Economic burden of nosocomial infections caused by vancomycin-resistant enterococci

Although vancomycin-resistant enterococci (VRE) are organisms of low virulence and low pathogenicity, they frequently cause nosocomial infections [1, 2]. A significant increase in the VRE prevalence has been observed in many countries recently. For example the proportion of vancomycin-resistant Enterococcus faecium has increased rapidly from <5% in 2001 to 14.5% in 2013 in Germany [3]. In the USA surveillance data show that of all nosocomial infections reported to the National Healthcare Safety Network, 3% were due to VRE [4]. In Europe, Enterococcus spp. was isolated from 9.6% of all nosocomial infections. Furthermore vancomycin-resistance was reported in 10.2% of isolated enterococci (ECDC-Annual-report 2014). The incidence of nosocomial infections caused by VRE is particularly high on intensive care units (ICU) [1, 5]. Risk factors associated with colonization and subsequent development of nosocomial infections due to VRE are severe underlying health conditions such as liver transplantation, neutropenia, diabetes mellitus or renal dysfunction [6–8]. Several studies have shown that VRE bloodstream infections (BSI) are associated with a significantly higher mortality compared to BSI by vancomycin-susceptible enterococci (VSE) [9, 10].

VRE infections result in a greater number of invasive procedures, additional antimicrobial therapy, and an extended length of stay, all of which can increase the total hospital costs [8, 11–14]. Hospital costs due to VRE infections such as BSI varied between $9949 and $77,558 in a university-based teaching hospital in the USA (calculated in 2003) [9, 15]. Another study in the USA revealed that attributable costs for surgical site infections (SSI) due to VRE were about $12,766 [16]. In contrast to these findings other investigators suggested that VRE infections and VSE infections do not differ in attributable costs, length of stay (LOS) and mortality [17]. Nevertheless, the Centers for Disease and Control and Prevention (CDC) as well as the German Society for Hygiene and Microbiology (DGHM) have developed infection control measures that aim to prevent VRE transmission [18, 19]. VRE screening in high risk areas and isolation of patients with VRE are therefore well-established in many hospitals despite the costs for those precautions.

At present studies on the economic burden of VRE infection in hospitalized patients have been performed in the USA and other countries but not in Europe, with its highly developed health care system. In the current study we compared hospital costs between patients with VRE infection and patients with VSE infections to determine the costs that are directly attributed to the vancomycin resistance. In addition, we determined factors that may be associated with increased costs of VRE patients in comparison to VSE patients.

The initial search of the microbiological database provided 504 patients with VRE. 403 patients were excluded because they had no signs of infection, or infection was not nosocomial. 101 cases were included as cases. VSE were detected in 5502 patients, of whom 936 patients had nosocomial infections. Matching was possible for 42 cases, thereof 41 were affected by Vancomycin-resistant E. faecium and one by Vancomycin-resistant E. faecalis. Those were then compared to 42 controls with VSE infection, thereof 24 infected by E. faecium and 18 infected by E. faecalis.

Increasing antimicrobial resistance is a growing problem for patients as well as for health care systems with respect to continuously growing hospital costs. Accurate estimates of the financial outcome associated with nosocomial infections caused by multidrug resistant organisms are still rare but they remain important for evaluating the cost effectiveness of prevention strategies to reduce transmission.

Attributable hospital costs for vancomycin-resistance per patient were EUR 13,157. Other authors calculated attributable hospital costs for vancomycin-resistance ranging from $ 1546 to $ 77,558 [9, 14–17, 26]. Three studies looked at blood stream infections (BSI) [9, 14, 15], one study looked at wound infections [16], one study looked at wound infections, urinary tract infections, and blood stream infections (BSI) [17], and one study looked at wound infection, urinary tract infections, blood stream infections (BSI) and intraabdominal infections [26]. In these studies, the total hospital costs ranged from $ 33,224 to $ 124,257 for a patient with VRE, from $ 20,895 to $ 56,707 for a patient with VSE, or from $ 8192 und $ 18,863 for a patient with neither VRE nor VSE, respectively [9, 14, 15, 17]. The wide range of costs are probably caused by the difference in the study design, differences in matching criteria, the size of the study population, and differences in adjustment for confounders. For example, Pelz et al. compared costs between cases and unmatched controls [17]. Song et al., even though so they matched controls for age, year of admission, morbidity and length of stay prior to VRE infection, they did not consider other factors that affect costs such as overall length of stay or stay on the ICU [9]. However although costs for VRE infections vary considerably between published studies, the key message remains: vancomycin-resistance in enterococci is associated with increased costs when patients suffer from infections due to VRE.

Multivariable analysis in this study showed that a VRE-infection was independently associated with a 1.4-fold increase in total costs per patient. This confirms previous findings by Kaye et al., who found also a 1.4 fold increase in total costs per patient [16]. Other studies found similar multiplicative effects (ME) ranging from 1.2 to 1.6 [9, 26]. Carmeli et al., who looked at four different VRE infections, found a ME of 1.2 for blood stream infections (BSI), a ME of 1.3 for urinary tract infection, a ME of 1.5 for wound infection, and a ME 1.6 for intraabdomial infections [26]. We could also show that age > 60 years was associated with lower (!) costs. These findings are somehow contrary to the data of Webb et al., who found a positive correlation of age (>50 years) and increased costs [27]. This may at least in part be explained by mortality in our older patient population. An early fatal outcome will diminish the patient’s length of stay and, by this attributive costs. A similar phenomenon may apply to patients with severe underlying diseases. A high Charlson Comorbidity Index may lead to increased and in particular early mortality, which in turn will then reduce the overall costs.

Previous studies did not differentiate between costs before and after VRE infection, but calculated costs for the total length of stay instead. Here we looked at costs prior to VRE infection and after the patient had developed VRE infection. Our study clearly shows, that it is indeed the VRE infection itself that drives costs, as the costs between patients with VRE and with VSE did not differ before the infection with enterococci occurred (EUR 17,893 vs. EUR 16,600; p = 0.386), but only after they had developed the infection (EUR 37,971 vs. EUR 23,025; p = 0.049).

One may argue that the increased costs of VRE patients could in fact be due to comorbidities rather than the infection itself. Patients with VRE infections often show an increased morbidity, which influences clinical outcome and might indirectly affect costs [10, 27–30]. Several surrogate parameters such as the overall length of stay or various comorbidity scores are being used as a measure of morbidity [28]. In a study by Webb et al., differences between VRE and VSE patients became apparent when the authors stratified for a case-mix-index ≤1, whereas matching patients with a case-mix-index >1 did not generate significant differences in costs between the two groups [27]. Thus to avoid confounding, we applied the Charlson comorbidity index for matching an appropriate control group. We also matched for the time at risk to apply a second surrogate parameter for morbidity. Both surrogate parameters have been used as matching parameters in previous case-control studies looking at costs due to nosocomial infections [24, 31].

It is noteworthy that in contrast to other matched case-control studies comparing costs between VRE and VSE patients that found significant differences in mortality between the two groups [9, 29], in this study we did not detect significant difference in mortality between cases and controls (33% vs 26%, p = 0.634). The meta-analysis by Salgado et al. showed an increased mortality of VRE patients in comparison to VSE patients in most studies [10]. Only few studies, all of which had a small study population (6 to 46 VRE cases) did not show a difference in mortality. The small study size might be one reason why we failed to detect a significant difference in mortality between cases and controls in the present study. Beside the size of the study population, the quality of the matching criteria will also greatly influence the mortality for cases compared to controls, because the Charlson comorbidity Index was particularly developed to estimate the risk of death from comorbid disease [20]. In the present study the Charlson comorbidity Index (CCI) was identical between cases and controls (CCI = 3), which might provide another explanation for equal mortality between cases and controls.

Detailed analysis of costs between patients with VRE and VSE showed significant differences for pharmaceuticals, nursing staff, medical products, and for assistant medical technicians. The most significant differences turned out to be for pharmaceuticals (EUR 6030 vs. EUR 2801; p = 0.008). This might be due to treatment with second and third line antibiotics such as linezolid or quinupristin/dalfopristin. Similar stratifications are scarce in other studies looking at costs of VRE infections. Geahart et al. did not find differences when stratifying for costs for pharmaceuticals, but did find differences for the laboratory costs [29]. Butler et al. found significant differences in costs for room and board, pharmaceuticals, laboratory and radiology [15]. Differences in the costs are most probably due to the prolonged length of hospital stay after the onset of VRE infection. VRE cases and VSE controls were matched for the time at risk (before infection), but VRE cases presented with a significant longer hospital stay thereafter. This prolonged time largely attributes to the cost calculation including the use of medical products and technical staff.

Our study has some limitations. First, all patients included in the analysis stayed at the Hannover Medical School in Germany. Thus results of this single-centre study may not apply to other institutions. Second, the small sample size of 84 patients for cases and controls each limits this study. However, due to the application of strict matching criteria, we were not able to further increase the study population. Although we did match for the type of infection, we did not differentiate between types of infection in terms of cost calculation and calculation of the multiplicative effect due to the limited number of case patients as it has done by others before [32, 33]. On the other hand, fewer matching criteria would have led to less power of statistical analysis, and we consider our type of matching as the main strength of this study. Particularly using “time-at-risk” for matching cases and controls limited the sample size. But including “time-at-risk” was important because Cosgrove et al. showed a significant correlation between time-at-risk and financial outcome [28]. Third, we did not evaluate the financial impact of VRE infection from the perspective of the patient and the society, such as lost wages, or the impact on life-time-quality. Finally, recent evidence suggests that even matching on infection onset still provides slight overestimates of costs attributable to infection [34, 35], but this should only have had a minor impact on the overall results.

To our knowledge this retrospective case-control study of infections with VRE and VSE provides the first data of its kind in Europe. We demonstrated that attributable costs for vancomycin-resistance in enterococci in patients who develop infections with this pathogen are considerable, and that vancomycin-resistance is an independent predictor for the overall increase of hospital costs. Infections with VRE not only increased costs for the health care system, but, due to the reimbursement policy, even generates loss for the individual hospital.

These results emphasise the importance of strategies for the prevention of VRE spread in the hospital.