Background
Seasonal influenza is an acute viral respiratory infection that poses a substantial burden on individuals, populations, and health care systems every year. It can cause mild to severe illness and can present with symptoms such as fever, muscle aches, fatigue, headache, and respiratory symptoms. While most people recover from influenza illness within a few days [
1], some may develop severe complications including primary influenza virus pneumonia or secondary bacterial pneumonia [
2]. Infected children often suffer from AOM [
3,
4]. Severe non-pulmonary complications include cardiovascular and neurologic complications, which are, however, widely under-recognized since they are often not clearly linked to a previous influenza infection [
5]. People at high risk for complications include very young children, older adults, pregnant women, nursing home residents, and people with chronic medical conditions or a compromised immune system [
6].
Several vaccines are available to prevent influenza virus infections, and in many industrialized countries, high risk groups are targeted by national immunization programs or immunization recommendations [
7]. In Germany, influenza vaccination is recommended for all persons above 60 years of age, pregnant women, high-risk persons (e.g. diabetes patients) and health care professionals [
8]. The recommendation did not specify whether a tri- or tetra-valent vaccine should be used until 2018. Since then, a tetra-valent vaccine is recommended. The protection through the vaccination is also dependent on the vaccination rates. A decrease of the vaccination rates can be observed for Germany [
9]: In the 2008/09 season 47.9% of persons above 60 years of age were vaccinated while only 34.8% were vaccinated in the 2016/17 season.
In this context, estimating the influenza-related disease burden in terms of morbidity, mortality, and economic consequences is an important contribution to the decision-making process regarding policies around influenza immunization [
10]. In many countries, data on disease burden caused by seasonal influenza viruses are mainly gathered through national surveillance systems. Such systems often have a strong focus on assessing the role of influenza in acute respiratory illnesses (ARI) or influenza-like-illnesses (ILI) or report the proportion of cases with laboratory-confirmed influenza (LCI). In addition, using the example of the United States, combining routinely collected surveillance data with results of outbreak investigations and health care surveys allows for estimating symptomatic community illnesses, outpatient medical visits, hospitalizations, and excess deaths related to seasonal influenza [
11]. The estimation of some of these outcomes, particularly influenza-associated hospitalization rates and excess deaths, commonly involves the use of statistical models [
12]. The focus of the German surveillance system also includes estimated excess rates for influenza-associated outpatient medical visits, hospitalizations, cases of absenteeism, and deaths [
13]. However, these reports do not include estimates for complications such as AOM or CAP and provide therefore rather partial insights in the burden of seasonal influenza. Furthermore, the German surveillance reports provide no information on the related direct and indirect cost of influenza. To close this evidence gap the aim of this study is to estimate the epidemiology of influenza and associated costs using claims data of a German sickness fund.
Methods
Data
Claims data from a large German statutory sickness fund (2012: 8,038,608 insured, 2013: 8,412,199, 2014: 8,849,736) covering a 3-year period (January 2012 to December 2014) was used to conduct the analysis. During this period, influenza strains were evenly distributed among the samples taken for the virological surveillance of the National Reference Center for Influenza for the 2012/13 season (A (H3N2): 31%; A (H1N1): 34; B-lineage: 35%). In the two following seasons, A (H3N2) was responsible for most cases (61 and 62%, respectively) with A (H1N1) being found in 30 and 15% and the two strains from the B-lineage in 9 and 23% of the cases. Yamagata has been the dominant B-strain in all seasons and has also been included in the trivalent vaccine [
14]. The claims data set covers a broad range of medical and economic information from insured such as diagnoses, resource use and cost data, but primarily serves sickness funds for accounting purposes. Statutory sickness funds are covering 90% of the German population and cover the expenditure for health services from inpatient and outpatient sector as well as pharmaceuticals, remedies and aids and other services.
Due to the nature of the dataset, no laboratory-confirmed definition of influenza cases could be used. Instead, an influenza or ILI case was defined via ICD-10 codes J09 to J11 which were either documented in the dataset as a main inpatient diagnosis or a “secured” or “suspected” outpatient diagnosis. Influenza incidence estimates were based on all insured, while complication rates, costs, and work days lost were analysed via a matched case-control design. A sample of 100,000 persons being representative by age group, sex and insurance status for all insured diagnosed with influenza were drawn from the total population of insured, and controls without influenza diagnosis were matched one-to-one with replacement by 27 age groups, sex, category of insured persons (e.g. unemployed, employed, student), and outpatient pharmaceutical costs of the previous year with a caliper of ±10% as a proxy for comorbidity. Potential complications of influenza were identified using the ICD codes H65, H66, and H67 for ear infections or AOM and J10.0, J11.0, and J12 to J18 for CAP.
Analysis
The calculation of the incidence was based on all insured of the sickness fund. All other analyses were based on the data of those insured for which a match could be identified. As outpatient diagnosis information in German claims data are usually documented on a quarterly base, the exact day of the influenza diagnosis cannot be determined. Therefore, the quarter of the initial influenza diagnosis (index quarter) and the following quarter were used to compare the documented diagnoses, resource use, and costs of the influenza patients and their matched controls. This means that the resource use and costs are calculated for half a year. To minimize the influence of outliers and to ensure meaningful results for rare events (e.g. influenza-related hospitalizations), the initial 27 age groups that were used for matching purposes were merged into six groups (“0 to 1”, “2 to 5”, “6 to 9”, “10 to 17”, “18 to 34”, “35 to 59”, “60 plus”) for the actual analysis. The choice of the age groups was based on to the different levels of the German educational system.
For the calculation of complication rates, the number of persons with at least one diagnosis of AOM or CAP during the index quarter and the following quarter was counted. Age and year-specific complications rates were calculated for both the influenza group and the control group. Influenza-attributable complication rates were then derived from the difference between the groups.
The calculation of influenza-attributable resource use and costs differed by health care sector. For inpatient treatment and sick leave, the corresponding principal diagnosis was identified directly from the claims data, i.e. ICD codes were available for each item of resource use on a daily base. Hence, there was no need of using the control group data for inpatient costs and sick leave. Inpatient costs could directly be determined by the payments of the sickness fund to the hospitals. For the indirect costs due to sick leave, the number of days absent from work was retrieved from the claims data. As income was not included in the dataset, the age-specific, average daily income of an employee (including part-time employees) [
15] was used and multiplied with the number of days of the sick leave. Indirect costs were not calculated for the three youngest age groups and only for employed persons in the other age groups.
For outpatient physician consultations and prescribed pharmaceuticals, a direct link between diagnosis information and resource use or cost data was not possible. Therefore, respective influenza-attributable resource use had to be calculated via the matched case-control approach, i.e. excess resource use and costs were estimated by calculating the difference between the influenza group and the control group for the index quarter and the following quarter. The analyses of pharmaceuticals were further restricted to a list of relevant ATC codes for the treatment of influenza and related complications, which was adopted from Ehlken et al. [
16]. The relevant ATC-codes are listed in Table
1.
Table 1
ATC-codes related to the treatment of influenza and included in the analyses
Analgesics and antipyretics | N02B |
Anti-inflammatory and anti-rheumatic products, non-steroids | M01A |
Nasal decongestants | R01AA, R01AB |
Throat preparations | R02A |
Cough and cold preparations | R05 |
Antibiotics | J01A, J01C, J01D, J01F, J01 M |
Antivirals | J05AH |
The analyses were stratified by age groups to reflect the different risk-groups and by year to account for varying influenza activity in different seasons. Detailed results for years and age groups can also be found in a table in the Additional file
1.
Statistics
Cost data from 2012 and 2013 were adjusted to the price level of the reference year 2014. The health-specific consumer price index of the German federal statistical office was used to adjust prices between 2012 and 2013 (− 3.7%) and between 2013 and 2014 (2.0%). Differences between the influenza group and the control group were tested using the two-sample, two-sided t-test for samples with unequal variances. Tests across k > 2 groups (e.g. comparison across years) were corrected for multiple testing by adjusting the alpha level by alpha/k according to the Bonferroni correction [
17]. Analyses were performed with R (Version 3.4.1) and Microsoft Excel (Version 2013). We followed the German guidelines on routine data analysis [
18].
Discussion
Our results confirm well-established differences in the incidence of seasonal influenza by season and by age groups. Furthermore, our analyses show that complication rates for influenza-attributable AOM and CAP are highly age-dependent. We found no significant differences of the complication rates between the three seasons included in this analysis. We were also able to show that the occurrence of both complications in the same patient is significantly more frequent than it could be expected from the marginal probabilities of the complications. The youngest and oldest age groups carried the highest burden of influenza measured by incidence and complication rates. Regarding the economic burden, the presented results suggest no strong cost differences per case between the years. Analogously to the disease burden, the youngest and oldest age group cause the highest mean costs among all age groups. Costs for outpatient care and pharmaceuticals are relatively similar for all age groups. Cost differences result from higher costs of inpatient care for influenza itself and influenza-related complications, respectively.
With regard to the incidence estimates, our results are comparable to the findings of the German Influenza surveillance at the Robert Koch-Institute. The four reports for the seasons 2011/12, 2012/13, 2013/14 and 2014/15 [
19‐
22] estimate the number of influenza-related excess consultations per 100,000 inhabitants also to be highest among the youngest age-group and to be the lowest in the age-group of people above 60 years of age. Our findings are also in line with the severity of the different seasons, i.e. the reports show the highest number of excess consultation in the year 2013 and a fairly equal number in the years 2012 and 2014.
The results presented on resource use and costs in this study can only to some extent be compared to previous studies for Germany. The only two recent studies on influenza-associated disease burden and costs that were identified for Germany differed in their scope and the applied methods [
16,
23]. While Ehlken et al. did not calculate incidence rates, our incidence estimates of 0.93% in 2012 and 1.73% in 2013 are comparable to the estimate of 1.7% estimated by Haas for the 2012/13 season. The complication rates for AOM and CAP calculated in our study show a similar trend across the age groups as compared to the figures presented by Haas et al. but each with a higher level. For example, Haas et al. estimated the influenza-attributable CAP to be 2.3% compared to 4.28% in our study. Ehlken et al. reported a complication rate of 1.6% for CAP in children (5.86% in our study) and 5.4% in adults (3.82%). The difference between the studies may be explained by relatively low numbers of CAP cases in all three studies. For AOM, Ehlken et al. found higher rates for children (13.4% vs. 7.09% in our study) but lower rates for adults (0.9% vs 2.20%). The complication rates for AOM as found by Haas et al. correspond well across the age groups, although on a lower level. For instance, for the age group “0 to 1”, the authors found a difference of 11.0% between influenza and control group compared to 14.41% in our study. This might be caused by different definition of the complications with regard to the selected diagnoses in the inpatient and outpatient sector documented in German routine data. Differences to Ehlken et al. might also be caused by differences in season-specific influenza epidemiology, e.g. the dominant strain circulating or the match of the seasonal vaccine, as the studies cover different seasons.
Regarding the cost estimates, the study by Ehlken et al. estimated the cost of one influenza episode to be €514 for adults and €105 for children from a societal perspective and €59 and €66 from a third-party payer perspective. Unfortunately, Haas et al. did not provide total cost estimates per influenza case or episode, but their estimates on the costs for outpatient care exceed the estimates of Ehlken et al. and our study fourfold (€224 vs. €34.51 vs. €53.95, respectively). Also, the estimates of Haas et al. for the costs per inpatient influenza case exceed our estimates (€5832 vs. €2033). For the outpatient sector, this may be due to the missing control group in the study by Haas et al. The corresponding costs of just the influenza group in our study are €259.75. The difference in outpatient costs might be explained by outliers.
Our study has several limitations. Firstly, the time period covered by the data did not coincide with the start and the end of the influenza seasons. Hence, the year 2012 contains the later part of the 2011/12 season and the first part of the 2012/13 season. By the nature of claims data, our dataset only contains cases of influenza seeking medical attendance. Thus, we might underestimate the burden of the disease as symptomatic cases who do not seek medical attention are not considered in the analysis. More generally, the estimation of the burden of disease of influenza is challenging to assess, since definition and diagnostic procedures differ, information on seek of medical attendance partly lacks, and virus circulation and vaccine fit differ almost every year.
Further, no out-of-pocket payments that are not reimbursed by the German sickness funds are part of our analysis. In this context, it is important to mention that the definition of the influenza group – including only patients with a J09, J10 or J11 diagnosis – is rather conservative. In this context, the high percentage of controls with influenza-related medication might be an indicator although many of the pharmaceuticals listed in Table
1 are used for various other medical conditions, e.g. ibuprofen or other analgesics. Additionally, it is not possible to exclude biases that may arise from the population of the sickness fund which was used for the analysis in this study. However, the sickness fund covers nearly 9.98, 10.42 and 10.90% of the general German population in the years 2012, 2013 and 2014, respectively. Finally, we did not include cardiovascular or neurological complications for which the link to influenza is not firmly established. To detect those complications and estimate their economic burden, a larger dataset would be necessary. Hence, our estimates can be seen as conservative.
Conclusions
In summary, our study suggests that the economic burden of influenza on a population level corresponds to the high non-economic burden of disease of influenza found for Europe [
24]. Even if only direct costs are considered, our estimates of €78,278,429 per year are nearly as high as the estimates for herpes zoster and postherpetic neuralgia with €105 million [
25] and amount to one third of the yearly spending on prostate cancer of €244 million in Germany [
26]. Costs are especially high for persons being defined as “high risk” from a clinical perspective and among children 2 to 5 years old. Further research could focus on the cost-effectiveness of a general, seasonal vaccination in dependency of the age at administration of the vaccine.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.