Incidence of Hospital-Acquired Pneumonia, Bacteraemia and Urinary Tract Infections in Patients with Haematological Malignancies, 2004–2010: A Surveillance-Based Study

Objective This study charted incidence trends of hospital-acquired (HA) pneumonia, bacteraemia and urinary tract infections (UTI) in a haematology department. Methods Prospective surveillance of hospital-acquired infections (HAI) was undertaken in a 42-bed haematology department of a university hospital. All patients hospitalized ≥48 hours between 1st January 2004 and 31st December 2010 were included. Definitions of HAI were based on a standardized protocol. The incidence was the number of events per 1000 patient-days at risk; only the first HAI was counted. Multivariate Poisson regression was fitted to assess temporal trends. Results Among 3 355 patients (58 063 patient-days at risk) included, 1 055 (31%) had HAI. The incidence of HA pneumonia, HA bacteraemia and HA UTI was respectively 3.3, 12.0 and 2.9 per 1000 patient-days at risk. HA bacteraemia incidence increased by 11% (95% confidence interval: +6%, +15%, P<0.001) per year, independently of neutropenia, central venous catheterization (CVC) and haematological disease. The incidences of HA pneumonia and HA UTI were stable. The most frequently isolated pathogens were Aspergillus spp. (59.2%) for pneumonia, coagulase-negative Staphylococcus (44.2%) for bacteraemia and enterobacteria (60%) for UTI. Conclusion The incidence of bacteraemia increased, indicating that factors other than CVC exposure, including chemotherapy with its impact on the immune system, could explain this trend. Further analytic studies are needed to explore the factors that could explain this trend.


Introduction
Patients with haematological malignancies are at high risk of hospital-acquired infections (HAI) because the severity of their underlying illness often requires aggressive treatment: chemotherapy, bone marrow or peripheral blood stem cell transplantation. This could lead to severe and prolonged immunosuppression, increasing the risk of infection and possibly worsening the prognosis [1,2]. Mortality attributable to hospital-acquired (HA) bacteraemia in patients with cancer ranges between 10 and 20% [3][4][5], whereas mortality from HA pneumonia is much higher, between 40 and 60% [6][7][8]. Over the past few decades, major progress has been made in the curative treatment of haematological malignancies which has resulted in reduced overall mortality [9][10][11][12], but also longer neutropenia and higher risk of infectious complications [13,14].
HAI are important adverse events in the disease history of patients with haematological malignancies, sometimes culminating in death; they are also responsible for longer hospital stay and increased healthcare costs [15,16]. However, few studies have examined the incidence of HAI in this adult patient population as most surveillance have focused only on bone marrow transplant patients or children [13,[17][18][19][20][21][22][23][24][25]. For example, Engelhart et al. noted an incidence of 25.3 HAI per 1000 patient-days at risk [20], and Urrea et al. reported 17.7 HAI per 1000 patient-days among paediatric haematology/oncology patients [21]. To the best of our knowledge, trends in HAI incidence in the general population of adult patients with haematological malignancies have been poorly investigated. The objectives of this prospective surveillance were: 1) to track trends in the incidence of HA pneumonia, bacteraemia and urinary tract infections (UTI) in patients with haematological malignancies; 2) to identify the common etiological pathogens for these infections.

Setting
Prospective surveillance of HA pneumonia, bacteraemia and UTI was implemented at the 42-bed haematology department of Edouard Herriot University Hospital in Lyon, France, between 1 st January 2004 and 31 st December 2010. This 1000-bed hospital is composed of 3 haematology adult acute care units. Each of these units can accommodate 14 patients in single rooms. In total, 16 rooms are equipped with laminar airflow (LAF), 14 rooms with positive pressure isolation and high-efficiency particulate air (HEPA) filtration, and 12 have no specific air system. Patients with acute myeloid leukaemia (AML) or acute lymphoid leukaemia (ALL) were selectively admitted in the department, whereas patients with lymphoma were more often admitted to another hospital.

Data Sources
All patients hospitalized $48 hours in the haematology department between 1 st January 2004 and 31 st December 2010 were included in the surveillance. A standardized data sheet was filled for each patient, based on clinical observation and microbiological results. It recorded demographic and medical data: patient age, gender, dates of admission and discharge, diagnosis of the haematological disease, reason for admission. The main risk factors of infection were also recorded: neutropenia (,500 neutrophils/mm 3 ) and its length, exposure to central venous catheterization (CVC), length of stay, hospitalization in non protected room. Surveillance was based on a local, standardized protocol derived from the French national protocol for HAI surveillance [26].

HAI Diagnosis
The case definition of HA pneumonia was: 1) chest X-rays or chest computed tomography-scans exhibiting lung infiltrates; and 2) temperature .38uC or leukocyte count .12 000/mm 3 or ,4 000/mm 3 ; and 3) at least 1 of the following: (a) sputum change, (b) suggestive auscultation, (c) low oxyhaemoglobin saturation, or (d) increased pulmonary oxygen consumption; and 4) confirmation of pneumonia diagnosis by: (a) the practitioner in charge of the patient; and/or (b) microbial identification of the potential causal microorganism. The etiological agents were isolated from sputum, or directed bronchoalveolar lavage, or protected specimen by telescopic brush or protected distal tracheal specimen by catheter, or histological sample [27].
The case definition of HA bacteraemia was: 1) at least 1 positive blood culture justified by clinical signs for most microorganisms; or 2) at least 2 positive blood cultures justified by clinical signs, at 2 different times, separated by less than 48 hours from each other, when the microorganism was 1 of the following: coagulasenegative Staphylococci, Bacillus spp. (except Bacillus anthracis), Corynebacterium spp., Propionibacterium spp., Micrococcus spp. [27].
The case definition of HA UTI was: 1) at least 1 of the following clinical signs/symptoms: fever (.38uC), urgency, increased urinary frequency, dysuria, burning on urination or pain in the lower abdomen, with no other recognized cause; and 2) a positive urine culture ($10 5 microorganisms/mm 3 ) with 1 or 2 different pathogens isolated in case of urinary catheter exposure in the previous 7 days, or leukocyturia ($10 4 leukocytes/mm 3 ) and positive urine culture ($10 3 microorganisms/mm 3 ) with 1 or 2 different pathogens isolated if any urinary catheter was in place in the previous 7 days [27].
The causative microorganism for each infection was recorded. Only the first HAI per hospital stay that occurred $48 hours after patient admission was considered. Also, patient-days at risk were censored at first HAI, if it occurred.

Statistical Analysis
Discrete variables were described as number and percentage, and continuous variables, as mean and standard deviation (SD). The HAI attack rate was the number of HAI per 100 patients, HAI incidence was the number of HAI per 1000 patients-days of hospitalization at risk. Poisson regression was fitted to assess temporal trends of HA pneumonia, bacteraemia and UTI incidence. The number of HAI was the dependent variable and the number of patient-days at risk was the offset. The independent variable of interest was the year of hospitalization. Potential confounders were: patient age, gender, neutropenia during hospital stay, CVC during hospital stay, diagnosis at admission, and main treatment received during hospitalization. Variables with P,0.15 after univariate analysis were entered in the initial multivariate model; then, a backward step-wise process was initiated, the models were compared with the likelihood ratio test. The significance level was P,0.05; all tests were 2-tailed. Statistical analysis was performed with Stata 10.0 (Stata Corp.). Data were recorded with EpiInfo, version 6.0. All data were analyzed anonymously.

Discussion
The study's objective was to describe trends in HA pneumonia, bacteraemia and UTI incidence between 2004 and 2010 as they were the most frequent and lethal sites of infection in patients with haematological malignancies [17,20,22,28]. No significant trend was found in HA pneumonia and UTI incidence. HA bacteraemia incidence increased by 11% per year independently of neutropenia, CVC and patient's haematological disease. Table 3 compares our results with 3 previous surveillance studies about neutropenic cancer patients [13] and patients with haematological malignancies [19,20].
Engelhart et al. reported data on HAI and fever of unknown origin among adult haematology and oncology patients [20]. They discerned a higher pneumonia incidence (9.7 per 1000 patientdays), while most rooms in their department were not equipped with LAF or positive pressure isolation and HEPA filtration. Consequently, protection against Aspergillus, the most common pathogen in HA pneumonia in this patient group, would be less efficient [29,30]. In our study, the globally-stable incidence of HA pneumonia could be balanced by decreased invasive aspergillosis during the same period [31]. Consequently, it is not clear whether pneumonia of other origins was truly stable. However, we found a high rate of Aspergillus pneumonia (59.2% of the causative agents in pneumonia) although 56.9% of the patients were hospitalized in a protected room. Aspergillosis remained a large concern among our patients and Aspergillus pneumonia occurred mainly in immunosuppressed patients like AML or ALL patients [31,32]. Moreover, in our survey, incidence of HA-pneumonia appeared higher among AML patient compared with ALL patients or other patients.
The incidence of bacteraemia rose during the study period. Bacteraemia are mostly of endogenous origin, caused by pathogens or saprophytic cutaneous flora, such as Staphyloccocus spp. [33]. Over the years, we noted a higher proportion of patients with CVC, well-known to be a major risk factor for HA bloodstream infection [23]. The length and depth of the neutropenic phase, especially during chemotherapy, could also be related to incidence of bacteraemia. Nevertheless, the increased HA bacteraemia incidence was independent of CVC-exposure or neutropenia. Other factors could explain this increase, like more intense chemotherapy over the years, changes in antibiotic prophylaxis or hygiene practices. However, a previous study reported a protective impact of hygiene measures [34]. Then, the increase of incidence might be mostly related to the severity of patients or to their underlying diseases. Carlisle et al. recorded a lower HA bacteraemia rate than what we observed in our department [13]. They implemented 42-month surveillance and detected 13.5 bacteraemia per 100 neutropenic days. However, their study focused on neutropenic cancer patients and not on a specific population hospitalized with haematologic malignancies, the last population could have more severe immune system dysfunction related to their disease and not only chemotherapy. On the other hand, we found a similar incidence of bacteraemia as Engelhart et al. (11.2 per 1000 patient-days at risk) who investigated patients with haematological-oncologic diseases [20].
UTI incidence was stable. We encountered a similar incidence as Carlisle et al. (5.7 per 100 neutropenic patients) [13] and Engelhart et al. (6.0 per 100 patients) [20]. Dettenkofer et al. reported a much lower UTI rate (3.0 per 100 patients, 0.8 per 1,000 patient-days) in transplant patients [19], but all of them received antimicrobial prophylaxis, which could have reduced the risk of infection in prolonged neutropenia cases [35][36][37]. Our study had some limitations. First, the study population was mainly composed of AML and ALL patients. Our findings cannot be generalized to all haematological populations. Secondly, no post-discharge surveillance was undertaken, which could have led to underestimation of the incidence because infections with long incubation periods could have been missed. However, it did not affect trends because possible underestimation could be constant over the years. Furthermore, some potential risk factors for infection were not collected in our surveillance database, like length of CVC exposure, mechanical ventilation, type of stem-cell transplant (autologous or allogenous).
Its main strength was prospective data collection, with standardized case definitions. Ours was a single-centre study, which permitted us to compare trends and reinforced internal validity. Moreover, the true incidence was assessed because all patients were followed until their discharge. Few investigations have analyzed data over such a long time period and in such a large population, which increased the study's power calculation. Finally, major confounding factors were taken into account with multivariate analysis.
In summary, we observed that the incidence of HA pneumonia and HA UTI remained stable. This could be explained by improvement of infection control measures, and took account of patient exposure to more aggressive chemotherapies. The incidence of bacteraemia increased independently of CVC exposure or neutropenia, which are known risk factors of bacteraemia. Further etiological studies are needed to explore in depth the factors that could explain this trend.