In this 10-year retrospective study, we aimed to highlight the predictors at the early onset of Kp-BSI and their predictive values for mortality. We have identified that a history of preterm birth, hospitalization within the previous six months, intra-abdominal source of bloodstream infection, presence of organ failures, and high levels of biomarkers, including C-reactive protein, albumin and lactate, were independent risk factors for the 90-day mortality. In particular, we have demonstrated for the first time that adding two risk factors (prior history of hospitalization and intra-abdominal source of bloodstream infection) into the SOFA score at the early onset of Kp-BSI had satisfactory predictive power for 90-day mortality with an AUC value of 0.89. Particularly, the specificity was elevated from 0.68 to 0.84 compared with that of the original SOFA score. This Kp-specific SOFA score represented an ability to discriminate patients at risk of death.
In recent years, the global prevalence of K. pneumoniae has rapidly increased. A multi-center study in the US reported a 3.8% point prevalence of K. pneumoniae [23]. In China, the prevalence of K. pneumoniae was found to be 12.0% among 244, 843 clinical isolates, and 19.1% of the K. pneumoniae isolates were collected from bloodstream [24]. The overall mortality reported in the previous literature for adult patients was as high as 29.3% [25]. In our pediatric study, the 90-day mortality was found to be 22.6%, higher than the mortality in a tertiary children’s hospital in China (12.3%). The common presence of organ failures at the onset of Kp-BSI in our study population may explain the high mortality rate. As K. pneumoniae is a formidable pathogen evading the early activation of innate immune reactions, the clearance of systemic toxicities is challenging, resulting in the presence of multi-organ failures [26]. Septic shock and mechanical ventilation at the early stages of Kp-BSI have been considered strong risk factors for mortality [6, 27]. In our study, the respiratory tract was the most common organ to fail at the onset of Kp-BSI, but was not related to the fatal outcome. Altered mental status, coagulation disorder, hepatic failure, shock/hypotension, and renal failure were significantly associated with mortality. We demonstrated that the risk of mortality increased with elevated SOFA scores at the onset of Kp-BSI. This highlights the importance of systematic evaluation of clinical parameters relevant to organ functions at the early onset of Kp-BSI.
Apart from organ failures, there was a difference in the 90-day mortality between patients with- and patients without- prior hospitalization within the previous six months, though significant difference in the specific categories of comorbidity was not observed. The impaired host defenses of children with severe underlying conditions are responsible for severe sepsis and poor prognosis [28]. Additionally, complicated intra-abdominal infections, including necrotizing enterocolitis, spontaneous bacterial peritonitis, and intra-abdominal sepsis associated with bowel perforation, are serious complications in premature babies and children in post-operative status or with other serious underlying conditions. The gastrointestinal system is considered critical to the host resistance to sepsis, with the bacterial translocation and microbiota disturbances in critically ill patients supporting the concept of impaired communication across the gut-organ axes [29, 30]. Meanwhile, the damaged tissue related to intra-abdominal sepsis can be a source of pro-inflammatory cytokines that induce distant organ failure [31]. The compromised gut-barrier functions also lead to a problem in the clearance of bacteria and endotoxins in the systemic circulation [32]. In this study, the intra-abdominal source of bloodstream infection was associated with the worst prognosis. However, the impact of primary infection site on mortality in Kp-BSI is controversial according to previous studies and needs to be further studied [33, 34].
Additionally, we explored several biomarkers that were not involved in the SOFA sub-scores but commonly measured in the intensive care unit. The addition of serum albumin and C-reactive protein to the SOFA score also improved risk prediction of mortality. Albumin, which is exclusively synthesized by the liver and reflects the physiological response to injury and infection, has been identified as an independent predictor of mortality in patients with Kp-BSI [35]. On the other hand, C-reactive protein is an acute-phase reactant with prognostic significance in septic patients [10], and its response to infection is determined by the invasive mechanisms as well as host predisposition [36]. Although the measurement of multiple biomarkers may improve the identification of patients at risk, it may increase the complexity in clinical application and the expense of diagnosis. Moreover, there was a marginal advantage of the addition of C-reactive protein and albumin to the risk model compared with the Kp-specific SOFA according to the integrated discrimination improvement analysis.
In this study, we validated the pediatric SOFA updated in 2017 for prognosticating the outcomes of children with Kp-BSI to fill the gap of knowledge in verifying the prognostic value of the SOFA score in a specific infection, and it achieved an AUC value of 0.80. Based on this finding, we developed a Kp-specific SOFA score with a significant increase in AUC value by adding two risk factors as sub-scores, which addresses the urgent need of studies concerning the prognosis of children with Kp-BSI. The Kp-specific SOFA had a high negative predictive value (0.95), suggesting that if extensively applied in the clinical field, it may be a reasonable reference to rule out pediatric patients at high risk. Most of our study population consisted of patients with healthcare-associated infection and with different comorbidities, which is representative of children with Kp-BSI in clinical practice. Therefore, our results can be extrapolated to the pediatric populations with Kp-BSI. In the context of the different etiology beyond Kp-BSI, the addition of risk factors into the SOFA score may further strengthen its predictive accuracy in identifying high-risk patients with sepsis caused by specific pathogens and other original sites of infection [37].
This study has some limitations. Firstly, it was a retrospective study, and the recall bias must be a problem. Prospective validation was not conducted for the risk score derived from our study population. Nevertheless, the 146 enrolled patients were from multiple departments in a tertiary hospital in China and underwent follow-ups regularly within 90 days after the onset of Kp-BSI with electronic medical records, and thus the recall bias has been reduced. Moreover, we did not consider factors concerning therapeutic interventions after the diagnosis of Kp-BSI in our analysis, nor did we investigate whether the mortality resulted from treatment failure. Because the objective of this study was to assess the predictive power of clinical parameters at the early onset of Kp-BSI, we did not consider treatment factors as potential confounders. Besides, the internal consistency in treatments was guaranteed by the guideline-adherent care provided by the same medical team. Thirdly, it was a single-center study with potential selection bias, and future multi-center studies are needed to determine the best clinical use of the Kp-specific SOFA at the early onset of Kp-BSI.