Acute gastroenteritis (AGE) caused by rotavirus (RV) is the most frequent cause of serious diarrhea affecting toddlers and infants worldwide. It accounts for 111 million infections, 25 million outpatient treatments and the average annual death of 440,000 children less than 5 years of age.1 Although AGE caused by bacteria and parasites is decreasing because of improved hygienic conditions in many developed countries, the number of AGE cases because of viral infections has not shown a similar trend.2 Particularly, no significant decline in cases caused by RV gastroenteritis (RVGE) has been recorded in the past 20 years. According to estimates, more than 3.6 million episodes of RV infection occur annually resulting in nearly 700,000 outpatient visits, more than 87,000 hospitalizations and 231 deaths among 23.6 million children younger than 5 years of age in the European Union-25. It is 3 times more likely for children to be hospitalized because of RVGE compared with other pathogens known to cause AGE.2 In addition, many RV-infected patients receive only homecare. These RVGE cases (estimated between 25% and 51% of patients) are not recorded, thus, the incidence rate is significantly underestimated.3 According to a German surveillance study conducted between 2001 and 2008, the average annual incidence of RV infection in children below the age of 5 is 115 of 10,000; approximately 50% of these patients are hospitalized.4 Although RVGE is rarely fatal in developed countries such as the European Union and the US, the disease burden for society and the patient is similar to that found in the developing world.2 RVGE still challenges European healthcare systems, accounting for 56.2% of hospitalizations and 32.8% of emergency department visits caused by community-acquired AGE in children less than 5 years of age.5
Since 2004, the diagnosis-related group (DRG) system is mandatory for reimbursement of hospitalized cases by health insurance companies in Germany. Each hospital must submit relevant data, for example, main and secondary diagnoses, procedures and the invoice amount of all in-patient cases, to the Federal Statistical Office (FSO). Analysis of these data allows acquisition of nearly all hospitalized cases from German hospitals. Rehabilitation and psychiatric cases are excepted examples.
RV vaccines were licensed in Europe in 2006; since April 2009, the advisory group of experts on immunizations (World Health Organization) has recommended vaccinating all children less than 2 years of age worldwide against RVGE.6 In 2013, the official German Standing Committee on Vaccinations updated their recommendation to include an oral RV vaccine for infants.7 Until this recommendation became universal, the statutory health authorities in only 5 German federal states included RV vaccination in their local recommendations: Saxony in January 2008, Brandenburg (BB) in January 2009, Mecklenburg-Western Pomerania (MV) in July 2009, Thuringia in October 2009 and Schleswig-Holstein (SH) in March 2011. Vaccination rates have increased slowly, however, because these recommendations have not led to automatic reimbursement for the vaccine through sick funds.
In this retrospective study, FSO data for children (0 to 10 years of age) hospitalized during the years 2005 to 2010 were analyzed. This analysis benefits from the high quality of data obtained from the nationwide representative DRG system and provides demographic characteristics of RVGE disease in Germany. Moreover, it establishes a valuable baseline for assessing the impact of RV vaccination by comparing the rates of hospitalization during the prevaccination and postvaccination phases in 5 federal states that recommend early immunization.
MATERIALS AND METHODS
Data obtained from the official nationwide DRG hospital statistics, provided yearly by the FSO and intended for scientific research, were used. These data are aggregated and anonymized. With few exceptions, for example, institutions for rehabilitation or mental disorders, the use of the current DRG system is mandatory for all hospitalized cases in order to recover reimbursements from the national health insurances. Currently, about 1700 hospitals (80% of all hospitals in Germany, accounting for 97% of all discharges) are reimbursed by DRG-type hospital payment.
According to G-DRG coding rules, all patients discharged from hospitals are assigned a specific DRG based upon a grouping algorithm using the in-patient hospital discharge data set. To determine the DRG, the algorithm considers main diagnosis, procedures, secondary diagnoses, and patient characteristics (eg, age, sex and weight of newborn). Usually, the main diagnosis is responsible for the grouping in the major diagnostic category, corresponding to the first letter of the DRG code (in our study: G = diseases of the digestive organs). Thus, the major diagnostic categories correspond to a single-organ system or etiology. DRG is determined directly by the procedure code only in high cost cases (eg, special transplantations).
Because AGE in children is usually caused by viruses, our data request at the FSO was specified as follows: all cases with codes from sections A08* to A09* of the valid ICD-10 catalog, that is, virus-related and other specified intestinal infections. Transferred data, which included frequency of hospitalizations, direct costs, length of stay and in-hospital mortality, were stratified by the following variables: year; season [first quarter (Q1), second quarter (Q2), third quarter (Q3) and 4th quarter (Q4)]; sex; age (<1, 1–3 and 4–10 years); ICD-10 code (A08.0 = rotaviral enteritis, A08.1 = acute gastroenteropathy because of Norwalk agent, A08.2 = adenoviral enteritis, A08.3–A08.5 = other viral enteritis, unspecified viral intestinal infection and other specified intestinal infections, A09/A09.0–9 = diarrhea and gastroenteritis of presumed infectious origin); DRG code (G67A, G67B, G67C, G67D and G67E) and federal state [Baden-Württemberg, the Free State of Bavaria, Berlin (BE) and BB, Hesse, MV, Lower Saxony and the Free Hanseatic City of Bremen, North Rhine-Westphalia, Rhineland Palatinate, Saarland, the Free State of Saxony (SN), Saxony-Anhalt (ST), SH and the Free Hanseatic City of Hamburg (HH), the Free State of Thuringia (TH)]. Because of a paucity of data, the 3 city states of Hamburg, Bremen and BE were merged with the neighboring federal state. Information for seasonal rates was only available for the years 2009 and 2010. Information for <5 observations was deleted because of data protection regulations. Data for the number, time and sold vaccine doses of each federal state were obtained from the IMS-Pharmascope database (IMS Health GmbH & Co. OHG, Frankfurt am Main, Germany).
During the course of the entire study period, more detailed information on hospitalizations because of AGE was collected in 4 German pediatric hospitals (University Medical Center Mainz, University Medical Center Mannheim, Faculty of Medicine Dresden and Dr.-Horst-Schmidt-Clinics in Wiesbaden). To validate further the quality of DRG coding, we took a random sample (N = 1003) of the hospitalization records from Mainz and compared the appropriate ICD-10 code with results of the enzyme-linked immunosorbent assay (ELISA) tests on RVGE and AGE because of Norwalk and adenovirus enteritis, which were performed at the hospital and readily available to the staff. Changes in the codes of the current DRG and ICD catalogs over time were considered. This study was reviewed and approved by the Ethics Review Board of the University Medical Center of the Johannes Gutenberg University Mainz.
The analyses included all children in Germany between the ages of 0 and 10 years who were hospitalized with AGE during the years 2005 to 2010. The number of cases, the crude rate (CR) per 100,000 and, when possible, the age-standardized rate (AR) per 100,000 were evaluated. The CR was defined as the frequency of new cases in a period of time:
where Npts equals the number of new cases and PY equals the person-years at risk over a specific follow-up period of time. Person-years at risk were calculated as follows: PY = Npop* follow-up time; Npop represents the average size of the disease-free cohort during the time period of interest. We assumed that the population was composed of a stable dynamic cohort that underwent no major demographic shifts and had relatively few persons who became in-patient treated for the disease of interest. Therefore, the average size of Npop could be estimated from the size of the entire population based upon the FSO data available during the study period.
AR or CR was evaluated for interregional and intertemporal comparisons. A standard population was created using the 1987 census statistics of the Federal Republic of Germany. AR was defined as follows:
where ai = (ni/pyi)×1000,000 equals the age-specific rate for age class i, A equals the number of age classes, ni equals the number of cases in the ith class, pyi equals the person-years at risk in the ith class and wi equals the fraction of the standard population in the ith class. Confidence limits for the AR were calculated according to Armitage and Berry as follows:
and Zα/2 is the standard normal deviate. CR and AR of RV infections stratified by federal states were analyzed because the local health authorities recommended RV vaccination at different times.
A McNemar test was performed to evaluate the agreement between the ICD-10 codes and the results of ELISA testing for RV. Sensitivity, specificity, positive predictive value and negative predictive value were also calculated in an effort to determine how the probabilities and proportions of the ICD-10 coding compared with the ELISA tests using the detailed data obtained from the pediatric hospitals.
Poisson regression was used to estimate rate ratios (RR) and 95% confidence intervals (95% CIs) for the seasonal effects and immunization recommendations (IRs) adjusted for year and federal state. The variable status of IR, coded as 1 for children living in a state with IR and 0 for children living in a state without IR, was generated as a consequence. In addition, the elapsed time because of IR was measured in years. Children could live in a state without IR or in states with IR for 1 or 2 years. To analyze the effect of vaccinated children, a Poisson regression was performed in which each unit increase corresponded to 1000 vaccinated children. The number of vaccinated children for each federal state and year was evaluated by the number of sold vaccine doses, whereas it has been taking into account that for obtaining a full immunization 2 (Rotarix) or 3 (RotaTeq) doses of the vaccine must be administered. The calculation based also on the assumption that each sold dose has been administrated. Analyses of IR were restricted to children less than 1 year of age. All statistical analyses were performed using SAS/STAT software version 9.3 (SAS Institute, Inc., Cary, NC) and all maps were generated by EasyMap 10.0.
A total of 5,843,730 hospitalizations because of diseases that occurred in 8 million children (ages 0 to 10 years) during the period 2005 to 2010 were analyzed (data flow shown in Fig. 1). This represented a total of 48 million person-years with an AR of in-patient treatment that ranged from 12,095 in 2009 (95% CI: 12,070–12,120) to 13,195 in 2006 (95% CI: 13,169–13,221). Boys were affected slightly more often than girls: 13,823 (95% CI: 13,808–13,839) versus 11,240 (95% CI: 11,226–11,254), respectively. Hospitalizations because of AGE (520,606 cases) accounted for 8.91% of all hospitalizations recorded during the study period.
The CR of children hospitalized because of AGE of any etiology was highest in boys and girls less than 1 year of age (Table 1). CR was relatively decreased in boys and girls at ages 1 to 3 years old, and was the lowest in both boys and girls in the 4-year-old to 10-year-old age group. CR for RVGE in children less than 1 year of age was slightly higher in boys compared with girls and decreased with age. Nearly a third of the total AGE hospitalizations were because of RV infections equating to 25,440 cases per year and an AR of 302 hospitalizations per 100,000 person-years. The remaining cases were of norovirus, adenovirus, unidentified virus or nonspecific etiology (see Table, Supplemental Digital Content 1, https://links.lww.com/INF/C295). Such cases showed an age and gender distribution similar to that associated with RV infection.
The seasonal distribution of hospitalizations because of various pathogens expressed in terms of the CR per quarter for the years 2009 and 2010 is shown in Figure 2. There was an increase in hospitalizations for RVGE between Q4 of 2009 and Q2 of 2010, followed by a strong decrease in Q3. Compared with Q3 and adjusted for year, the RR and 95% CI for the seasons were 8.38 (8.04–8.74), 7.96 (7.63–8.30) and 1.69 (1.61–1.78) for quarters 1, 2 and 4, respectively. Although hospitalizations because of RV, nonspecific and other AGE fluctuated only slightly during the entire 6-year study period, hospitalizations because of norovirus and adenovirus increased (Fig. 3).
Figure A, Supplemental Digital Content 2, https://links.lww.com/INF/C296, shows the CR of RV infections per 100,000 person-years for children less than 1 year of age during the period 2005 to 2010 listed by federal state; Figure B, Supplemental Digital Content 2, https://links.lww.com/INF/C296, illustrates the number of vaccine doses sold. The number of sold doses was higher for BB combined with BE, MV, SN, TH and ST compared with the other federal states. Immunization recommendations (IRs), adjusted for federal state, and an interaction term of year and federal state, showed decreased RR for children below the age of 1 year: RR and 95% CI: 0.70 (0.65–0.76; data not shown). One thousand vaccinated children, adjusted for federal state, and an interaction term of year and federal state, exhibited a similar decrease: RR and 95% CI: 0.984 (0.977–0.991; data not shown).
In 2005, the AR of RVGE was higher in MV, SN, ST and TH compared with other federal states. This observation was true, however, for hospitalizations in general (data not shown). Figure 4 illustrates the change in AR of RV infections per 100,000 person-years in the federal states between 2005 (Fig. 4A) and 2010 (Fig. 4B). AR increased in several western federal states, but decreased in the east. After stratification by year, only SH with HH, BB with BE, MV, SN, TH and ST showed a decline AR between the years 2005 and 2010, whereas AR values increase in the other federal states (see Table, Supplemental Digital Content 3, https://links.lww.com/INF/C316).
To validate these findings and the quality of DRG coding, the appropriate ICD-10 code (included in the DRG code) was compared with the results of ELISA tests for representative samples (N = 1003) reported in the records obtained from the Mainz study center. The ELISA test results were strikingly consistent with the ICD-10 codes (Table 2).
This retrospective analysis of the DRG codes of nearly all German hospitals provides real-life data on 0-year-old to 10-year-old children hospitalized with AGE and RVGE during a 6-year period. A total of 152,636 hospitalizations, 29.3% of all children diagnosed with an AGE, were coded as RVGE. This equates to 25,440 hospitalizations per year substantiating the enormous impact of RVGE on the German healthcare system.
Studies that employ data collected retrospectively often emphasize the importance of data quality. In this regard, data based exclusively on the hospital discharge diagnoses according to the current ICD-10 catalog probably underestimate incident rates.8 Moreover, it can argue that the use of data intended solely for reimbursement has limitations, but it also has an advantage. Detection and identification of a specific pathogen, for example, RV, norovirus or adenovirus, and pathogen-specific coding (eg, A08.0 for RVGE) result in significantly higher hospital revenues. Although the identity of the pathogen usually does not change the course of therapy, hospitalized AGE cases in Germany are generally tested and pathogen-specific codes are reported.
The quality of the data presented herein is putatively better than that reported in previous studies.4,9 Although the data source is the same, the data sets, statistical methods, results and conclusions differ significantly; as such, these studies are not directly comparable. Uhlig et al,9 for example, only reported raw data that lacked CIs and, in the absence of age-standardized infection rates and regression models, failed to adjust for the effects of a specific year and federal state. Moreover, the quality of DRG coding was validated in this study by random sampling the hospitalization records of the Mainz study center; nearly all ICD-10 codes correlated with the ELISA test results. These data are rare inasmuch as they come from a fixed and established statutory system covering the whole country and, thus, represent the entire German pediatric population.
The results of many epidemiological studies are limited because their data are based upon active surveillance systems. Other studies only include a small, nonrepresentative number of hospitals or study populations.5,10 Extending the data sources by meta-analysis is difficult because of variations in study design, for example, differences in age groups or detection methods. It is not surprising, therefore, that meta-analysis of 76 studies conducted in 16 European countries reported that the cases of AGE caused by RV in children less than 5 years of age ranged from 25.3% to 63.5%.3 Hospitalization rates among infected children varied between 7% and 81%, dependent upon the country.
The analysis reported here indicates that RV caused 29.3% of AGE cases in children aged 0–10 years (ICD-10 code A08.0). The actual proportion of RVGE was probably higher because the facilities for pathogen-specific ELISA testing are neither standardized nor available nationwide. Thus, it is assumed that a significant proportion of cases coded as nonspecific viral AGE (n = 307,942; Table 1) were RVGE if they were untested. Regardless, we observed a considerably higher number of children (25,440 aged 0 to 10 years) hospitalized because of RVGE each year compared with previous studies conducted in Europe. Soriano-Gabarró et al2, for example, reported the annual hospitalization of 13,573 children younger than 5 years of age because of RVGE; Koch et al4 found an average annual number of 17,636 RV-related hospitalizations in this same age group in Germany. Because 95% of children experience RVGE by the age of 5 years, extending the age of the group considered up to 10 years could contribute only partially to the higher incidence of hospitalizations reported herein.11 Rather, it seems far more likely that the use of the large data set produced a more realistic picture of the RV infection rate in Germany.
The seasonal distribution of RVGE determined in this study is based upon the rate of hospitalization. This is not seen as a limitation because almost half of all children with RVGE in Germany are hospitalized.4 In fact, the results presented here correspond to the data of Patel et al12, which showed RV prevalence in the European region peaked in Q1 and reached its lowest level in Q3.
Six federal states, that is, SH with HH, BB with BE, MV, SN, TH and ST, showed a slightly lower AR in 2010 compared with 2005, whereas AR increased in the other federal states. This trend reflects, in part, the consequences of introducing routine RV immunization in 4 eastern federal states (SN, BB with BE, MV and TH) in 2008/2009. Indeed, RV surveillance conducted in MV to assess vaccine effectiveness in children less than 2 years of age revealed significantly fewer hospitalizations among vaccinated (23%) than unvaccinated children (61%).13 Notably, vaccine recommendations are more easily implemented in former Eastern Germany than the western part of the country due, undoubtedly, to the existence of mandatory vaccination programs. Socioeconomic differences between the east and west, which are known to influence susceptibility to childhood infections, probably also contribute to the elevated baseline infection rate observed. Significantly more hospitalizations in the federal states of former East Germany are known.14 Contrariwise, the AR of RV infections in 2010 was lower in many western states, for example, Bavaria, Hesse, SH, although vaccination was already implemented in the east (see Table, Supplemental Digital Content 3, https://links.lww.com/INF/C316). General reporting differences, that is, hospitals in the eastern federal states seem more eager than the western clinics to report RV infections via DRG, account for this disparity.15
An inverse correlation between RVGE and vaccine doses sold substantiates the positive influence of immunization. This finding correlates with a previous study demonstrating a >60% decline in both community and hospital-acquired RV infections in the US after implementing routine immunization of infants.16 Moreover, population-based RV surveillance in 3 US counties estimated that vaccination could reduce total annual medical costs by 187 million dollars.17 In Australia, the adjusted incidence of RV disease was 23.1 and 6.5 per 1000 hospitalizations in children less than 5 years of age prevaccination and postvaccination, respectively.8 A significant reduction in RV notifications was also found in Germany after the introduction of a vaccine. In areas that attained 64% vaccine coverage, RV-related hospital admission of children less than 1 year of age decreased by 60% compared with a 19% decrease in areas characterized by low vaccine coverage.9 A similar correlation was found between vaccine coverage and RV cases in children younger than 2 years.14 The overall RV vaccination coverage in Germany is still low, however; coverage in children between 0 year and 1 year of age only increased from 3% in 2007 to 26% in 2010.14
In addition to benefiting the vaccinated child, immunization with live attenuated vaccine can reduce RV transmission and induce herd protection.18 A US study found, for example, an 83% decrease in hospitalizations because of RV in vaccine-eligible children between 1 and 23 months of age, as well as, a 70% decrease in nonimmunized older children up to 18 years of age.19
Results reported here must be interpreted with care because exposure assessment was crude, and the design to evaluate IR was limited to an ecological study. Thus, bias and unknown confounding factors cannot be excluded. Additional limitations must be considered when assessing the effect of vaccination recommendations in eastern Germany, for example: (1) the period of time between the recommendation and the study end was relatively short, (2) validation of correct coding in each study center was impossible and (3) ELISA tests were not equally available and varied in specificity and sensitivity. Moreover, improved hygienic conditions alone could have contributed to the overall decline in AGE incidences. Nonetheless, provided reasonable vaccine prices, generalized immunization against RV could significantly reduce the financial burden of the German healthcare system.20 A detailed analysis of the economic data generated in this study is in preparation.
Although mortality because of RVGE in Germany is low, the disease burden for society and the individual child receiving in-patient treatment is high. Incidence rates decrease with age and differ by federal state, year and season. RVGE cases decreased after recommended immunization and a negative correlation exists between AR and vaccine doses sold. Therefore, we conclude that high RV vaccination coverage will reduce the risk of in-patient treatment and the costs associated with RVGE.
This work was supported by the Sanofi Pasteur MSD GmbH. Data were generated in cooperation with the Federal Statistical Office (Destatis) Wiesbaden, Germany. The authors thank Drs. Frank Erdnüß (Wiesbaden) and Stephen H. Gregory (Providence, RI) for their help in preparing this manuscript.
1. Parashar UD, Hummelman EG, Bresee JS, et al. Global illness and deaths caused by rotavirus disease in children
. Emerg Infect Dis. 2003;9:565–572
2. Soriano-Gabarró M, Mrukowicz J, Vesikari T, et al. Burden of rotavirus disease in European Union countries. Pediatr Infect Dis J. 2006;25(1 Suppl):S7–S11
3. Ogilvie I, Khoury H, Goetghebeur MM, El Khoury AC, Glaquinto C.. Burden of community-acquired and nosocomial rotavirus gastroenteritis in the pediatric population of Western Europe: a scoping review. BMC Infec Dis. 2012;12:62
4. Koch J, Wiese-Posselt M.. Epidemiology of rotavirus infections in children
less than 5 years of age: Germany, 2001-2008. Pediatr Infect Dis J. 2011;30:112–117
5. Forster J, Guarino A, Parez N, et al.Rotavirus Study Group. Hospital-based surveillance to estimate the burden of rotavirus gastroenteritis among European children
younger than 5 years of age. Pediatrics. 2009;123:e393–e400
6. World Health Organization. . Rotavirus vaccines WHO position paper: January 2013 – Recommendations. Vaccine. 2013;31:6170–6171
7. . Standing Committee on Vaccination (STIKO) Recommendation for routine rotavirus vaccination in Germany. Bundesgesundheitsbl. 2013;56:955–956
8. Jayasinghe S, Macartney K.. Estimating rotavirus gastroenteritis hospitalisations by using hospital episode statistics before and after the introduction of rotavirus vaccine in Australia. Vaccine. 2013;31:967–972
9. Uhlig U, Kostev K, Schuster V, Koletzko S, et al. Impact of rotavirus vaccination in Germany: rotavirus surveillance, hospitalization, side effects and comparison of vaccines. Pediatr Infect Dis J. 2014;33:11:e299–e304
10. Hsu VP, Staat MA, Roberts N, et al. Use of active surveillance to validate international classification of diseases code estimates of rotavirus hospitalizations in children
. Pediatrics. 2005;115:78–82
11. Bernstein DI.. Rotavirus overview. Pediatr Infect Dis J. 2009;28(3 Suppl):S50–S53
12. Patel MM, Pitzer VE, Alonso WJ, et al. Global seasonality of rotavirus disease. Pediatr Infect Dis J. 2013;32:e134–e147
13. Adlhoch C, Hoehne M, Littmann M, et al. Rotavirus vaccine effectiveness and case-control study on risk factors for breakthrough infections in Germany, 2010-2011. Pediatr Infect Dis J. 2013;32:e82–e89
14. Uhlig U, Kostev K, Schuster V, Uhlig HH.. Rotavirus vaccination in Germany: analysis of nationwide surveillance data 2006 to 2010. Pediatr Infect Dis J. 2011;30:e244–e247
15. Robert Koch-Institut. . Rotavirus-gastroenteritis. Infektionsepidemiologisches Jahrbuch meldepflichtiger Krankheiten für. 2013;2014:169–175
16. Anderson EJ, Rupp A, Shulman ST, Wang D, Zheng X, Noskin GA.. Impact of rotavirus vaccination on hospital-acquired rotavirus gastroenteritis in children
. Pediatrics. 2011;127:e264–e270
17. Kilgore A, Donauer S, Edwards KM, et al. Rotavirus-associated hospitalization and emergency department costs and rotavirus vaccine program impact. Vaccine. 2013;31:4164–4171
18. Paulke-Korinek M, Kundi M, Rendi-Wagner P, et al. Herd immunity after two years of the universal mass vaccination program against rotavirus gastroenteritis in Austria. Vaccine. 2011;29:2791–2796
19. Chang HGH, Smith PF, Tserenpuntsag B, et al. Reduction in hospitalizations for diarrhea and rotavirus infections in New York state following introduction of rotavirus vaccine. Vaccine. 2010;28:754–758
20. Aidelsburger P, Grabein K, Böhm K, et al. Cost-effectiveness of childhood rotavirus vaccination in Germany. Vaccine. 2014;32:1964–1974