Burden of Six Healthcare-Associated Infections on European Population Health: Estimating Incidence-Based Disability-Adjusted Life Years through a Population Prevalence-Based Modelling Study

Background Estimating the burden of healthcare-associated infections (HAIs) compared to other communicable diseases is an ongoing challenge given the need for good quality data on the incidence of these infections and the involved comorbidities. Based on the methodology of the Burden of Communicable Diseases in Europe (BCoDE) project and 2011–2012 data from the European Centre for Disease Prevention and Control (ECDC) point prevalence survey (PPS) of HAIs and antimicrobial use in European acute care hospitals, we estimated the burden of six common HAIs. Methods and Findings The included HAIs were healthcare-associated pneumonia (HAP), healthcare-associated urinary tract infection (HA UTI), surgical site infection (SSI), healthcare-associated Clostridium difficile infection (HA CDI), healthcare-associated neonatal sepsis, and healthcare-associated primary bloodstream infection (HA primary BSI). The burden of these HAIs was measured in disability-adjusted life years (DALYs). Evidence relating to the disease progression pathway of each type of HAI was collected through systematic literature reviews, in order to estimate the risks attributable to HAIs. For each of the six HAIs, gender and age group prevalence from the ECDC PPS was converted into incidence rates by applying the Rhame and Sudderth formula. We adjusted for reduced life expectancy within the hospital population using three severity groups based on McCabe score data from the ECDC PPS. We estimated that 2,609,911 new cases of HAI occur every year in the European Union and European Economic Area (EU/EEA). The cumulative burden of the six HAIs was estimated at 501 DALYs per 100,000 general population each year in EU/EEA. HAP and HA primary BSI were associated with the highest burden and represented more than 60% of the total burden, with 169 and 145 DALYs per 100,000 total population, respectively. HA UTI, SSI, HA CDI, and HA primary BSI ranked as the third to sixth syndromes in terms of burden of disease. HAP and HA primary BSI were associated with the highest burden because of their high severity. The cumulative burden of the six HAIs was higher than the total burden of all other 32 communicable diseases included in the BCoDE 2009–2013 study. The main limitations of the study are the variability in the parameter estimates, in particular the disease models’ case fatalities, and the use of the Rhame and Sudderth formula for estimating incident number of cases from prevalence data. Conclusions We estimated the EU/EEA burden of HAIs in DALYs in 2011–2012 using a transparent and evidence-based approach that allows for combining estimates of morbidity and of mortality in order to compare with other diseases and to inform a comprehensive ranking suitable for prioritization. Our results highlight the high burden of HAIs and the need for increased efforts for their prevention and control. Furthermore, our model should allow for estimations of the potential benefit of preventive measures on the burden of HAIs in the EU/EEA.

The study by Pépin et al (2005), conducted during an epidemic caused by C. difficile PCR ribotype 027 reported the highest estimates for LOS and attributable mortality. As the study setting was not representative of a setting where CDIs generally occur, after consultation with the experts we decided not to include the study.
The transitional probabilities for the outcome 'postcolectomy state' ranged between 0.2 and 3.8% (Bhangu, 2012). The duration of this outcome was lifelong.
Calculation of a point estimate by pooling incidences did not seem appropriate without knowing the baseline risks for colectomy within the studies. In the majority of articles, the characteristics of the study population only referred to patients with colectom y because the selected studies focused on outcomes after colectomy.
Since the systematic search for sepsis did not reveal any eligible studies, this outcome was omitted.
From a clinical point of view, recurrence is a major issue in patients with CDI. Nevertheless, inclusion of recurrence (e.g. by introducing one or more loops in the outcome tree) may have led to double counting of cases, and subsequently to an overestimation of the burden of disease, depending on how the issue of recurrence was dealt with by the prevalence data fitted into the tree. Therefore, we chose not to include recurrence in the outcome tree. Healthcare-associated pneumonia (HAP)

Risk of symptomatic infection and duration of disease
Healthcare-associated pneumonia (HAP) was defined in accordance with the ECDC case definition (ECDC, 2012). A systematic review of the literature was performed to estimate the sequelae and the probability of developing these sequelae following HAP. This review took into consideration the role of co-morbidities by estimating the attributable impact of HAP: attributable mortality, attributable risk of developing sequelae and its duration (effect on the length of hospital stay).
The extracted literature was limited to ventilator-associated pneumonia (VAP) because no specific literature was found for the more general search terms 'pneumonia' and 'lower respiratory tract infection'.
Attributable ICU-mortality due to VAP varied from 0.1% to 9% based on the studies by Aybar Türkoglu et al. (2008) and Rello et al. (2002). The report on surveillance of healthcare-associated infections in intensive care units in Europe, 2008-2012, stemming from the HAI-Net ICU surveillance (ECDC, 2016), found an overall attributable case fatality proportion for pneumonia patients of 3.5% that was used as the median estimate in our model.
Attributable length of stay in an intensive care unit (ICU) due to VAP varied from 2.03 to 7 days whereas attributable length of hospital stay following VAP varied from 7 to 11.5 days. The latter range is included in the outcome tree (AybarTürkoglu, 2008; Rello, 2002.
Sepsis and acute respiratory distress syndrome (ARDS) are generally considered to be frequent consequences of VAP. However, only one study provided data on the transitional probability which amounted to 39% (proportion of patients suffering from severe sepsis and/or septic shock) for sepsis/ARDS as a health consequence of VAP (Damas, 2011). Duration of severe sepsis or septic shock was 9.9-13 days (Olaechea, 2013; Renaud, 2001. The long-term health outcomes following sepsis and ARDS were taken from the model developed for healthcare-associated primary bloodstream infection. Healthcare-associated neonatal sepsis

Risk of symptomatic infection and duration of disease
Neonatal sepsis is characterised in accordance with the definition set out in ECDC's point prevalence survey of healthcare-associated infections and antimicrobial use in European acute care hospitals with the code NEO-LCBI -laboratory-confirmed bloodstream infection in neonates (ECDC, 2012).
The entire disease model was based on the systematic review of the natural history of neonatal sepsis, conducted for the purpose of estimating burden of disease by Haller et al. (2015, in press). Given the lack of studies referring to all sepsis in term infants, this outcome tree represents very low weight birth (VLBW) infants. This must be taken into account when inputting the number of cases.
Model input summary Healthcare-associated Primary blood-stream infections (P-BSI)

Risk of symptomatic infection and duration of disease
A healthcare-associated primary bloodstream infection (HA primary BSI) in adults was defined in accordance with the ECDC case definition (ECDC, 2012).
For some of the long-term complications, we did not identify any relevant literature for cases with a health status of sepsis or BSI, although information was found for Acute Respiratory Distress Syndrome (ARDS). As ARDS frequently occurs following severe sepsis, we agreed with the clinical experts that ARDS could be used as a proxy for complicated sepsis.
Two studies were identified that reported on the risk of developing cognitive impairment following BSI, but the reported values differed significantly: 11% for  and 47% for  . Major methodological differences included different ascertainment methods and study participants' health states at the inclusion phase.
A certain degree of physical impairment was reported in all patients during follow-up after severe sepsis/septic shock (Hopkins, 2005; Herridge, 2003; Hofhuis, 2008. We assumed that the risk of chronic/long-term renal replacement therapy (RRT) was the same for all instances of acute renal failure, irrespective of cause. A total of 8% of BSI patients needed RRT at the onset of BSI . In accordance with the American College of Chest Physicians/Society of Critical Care Medicine (ACCP/SCCM) definition, we classified BSI patients with acute renal failure as "complicated sepsis". Thus, the percentage of BSI patients needing RRT (8%) was weighted with severity proportions (complicated sepsis in 33.7-47.1%) and a range of 16.6-23.2% was estimated for patients with complicated sepsis that require RRT at the onset of BSI. In conclusion, 16.6-23.2% of all patients with complicated sepsis patients were estimated to require RRT in the short-term and of those 5.4% were estimated to require long-term maintenance RRT. Assuming a background risk of zero, the risk of requiring long-term RRT after severe sepsis was 0.9-1.3%. In an attempt to provide an outcome tree reflecting the natural history of healthcare associated SSI, a systematic review of the literature was undertaken. During the initial search strategy, it became clear that it was impossible to obtain a full picture of all attributable deaths, complications, sequelae and length of hospital stay due to the heterogeneous nature of deep, organ-space and superficial SSIs.

Model input summary
We therefore decided to group all SSIs and focus on one sequelae: the proportion of case fatalities. The latter was set to 0.9% for patients aged under 65 years and to 3.6% for patients 65 years and over (Astagneau, 2001). However, this approach underestimated the burden of healthcare-associated SSI which would need to include long-term sequelae, such as mobility restrictions and reoperations following joint replacement surgery.
Model input summary  Healthcare-associated urinary tract infection (HA UTI)

Risk of symptomatic infection and duration of disease
Healthcare-associated urinary tract infection (HA UTI) is defined in accordance with the ECDC case definition (ECDC, 2012). A systematic review of the literature was performed to estimate the sequelae and the probability of developing these sequelae following a UTI. This review took into consideration the role of co-morbidities by estimating the attributable impact of a UTI: attributable mortality, attributable risk of developing sequelae and duration (effect on length of hospital stay).
The extracted literature was limited to catheter-associated UTIs because no specific literature was found for more general search terms.
Attributable ICU and hospital mortality due to UTIs was not statistically significant and we therefore did not attribute a case-fatality rate to UTIs directly. The attributable length of stay for patients with urinary tract infections was 2.6 days (Chant, 2011).
According to the ECDC surveillance definition for UTI used in the point prevalence study, which was the source of the incidence data, both the presence of bacteruria and symptoms are required for the diagnosis of HA UTI. From the systematic literature review, we could not find evidence of the risk of secondary bacteraemia or urosepsis among patients with symptomatic HA UTI, in order to calculate the transitional probability related to complications and death. Therefore we indirectly estimated this risk by combining the risk of developing secondary bacteraemia/urosepsis in critically-ill patients with catheter-associated bacteruria, 1.3-4.8% (Laupland, 2002; Laupland, 2005; Clec'h, 2007, and the risk of developing a urinary tract infection in patients with bacteruria, 24% (Saint, 2000). The resulting transitional probability was set to 5.42-20%.
We assumed that all patients developing bacteraemia/urosepsis will have the same outcomes and relative risks as those developing sepsis. For details, please refer to the bloodstream infection outcome tree.