Pediatric Severe Sepsis: Current Trends and Outcomes From the Pediatric Health Information Systems Database

Amanda Ruth, MD; Courtney E. McCracken, PhD; James D. Fortenberry, MD, MCCM; Matthew Hall, PhD; Harold K. Simon, MD, MBA; Kiran B. Hebbar, MD, FCCM


Pediatr Crit Care Med. 2014;15(9):828-838. 

In This Article


This review represents the largest reported cohort analysis of PICU patients with severe sepsis to date. The overall prevalence of PSS in PICU patients was 7%, and the yearly prevalence rates increased over the study period. This rise is consistent with the noted increase in prevalence of patient comorbidities. In spite of rising comorbidities, a concomitant 37% reduction in annual PSS mortality rate occurred over time. Unlike previous pediatric sepsis reviews, the 1- to 4-year-old age group represented the largest proportion of our cohort (24.8%). This analysis focused on PICU patients with PSS, excluding patients who were in the NICU or who had been in the NICU during the same hospitalization. However, this study included former NICU patients who had been discharged and required readmission to a PICU. This subgroup of former preterm infants remains at a higher risk of infection during the first year of life.[23,24] Consistent with previous data analyses, younger age, presence of comorbidities, and multiple organ dysfunction conferred increased mortality risk in PSS. Mortality was highest in infants less than 1 year. Neonatal comorbid conditions such as chronic lung disease and delayed recognition and treatment of septic shock could account for increased relative risk of death in infants.

Subgroup analysis did demonstrate that patients meeting PSS definition based on both modified Angus criteria definition and at least one ICD-9 sepsis code identified a smaller subgroup with significantly worse severity of illness and higher mortality. We chose to perform analysis of the broader category of patients with PSS to better capture the breadth of severe sepsis in PICU patients. Further analysis of these case definition subgroups is outside the scope of this study, but focus on the combined subgroup could be useful for more specific studies on high-risk patients.

Staphylococcus species were the most commonly coded causative organisms for PSS. The prevalence of methicillin-resistant Staphylococcus aureus (MRSA) was low; however, specific ICD-9 coding for MRSA did not exist until 2009, making determination of true prevalence difficult. Prevalence of Haemophilus influenzae and Streptococcus pneumoniae infection was low over the time period.

This analysis differs from previous epidemiologic studies of pediatric sepsis in several aspects. Previous studies evaluated sepsis prevalence through state discharge data encompassing all hospital systems and NICUs. The focus of the PHIS database review was to provide specific assessment of prevalence and characteristics of PSS in a large cohort of nonneonatal pediatric patients from dedicated children's hospitals. These hospitals likely have a higher proportion of patients with comorbidities and more severe illness than other series utilizing broad state discharge data.[3,4] Thus, the higher prevalence of PSS (7%) in the current study is difficult to compare to that of other pediatric database reviews. Previous studies based on nonfederal state data have also included NICU patients. We elected to specifically evaluate sepsis only in patients admitted to a PICU. Further review of neonatal sepsis data from PHIS is in progress. Finally, this review evaluates the most recent data possible from up to 2012, compared with previous data only available to 2005.

The nature of this review does not allow determination of specific causes for decreased mortality over time. Possible factors could include better overall recognition of sepsis, improved institution of goal-directed therapy,[7] and utilization of Surviving Sepsis Campaign Guidelines and the ACCM pediatric sepsis guidelines.[13,25] Given the increase in comorbidities that might have been expected to increase mortality, the actual decrease in death rates is both encouraging and suggestive of the benefits of the above changes. Of note, implementation of ACCM sepsis guidelines in adults was associated with significant reduction in mortality.[26] Additional explanations for apparent increases in PSS prevalence rates over time could include increasing use of sepsis codes for less critically ill children than in previous years. Increased PICU bed availability and/or lowered admission thresholds with resultant increases in less severely ill ICU pediatric patients with sepsis could also potentially impact PSS prevalence rates. Of note, the proportion of patients with PSS and multiple organ dysfunction did demonstrate a slight decrease. This could be related to either increased recognition of PSS or improvements in recognition and care to lower organ failure progress. However, comorbidities (generally associated with higher risk of mortality) significantly increased through the period as well. These findings provide potentially conflicting supporting evidence and emphasize the potential value of more specific severity of illness scoring in administrative databases to allow for more consistent comparisons.

We found significant variability in PSS prevalence, outcomes, and costs between hospitals. Notably a significant negative correlation between hospital PSS prevalence and mortality was also seen between center volume and mortality. Reasons for this volume/outcome relationship are uncertain but could be related to experience in sepsis management, available resources, or other undetermined factors. Higher PSS cost was associated with higher mortality, suggesting that greater resource utilization was not associated with improved outcomes. Of note, the 75th percentile for LOS in nonsurvivors was much higher than survivors (47 vs 35 d), suggesting higher intensity of care and possibly explaining the higher overall costs of care associated with nonsurvival.

Finally, this review underscores the financial burden of PSS, as demonstrated by a cumulative annual cost of over 730 million dollars for the 33 selected children's hospitals alone.


This analysis is limited by the inherent features of an administrative database. Data to allow calculation of severity of illness measures such as Pediatric Risk of Mortality III scores are not available, nor are some therapeutic interventions such as specific antibiotic or fluid use. Site of infection and causative organisms were determined based on coding and not on evaluation of actual culture reports, potentially leading to incomplete coding or underestimation of contributing organisms. Differences in coding approaches could account for the higher prevalence of identified sites and organisms compared with studies using state health database reporting.[4] Other variables such as vaccination status, race, and birth weight were not consistently captured and thus were excluded in analysis. Efforts to provide shared data use between administrative databases and clinically driven pediatric critical care databases, such as the Virtual PICU database registry,[27] could provide desirable synergy for future work in sepsis and other disease processes.

The PHIS database also captures patients in tertiary and quaternary care children's hospitals. It is possible that hospitals contributing to this database have greater resources that led to better pediatric outcomes than centers without full pediatric services, limiting the ability to generalize to all facilities caring for pediatric patients. Although robust in nature, patients from institutions in the PHIS database only account for 15% of all U.S. pediatric hospitalizations, making generalization similarly limited. However, it is likely that this database represents a higher overall complexity and severity of illness than that of non-CHA institutions.