Characteristics and Outcomes of Sepsis Presenting in Inpatient Pediatric Settings Free

This retrospective cohort study consisted of cases from the IPSO sepsis data registry between January 1, 2017 and December 31, 2019 that fulfilled IPSO sepsis criteria. The 46 hospitals in the IPSO collaborative in this study had a median of 222 pediatric beds (interquartile range [IQR] 117–341), with 8 (17.4%) having <100 beds, 22 (47.8%) having 100 to 299 beds, and 16 (34.8%) having ≥300 beds. Among the 46 hospitals, 22 (47.8%) were freestanding children’s hospitals and 21 (45.7%) were children’s hospitals within adult hospitals. Most (n = 33, 72%) had a separate pediatric oncology unit; 13 (28%) did not. Most hospitals cared for and contributed data for patients >18 years of age; 9 (20%) did not contribute data on these patients. The hospitals had been participating in the collaborative for a median of 43 months at the time of data collection.

The IPSO collaborative developed sepsis definitions adapted from the previously published Goldstein definitions for the purposes of quality improvement (QI) and for ease of extraction from the electronic medical record.^15–18  Patients fulfilled IPSO sepsis criteria in several ways (see Fig 1) which would include recognition and treatment elements. Further details of IPSO sepsis definitions and metrics are described in our previous works.^15,16,19  A positive sepsis huddle result was defined by each individual hospital but generally was described as a team-based discussion that reviewed the clinical findings and determined that severe sepsis was either present or developing. Treatment criteria included the receipt of an intravenous antibiotic and 2 intravenous fluid boluses or 1 bolus and a vasopressor, all within 6 hours, in addition to a blood culture within 72 hours. Smaller bolus volumes were included to capture patients who may have underlying illnesses which limit their fluid tolerance. A vasopressor can be any new initiation of a vasoactive agent in the specified time frame. A smaller proportion of the sepsis cohort fulfilled the criteria for IPSO critical sepsis, defined by the receipt of a third intravenous fluid bolus or vasopressor, and met the treatment criteria (Fig 1).

Functional time zero (FTZ), used to represent the time of sepsis recognition (see Supplemental Information) was defined as the time of 1 of 5 events: (1) time of positive sepsis screen result, (2) time of positive sepsis huddle result, or (3) sepsis order set if sepsis screen not used or (4) time of intravenous antibiotic or (5) fluid bolus initiation if none of the previous were used.¹⁵  Individual hospitals developed and trained their teams to use sepsis screens that were shared within the QI collaborative; however, not all hospitals shared similar electronic medical records or screening capabilities.

The IPSO sepsis registry included 41 938 submitted cases. Cases were excluded from the dataset if FTZ could not be determined (n = 889), data from the submitting hospital were incomplete (n = 4042),¹⁵  the date of birth was missing (n = 1775), or FTZ occurred in a care setting that was not the ED or inpatient (n = 8377). The final cohort for this analysis included 26 855 sepsis cases from 46 hospitals (see Supplemental Information). Patients with sepsis identified in the ED and then admitted to inpatient or intensive care were included in the ED group only. Patients who had sepsis identified in the inpatient unit and were then transferred to intensive care were included in the inpatient group. Patients with multiple episodes of sepsis in 1 hospitalization had their first episode of sepsis treatment reported to the data registry. Patients with sepsis first identified in intensive care or a dedicated oncology unit were excluded.

Clinical characteristics included age, comorbid conditions, admission type, and proportion meeting criteria for critical sepsis. Comorbid conditions are outlined in Table 1.¹⁹  Patients with FTZ occurring in the ED setting were included in the ED cohort, and patients with FTZ occurring in the inpatient unit were included in the inpatient cohort. The admission source was reported as either admitted from an outside hospital, admitted via ED, both, or unknown. Cases that were both from an outside facility and via ED came from an outside hospital to the IPSO hospital’s ED. Transfers from an outside hospital were only included if they occurred in the 24 hours before FTZ. Process measures included sepsis screen activations, sepsis huddle activations, order set use, time to first antibiotic, and time to first bolus. Time from hospital arrival to FTZ was assessed for the inpatient and ED cohorts. Clinical outcomes included 30-day sepsis-attributable mortality, all-cause mortality, hospital days after FTZ, vasopressor days after FTZ, ICU days after FTZ, and hypotension in the first 24 hours after FTZ. Each hospital was asked to use what their institution defined as clinical hypotension. Sepsis episode reporting and, therefore, hospital days were limited to a maximum of 30 days after FTZ. Readmissions for other causes were not reported. Rehospitalization with another episode of sepsis after hospital discharge was reported as a second episode. Reported mortalities were reviewed and assessed as sepsis-attributable or not sepsis-attributable by each hospital.¹⁹ 

We evaluated the inpatient cohort for 30-day sepsis-attributable mortality associated with compliance to a treatment bundle consisting of (1) the use of any sepsis recognition method (screen, huddle, or order set), (2) delivery of a fluid bolus within 60 minutes, and (3) delivery of an antibiotic within 180 minutes. Outcomes were not risk-adjusted because of infrequent reporting of risk of mortality scores, such as the pediatric index of mortality score and the pediatric risk of mortality score.¹⁵ 

Clinical characteristics, process measures, and outcomes were compared between the inpatient and ED cohorts by using χ² tests for categorical variables and Wilcoxon rank tests for continuous variables. A secondary analysis of inpatient cases was completed by using a log-linear model to identify factors associated with geometric mean time in hours between arrival and FTZ. Covariates in the log-linear model included age group, presence of comorbid conditions, transfer status, and random hospital effect to account for the clustering of sepsis episodes receiving care at the same hospital. For each covariate, comparisons to a reference group were made by using a ratio of geometric means with 95% confidence intervals (CI). Finally, multivariable generalized linear mixed modeling of the inpatient population was done by using 30-day sepsis-attributable mortality as the outcome measure and bundle compliance as the primary exposure. The model adjusted for age, high-risk conditions, IPSO-critical sepsis, time of sepsis episode in the months after the start of the study period, and clustering of episodes within a hospital. Analyses were completed by using SAS 9.4 (SAS, Cary, NC) with P <.05 considered statistically significant.

All participating hospitals signed a participation agreement for the transfer of limited data sets, and institutional review board approval was obtained where required.¹⁹  The primary author’s university institutional review board deemed the analysis of the limited dataset for the current study exempt from review.

The sepsis cohort after exclusions included 26 855 cases from 46 hospitals, with 2254 (8.4%) having FTZ in the inpatient unit and 24 601 (91.6%) having FTZ in the ED. Table 1 reveals characteristics of the population by setting and Table 2 reveals sepsis process measures and outcomes by setting. Compared with ED patients, inpatients had significantly greater 30-day sepsis-attributable mortality, significantly more hospital days after sepsis recognition, more ICU days, and more patients with hypotension. Patients aged 18 years and older represented 7.5% of the combined cohort. If these patients were excluded, the significant differences in clinical outcomes in Table 2 remain unchanged, and, in addition, the differences seen in “any ventilator use” of 482 (34.2%) in the inpatient group as compared with 4696 (31.2%) in the ED group (P <.021) would reach statistical significance. When we isolated the population of infants <1 year of age with IPSO sepsis, 30-day sepsis-attributable mortality was not significantly different between inpatient and ED, 0.8% compared with 1.8%, respectively (P = .124).

Critical sepsis mortality in the inpatient group was 3.5% compared with 2.9% in the ED group (P = .483). Figure 2 reveals the distribution of time to FTZ from arrival for the inpatient cohort with >40% within 12 hours and 21% after 96 hours. In the critical sepsis inpatient population, those who died had a significantly longer time from hospital arrival to sepsis recognition than those who survived, 64.4 hours (IQR 26.3–161.0) versus 22.0 hours (IQR 6.9–87.4; P = .043), respectively. For the inpatient cohort, there was a significant reduction in mortality over time (from 3.4% to 1.2%, P = .045) concurrent with a significant increase in bundle compliance (from 30% to 62%, P = .024) over time (Fig 3). Results of multivariable modeling (Supplemental Table 5) revealed that bundle compliance was significantly associated with lower mortality (P = .01). Mortality did not change meaningfully in association with later time of sepsis episode in months after the start of the study period (P = .11). Sepsis case volume increased over the study period (Table 1); however, the percentage of inpatients with critical sepsis did not change significantly: from 25.0% in 2017 to 20.3% in 2018 to 21.3% in 2019 (P = .10).

In the adjusted geometric means analysis (see Supplemental Table 3), patients in the inpatient cohort aged 18 years and older and those with comorbid conditions had a longer time from hospital arrival to FTZ of 20.7 (95% CI 13.8–31.1) and 18.6 (95% CI 14.2–24.5) hours, respectively. Patients who arrived as a transfer from an outside hospital had a shorter time from arrival to FTZ of 7.2 hours (95% CI 5.3–9.8) than patients who did not arrive as a transfer from an outside hospital of 25.7 (95% CI 20.1–32.9).

In this study, we have described inpatient sepsis and isolated sepsis-attributable mortality from all-cause mortality. We found higher sepsis-attributable mortality and increased length of stay, ICU admissions, and ICU days compared with ED patients. Like other sepsis studies, we found that improved adherence to treatment bundles improves mortality.^13,20  There were more patients captured in the database over time without an increase in the percentage of critical sepsis for inpatients. If changing mortality rates were a result of denominator dilution rather than increasing bundle compliance, we would expect the results from the multivariable model to reveal an association with the month of onset after the start of the study period and not an association with bundle compliance.

We found that the inpatient cohort had more infants and teens, but fewer young adults >18 years than the ED cohort. This has not been the case in other studies of pediatric sepsis.^3,7  In addition, it was unexpected that more children with comorbid conditions would be in the ED cohort, given that children with comorbidities account for most pediatric hospital days in the United States.²¹  Previous studies have revealed that children with comorbid conditions hospitalized with sepsis have poorer outcomes and higher health care costs.^21–23  The only subgroup of comorbidities that differed, however, was the immunosuppressed/immunocompromised/solid organ transplant/bone marrow transplant/malignancy/asplenia subgroup. We suspect immunocompromised patients presenting with sepsis to the ED were admitted to either an ICU or a dedicated oncology unit. The IPSO collaborative plans to analyze treatment patterns and outcomes of immunocompromised sepsis patients in the future. We also found that inpatients with comorbidities developed sepsis later in their hospital course than those without comorbidities, suggesting those patients were admitted to the hospital for other reasons and then developed sepsis during their hospital stay. Because the IPSO sepsis registry does not collect information regarding the source of infection, it is unknown if any of the late onset inpatient sepsis cases were due to nosocomial infection.²⁴ 

In our study, there were more patients who arrived by interfacility transfer in the inpatient cohort than in the ED cohort. In the adjusted geometric means analysis of only the inpatient cohort, sepsis occurred much earlier in the hospital course of patients transferred from outside hospitals than nontransferred patients. Previous studies indicate that longer length of stay, more ICU use, and higher mortality occur in patients whose care involved interfacility transfer which is consistent with worse outcomes seen in this study.²⁵  It is unknown if these patients had delays the in recognition or treatment of sepsis.

Sepsis recognition remains a focal point of QI efforts. The 2020 Surviving Sepsis Campaign Guidelines state “Although the optimal method or tool for screening is unclear, we suggest that screening tools be adapted to the type of patients, and resources and processes within each institution.”²⁰  Optimizing electronic sepsis screening for the inpatient setting requires significant time and resources, although with an unclear patient benefit.^26,27  In the era of electronic medical records, there remains opportunities for predictive analytic data models to detect changes in patient status before, or at the same time as, an alert clinician.^11,28,29  Hospitals seeking to improve screening for inpatient sepsis may be informed by differences in patient populations shown in this analysis as well as by internal patient populations and workflows.^11,12,20  For example, in our analysis, sepsis occurred in the first 12 hours after arrival in >40% of inpatient cases, potentially making these first 12 hours a higher-yield timeframe for screening. Additionally, longer length of stay patients and those with comorbidities would benefit from a tailored screening approach because 21% of IPSO sepsis in this cohort occurred after 96 hours, which is longer than the average pediatric hospitalization in US hospitals.³⁰  Further analysis is needed to understand the reasons for sepsis occurring either early or late in a hospitalization.

In this study, inpatients received smaller fluid volumes than ED patients. A recent study on ED patients found no difference in mortality associated with smaller-volume sepsis resuscitation making it unlikely that the mortality difference is due to volume resuscitation differences alone.³¹  The sepsis literature has not yet revealed conclusive evidence regarding the optimal volume of fluid resuscitation for ED patients, or for inpatients. This is an area that requires additional study.

Inpatients also had longer times to antibiotic administration than ED patients. On average, both were <180 minutes, consistent with SCC guidelines.²⁰  There are likely multiple reasons for longer times to antibiotics in the inpatient group requiring exploration if improvements are to be made. First, sepsis was less likely to present as critical. Second, screening based on vitals may be tied to an every 4 hours vitals cycle. Third, inpatients may have already received intravenous fluids and antibiotics at the time of their sepsis recognition. Finally, in the hospital, the use of diagnosis-based treatment pathways may present a barrier to early sepsis recognition because of cognitive bias.³² 

This study has several limitations. The submitted data included 44 variables, 40 of which could be reliably abstracted automatically from the electronic medical record; however, some fields may have been left as unknown (see Supplemental Table 4). Risk score and risk score method were frequently unreported;¹⁵  therefore, we were unable to perform a risk-adjusted analysis.

Regarding definitions, the collaborative designed the inclusion criteria to be sensitive and miss as few cases as possible. The IPSO critical definition is intended to have greater specificity.¹⁵  It is impossible to know from this data retrospectively if the team at the bedside was in fact providing the interventions as treatment of sepsis. Thus, there may be cases included in the sepsis cohort that did not actually have sepsis.

Individual hospitals developed sepsis screening methods and shared them within the collaborative. Those screens were not validated across hospitals, so each hospital team was responsible for monitoring the performance of their sepsis screen and other bundle elements.

Cases of sepsis recognized in oncology units were excluded; however, 13 hospitals (28%) did not have a separate oncology unit. We were unable to distinguish step-down units as separate entities, so it was left to the individual hospitals to report those patients either as ICU or as inpatient. Most hospitals reported data on patients >18 years of age. There is no upper age limit for IPSO reporting, so the dataset includes patients outside the typical pediatric age range; however, the inclusion of this age group did not affect most of the aggregate results.

Only the first episode of sepsis in a hospitalization is reported, so this data registry does not include subsequent sepsis episodes within the same hospitalization. The care received at outside facilities before transfer was not captured in this analysis and may have an important impact on the patient outcomes after interfacility transfer. Patients were reported as transfers if they occurred within 24 hours before FTZ. Reason for transfer, admission to the inpatient unit before sepsis onset, hospital staffing model, other human factors, and source of sepsis are not reported to this registry, making it impossible to draw conclusions on factors upstream in their care that might have led to sepsis.

Fewer sepsis cases originate in pediatric inpatient units compared with the ED; however, these cases have higher sepsis-attributable mortality. Mortality improved with improved treatment adherence. Further work is needed to refine sepsis screening and improve care delivery for inpatients with sepsis.

Mary Headley, Jayne Stuart, and Lowrie Ward of Children’s Hospital Association for their assistance with this work.

The Improving Pediatric Sepsis Outcomes Collaborative Investigators include:

Mashael F. Alqahtani, Department of Pediatrics, University of Central Florida, Orlando, FL, United States; Jeffery J. Auletta, Department of Pediatrics, The Ohio State University College of Medicine, Hematology/Oncology/BMT & Infectious Diseases, Nationwide Children’s Hospital, Columbus, OH, United States; Audrey H. Barnett, Quality & Patient Safety, Children’s Memorial Hermann Hospital, Houston, TX, United States; Richard J. Brilli, Nationwide Children’s Hospital, Columbus, OH, Department of Pediatrics, The Ohio State University College of Medicine; Deborah R. Campbell, Quality Program, Kentucky Hospital Association, Louisville, KY, United States; Pearl W. Chang, Department of Pediatrics, Seattle Children’s Hospital/University of Washington, Seattle, WA, United States; Ashmita Chatterjee, Department of Hospital Medicine, University of North Carolina at Chapel Hill, Division of Internal Medicine and Pediatrics, Chapel Hill, NC, United States; Theodore K. M. DeMartini, Department of Pediatrics, Penn State University, Division of Pediatric Critical Care, Penn State Health Children’s Hospital, Hershey, PA, United States; Holly Depinet, Division of Emergency Medicine, Cincinnati Children’s Hospital Medical Center, University of Cincinnati College of Medicine, Cincinnati, OH, United States; Jill B. Dykstra-Nykanen, Department of Pediatrics, Orlando Health Arnold Palmer Hospital for Children, Orlando, FL, United States; Guliz Erdem, Department of Pediatrics, The Ohio State University and Nationwide Children’s Hospital, Section of Pediatric Infectious Diseases, Columbus, OH, United States; Julie C. Fitzgerald, Department of Anesthesiology and Critical Care, The University of Pennsylvania Perelman School of Medicine, Department of Anesthesiology and Critical Care, Children’s Hospital of Philadelphia, Philadelphia, PA, United States; Meg Frizzola, Department of Pediatrics, Nemours Children’s Health, Sidney Kimmel Medical College at Thomas Jefferson University, Wilmington, DE, United States; Rachel E. Graham, Pediatric Progressive Care Unit, University of MD Children’s Hospital, Baltimore, MD, United States; Amy R. Grant, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH, United States; Hana Hakim, Department of Infectious Diseases, St. Jude Children’s Research Hospital, Memphis, TN, United States; Baila S. Harris, Department of Pediatrics, Spectrum Health, Department of Pediatric Hospital Medicine, Helen DeVos Children’s Hospital, Grand Rapids, MI, United States; Leslie A. Hueschen, Department of Pediatrics, University of Missouri, Kansas City, University of Kansas, Section of Pediatric Emergency Medicine, Children’s Mercy Hospital; Kansas City, MO, United States; W. Charles Huskins, Department of Pediatric and Adolescent Medicine, Mayo Clinic; Mayo Clinic College of Medicine and Science, Rochester, MN, United States; Jennifer Jacob-Freese, Nursing and Clinical Care Services Department, Children’s Hospital of Philadelphia, Philadelphia, PA, United States; Raed M. Khoury, Division of Quality & Patient Safety and Medical Affairs, Valley Children’s Healthcare, Madera, CA, United States; Gitte Y. Larsen, Department of Pediatrics, University of Utah, Pediatric Critical Care, Intermountain Primary Children’s Hospital, Salt Lake City, UT, United States; Vincent M. Laurich, Department of Pediatrics and Emergency Medicine, University of Connecticut School of Medicine, Division of Emergency Medicine, Connecticut Children’s, Hartford, CT, United States; Hadassah L. Little, Department of Quality and Safety Children’s, MUSC Shawn Jenkins Children’s Hospital, Charleston, SC, United States; Jeremy M. Loberger, Department of Pediatrics, University of Alabama at Birmingham, Division of Critical Care Medicine, Birmingham, AL, United States; Charles G. Macias, UH Rainbow Babies and Children’s Hospital, Case Western Reserve University, Cleveland Ohio; Elizabeth H. Mack, Department of Pediatrics, MUSC Children’s Health, Division of Pediatric Critical Care, MUSC Children’s Health, Charleston, SC, United States; Kate Madden, Department of Anesthesiology, Critical Care and Pain Medicine, Boston Children’s Hospital, Department of Anesthesiology, Harvard Medical School, Boston, MA, United States; Eliana M. Maldonado, Surgical Care Unit, Children’s National Hospital, Washington, DC, United States; Megan Morgan, Department of Pediatrics, Advocate Children’s Hospital, Park Ridge, IL, United States; Ashish Nagpal, Department of Pediatrics, Oklahoma University College of Medicine, Section of Critical Care Medicine, Oklahoma Children’s Hospital, Oklahoma City, OK, United States; Matthew Niedner, Matthew Niedner; Unaffiliated; Ann Arbor, MI, United States; Lori A. Parrott, Pediatric Intensive Care Unit, Cook Children’s Medical Center, Fort Worth, TX, United States; Nikhil S. Patankar, Beacon Children’s Hospital, Beacon Health System; Department of Pediatrics, Indiana University School of Medicine, South Bend Campus, South Bend, IN, United States; Eduardo Pino, Department of Pediatrics, Joan C. Edwards School Marshall University School of Medicine, Division of Pediatric Critical Care Medicine, Hoops Children’s Hospital, Huntington, WV, United States; Gregory P. Priebe, Department of Anesthesiology, Critical Care and Pain Medicine, Division of Critical Care Medicine, Boston Children’s Hospital, Department of Pediatrics, Division of Infectious Diseases, Boston Children’s Hospital, Boston, MA, United States; Vikram K. Raghu, Department of Pediatrics, University of Pittsburgh School of Medicine, Pittsburgh, PA, United States; Prabi Rajbhandari, Department of Pediatrics, Akron Children’s Hospital, Division of Hospital Medicine, Akron, OH, United States; Joyce M. Ramsey Coleman, Department of Quality and Safety, Texas Children’s Hospital, Houston, TX, United States; Nora E. Raynor, Atrium Health Levine Children’s Hospital, Charlotte, NC, United States; Faisal Razzaqi, Pediatric Hematology and Oncology, Valley Children’s Healthcare, Madera, CA, United States; Elise K. Rolison, Children’s Hospital of Colorado, Aurora, CO, United States; Jane R. Rosenman, Department of Pediatric and Adolescent Medicine, Mayo Clinic, Division of Pediatric Hospital Medicine, Mayo Clinic, Rochester, MN, United States; Ranna A. Rozenfeld, Department of Pediatrics, The Warren Alpert Medical School, Brown University, Division of Pediatric Critical Care Medicine, Hasbro Children’s Hospital, Providence, RI, United States; Lori Rutman, Department of Pediatrics, University of Washington, Division of Pediatric Emergency Medicine, Seattle Children’s, Seattle, WA, United States; Carly A. Schwab, Department of Nursing, Ann & Robert H. Lurie Children’s Hospital of Chicago, Chicago, IL, United States; Jonathan A. Silverman, Department of Emergency Medicine, VCU School of Medicine, Division of Pediatric Emergency Medicine, Children’s Hospital of Richmond at VCU, Richmond, VA, United States; Brittany M. Slagle, Department of Pediatrics, University of Arkansas for Medical Sciences, Section of Pediatric Hospital Medicine, Arkansas Children’s Hospital, Little Rock, AR, United States; Erika L. Stalets, Department of Pediatrics, University of Cincinnati College of Medicine, Division of Critical Care Medicine, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH, United States; Nathan E. Thompson, Department of Pediatrics, Medical College of Wisconsin, Section of Pediatric Critical Care Medicine, Children’s Wisconsin, Milwaukee, WI, United States; Roopa Thukaram, Department of Pediatric Critical Care Medicine, Saint Francis Children’s Hospital, Section of Pediatric Critical Care Medicine, Tulsa, OK, United States; Beth Wathen, Children’s Hospital Colorado, Aurora, CO, United States; Jeffrey D. Williams, Department of Pediatrics, Cohen Children’s Medical Center, Section of Pediatric Hospital Medicine, New Hyde Park, NY, United States.