Burden of seasonal influenza in the Swiss adult population during the 2016/2017–2018/2019 influenza seasons

Evidence on the burden of seasonal influenza in Switzerland is scarce, yet it is critical for the design of effective prevention and control measures. The objective of this study was to assess influenza‐related resource utilization, health care expenditures and quality‐adjusted life‐years (QALYs) lost in Switzerland across the 2016/2017–2018/2019 influenza seasons.

including but not limited to cardiovascular, renal, and neurological complications and morbidity. 2though many influenza disease burden studies have been conducted worldwide, [3][4][5][6][7][8][9][10] evidence on the burden of seasonal influenza in Switzerland is scarce and limited in scope.Muery et al. 11  Mombelli et al. 13 assessed the epidemiology and outcomes of influenza in a Swiss nationwide cohort of transplant recipients during eight consecutive influenza seasons.The latest estimate of the direct medical costs of influenza in Switzerland dates back to 2003 and heavily relies on expert input and a global burden of illness model instead of local data. 14A recent study assessed the lost production due to influenza in Switzerland but did not estimate direct medical costs. 15To our knowledge, no study has used local real-world data and combined multiple data sources to describe the burden of influenza in Switzerland among the overall adult population.
Comprehensive, locally representative, and accurate data are key to informing policy and public health decisions for the treatment and prevention of influenza.To improve the understanding of influenza burden in Switzerland, we conducted a retrospective analysis of real-world data to describe the burden of influenza among Swiss adult residents over three influenza seasons (2016/2017-2018/2019), including epidemiological outcomes, health outcomes, healthcare resource utilization, quality-adjusted life-years (QALYs) lost due to influenza, and associated direct medical costs.

| METHODS
We retrospectively analyzed multiple real-world data sources to calculate epidemiological and health outcomes, QALYs lost, and direct medical costs due to influenza in the Swiss adult population (overall and stratified by age: 18-49, 50-64, and 65+ years) over three influenza seasons: 2016/2017, 2017/2018, and 2018/2019.Due to the nature of this study, no approval from an ethics committee was needed.A season was defined as the period between Week 26 of the first year and Week 25 of the second year.First, we estimated epidemiological outcomes, health outcomes, and healthcare utilization for the treatment of influenza (see Section 2.2).Second, we estimated life-years (LYs) lost at premature death due to influenza and QALYs lost at premature death due to influenza and during influenza episodes (see Section 2.3).Third, we estimated the unit costs of medical services used and total direct medical costs.We assessed the uncertainty of the results using probabilistic sensitivity analysis (see the Probabilistic sensitivity analysis section in Appendix S1 for details) and scenario analysis (see the Scenario analysis section in Appendix S1 for details).
The results of the probabilistic sensitivity analysis are reported as expected values and 95% credible intervals (CIs).

| Data sources
We obtained data on the permanent resident population by age and life tables from the Swiss Federal Statistical Office. 16Data on weekly incidence rates of general practitioner (GP) visits for influenza-like illness (ILI) as well as weekly virology data of cases that were analyzed at the National Reference Laboratory for Influenza in Geneva were obtained from the Swiss Sentinel monitoring system. 17The system includes information from 150-250 volunteer GPs and pediatricians who may change every season.Data on risk factors and vaccination rates were obtained from the Swiss Health Survey 18 and the Statistics of Sociomedical Institutions. 19We obtained data on hospitalizations with a main diagnosis of influenza from the Medical Statistics of Hospitals. 20The Medical Statistics of Hospitals is a complete registry of all inpatient stays in acute care, rehabilitation, and psychiatric hospitals in Switzerland.Data on the cost of hospitalizations were obtained from the SwissDRG batchgrouper. 21Weekly data on the total number of deaths were obtained from the Causes of Death Statistics. 22Daily temperature measurements used in the statistical analysis of influenza-related excess deaths were obtained from the Swiss Federal Office of Meteorology and Climatology. 23

| Epidemiological and health outcomes and quantity of services used
We combined the estimates of total GP visits for ILI from Swiss Sentinel monitoring with Swiss case-to-visit ratios of 57.7% (2016/2017), 62.46% (2017/2018), and 63.79% (2018/2019) from Richard et al. 24 to estimate the total number of ILI cases.While these case-to-visit ratios were collected in a population that is not representative of the Swiss population, 24 we still believe that local data are more meaningful than foreign estimates.7][28] Because these proportions were only reported for all visits combined, we assumed identical proportions for all age groups.0][31] The prevalence of risk factors in the population was estimated using the prevalence of medical risk factors in the Swiss Health Survey (see Appendix S1 for detailed results on risk factors).The estimated proportion of the population at high risk of influenza complications was used in the calculation of costs of GP visits for ILI as clinical experts suggested that these visits included more services for high-risk individuals (see Tables 6 and 7 in Appendix S1).The Swiss Health Survey was also used to estimate the proportion of individuals who were vaccinated in the last 12 months.
We validated these self-reported results by previous Swiss estimates of vaccination uptake.The total number of hospitalizations with influenza was estimated using the Medical Statistics of Hospitals. 20We identified cases of seasonal influenza by ICD-10-GM main diagnoses J10 and J11.The ICD-10-GM is the German modification of the ICD-10 classification. 32We excluded cases with ICD-10-GM main diagnosis J09 and/or secondary diagnoses U6920 or U6921, as these identify zoonotic or pandemic influenza cases in the ICD-10-GM classification.*We further used the Medical Statistics of Hospitals to identify the proportion of acute care hospitalizations with inpatient rehabilitation, follow-up GP visits, in-hospital death, emergency department (ED) admission, intensive-care unit stay, and respiratory, cardiac, or cardiorespiratory complications.Total influenza-related excess mortality was estimated using a negative-binomial regression analysis of weekly all-cause mortality from the Causes of Death Statistics. 22Explanatory variables included weekly numbers of GP visits for ILI as a proxy for influenza activity, weekly temperature data, and lagged all-cause mortality.The independent variable of interest was GP visits for ILI.The dependent variable was weekly all-cause mortality.The model was specified and fitted according to prespecified criteria and statistical tests.Specifically, we ran a modified Park test and modified Hosmer-Lemeshow tests and specified the right-hand side based on the Akaike information criterion (see Appendix S1 for a detailed description of the methods used).The number of excess deaths was estimated as the difference between model predictions with observed GP visits for ILI and model predictions with GP visits for ILI set to zero.A sieve bootstrap 33 was used to assess the uncertainty of influenza-related excess mortality.

| LYs lost and QALYs lost
We used life tables from the Swiss Federal Statistical Office to estimate age-specific life expectancies at premature death due to influenza based on the area under the curve of extrapolated, half-cycle corrected overall survival curves over years since premature death. 34e average life expectancy was calculated as the average of agespecific life expectancies weighted by the observed numbers of deaths at each age in the general population.This approach assumes that the relative increase in the probability of death due to influenza is constant across age and that the absolute risk of dying from influenza increases with age.The estimation of QALYs lost followed the same approach as total LYs lost, but each annual survival probability was weighted by the EQ-5D utilities in the Swiss population, which we obtained from Perneger et al. 35 Future QALYs were discounted at 3% per annum.We estimated QALYs lost during an influenza episode for healthy adults and elderly and high-risk individuals based on data from Turner et al. 36 who pooled multiple randomized controlled trials comparing oseltamivir with a placebo.The average utility loss was calculated as the cumulative difference between daily quality of life and the last value at Day 21 as a proxy for pre-influenza quality of life as no pre-influenza quality of life was available from the pooled randomized controlled trials.This approach ignored long-term decreases in quality of life after influenza episodes.Total LYs and QALYs lost because of premature death due to influenza were obtained by multiplying the estimated number of excess deaths by the average amount of LYs and QALYs lost in cases of premature death.The total amount of QALYs lost during influenza episodes was computed as the product of influenza episodes and average QALYs lost during an influenza episode.

| Direct medical costs
All costs were computed in Swiss francs at 2022 prices and tariff rates and converted to euros using an exchange rate of 0.9767 euro per Swiss franc. 37We estimated the unit cost of GP visits for ILI based on two sample invoices that we created based on expert input using the Swiss fee-for-service outpatient tariff system TARMED. 38As the TARMED tariff system has special rates for examinations of elderly and comorbid patients, one invoice was compiled for patients younger than 75 years of age who are not at high risk of influenza complications and one for patients over 75 years of age or patients of any age who are at high risk of influenza complications (the sample invoices can be found in Appendix S1).The cost of a follow-up GP visit for ILI after hospitalization was assumed to be identical to the cost of an initial GP visit for a low-risk patient.We multiplied the population in each age group by incidence rates of GP visits for ILI, the proportion of positive sample from GP visits for ILI, the proportion of the population at high risk, and the cost of the appropriate sample invoice to obtain the total direct costs of GP visits for influenza.
The price of antiviral medication during a GP visit for ILI was set to the oseltamivir sales price of EUR 55. 34 for ten 75 mg capsules. 39Oseltamivir had not been on the Swiss list of pharmaceutical specialties during the observation period of our study.This price is very similar to the last list price of Tamiflu ® (EUR 54.37) before it was taken off the list of pharmaceutical specialties on March 1, 2010.We multiplied the estimated number of GP visits for ILI by the proportion of GP visits for ILI with an antiviral prescription and the price of antiviral drug medication to obtain the total direct medical costs of antiviral prescriptions.We assumed that all patients who received antiviral medication had influenza and not just ILI.
The cost of hospitalizations with a main diagnosis of influenza was estimated based on the SwissDRG per-case payment system as the product of a case-specific cost weight and the treating hospital's base rate.Case-specific cost weights were obtained by uploading the dataset of cases with a main diagnosis of influenza to the SwissDRG batchgrouper. 21The base rates were taken from the canton of Zurich in 2022. 40We included the costs for inpatient rehabilitation, calculated as the product of length of stay in inpatient rehabilitation with an influenza main diagnosis and an average per diem rate in the Canton of Zurich in 2022. 40CD-10-GM diagnosis codes J09 constituted only 3.0% of all influenza main hospital diagnoses in the 2018/2019 season, and the exclusion of these cases does not lead to a large underestimation of the number of inpatient cases if J09 was also used for seasonal influenza.
The total direct medical costs were obtained by summing the estimated total costs of GP visits for influenza, antiviral prescription medications, and hospitalizations.
We further estimated an average of 3207 (se ± 380) antiviral medication prescriptions and 1355 (se ± 169) excess deaths per season.The 2016/2017 and 2018/2019 seasons were A-dominant, with 96.05% and 99.5% of all tested strains being A-type, while the 2017/2018 season was B-dominant, with 70.51% of all tested strains belonging to the B-type.Estimated vaccination rates were 6 The probabilistic sensitivity analysis showed that the uncertainty about the total amount of QALYs lost was quite high.The scenario analysis showed that the main driver of overall uncertainty was variability in the number of influenza-related excess deaths (detailed results of the scenario analysis can be found in Appendix S1).

| Direct medical costs
The total direct medical across all three seasons amounted to 168.0 (95% CI 167.8; 168.8) million euros (Figure 3).A total of 18.1 (95% CI 17.9; 18.9) million euros were spent on GP visits for influenza; 0. Most of the costs were accrued in the 65+ age group, accounting for an average of 66.34% of total direct medical costs.

| DISCUSSION
This study is the first in 20 years to estimate the burden of influenza on health care payers and patients in Switzerland and to use such a wide range of data sources. 14We obtained most input parameters from de novo analyses of Swiss real-world data sources and combined the results to obtain a comprehensive assessment of the health effects and costs of influenza in the Swiss adult population.Notably, our findings highlight the significant disease burden among older adults aged 65+ years, the group for whom rates of hospitalizations and deaths were consistently the highest across the three observed seasons.Moreover, we also found elevated disease burden among the 50-64-year-old population versus the younger adult population (<50 years of age), where rates of hospitalizations were approximately three-fold greater.Loss of QALYs and increased direct medical costs further demonstrate the impact of this increased disease burden.
Our epidemiological results are largely similar to estimates for other European countries.Paternoster et al. 41   sons, which is quite high in comparison to our estimates. 42Because the present study covers different seasons, it is not clear if these differences in excess deaths are driven by actual differences in mortality, different modeling decisions, or occur simply because of differences in timing and settings.
Total medical expenditures varied considerably across seasons and accounted for 0.04% to 0.09% of total Swiss healthcare expenditures 43 or 0.47% to 1.07% of medical expenditures for communicable diseases. 44The largest proportion of medical expenditures was observed among residents aged 65+ who exhibited much higher hospitalization rates than younger patients.
In addition to the financial burden, influenza causes a significant humanistic burden.Premature death was the main driver of QALYs lost, and most deaths occurred in individuals aged 65+.
The strengths of our study include the combination of a wide range of data sources.This leads to a multitude of outcomes that can serve as a first step in continuous monitoring of the burden of influenza in Switzerland.This seems especially relevant given the uncertainty on the epidemiology of influenza after the COVID-19 pandemic.
However, this study also has limitations.First, the study focuses on the adult Swiss resident population and does not include minors.
This focus on the adult population was prespecified based on a preliminary assessment of the available data and its external validity for the Swiss population.We decided that a study for both adult and minor populations using the same data sources and data analytic approaches would have limited the external validity, and we deemed it more robust to focus on adults and tailor the methods to this population.Second, the estimates of influenza incidence are uncertain because they depend on projected numbers of GP visits for ILI and published results from the literature.However, our results seem plausible compared with results from other countries.Third, the estimation of excess deaths related to influenza relies heavily on modeling decisions, and our estimates might not exactly capture the true values.
Nevertheless, we specified the regression model in an empirical manner based on prespecified procedures and achieved a better fit with our model than with alternative models.Fourth, the identification of influenza-related hospitalizations by influenza main diagnoses led to an underestimation of the costs of hospitalizations as cases who were hospitalized for influenza-related secondary acute events were not classified as influenza-related hospitalizations.However, the inclusion of all cases with influenza secondary diagnoses and attribution of the full hospitalization costs to influenza would have led to a considerable overestimation of the hospitalization costs, and we decided to focus on influenza main diagnoses in favor of a conservative approach.This decision only affected hospitalization incidence and costs but not influenza epidemiology which was obtained from the Sentinel monitoring system.Fifth, the calculation of LYs lost at premature death due to influenza assumed that patients who die due to influenza have the same life expectancy as members of the general population of the same age.This might lead to an overestimation of LYs lost if individuals who die due to influenza are more vulnerable than members of the general public.We counteracted this overestimation of LYs lost by invoking the assumption that influenza has a proportional effect on age-specific mortality rates.Sixth, we are only able to detect hospitalizations that are coded with the specified ICD-10-GM diagnoses.Thus, we may underestimate the true number of hospitalizations due to influenza.Seventh, the samples tested at the National Reference Laboratory for Influenza were not randomly selected and might not reflect the true proportion of GP visits for ILI that actually were influenza.

| CONCLUSION
Our results show that seasonal influenza poses substantial burden on the Swiss healthcare system and population.Policy interventions to increase vaccination rates and uptake of more effective vaccines are needed to reduce influenza burden in future seasons.Continuous monitoring of the burden of influenza for the directing of public health interventions is particularly important as well in the uncertainty of the years following the COVID-19 pandemic.
previously described clinical characteristics among a small sample (n = 60) of children and adolescents hospitalized with influenza infection in a University Children's Hospital in Switzerland during the 2002/2003 and 2003/2004 seasons.Brinkhof et al. 12 used Poisson regression modeling to estimate influenza-attributable mortality among Swiss residents in 1969-1999 among adults 60 years of age and older.

F I G U R E 2
Quality-adjusted life-years (QALYs) lost by age group and season.The error bars show ±1 standard deviation of the total QALYs lost.F I G U R E 3 Direct medical costs in euros by age group and season.The error bars show ±1 standard deviation of the total direct medical costs.deaths are generally lower than estimates from other countries.An analysis of influenza-related excess mortality in the area of Vienna (Austria) finds an average number of 140.1 excess deaths per 100,000 residents over 60 years of age in the 1999/2000-2008/2009 sea- Average per-season results across all three influenza seasons.
visits for ILI was 1975 in the 65+ age group compared with 3258 and 3036 in the 18-49 and 50-64 age groups, respectively.Hospitalizations per 100,000 residents were much higher in the 65+ age group with 241 compared with 13 and 41 in the 18-49 and the 50-64 age groups.The same is true for excess deaths, with an The number of QALYs lost due to premature death was highest in the 2017/2018 season, with 9801 (95% CI 8007; 12,783) QALYs compared with 7401 (95% CI 6071; 9630) and 6773 (95% CI 5450; 8890) in the 2016/2017 and 2018/2019 seasons, respectivelyF I G U R 1 (A) Total cases of influenza-like illness (ILI) and general practitioner (GP) visits for ILI and influenza per age group and season.(B) Total hospitalizations, excess deaths and in-hospital deaths per age group and season.Error bars show ±1 standard deviation of total cases.(Figure 2).Loss of quality of life during influenza episodes was also comparatively high in the 2017/2018 season, with a total loss of 1378 (95% CI 915; 1996) QALYs compared with 1028 (95% CI 683; 1487) and 928 (95% CI 608; 1342) QALYs in the 2016/2017 and 2018/2019 seasons, respectively.Most of the QALYs lost due to premature death occurred in the 65+ age group with an average of 5936 (95% CI 3760; 6945) QALYs lost compared with 1122 (95% CI 814; 1305) and 933 (95% CI 528; 1183) QALYs lost in the 18-49 and 50-64 age groups, respectively.
Our estimates of influenza-related excess † Hospitalization costs have a very low uncertainty because the Medical Statistic of Hospitals includes every case that was hospitalized and the SwissDRG batch grouper provides the actual cost weights that were used to reimburse the hospitals.