclusions The development or modification of seasonal influenza vaccination strategies requires the recognition that the incidence is higher in preschool‐ and school‐aged children. Although seasonal influenza circulates annually in Shandong, the predominant virus strain circulation pattern is extremely unpredictable and strengthening surveillance for the predominant virus strain is necessary. Lower longitude inland regions need to take nonpharmaceutical or pharmaceutical interventions in advance during influenza high‐occurrence seasons.


| INTRODUCTION
Influenza A virus and influenza B virus are two common types of influenza viruses that cause human respiratory disease. The outbreak of seasonal influenza usually occurs in autumn and winter in the Northern Hemisphere and leads to a huge disease burden, particularly in children. 1   China. 8,9 According to the guidance, the ILI case was defined as an individual with fever (≥38 C) and other nonspecific symptoms, including cough, and sore or dry throat. Influenza-confirmed cases require isolation of influenza virus by culture or detection of influenza virus by nucleic acid testing from a nasopharyngeal swab. The confirmed influenza cases were classified as for influenza A and B infection.

| Specimens collection and laboratory testing
All sentinel hospitals collected the required number of clinical specimens every week or every month. Sentinel hospitals in Shandong were required to collect 10-15 nasopharyngeal swabs every week from October to March of the next year and 10-15 nasopharyngeal swabs every month from April to September. 8 The collected specimens were immediately placed in viral transport medium (VTM) and stored at 2-8 C at local hospitals. These specimens were transferred to the closest influenza network laboratory within 48 h of collection to identify the type/subtype of influenza virus by real-time reverse transcription polymerase chain reaction (PCR) or hemagglutination inhibition after virus isolation. If the specimens could not be sent to the laboratory within 48 h of collection, they were stored at <70 C and sent to the laboratory within 1 week. Any specimens that were not identified as serotypes by RT-PCR or cytology were sent to the national influenza center for review.  were studied. The age-specific and yearly infection rates of influenza were estimated. We also categorized the 17 cities into two zones (inland zone and coastal zone) on the basis of whether the city had a coastline. The coastal zones contain 7 coastal cities, and the inland zones contain 10 inland cities ( Figure S1).

| Data analysis
A previous study used a multiplier model approach to estimate the ILI incidence rate. 12 We also used the same models to estimate the incidence rate of ILI. We first estimated the number of ILIs and then divided by the population size from the Shandong Provincial Bureau of Statistics. The total number of ILIs and the equation are described in where T ili are the year number of influenza-associated outpatient visits, the N a was an age-specific reported number of ILI consultations from ILI surveillance, the R a was the age-specific proportion of positive cases from influenza virus surveillance, and the P s was the proportion of symptomatic cases among influenza infections. The parameter P s was obtained from a review study and ranged from 58.3% to 74.5%, 13 the Q was the proportion of ILI cases among symptomatic infections, 12 the S was the success rate for sampling pharyngeal swab specimens, which ranged from 80% to 90%, 12   was the test sensitivity of detection of PCR, which ranged from 95% to 100%. 12 We used Monte Carlo simulation to estimate the incidence rate 95% confidence intervals (CIs) by the mean AE 1.96 times the standard deviation.
To analyze the periodicity of influenza cases in Shandong, we conducted wavelet analysis based on the weekly number of influenza A and B infected cases using the Morlet function. 14 In addition, we further studied the seasonal parameters following Fourier analysis. 15 The linear regression model estimates the peak timing and amplitude of the annual and semiannual periodicities of influenza activity in each city, as described in where flu i t ð Þ are the weekly standardized number of influenzapositive A or B in cities i; t is a running index for week; and and e i are the intercept and seasonality coefficients to be estimated from the data.
Specifically, the amplitude of the annual periodicity is estimated , the annual peak timing is estimated as the amplitude of the semiannual periodicity is estimated as AnnPeakTiming i ¼ Àatan c i =b i ð Þ, and the semiannual peak timing is estimated as To predict the effects of influenza seasonality in Shandong geographic regions, we fitted seasonal regression models in all 17 cities using a dependence variable for longitude. In seasonal regression models weighted by the weekly number of ILI cases, we were able to reduce the potential for systematic differences to be caused by the city's population.  Table 1). The incidence rates of ILI were inversely associated with age (  April of the next year. A single annual peak of influenza was observed in Shandong province ( Figure 1A). (58.79%), respectively ( Figure 3A). The pathogen proportion was significantly different in different age groups (P < 0.001). A/H1N1 was the predominant pathogen in all age groups (range: 33.42%-55.76%; Figure 3B). H3N2 was the other important influenza virus that infected many cases. Although influenza A predominated in the majority of the seasons considered, influenza B infection also played a significant role in seasonal epidemics.

| Local influenza dynamics
The results of wavelet analysis showed that the largest powers were at the period of 1 year, and a significant annual periodicity was observed for influenza A and B epidemics ( Figure 4A However, the semiannual peak time was also significantly related to longitude. The influenza B epidemic in inland cities reached a peak earlier than that in coastal cities ( Figure S2B,D).

| DISCUSSION
To our knowledge, this is the first comprehensive study of ILI burden and epidemiological characteristics of influenza in Shandong. Our study revealed that generally influenza A is relatively more active than influenza B, and seasonal epidemics occur between August and April of the next year. We estimated that the average incidence rate of ILI The ILI incidence rates in Shandong during 2009-2018 ranged from 785.35 to 11,841.97 per million. Globally, the ILI incidence rate was much lower than that of Guangdong and Beijing. 16,17 Our age profile of the ILI incidence rate is in agreement with reports from other studies; the incidence rate peaked at <5 years of age and declined with increasing age. 12,18 Findings from Beijing also showed that preschool-and school-aged children had the highest incidences of morbidity due to seasonal influenza. 12 Therefore, seasonal influenza vaccination strategies should be considered to target preschool-and school-aged children. Regrettably, in China, influenza vaccination is still mostly self-paid and not included in the National Immunization Program. Influenza vaccination is rarely included in the government finance-reimbursed policy in China. 19   For seasonal patterns of influenza in Shandong, we demonstrated that longitude rank is a significant predictor. The annual amplitude of influenza A and B epidemics increased with decreasing longitude, and the peak time occurred early in the western regions. We also found evidence of a West-to-East spatiotemporal epidemic peak of spread of seasonal influenza A and B across Canada, America, and Europe. [22][23][24] The special spread pattern was influenced by many factors. In one study, analysis of the spatiotemporal spread of influenza in the United States found a consistent early onset of the epidemic in California, which is the most populous state in the United States. 23 Similarly, the western regions of inland cities in Shandong are the most populated part and that may also contribute to spatiotemporal transmission. In addition, the spread may be related to climate factors, such as mountain ranges, plains, lakes, and predominant wind direction, which may drive early epidemic activity. 24 However, the complex spatiotemporal spread of influenza activity is not clearly understood. Our study has some limitations. First, we were not able to fully assess the disease burden due to lack of mortality and influenza-associated hospitalization data. Second, for any common, self-limiting illness, surveillance is not complete, and most cases go undetected because the disorder is asymptomatic, and the patient does not seek formal care or is not diagnosed and reported. Third, our study data were from a passive surveillance system of influenza, and the patients who did not visit the hospitals could not be captured.

| CONCLUSIONS
We sufficiently documented the seasonal influenza burden and the predominant virus and spatiotemporal spread of influenza transmission. We attempted to illustrate both the detection of influenza epidemic trends and the explanation of their interaction with spatiotemporal variations. We anticipate that other epidemic studies based on the combination of genetic and climate data will be a more powerful way to elucidate intriguing spatiotemporal movements in the future.

ACKNOWLEDGMENTS
We thank the doctors of surveillance hospitals for information and specimens collection.

PEER REVIEW
The peer review history for this article is available at https://publons. com/publon/10.1111/irv.12959.

DATA AVAILABILITY STATEMENT
Research data are not shared.