The present study clearly demonstrates that UMV not only led to reduction of RV-GE-associated hospitalizations by 73.9%, but also to a pronounced reduction of nosocomial RV infections by 92.5%. The possible indirect effect of UMV on nosocomial RV infections may be an important aspect considering severe disease courses in hospitalized patients with co-morbidities. This was corroborated by the fact that mortality in our pre-vaccination cohort was attributed to deterioration of the underlying disease by nosocomial RV infection. A reduction of nosocomial RV infection was also seen in the US shortly after introduction of RV vaccination .
Also, in our study, secondary BSI was, in the majority of cases, linked to nosocomial RV-GE. In infants, RV infection has been described as a cause of pneumatosis intestinalis, hemorrhagic gastroenteritis, necrotizing enterocolitis and secondary BSI with mainly pathogens belonging to the intestinal microflora and the members of the Enterobacteriaceae family [24–29, 33, 34]. In our cohort, many of the 20 patients with secondary BSI were preterm infants or patients with compromised immune system and showed BSI due to Staphylococcus aureus. These findings allow us to hypothesize that not only intestinal mucosa dysfunction due to RV-GE promotes transition of intestinal bacteria, but that also a fatal combination of severe underlying diseases with dehydration and malnutrition in succession of RV-GE could have made patients more prone for secondary BSI caused also by non-intestinal bacteria. Although both RV vaccines cover the most important serotypes even in Austria , so far we are not able to know whether a shift to other RV serotypes will take place in the future causing inefficiency of the vaccine serotypes [35, 36] and whether other gastroenteritis pathogens will take over.
The reduction of community-acquired hospitalized RV-GE cases found in our study confirms the data of recently published studies [5, 11, 12] which showed a significant reduction of RV cases reported by sentinel hospitals in the vaccination period in all age groups. In our study, the age group 0–11 months had the highest benefit of vaccination, highlighting the importance of starting the vaccination as early as possible . A clear reduction in the older individuals indicates the presence of herd protection in the population by reduction of RV transmission [12, 37, 38]. However, also transmission of the attenuated RV types from vaccinated children to unvaccinated individuals may induce some immunity against RV in the unvaccinated older population [39–41].
In the US, a 50% decrease in RV-positive laboratory tests has been found after recommendation of RV vaccination for routine use in 2006, showing a delay in seasonal onset of the RV season 2007–2008 by 2–4 months . These data are in contrast to our data and previous data from Austria  which could not detect a shift of RV-associated hospitalizations to later months. However, the findings of the US study are limited by missing data from the end of the RV season 2007–2008 and by the fact that RV was tested based on the discretion of the physicians and local policies.
In our cohort, we could demonstrate a reduction of mean annual real costs by 72.7% between the pre-vaccination and the funded vaccination period, although it was not possible to include indirect costs, such as work loss by parents staying with the hospitalized child and supervision of siblings staying at their own. Considering the direct and indirect costs of hospital admissions due to RV-GE, UMV programs have been shown to be cost-effective  and would lead to a reduction of costs for RV-GE-related hospitalizations/emergency visits by 83% and for medical consultations by 75% . In our study, in the funded vaccination period, hospital stays were about half a day shorter than in the pre-vaccination period which may also be accounted for by milder disease courses .
Limitations of our study exist in the fact that only hospitalized cases were counted and that local health-seeking behaviors of the population and hospital-specific guidelines for admission may influence the hospitalization rates . We here reported the results of a single, large, tertiary care center and might have consistently missed mild cases of RV-GE which were treated at home. One disadvantage of a single-center experience also exists for differences in incidence oscillations by seasonal forcing and demographic forcing that may cause different patterns of hospitalizations in different places. Certainly not all RV cases would have been detected and might have been influenced by different detection systems and coding practices. The detection system was changed in 2005 to a more sensitive method but with similar specificity. Additionally, the nature of the disease itself has to be taken into account, as wild-type RV infection does not prevent re-infection, or does not consequently enhance protection against recurrent infection [45–47]. However, the relative risk of re-infection is lower after one RV infection (0.62), two infections (0.4) or three subsequent infections (0.34) . Subsequent infections are often subclinical or mild  and will not consequently lead to hospitalization of the children. Milder and asymptomatic cases will be less infectious, both as a result of reduced viral shedding rate and a shorter duration of infection. Thus, vaccination will cause fewer infections by making vaccinated individuals less able to become re-infected, if vaccination elicits the same immunological response as natural infections, and less able to transmit RV to others as predicted by transmission dynamic models . In addition, also the numbers of older children may be underestimated, as most of them will suffer from mild infection without need for hospitalization . Hospitalized re-infections were not assessed in our study, as double-counters were excluded to avoid biasing our results. Interestingly, in our study, 557 patients were double-counters, however, most of them were repeated detections of RV antigen in the stool samples during the same hospital stay or in immunocompromised patients showing a slow clearance from intestinal RV replication . Unfortunately, due to the retrospective character of our study, the vaccination status was not sufficiently documented in the investigated patients, thus, despite being certainly of interest, break-through RV infections could not be assessed. Natural epidemiological oscillations are other factors which might have influenced our results. However, epidemiological studies since 1997 based on a sentinel system showed that fluctuations of RV-GE numbers were always less than the decrease observed since introduction of UMV . An unrecognized change in the prevention of pathogen transmission was excluded in our study, as similar hygienic rules and introduction of nurses, parents and visiting persons were applied during the whole study period, interventions which have been shown to be crucial in reduction of transmission .