Which Factors Predict Hospital-Acquired Late-Onset Neonatal Sepsis?

James W. Gray


Pediatr Health. 2008;2(4):477-484. 

In This Article

Predicting Risk Factors for LOS

The identification and assessment of risk factors for neonatal sepsis is complicated. The first consideration is the difficulty in diagnosing sepsis in a group of patients whose clinical response to infection may be different to other age groups,[11] and in which the common pathogens are also common contaminants. Even measurement of inflammatory markers, such as C-reactive protein, may not help in interpreting the significance of single blood-culture isolates of equivocal pathogens, such as coagulase-negative staphylococci. Bloodstream infections may be over-diagnosed where techniques for collecting blood cultures or interpreting positive culture results are suboptimal or under-diagnosed where inadequate sample volumes are collected.[1] Attempts to standardize blood-culture practices across centers have not always been unequivocally successful.[8] The limitations of conventional international definitions, such as the CDC definitions for sepsis in children aged less than 1 year, have recently been highlighted.[12] Thus, studies of neonatal sepsis may not be comparable and, in particular, there may be a tendency to under-diagnose neonatal sepsis.

The next consideration is to determine the relative importance of different risk factors. Potential risk factors for infection may be linked, meaning that multivariate analysis is necessary to demonstrate independent risk factors for infection.[9] In addition, one needs to consider the fact that presence of risk factors will vary from day-to-day in the same neonate, and even fixed intrinsic risk factors may have different influences at different postnatal ages. Statistical packages that allow adjustment for the presence or absence of risk factors on a day-by-day basis can be used, albeit that collection and analysis of these additional data will add to the complexity and costs of a surveillance program. However, that still leaves the question of how to equate the timings of changes in risk factors with the incubation period for infection. Incubation periods themselves may be quite variable, depending on factors such as the site of infection and the abundance and virulence of the infecting microorganism. Holmes et al. assessed risk factors daily,[9] and deemed them applicable if present during any of the 3 days prior to an infection being diagnosed, whereas others have only considered risk factors present in the 24 h preceding diagnosis of infection.[12] Further studies are required to determine how the daily presence of risk factors should be analyzed. Another consideration is that there is evidence that the rate of sepsis increases with the number of risk factors present. Holmes et al. found that the rate of neonatal bloodstream infection was 40-times higher in infants with the two independent risk factors that they identified, compared with those with neither.[9] Again, further work is required to understand how this observation might affect risk-factor stratification.

Many studies of risk factors for neonatal sepsis have been confined to a single NICU or small numbers of units,[9,12,13,14,15,16] which makes it difficult to assess whether or not the findings in individual studies are generally applicable. Comparisons can be further complicated by methodological differences between studies. Some studies have considered risk factors for sepsis of any cause,[7,9,10,13] whilst others have focused on risk factors for infection with specific microorganisms.[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32] Several studies have investigated endemic levels of infection on NICUs, whereas others have been undertaken in response to outbreaks of infection.[32] The determination of risk factors for all-cause sepsis is the most useful approach, where the aim of surveillance is to compare intra- or inter-unit infection rates, since the predominant microbial flora in units varies. The only exception to this might be coagulase-negative staphylococci, which account for the majority of blood culture isolates in NICUs.[1,22,23] Infection rates with these bacteria might, therefore, represent a reasonable proxy for overall infection rates. However, such an approach might mask important differences in rates of infection with less common, but more virulent, pathogens. Studies of risk factors for sepsis in outbreak settings can be useful to other units faced with controlling outbreaks of infection with the same or similar microorganisms.

Risk factors for LOS fall into three main categories:

  • Intrinsic risk factors that can not be influenced, and relate to the patient's biological status. Some such risk factors are static, for example, birthweight, gestational age and condition at birth, whereas others, for example, postnatal age, vary daily;

  • Extrinsic risk factors that relate to treatment that the individual patient receives. These risk factors may vary from day to day;

  • Risk factors that relate to infrastructural considerations, such as access to handwashing and isolation facilities, environmental cleanliness and staffing numbers. These consist of a mixture of static and variable elements.

Intrinsic Risk Factors for LOS

It is universally agreed that the incidence of LOS is inversely proportional to birthweight and gestational age.[3,12] For example, incidences of sepsis in infants of birthweight below 750g have been reported in 50% if these infants, compared with 10% in those weighing 1000-1500g.[3,12] However, it is incorrect to assume that the relationship between birthweight and gestational age is linear; in a recent UK study, univariate regressions demonstrated both birth gestational age and birthweight to be risk factors for bloodstream infection, but only a gestational age of less than 26 weeks was an independent risk factor (incidence rate ratio [IRR]: 2.5; 95% CI:1.7-3.8; p<0.001).[9] Other workers have also found that birthweight is not an independent risk factor for LOS.[13] Nevertheless, Holmes et al. suggested that because birthweight is easier to capture accurately, it may be an acceptable alternative.[9]

Postnatal age has also been shown to be a significant risk factor for LOS, although there are differences in experiences. Holmes et al. found that the incidence of bloodstream infection was highest in the first 10 days of life (IRR: 4.9),[9] followed by between the age of 11 and 28days (IRR:2.5), whereas other authors have found the peak infection rate to occur beyond 10 days of age.[3] However, there is general agreement that infection rates fall after 30 days of age.[3] Differences in extrinsic risk factors about the intensity of care are likely to have a significant contribution to the infection rates at different postnatal ages and, therefore, it is not surprising that postgestational age has often been found to not be an independent risk factor for LOS.

With any intrinsic risk factor for LOS, there remains some arbitrariness over which age or weight thresholds should be used for stratification.

Extrinsic Risk Factors for LOS

These are important risk factors because as well as having a potential impact on risk stratification, they may be modifiable. There have been many studies in this area that have identified several interventions which are common on NICUs as potential risk factors for sepsis. However, not all studies have identified strong risk factors that would allow interventions to reduce infection rates.[12]

Ventilatory support and use of central venous devices are the two most common invasive-treatment modalities used in NICUs. A substantial proportion of infections associated with these devices may be preventable through general good practice, such as high hygiene standards whilst devices are insitu, and removal of devices as soon as they are no longer clinically required.[33,34] However, neonates with these devices are often sicker and, thus, predisposed to infections. Not surprisingly, many studies have found that assisted ventilation and the presence of central venous devices are risk factors for LOS,[13,15,16,20,35,36] whilst others have found the duration of use of such devices to contribute to the risk of nosocomial infection.[13] However, not all studies have identified these therapeutic modalities as independent risk factors for LOS,[9] and it may be that differences in the pattern of device-related infections partly reflect differences in local patient populations. It is important to understand whether use of invasive devices are truly independent risk factors for infection, because stratification for risk factors that could be influenced by local standards of care might mask important remediable shortcomings in care.

Parenteral nutrition (PN) is increasingly being used, and often at an earlier age, for preterm neonates before full enteral feeding can be administered.[12] Unlike the presence of central venous devices, there is almost universal agreement that the administration of PN is a major independent risk factor for neonatal sepsis. Several possible physiological reasons for this association have been suggested.[9,12] Holmes et al. reported a very strong link between administration of PN and bloodstream infection, with an IRR of 14.2.[9] Other studies have also highlighted the importance of PN, reporting relative risks for LOS of 4.0-5.7.[12,35,37]

Risk factors related to the GI tract have also been investigated. A systematic review concluded that there was insufficient evidence of human milk in preventing infection in preterm, very low birthweight babies.[38] In a more recent study, the number of days without establishment of full enteral feeding with human milk was the most important influential risk factor for LOS, with an adjusted relative risk of 3.7 (2.0-6.9) for LOS, if feeding was not established within the second week of life.[39]

There is growing interest in the role of drug treatments as contributors to the risk of LOS. The drugs that have most often been implicated include drugs to prevent acid production (especially the H2-receptor antagonists such as ranitidine),[40,41] corticosteroids[42] and antibiotics.[12] These are drugs where evidence for the benefits of their use are not always clear. Ranitidine has been found to be associated with a sevenfold increased risk of LOS (95% CI: 3.78-12.94).[40] Stoll et al. found that a 14-day course of dexamethasone treatment was associated with an increased risk of sepsis.[42] Prior exposure to broad-spectrum antibiotics has mainly been identified as a risk factor for sepsis in general,[12] and especially for sepsis with Gram-negative bacteria[20] and candida.[17,19] It is unlikely that it would be appropriate to stratify for these therapies in intercenter comparisons, because there are no widely followed protocols for their use. Nevertheless these associations are important, because they may help to explain some of the differences in standardized sepsis rates between units.[3]

Infrastructure-related Risk Factors for LOS

A small number of studies have examined the relationship between neonatal sepsis and organizational and structural factors on the NICU. Cimiotti et al. investigated the impact of staffing on infection rates.[43] After controlling for neonates' intrinsic and extrinsic risk factors and bloodstream infections rates, they found that nosocomial bloodstream infection rates were inversely related to the number of hours of care provided by registered nurses in one of the two NICUs studied. Other authors have reported outbreaks of infection with specific microorganisms attributed to understaffing, overcrowding and movement of infants between rooms.[44,45] A study by the UK Neonatal Staffing Study Group found that the incidence of nosocomial bloodstream infection was reduced in units with a dedicated infection control nurse and where there were adequate handwashing facilities, whereas sepsis rates were unaffected by the floor area of the unit.[46] One of the strengths of this study was the large number of units surveyed, but there were several opportunities for bias in that no allowance was made for difference in intrinsic and extrinsic risk factors between units, and no attempt was made to identify blood-culture isolates that were significant from those that were contaminants. Few definite conclusions can be drawn from the present studies of these aspects of care, but it is important to establish the importance of infrastructural risk factors to inform the cost-effective design and operation of NICUs.

Risk Factors for Specific Infections

A number of generally small and single-center studies have investigated risk factors for specific infections. These studies are of little or no value in benchmarking infection rates in different units, but may be helpful in directing infection prevention and control measures where such bacteria are unusually prevalent in a particular NICU.

Despite the fact that staphylococci are the most common cause of neonatal sepsis, there have been few good quality studies specifically examining risk factors for infections with either coagulase-negative staphylococci[23] or Staphylococcus aureus.[22,47] Gram-negative bacteria are less common, but often a more serious causes of neonatal sepsis. Moreover, without control measures, particular species of Gram-negative bacteria can become established in NICUs, possibly adding to the overall burden of nosocomial infections. Graham et al. reported a high level of genetic concordance between Gram-negative bacteria in paired blood cultures and GI tract isolates from low birthweight babies, suggesting that the GI tract is the main focus of Gram-negative sepsis in this population.[21] This would be consistent with our own observation that respiratory tract and cutaneous surveillance cultures are of limited value in predicting bloodstream infections with Gram-negative bacteria.[48] In another study, Graham et al. found that risk factors for Gram-negative sepsis overlapped those for sepsis in general, with a greater contribution from GI tract-related risk factors[20]. Others have investigated risk factors for sepsis with single species of Gram-negative bacteria, and have achieved similar findings.[18,32] Perhaps surprisingly, antibiotic exposure was not found to be a risk factor for colonization with antibiotic-resistant Enterobacteriaceae.[49]

Neonatal candidal sepsis is associated with significant morbidity and mortality. The incidence of candida sepsis is increasing in NICUs, although in most centers it remains much less common than bacterial sepsis. A large number of studies have investigated risk factors for candida sepsis, and again considerable overlap with risk factors for all-cause LOS has been found. However, there does appear to be a particular association with previous exposure to broad-spectrum antibiotics, and it has been suggested that differences in antibiotic-prescribing practices may be a major contributor to the wide variation in the incidence of candida sepsis. Cotton et al. noted that the incidence of invasive candidiasis in very low birthweight babies in 12 centers ranged from 2.4 to 20.4%.[19] However, there may be other explanations for differences in the incidence of neonatal candidiasis, including differences between NICUs in the use of antifungals as prophylaxis or empiric treatment.[50]


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.