Discussion

Culture-negative sepsis remains an entity to be elucidated. The majority of adult literature focuses on characteristic and outcome comparisons of culture-negative and culture-positive sepsis with conflicting results. Gupta et al found that patients with culture-negative sepsis were more likely to have co-morbidities, acute organ dysfunction, and higher mortality. ^{Reference Gupta, Sakhuja and Kumar8} Phua et al. found patients with culture-negative sepsis to have fewer co-morbidities and shorter duration of hospitalization. ^{Reference Phua, Ngerng and See9} Yet, additional studies suggest similar characteristics and outcomes between culture-positive and culture-negative sepsis events. ^{Reference Kethireddy, Bilgili and Sees10,Reference Sigakis, Jewell and Maile11}

When it comes to treatment of culture-negative sepsis, data are even more limited, especially in the pediatric population. Our results suggest that treatment of culture-negative sepsis for less than or equal to 7 days may be adequate for patients in the pediatric ICU based on the outcomes of mortality and length of stay.

Interestingly, patients who have been hospitalized for a longer period before the sepsis event, those on mechanical ventilation, and underlying pulmonary disease were more likely to be treated with a longer course of antibiotics. However, mechanical ventilation and underlying pulmonary disease were not shown to be confounding variables, however, in our primary outcome of hospital length of stay suggesting that length of antimicrobial treatment could be a contributing factor in prolonging hospitalization.

We hypothesize that there is a perception that patients with longer prior hospitalization, requiring mechanical ventilation, or underlying pulmonary disease are more sick, not improving, or higher risk for a serious bacterial infection persuading providers to extend the course of antibiotics.

Despite similar severity of illness scores between the two groups, it is possible that these patients had more severe illness at the time of the sepsis event. This may be supported by the increased overall mortality rate in the longer course group. The PIM-2 and PRISM 3 scores are reflections of the patients at admission. Given the longer course group was more likely to be hospitalized for a longer period of time prior to the sepsis event, these mortality risk scores may have changed since admission.

If the perception of severity of illness was higher in the long-course treatment group, one might expect the selection of antibiotics to change. However, in our study, the choice of antibiotics was similar between the two groups. Upon further analysis, there was a difference in DOTs between the two groups (Supplemental Figure 1). This may suggest that providers were hesitant to narrow antibiotics based on the perceived severity of illness, despite no identified pathogen or specific focus of infection. The assessment of unnecessary antimicrobials has been identified as a knowledge gap in the surviving sepsis campaign. ^{Reference Weiss, Peters and Alhazzani2}

One of the largest pediatric studies assessing the prevalence and outcomes of patients with severe sepsis found only 65% of patients had an identified organism on culture. ^{Reference Weiss, Fitzgerald and Pappachan15} In culture-negative sepsis, patients often remain on broad-spectrum antibiotics for prolonged periods of time because there is no organism for which to tailor therapy. Since unnecessary or prolonged antibiotic exposures contribute to the continued threat of MDROs, we assessed MDRO acquisition for our secondary outcome. Although our study sample was insufficiently powered for MDRO acquisition, notably, there were double the cases of new MDROs in the group treated with a long course of antibiotics. Like us, Tamma et al also observed an increase in MDRO acquisition among patients receiving longer durations of therapy. ^{Reference Tamma, Turnbull and Milstone14} Furthermore, Tashome et al reported that additional antibiotic exposure can contribute to the development of resistance, with each additional day of exposure conferring a 4%–8% risk. ^{Reference Teshome, Vouri, Hampton, Kollef and Micek16} Given the clinical risks of MDRO development, we believe identifying the shortest, effective duration of therapy for the treatment of culture-negative sepsis remains an important area of study with a larger population. Based on our current data, we calculated a sample size of 298 subjects needed to achieve 80% power.

As a single-center, retrospective study, there were inherent limitations. Identification of patients with culture-negative sepsis was difficult, but we did follow a validated data collection method for the identification of sepsis in hospitalized patients. ^{Reference Larsen, Brilli and Macias17} ICD 9/10 codes are not a reliable source for routinely identifying culture-negative patients, and the rationale for antibiotic courses was not always clearly documented. Unlike other studies, we did exclude patients documented to have aspiration pneumonia with X-ray findings, pneumonia, cellulitis/skin/soft tissue infections, toxic shock syndrome, and endocarditis, despite negative cultures given current literature to aid in treatment duration for these conditions. Although this strengthens our study by focusing on patients with concerns for infection despite no obvious source, it does limit our sample size. One major limitation in our study is the variation in length of time patients were hospitalized prior to sepsis case, making it challenging to interpret hospital length of stay exclusively from sepsis event between the two groups. In addition, as a retrospective study, not every patient in our study had follow-up cultures further limiting our evaluation for subsequent MDRO acquisition. Also, we did not evaluate differences in treatment duration related to providers. Though there may be a provider effect, our large critical care medical staff (including 24 physicians, 18 advance practice providers, and 12 fellows) likely balances out any effect from one single provider. Furthermore, we were not able to include data related to mechanical ventilation or underlying pulmonary disease, both of which could contribute risk for infection, due to limitations with our data set. However, it is not standard practice at our institution to start antibiotics solely based on chest X-ray or ventilator changes and was more likely related to a constellation of symptoms, strengthening the real-world application of our findings.

In conclusion, there are limited data to guide the treatment of culture-negative sepsis, and thus, it is identified as a knowledge gap within the pediatric surviving sepsis campaign. ^{Reference Weiss, Peters and Alhazzani2} Our single-center, retrospective study suggests patients treated with shorter courses of antibiotics do not have worse outcomes to those treated with longer courses and may have the benefit of reducing future development of resistant organisms. Future prospective studies with a larger population are warranted.