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


Prevalence of Severe Sepsis

During the study period, 43 participating hospitals reported a total of 636,842 patient admissions to the PICU. Of these, 49,153 patients met PSS definition based on presence of either modified Angus criteria or an ICD-9 sepsis code, for an overall prevalence of PICU PSS of 7.7%. Defining PSS by modified Angus criteria alone identified 39,372 patients (prevalence 6.2%). PSS as defined solely by either ICD-9 code 995.92 or ICD-9 code 785.52 identified 19,776 patients (3.1% prevalence) (Fig. 1). Of the total, 9,995 patients were identified as PSS by both modified Angus criteria and use of at least one ICD-9 sepsis code.

Figure 1.

Patient cohorts with pediatric severe sepsis (PSS) from Pediatric Healthcare Information Systems as identified by various criteria. Total patients identified by three different criteria are noted in legend. Patient numbers identified by overlapping diagnostic criteria are noted within each area.

We performed a limited analysis comparing subgroups of patients with PSS identified by 1) ICD-9 sepsis codes alone (n = 9,781), 2) modified Angus criteria alone (n = 29,377), and 3) both modified Angus criteria/at least one ICD-9 sepsis code (n = 9,995). Children identified by Angus criteria alone or both Angus/ICD-9 criteria were more likely to have comorbidities compared with children identified using only ICD-9 codes (76.5% vs 73.0% vs 66.4%, respectively; p < 0.001). Median LOS was significantly longer in patients identified by modified Angus criteria (median, 18 d) and combined criteria (20 d) compared with those identified by ICD-9 code criteria alone (13 d) (p < 0.001 for both comparisons). Rates of extracorporeal membrane oxygenation (ECMO) and continuous renal replacement therapy (CRRT) utilization were higher in patients identified by both Angus/ICD-9 criteria compared with Angus criteria and ICD-9 criteria alone (ECMO, 6.0%, 3.6%, and 2.5%, respectively; p < 0.001; CRRT, 10.2%, 3.6%, and 1.9%, respectively; p < 0.001). Patients with PSS identified by both modified Angus and ICD-9 criteria had significantly higher mortality (26.4%) compared with patients identified only by modified Angus criteria (11.2%) or only by ICD-9 sepsis codes (11.7%) (p < 0.001). No other clinically significant differences in patient characteristics were seen between PSS definition subgroups.

Further analysis was performed on the total PSS cohort (n = 49,153).

Clinical Characteristics

PSS clinical characteristics are noted in Table 1. Children aged 1–4 years made up the largest group (24.8%) in the cohort, followed by children younger than 1 year (23.6%). Median ICU LOS was 7 days (IQR, 2–17 d). Median hospital LOS was 17 days (IQR, 8–36 d).

Seventy-four percent of patients had one or more underlying comorbidities, of which cardiovascular disease was the most common (26.6%). Patients with PSS and with at least one comorbidity increased steadily from 64.9% in 2002 to 76.6% in 2012 (p < 0.001). Most patients had at least one organ dysfunction (91.1%), and over 53% of patients with PSS had at least two organ dysfunctions.

Sites and selected organisms associated with infection are described in Table 2. A presumed site of infection based on ICD-9 code description was noted in 91.5% of patients with PSS, with two or more sites of infection noted in 60%. Most common sites noted were bloodstream (67.8%), respiratory tract (57.2%), and genitourinary system (21.6%). The most common causative organisms noted were Staphylococcus species (9.9%) and Streptococcus species (5.4%). The prevalence in the percentage of PSS patients with influenza increased from 2008 (2.9%) to 2009 (7.8%), with prevalence returning to 2.1% in 2010. This was concurrent with a rise in the PSS prevalence rate in 2009 (Fig. 2A). No statistically significant trends of infection rate were seen in other organisms. Bacterial or fungal coinfection was noted in 16% of patients with a documented organism.

Figure 2.

A, Change in overall prevalence and mortality rates of children with pediatric severe sepsis (PSS). Data are from 33 hospitals for which continuous data were available for the time period. Prevalence rates significantly increased (p < 0.001) and mortality rates significantly decreased (p < 0.001) over time. B, Hospitalspecific prevalence and mortality for pediatric patients with severe sepsis. Diamonds represent prevalence, triangles represent 95% CIs, and circles represent mortality rates. There is significant negative correlation between prevalence and mortality (r s = –0.48; 95% CI, –0.70 to –0.15; p = 0.005).

Patient Survival

During the time period, 20,655 PICU deaths were reported (PICU mortality rate 3.2%). Of patients meeting PSS criteria during PICU admission, 7,074 patients died, for a mortality rate in PSS patients of 14.4% (95% CI, 14.1–14.7%). Of total PICU deaths during the time period, 34.2% of patients had an associated PSS diagnosis. Mortality was highest in children less than 1 year (19.2%), followed by children 1–4 years old (13.8%). Children with a at least one comorbidity had a higher mortality rate (15.8%) than children without any reported comorbidities (10.4%) (p < 0.001). Among patients with PSS with specific comorbidities, mortality was highest in patients with malignancies (22.4%), hematological or immunological disorders (20.3%), and cardiovascular disease (20.0%). Mortality rate in patients with a specific reported bacterial organism identified was 13.2%. Although fungal infection alone was rare (< 1%), PSS mortality was high in this group (20.1%) relative to the overall mortality rate.

Hierarchical logistic modeling identified several patient characteristics associated with increased odds of mortality (Table 3). Patient age was associated with mortality; odds of death in children younger than or 1 year old was 1.4 times higher than for children 10–19 years old (p < 0.001). Odds of mortality increased with increasing number of organs with dysfunction and was highest in children with five or more systems affected (odds ratio [OR] = 20.1; 95% CI, 16.2–25.1; p < 0.001). After adjusting for age and organ dysfunction, children with malignancies had greater odds of mortality compared with children without malignancies (OR = 1.9; 95% CI, 1.8–2.1; p < 0.001). Similarly, children with hematological disorders (OR = 1.5; 95% CI, 1.4–1.6; p < 0.001) and cardiovascular conditions (OR = 1.4; 95% CI, 1.3–1.5; p < 0.001) had increased odds of mortality.

Median individual PSS patient total hospital cost, adjusted for inflation, was $77,446 (IQR, $32,545–$183,458) with a median daily cost of $4,516 (IQR, $3,251–$6,442).

Trends and Correlations

Thirty-three hospitals reported data over the complete 9-year period. Trend analysis on yearly prevalence rates, mortality, and hospitalization costs was performed on this cohort, with 535,184 PICU admissions and 40,747 cases of PSS (7.6% prevalence) identified.

During the cohort period, PICU admissions increased from 2004 (n = 54,681) to 2012 (n = 63,569). Similarly, the number of PSS admissions also increased, from 3,408 to 4,889 as did the proportion of PICU admissions with a diagnosis of PSS (6.2–7.7%). The highest annual PSS prevalence rate occurred in 2009 (8.4%), and the largest year-over-year increase (0.7%) occurred from 2006 to 2007. In contrast to the rising PSS prevalence rate, overall PSS mortality in this group significantly decreased from 2004 (18.9%) to 2012 (12.0%) (Fig. 2A) (p < 0.001). The proportion of patients with PSS and multiple organ dysfunction (≥ 2 organs) slightly decreased over the time period (56.5–51.3%; p < 0.001), whereas the proportion of PSS patients with associated comorbidities increased over the time period, from 69.4% in 2004 to 76.6% in 2012 (p < 0.001).

Median total hospitalization cost, adjusted for inflation, decreased over the time period (from $86,156 in 2004 to $72,308 in 2012; p = 0.002). There were significant differences in the cost of PSS among hospitals (p < 0.001) (Fig. 3). Even after stratifying by LOS quartiles and by mortality, PSS cost still differed significantly between hospitals (p < 0.001). Estimated total cost of hospitalization was significantly correlated with LOS (r s = 0.90; p < 0.001). Median inflation-adjusted PSS cost/day also significantly decreased over the time period ($4,563–$4,413; p = 0.002). Median cost of hospitalization was significantly higher in patients who died (died vs survived, $142,501 vs $72,936; p < 0.001) despite no significant difference in median LOS between nonsurvivors (median 17 d) and survivors (17 d). However, there was a significant difference in LOS at the 75th percentile between survivors and nonsurvivors (35 vs 47 d, respectively; p < 0.001). In these 33 hospitals, the total burden of hospital cost for patients with severe sepsis from 2004 to 2012 was $6,575,327,932.

Figure 3.

Median total hospitalization cost of patients with pediatric severe sepsis by hospital. Diamonds represent aggregate median costs from 2004 to 2012 after inflation, and triangles are 25th and 75th percentiles. Dashed line represents overall median cost ($77,598). Hospital number correlates with Figure 2.

Individual institutional PSS prevalence and associated mortality rate varied significantly between the 33 hospitals (Fig. 2B). Individual center PSS prevalence ranged from 4.5% to 13.1%. Mortality ranged from 3.9% to 23.0%. Aggregated data from 2004 to 2012 for each of the 33 hospitals demonstrated a significant inverse correlation between individual hospital PSS prevalence and PSS mortality rate (r s = –0.48; 95% CI, –0.70 to –0.15; p = 0.005) (Fig. 2B). Analysis of individual hospital PSS patient volume revealed a similar negative correlation to mortality (r s = –0.38; 95% CI, –0.64 to –0.05; p = 0.02). Aggregated relationships among prevalence, mortality rates, and cost are demonstrated in Figure 4A. Higher individual hospital PSS cost was also positively correlated with higher mortality rates (r s = 0.36; 95% CI, 0.02–0.63; p = 0.04) (Fig. 4B). No correlation was seen between hospital PSS prevalence and PSS cost (–0.004; 95% CI, –0.346 to 0.339; p = 0.93).

Figure 4.

A, Association of individual center pediatric severe sepsis (PSS) prevalence, mortality, and cost. Horizontal axis represents median hospital prevalence, vertical axis represents median hospital mortality, and bubble size represents relative center median cost. Individual center prevalence was negatively correlated with mortality (r s = –0.48; 95% CI, –0.70 to –0.15; p = 0.005). B, Association of individual center PSS cost and mortality. Horizontal axis represents median hospital mortality, vertical axis represents median cost, and bubble size represents relative center prevalence. Higher cost was positively correlated with higher mortality rates ((r s = 0.36; 95% CI, 0.02–0.63; p = 0.04). PHIS = Pediatric Healthcare Information Systems.