Risk Factors for Healthcare-associated Infection in Pediatric Intensive Care Units: a Systematic Review

A systematic review of observational studies on risk factors for healthcare-associated infection in pediatric Intensive Care Units (ICU) was carried out. Studies indexed in MEDLINE, LILACS, Co-chrane, BDENF, CAPES databases published in English, French, Spanish or Portuguese between 1987 and 2006 were included and cross references added. Key words for search were " cross infection " and " Pediatric Intensive Care Units " with others sub-terms included. 11 studies were selected from 419 originally found: four studies had healthcare-associated infection as the main outcome without a specific site; three articles identified factors associated with lower respiratory tract infection (pneumonia or tracheitis); three articles were concerned with laboratory-confirmed bloodstream infection; and a single retrospective study analyzed urinary tract infection. The production of evidence on risk factors Paediatric ICU has not kept up the same pace of that on adult – there are few studies with adequate design and statistical analysis. The meth-odological diversity of the studies did not allow for a summarized measurement of risk factors.


Introduction
Substantial progress has been made in the surveillance of healthcare-associated infections (HAI), previously known as hospital infections.The identification of risk factors enables the development of preventive strategies 1,2 .Although the National Nosocomial Infection Surveillance System (NNIS) was established in 1970 in the United States, it was not until 1987 that Jarvis and colleagues published a study using data from the NNIS, stressing the higher incidence of healthcare-associated infection in children and adolescents as well as differences in the site and microorganisms involved 3 .Another ten years went by before the Centers for Disease Control and Prevention (CDC), in cooperation with the National Association of Children's Hospitals and Related Institutions (NACHRI), established the Pediatric Prevention Network (PPN) with the aim of determining the characteristics of hospital infection as well as developing and testing intervention strategies for reducing the occurrence of these events 4 .
The incidence or prevalence rates of healthcare-associated infections in pediatric Intensive Care Units (ICU) and their sites vary considerably in both individual and multicentric studies, but most authors have progressively adopted the standardized methodology of the NNIS 5,6,7 .Although indicators differed according to the type of adult ICU (cardiothoracic, surgical, medical, oncology, etc.), the NNIS system did not establish a classification for different types of pediatric ICU.Similarly, information reported on surgical patients in pediatric ICUs used to be combined with data on patients in adult ICUs in this surveillance system 8 .In the NNIS system, data were collected using standardized protocols that relied upon selected procedures such as urinary catheter, central line and ventilator utilization.Despite the advances made with this approach, these studies do not allow the establishment of associations other than those with the selected risk procedures.
Studies carried out to establish predictive models for healthcare-associated infection have investigated risk factors, both intrinsic and extrinsic, either separately or in combination.Intrinsic factors generally include age, gender, nutritional status, underlying disease and severity of the illness.To assess disease severity and to predict the risk of death, scores such as the Paediatric Risk of Mortality (PRISM), or its modified PRISM III version, and the Paediatric Index of Mortality (PIM) or its PIM II modification have been used 9,10,11,12,13,14,15 .Extrinsic factors such as aspects of treatment, available structure and the quality of care have been studied as well 16 .These include common invasive procedures (mechanical ventilation, central venous catheter, urinary catheter), the use of medication (antimicrobial agents, immunosuppressors, gastric acid blockers), parenteral nutrition and blood products, for example 14,15 .
A large number of studies have demonstrated associations between different risk factors and healthcare-associated infection in adult ICU.Such studies have been carried out with a certain degree of standardization, which allowed systematic and non-systematic literature reviews to support the production of guidelines 17,18,19,20,21,22,23,24 .However it is inadequate to simply extrapolate data on adult ICU to paediatric ICU, as children are not little adults 25 .Pediatric ICUs are unique entities and require specific prevention and control strategies based on the results of studies carried out locally.
Healthcare-associated infections in paediatric ICUs result in substantial illness, mortality and high costs and requires specific prevention measures.This justifies the present systematic literature review, the aim of which is to describe the variables studied and those identified as risk factors for healthcare-associated infection in pediatric ICUs.

Material and methods
A protocol for the systematic review was designed based on the recommendations of the Meta-analysis of Observational Studies in Epidemiology (MOOSE) study group 26 and the definitions, outcomes and study eligibility criteria were defined a priori.
Observational cohort or case-control studies published between 1987 and 2006 in English, French, Spanish or Portuguese investigating healthcare-associated infection as an outcome in children or adolescents (from one month to 12 years old) in paediatric ICUs were included.Healthcare-associated infection and its specific infection sites were defined by the CDC criteria.There are specific criteria only for children ≤ 12 months 5,6,7 .
The same authors independently assessed the full papers on the basis of a priori eligibility criteria.Reasons for exclusion at this stage included studies that simultaneously analyzed children and adults 32 , those with no multivariate analysis or with no presentation of association measures 33,34,35 , those in which the outcome was death due to healthcare-associated infection 36 or that reported healthcare-associated infection in paediatric patients analyzing "stay in the pediatric ICU" was one of the risk factors 37,38 .Thirteen papers were selected after consensus, but two of these studies were subsequently excluded, either because the results of the study had already been published on the same population and assessed the same outcome, differing only in sample size 41 , or because the authors performed a randomization of the cases for a validation study of the risk of healthcare-associated infection based on a previously published article 42 .
Data from the final 11 selected studies were extracted into a pre-defined, Microsoft Excel v. 8.0  Studies were assessed for quality using the "Newcastle-Ottawa Scale" 111 for cohort studies, which includes evaluation of selection (representativeness of exposed subjects, selection of non-exposed subjects, assessment of exposure and outcome not present at the beginning of the study), comparability and outcome (assessment, sufficient follow-up for the outcome to occur and percentage of losses).This scale is recommended by the Cochrane Non-Randomized Studies Methods Working Group available at the electronic address (http://www.ohri.ca/programs/clinical_epidemiology/oxford.htm).
Although the authors intended to construct a summarized measurement of the association between risk factors and healthcare-associated infection, this was not possible due to the diversity of outcomes and the manner in which the risk factors were defined or categorized.

Description and quality of studies
The 11 selected studies included 10 papers published in peer-reviewed journals and one dissertation.Age range was heterogeneous and it was unclear whether some studies included newborns.The upper age limit found was 18 years.In these studies, the main outcomes were hospital infection in any site 50,51,52,54 , laboratory-confirmed bloodstream infection 55,56,57 , lower respiratory tract infection 58,59,60 and urinary tract infection 49 .All but one article 49 achieved the maximum score of the Newcastle-Ottawa Scale for cohort studies.Table 1 presents the characteristics of the selected studies.

Healthcare-associated infection as outcome
Four selected studies had healthcare-associated infection without a specific site as the main outcome 50,51,52,54 .These reported an incidence of healthcare-associated infection episodes ranging from 10 to 33%.The healthcare-associated infection incidence density (not reported in the study by Singh-Naz et al. 50) ranged from 20.0 to 51.7 per 1,000 patients/days.
Among the intrinsic variables (displayed in Table 2) neither gender nor underlying disease remained associated to healthcare-associated infection in the final model in any of the four studies.Regarding disease severity as assessed by PRISM, Singh-Naz et al. 50and El-Nawawy et al. 52 found a significant association to healthcare-associated infection (the latter used the PRISM III).Singh-Naz et al. 50considered the mean PRISM score of the patients for the analysis, whereas Gilio et al. 51 , who categorized the scores into fourunit intervals, found no significant association.Being a patient in the post-operative period was a risk factor in the analysis of Singh-Naz et al. 50nd Figueiredo 54 .This variable was not significant in the final model of Gilio et al. 51 and was not studied by El-Nawawy et al. 52 .
Among the extrinsic factors, the time from pediatric ICU admission to the development of healthcare-associated infection was found to be significantly associated to healthcare-associated infection in the studies by Singh-Naz et al. 50and Gilio et al. 51 .El-Nawawy et al. 52 , on the other hand, found that healthcare-associated infection was associated to a longer mean length of hospitalization in the pediatric ICU.
The ratio of invasive procedures, defined as the relation between the total utilization time in days of each procedure (central venous catheter, urinary catheter and respirator) and length of stay in the pediatric ICU until healthcare-associated infection diagnosis, was associated to healthcare-associated infection in the studies by Singh-Naz et al. 50and Gilio et al. 51 .In contrast, El-Nawawy et al. 52 and Figueiredo 54 did not find a significant association between invasive procedures and HAI.
Parenteral nutrition was a risk factor for HAI in the studies by Singh-Naz et al. 50, Gilio et al. 51 and Figueiredo 54 .This variable was not studied by El-Nawawy et al. 52 .
Antimicrobial therapy and its duration were analyzed by Singh-Naz et al. 50and Gilio et al. 51 and only in the former a significant association was found.Being admitted to the pediatric ICU being referred from wards other than the emergency room was a risk factor observed by El-Nawawy et al. 52 but was not studied by others.

Bloodstream infection as outcome
Three studies assessed risk factors for bloodstream infection -Laboratory-confirmed bloodstream infection (BSI-LCBI) 55,56,57 , two of them included only patients with central venous access 55,56 .The BSI-LCBI rate (Table 1) ranged from 4.2 to 9.98%.Table 3 displays the set of variables associated to the outcome.Amongst intrinsic factors, underlying disease and age showed no statistical significance, whereas genetic syndrome remained in the model by Elward et al. 57 .Elward et al. 57 studied the use of central venous access as an extrinsic risk factor for bloodstream infection, whereas this variable was an inclusion criterion in the studies by Odetola et al. 55 and Almuneef et al. 56 .The use of multiple catheters and parenteral nutrition were found to be significant risk factors for Almuneef et al. 56 .Only Elward et al. 57 assessed the number of packed red blood cell transfusions and they found that a high number was a significant risk factor.Hemodialysis was studied either separately 57 or associated to blood filtration and no significant association was found with BSI 55 .Odetola et al. 55 were the only authors to examine extracorporeal circulation and determined it to be a risk factor for BSI.The duration of arterial catheter use was a risk factor for BSI in the study by Elward et al. 57 whereas the duration of central catheter use remained in the final model in the Odetola et al. 55 study.Almuneef et al. 56 did not study these variables.Changing the catheter with a guidewire represented a nearly five-fold greater risk for BSI in the study by Almuneef et al. 56 but this was not assessed in the other two studies.

Lower airway infection as outcome
Three studies assessed risk factors for lower airway infection 58,59,60 .The incidence of healthcarerelated pneumonia was 1.2%, 5.4% and 10.2%, respectively (Table 1).The latter two studies assessed ventilator associated pneumonia (VAP).Fayon et al. 58 also studied bacterial tracheitis, which had a cumulative incidence of 1.8%, with an incidence density for grouped bacterial nosocomial tracheitis and bacterial pneumonia of 6.5 per 1,000 days on a ventilator.VAP rates were 11.6 and 8.87 per 1,000 days of ventilation in the studies by Elward et al. 59 and Almuneef et al. 60 , respectively.
Factors related to underlying diseases were associated to bacterial tracheitis 58 (Table 4).With regard to bacterial pneumonia, the same study found an immunodeficiency status to be significant.Genetic syndrome was found to be a risk factor in the study by Elward et al. 59 for VAP.No intrinsic factors were significant in the study by Almuneef et al. 60 .
Regarding extrinsic factors, Fayon et al. 58 found that the use of immunosuppressor drugs or neuromuscular blockers had a 4-fold and 11-fold greater risk for healthcare-associated pneumonia, respectively.In the study by Elward et al. 59 , transporting the patient out of the pediatric ICU was a risk factor for VAP.Almuneef et al. 60 found risks for VAP with prior antibiotic therapy, continuous enteral nutrition and bronchoscopy.

Urinary tract infection as outcome
In a retrospective study, Matlow et al. 49 identified 25 patients who developed 27 episodes of urinary tract infection among 2,832 admissions over 20 months (0.95 per 100 admissions).Heart surgery was identified as a significant adjusted risk factor (odds ratio, OR = 2.67; 95% confidence interval, 95%CI: 1.13-6.32).

Discussion
Systematic reviews of observational studies have limitations that have been well described in the literature.However, such reviews are justified when there is interest in attempting to summarize association measurements between risk factors and illness outcomes that, for ethical rea-  # Singh-Naz et al. 73 -an increase of fi ve days in length of stay was associated to a four-fold greater risk of developing HAI (OR = 4.3; 95%CI: 3.8-4,8); Gilio et al. 51 categorized variable in fi ve-day intervals; ## Yes for both variables; ### > 0 but < 10 days of antimicrobial therapy and a four-point increase on the PRISM score; § Yes for parenteral nutrition and a fi ve-day increase in length of stay; § § Singh-Naz et al. 73 -an increase of four points of the PRISM score was associated to a 1.5-fold greater risk of developing HAI (OR = 1.6; 95%CI: 1.5-1.7); Gilio et al. 51 categorized variable in 4-point intervals; § § § An increase of 1 unit; ‡ Mean duration of hospitalization; ‡ ‡ Dependent variable -time elapsed until fi rst HAI episode; ‡ ‡ ‡ Categorized the number of invasive procedures (2 or less and more than 2).
sons, cannot be analyzed in clinical trials 111,112 .
Although the importance of critically appraising articles is well recognized, there is no consensus on valid indicators for the quality of observational studies 113 .In the analysis of quality proposed by the Newcastle-Ottawa Scale, studies can obtain the maximum score even when presenting a number of sources of heterogeneity and bias.
Factors related to external validity should be pointed out: one of the pediatric ICUs only offered care to children and dependents of military personnel 56,60 ; and one of the Brazilian studies was carried out in a private hospital, the patients of which are not representative of the general community 51 .
Regarding the records of exposure and the detection of outcomes, there was a degree of homogeneity in the NNIS/CDC definitions and procedures for healthcare-associated infection.The opposite occurred for the categorization of exposure levels (intervals of PRISM scores; length of stay in the pediatric ICU or catheter use).Multidisciplinary pediatric ICUs were described as clinical-surgical pediatric ICUs, or the post-operative status was included as the underlying condition in the methodology.Age range was also heterogeneous, and it was unclear whether some studies included newborns and whether the upper age limit was 12 or 18 years.Occupational conditions, turnover and number Table 3 Intrinsic and extrinsic risk factors for bloodstream infection according to authors of the articles selected in the systematic review on "risk factors for healthcareassociated infection (HAI) in pediatric Intensive Care Units (ICU)". of patients admitted in the pediatric ICUs also varied.

Study
Structural factors, such as area and distribution of beds, availability of isolation beds, number of sinks or dispensers of products for hand hygiene and ratio of patients to healthcare workers, were not part of the analysis of the studies selected, nor were factors related to care processes and compliance to infection control protocols.
Among the four studies that analyzed healthcare-associated infection regardless of site of infection, only the Figueiredo study 54 found age to be a risk factor.The other authors did not discuss the association of age to healthcare-associated infection.The heterogeneity of the populations studied, especially in relation to age, hinders a comparison of the results.The currently accepted paradigm that the frequency of healthcareassociated infection in pediatrics is inversely proportional to age (more common among children under 12 3,30 or 24 months 28,114 ) needs to be clarified, despite its biological plausibility: immunological immaturity could account for the higher incidence of healthcare-associated infection among younger patients 115 .
Many pediatric ICUs were clinical-surgical and catered for a large variety of conditions, making it difficult to establish associations between risk and underlying disease or severity of the condition.The post-operative period per se was a risk factor for healthcare-associated infection 50,54 .Severity, as assessed by the PRISM score, was significant in four studies, but differences in the scores employed (PRISM has 14 clinical and laboratory parameters, whereas PRISM III has 17) and in the manner of categorizing the variable did not allow for summarizing how the measurement is associated with healthcare-associated infection.
Among extrinsic factors, the length of stay in the pediatric ICU was found to be the only risk factor for healthcare-associated infection in three studies 50,51,52 .Pediatric ICU stay reflects the severity of the condition, requiring greater care, but also expresses a measure of time in which the patient is potentially exposed to diverse sources of infection.consensus among 3 "blinded" experts (2 intensivists and 1 bacteriologist), using Centers for Disease Control and Prevention criteria and analyzing evocative signs and symptoms (purulent secretions, fever, new consolidation on thorax radiography, etc.)In intubated patients, tracheal aspirations were collected through a tracheal probe or tracheostomy and immediately sent to the laboratory for gram determination and culture.The patient was considered colonized in the presence of normal fl ora of the upper tract or if more than two different bacteria were identifi ed in small quantities in the secretions without any formation of pus (<25 polymorphonuclears per fi eld); ** Congenital (ex: adenosine desaminase defi ciency) or acquired (ex: AIDS); *** Barbiturates in high dose (≥ 30mg/kg/day), cyclosporine, azathioprine and corticosteroids;

Study
# Elward et al. 59 -patient on mechanical ventilation for at least 48 hours, presenting new infi ltration on the thorax radiography 48 hours or more after initiating mechanical ventilation.Additionally, the patient must have 2 or more of the following: fever > 38ºC, leukocytosis (> 12,000) and the appearance of purulent sputum (> 25 leukocytes per fi eld in the GRAM).Associated organisms were designated as those recovered from the tracheal aspiration or bronchoalveolar lavage in patients with respirator-associated pneumonia; ## Moving the patient out of the pediatric ICU for surgery or imaging exams; ### Reported in the study by Almuneef et al. 83 ; the study by Elward et al. 59 only mentions reintubation; § Almuneef et al. 83 -patient on mechanical ventilation for at least 48 hours, presenting new or progressive infi ltrate, consolidation, cavitation or pleural effusion on the thorax radiography.Additionally, the patient must have: appearance of purulent sputum or change in the appearance of phlegm; microorganism isolated in the blood culture not related to other source of infection; isolation of pathogens of specimen obtained from tracheal aspiration, bronchoalveolar lavage or protected pulmonary brushing or biopsy; or histopathologic with evidence of pneumonia in pulmonary biopsy.primarily on time reduction in the use of invasive procedures.The diversity of the categorization of the variable "length of stay in the pediatric ICU" renders the summarization of the association measure unfeasible.It was also not possible to summarize the association measures between healthcare-associated infection and the invasive procedure ratio 50,51 or the use of parenteral nutrition 50,51,54 due to the unavailability of tabled data in the articles.One of the characteristics of healthcare-associated infection in the pediatric ICU is that BSI is the most frequent type of infection 6 .As in adults, BSI in the pediatric ICU is nearly always associated with the use of a central venous catheter 27,30,61 .Except in specialized ICUs for burn victims or newborns weighing less than 1,000g, the NHSN (which replaced the NNIS) reports that it is in the clinical-surgical pediatric ICU that BSI associated to a central venous catheter which has the highest incidences 8 .

Study
Regarding BSI-LCBI, the incidence density observed in the present review was as high or higher than NHSN references for clinical-surgical pediatric ICU 56,57 .This indicator currently includes only BSI associated to a central venous catheter with bacteriological confirmation, whereas, previously, clinical sepsis was included 8 .Without reporting an overall rate, Odetola et al. 55 analyzed an increasing incidence density according to the number of intravascular accesses and characteristics of the catheter type or use, finding rates that ranged from 4.0 to 22.1/1,000 catheterdays.
Regarding risk factors for BSI in the pediatric ICU, the use of an arterial or venous catheter, devices with multiple lumens, changing the catheter with a guidewire, central vein parenteral nutrition, extracorporeal circulation and packed red cell transfusions are risk factors that are potentially modifiable through educative measures.In turn, a genetic syndrome may be a marker of immune defects that impede the defense of the host against microorganisms.However, it should be stressed that in the three studies that analyzed BSI in the pediatric ICU, no common risk factors were found 55,57 .
In the prevention of infection associated to a central venous catheter, the "package" of interventions proposed in a campaign by American hospitals to save 100,000 lives addressed five components: hand hygiene, maximal sterile barrier, skin antisepsis with chlorhexidine, the choice of the subclavian vein as the preferred insertion site for non-tunneled catheters and daily checks with immediate removal if central access is no longer necessary 116 .Based on these recommendations, Lee & Johston 117 published a systematic review seeking the best evidence for handling central venous catheters and the prevention of bloodstream infections associated to catheters in the pediatric population and concluded that the quality and diversity of the articles in relation to the outcome did not allow the establishment of clear guidelines, such as those existing for adults.
The present systematic review, together with that by Lee & Johnston 117 and other narrative reviews, stress the fact that most of the current knowledge on bloodstream infection related to central venous catheters has principally emerged from data collected in adult ICUs 118,119 .The interventions currently recommended regarding the use of central venous catheters are the same from the 100,000 lives campaign 120 .Although the pediatric ICUs have high rates of BSI associated to central venous catheters, the mean percentage of use (approximately 50%) of this procedure is among the lowest when considering the different types of ICU 8,121 .Thus, recommendations regarding the identification of factors other than the frequency of central access per se seem to be particularly necessary.
Pneumonia is among the three main types of infection in the pediatric ICU and, as in adults, is more frequent among patients on mechanical ventilation 4,27,30,31,61 .The diagnosis of VAP, however, is challenging and may explain the differences in the diverse studies concerning the occurrence of this event as well as possible risk factors.Taking into account the difficulty in diagnosing nosocomial pneumonia or tracheitis, Fayon et al. 58 used a consensus between three researchers, two intensivists and one microbiologist.Agreement was good for pneumonia and weak for tracheitis.Recent criteria aimed at improving the specificity of this diagnosis were stipulated by the NNIS for both adults and children 122 , but need to be validated with regard to pediatric patients.The heterogeneity of the three studies on risk factors for lower airway infection impeded the summarization of the measures found 58,59,60 .The outcomes were different: tracheitis and pneumonia, in which endotracheal intubation was one of the risk factors 58 ; and ventilator-associated pneumonia, in which being on a ventilator was a criterion for inclusion 59,60 .The studies analyzing VAP found different independent risk factors and no risk factors that were common to both studies 59,60 .
In a systematic review of intervention studies based on risk factors for VAP, the authors graded the prevention recommendations according to the quality of the studies, but none was based on a pediatric population 123 .Recommendations were extrapolated for children and adolescents: elevation of the head between 30 and 45 degrees, daily verification of sedation and the possibility of extubation, prophylaxis for peptic ulcers and deep venous thrombosis 116 .
Urinary tract infection is the main cause of healthcare-associated infection in adults, but is scarcely studied in the pediatric ICU, despite being considered the third most frequent type of infection.Only two of the 84 studies in the present review discussed risk factors for urinary tract infection and one of these failed to present multivariate analysis 62 .The study by Matlow et al. 49 presents several methodological difficulties for being a retrospective study, including the impossibility of calculating the incidence density in relation to catheter use in the exposed and non-exposed populations, which is the main risk factor reported in the literature.Despite the relatively lesser use of urinary catheters in children in the pediatric ICU (mean use ratio of approximately 0.30), the rate of infection remains high, with a median value of 2.8 per 1,000 urinary cathetersdays 8 , indicating the presence of risk factors likely associated to urinary catheter use.EPIC 2 (Evidence based Practice in Infection Control) recommendations for the prevention of infection associated to the use of short-term urinary catheters only applied to adults and children over the age of one, and a number of reviews cited were of studies on adult patients 24 .
Children have anatomical and functional peculiarities, different underlying conditions and are submitted to different surgical procedures than adults.This justifies the conducting of specific studies on pediatric patients 59 .An important step toward determining specificities and improving the comparison of healthcare-associated infection surveillance in the pediatric ICU was the definition of different types of pediatric ICU considered in the new system adopted by the NHSN 8,124 .A number of reviews have sought to establish a consensus among the different medical societies and governmental surveillance agencies for the standardization of definitions regarding the types of infection -sepsis, pneumonia and urinary tract infection 125,126,127 .
The present systematic review demonstrates the scarcity of studies on risk factors for healthcare-associated infection in children hospitalized in pediatric intensive care units.The following were risk factors for a first episode of healthcare-associated infection: post-operative period, greater severity upon admission based on the PRISM or PRISM III, invasive procedures, the use of antimicrobial agents, parenteral nutrition, remaining more days or more than seven days in the pediatric ICU.However, being sent from the emergency ward is a protective factor.Risk factors identified for BSI-LCBI were: age, genetic syndrome, use of packed red cells, long duration of catheter use, parenteral nutrition through a catheter, change of catheter with a guidewire, multiple central lines and extracorporeal circulation.The development of bacterial pneumonia were associated to the following risk factors: young age, immunodeficiency, neuromuscular blocker and immunosupressor drugs; for tracheitis: age ≤ 28 months, respiratory failure, head trauma; for VAP: genetic syndrome, transport outside of the pediatric ICU, reintubation, bronchoscopy and continuous enteral nutrition.
Risk factors for healthcare-associated infection in PICUs stem from the dynamics of healthcare in child or adolescent units and are specific to these populations.
This systematic review demonstrated that the production of evidence on risk factors in paediatric ICUs has not kept up the same pace of that on adult ICUs.There are few studies on the subject with adequate design and statistical analysis.Furthermore, the methodological diversity of the studies requires a systematic analysis of the scientific evidence produced on this topic.Efforts should be made to obtain data on risk factors utilizing standardized prospective protocols to measure these different factors.Recommendations drafted from evidence produced in pediatric ICUs would be more adequate for contributing to control strategies of these adverse healthcare events, thereby ensuring the safety of children who require intensive care.
Infecção Hospitalar; Unidades de Terapia Intensiva Pediátrica; Assistência à Saúde Contributors M. J. G. Mello accessed the databases; read the selected summaries and complete articles; carried out the data extraction; analyzed the quality of the chosen articles and wrote the article.M. F. P. M. Albuquerque read the selected summaries and complete articles; participated in the data extraction; and helped in the write up.H. R. Lacerda participated in the design of the protocol for data extraction; analyzed the results of the selected articles; and helped in the write up.W. V. Souza analyzed the quality of the chosen articles, according to the "Newcastle-Ottawa Scale" protocol; analyzed the heterogeneity of the results and the possibility of summarizing the measures for associations found in the studies.J. B. Correia participated in the preparation of the protocol for data extraction and the article write up.M. C. A. Britto participated in the preparation of the protocol for data extraction and the article write up.

Figure 1 Flowchart
Figure 1Flowchart of study selection process in the systematic review of risk factors for healthcare-associated infection in pediatric intensive care units.

Table 1
Characteristics of the studies included in the systematic review of observational studies on healthcare-associated infection (HAI) in pediatric Intensive Care Units, published between 1987 and 2006.

Table 2
Intrinsic and extrinsic risk factors for healthcare-associated infection (HAI) in pediatric Intensive Care Units (ICU) identifi ed in the articles selected in the systematic review of observational studies published between 1987 and 2006.

Table 2 (
continued) * Score was estimated based on 14 physiological parameters during the fi rst 24 hours since admission to the pediatric ICU; ** Assessed upon admission: no risk in immune function (absence of the criteria to follow), minor risk (total lymphocytes < 1,500/mm 3 , albumin < 2g% or corticoid use), moderate risk (neutrophil count < 500/mm 3 , AIDS diagnosis), major risk (no granulocyte, bone marrow transplant or anti-neoplasm chemotherapy less than 6 months prior to admission).Maximum state of dysfunction as criterion for major risk; *** Ratio of invasive procedures = [central venous catheter -days] + [urinary catheter -days] + [mechanical ventilation -days]/Length of stay until day of HAI diagnosis;

Table 3 (
59ward et al.59, primary infection that began after 48 or more hours of internment in the pediatric ICU, with the isolation of bacterium or fungus in the blood culture and that were not present in incubation prior to hospital admission.Coagulase-negative Staphylococcus was only considered as an etiological agent if blood culture was positive at least twice on separate occasions; continued) 5682 Cad.Saúde Pública, Rio de Janeiro, 25 Sup 3:S373-S391, 2009 BSI: bloodstream infection; PRISM: Paediatric Risk of Mortality; CRBSI: catheter-related bloodstream infection; ECLS: extracorporeal life support; CRBSI: catheter-related bloodstream infection.***Odetolaetal.55,primarybloodstreaminfection in patient with vascular access at least 48 hours prior to the establishment of the infection; *** In the Odetola et al.55study refers to catheter with double or triple lumen for dialysis or blood fi ltration; # Almuneef et al.56, bacteremia or fungemia in patient with intravascular catheter, with at least one positive blood culture obtained in peripheral vein with clinical manifestation of infection and no apparent source of bacteremia other than the catheter.Coagulase-negative Staphylococcus was only considered as an etiological agent if blood culture was positive at least twice on separate occasions; ## In the Almuneef et al.56study refers to the use of other central venous catheters in comparison to a single central venous catheter.