Characteristics of the study population
During the study period (January 2009 to June 2019) there were 50,000 estimated admissions per year in pediatric wards at Xinhua Hospital, and 154 (0.03%) patients were diagnosed with Kp-BSI. Eight (5.5%) patients were lost to follow-up; a total of 146 patients were included in this retrospective study. Overall, 20 (13.7%), 25 (17.1%), and 33 (22.6%) patients died within 14, 30, and 90 days from Kp-BSI onset, respectively.
The median age of patients was 2.7 months (interquartile range: 1.0, 10.9 months, Table 1). Bloodstream infections were mainly secondary to pneumonia (50.0%), followed by intra-abdominal infection (20.5%) and indwelling central catheters (6.9%). Primary bacteremia occurred in 22.6% of patients (Table 2). Most of the Kp-BSI (84.9%) cases were healthcare-associated, and nearly half of the patients (48.6%) were infected with carbapenem-resistant K. pneumoniae . The most common organ failure within 24 h from Kp-BSI onset was respiratory failure (63.0%), followed by hepatic failure (52.7%), coagulation disorder (39.7%), shock/hypotension (29.5%), renal failure (22.6%), and altered mental status (15.1%). Only 11.6% of the patients had no organ failure at the time of Kp-BSI onset (Table 2).
Risk factors associated with 90-day mortality in patients with Kp-BSI
A total of 33 (22.6%) patients died within 90 days after the onset of Kp-BSI, and 113 (77.4%) were discharged from the hospital after reaching clinical stability. No significant differences in baseline characteristics (age, sex, admission year, and admission ward, weight-for-age z-score) were observed between survivors and non-survivors (Table 1). There were more patients with a history of preterm birth and a malignant disease as comorbidity among the non-survivor group compared to the survivor group. Non-survivors were characterized by a history of hospitalization in the last six months (P < 0.01, Table 1). Before the onset of Kp-BSI, no differences in invasive procedures during hospitalization were observed between survivors and non-survivors (Table 2). Intra-abdominal sources of infection were more frequent among non-survivors (48.5%) than survivors (12.4%), while bloodstream infections of patients alive at day 90 were more likely to result from primary bacteremia. The prevalence of carbapenem-resistant strains in the survivors and the non-survivors was 44.2% and 63.6%, respectively. Compared to the survivors, the non-survivors had lower levels of serum albumin and platelet counts, and higher levels of C-reactive protein, procalcitonin, total bilirubin, creatinine, and lactate. Additionally, the median SOFA score of non-survivors was approximately 2-fold higher than that of survivors. The increased number of organ failures at the onset of Kp-BSI was associated with a higher risk of mortality (Fig. 1). The proportion of non-survivors with abnormal function of at least three organ systems was 41.4% (24/58).
After adjusting for age, sex, admission year and admission ward, patients were nearly twice as likely to die if they were born prematurely, and prior hospitalization in the last six months was associated with a 2.4-fold increase in mortality (Table 3). For patients infected with K. pneumoniae from intra-abdominal sources, the risk of death was 6.8-fold higher compared to that of patients with primary bacteremia or catheter-related infections. Malignancies, carbapenem-resistant isolates, and respiratory failure were not significantly associated with mortality. Among the various types of organ failure, altered mental status was associated with the worst outcome (HR = 5.19, 95% CI 2.50–10.77), followed by coagulation disorder (HR = 4.54, 95% CI 2.12–9.72), hepatic failure (HR = 2.90, 95% CI 1.33–6.35), shock/hypotension (HR = 2.41, 95% CI 1.21–4.82), and renal failure (HR = 2.20, 95% CI 1.02–4.73). Low levels of albumin and high levels of sepsis-related biomarkers, including the C-reactive protein and lactate, were also significant risk factors for mortality (Table 3).
Antimicrobial treatments in patients infected with CR-Kp
Among the 71 patients with CR-Kp bloodstream infection, 40 (56.3%) received a carbapenem combined with another antibiotic as a definitive therapy, proven to be effective in vitro (including amikacin, fosfomycin, or polymyxin B), while the remainder received a high-dose extended or continuous infusion of a carbapenem. The mortality rate of patients receiving polymyxin B plus carbapenem or fosfomycin plus carbapenem was 19.4% (6/31), while that of patients treated with amikacin plus carbapenem and with carbapenem monotherapy was 44.4% (4/9) and 35.5% (11/31), respectively [see Table S1, Additional file 1]. Additionally, a subgroup analysis using Cox proportional hazards model showed no significant association between the type of definitive antimicrobial therapy and the 90-day mortality of patients with CR-Kp bloodstream infection [see Table S1, Additional file 1].
Development of a combined model on the basis of the SOFA Score
The pediatric SOFA score had a moderate predictive power for 90-day mortality, with an AUC of 0.80 (95% CI 0.71–0.89), along with a high sensitivity (0.84) and a low specificity (0.68). The predictive value of clinical characteristics and blood biomarkers was also evaluated. Each combination of SOFA with risk factors (including prior hospitalization, intra-abdominal source of infection, C-reactive protein, and albumin) improved the predictive performance of the model to different degrees [see Table S2, Additional file 1]. A modified SOFA model combining SOFA plus prior hospitalization and intra-abdominal source of infection was selected as the optimal multivariate predictive model (AUC = 0.89, 95% CI 0.82–0.96; sensitivity = 0.86; specificity = 0.84; positive predictive value = 0.64; negative predictive value = 0.95), outperforming the original SOFA score (DeLong’s test, P < 0.01, Fig. 2). According to the integrated discrimination improvement index, other combinations of SOFA with risk factors had no additional prognostic value over the selected model [see Table S3, Additional file 1]. The final multivariate model (Logit [mortality] = -5.86 + 0.53 × [SOFA score] + 1.81 × [prior hospitalization, 0 or 1] + 1.83 × [intra-abdominal source, 0 or 1]) was used to develop a scoring system (Kp-specific SOFA) including: SOFA (original score), prior hospitalization (4 points) and intra-abdominal source (4 points). The mean estimated risk score for the study population was 8 ± 4 (range 0–21).
The observed risk of mortality and the estimated probabilities based on the Kp-specific SOFA for three risk strata (low, medium, and high) are shown in Table 4. The 90-day mortality rate was 5.4%, 35.7%, and 72.0% for low- (score 0–8), medium- (score 9–11), and high-risk (score ≥ 12) patients, respectively. Compared to the low-risk group, patients with medium- or high-risk scores were more likely to die (HR = 8.36, 95% CI 3.60–27.83 for medium-risk and HR = 20.27, 95% CI 7.47–54.95 for high-risk; Table 4). This Kp-specific SOFA score was able to stratify patients into low-, medium-, and high-risk groups on the basis of the Kaplan-Meier survival curves (log-rank test, P < 0.01, Fig. 3).
Kp-specific SOFA in predicting 30-day mortality
The 30-day mortality rate was 17.1% (25/146). For predicting 30-day mortality, the AUC value of pediatric SOFA and Kp-specific SOFA was 0.79 and 0.85, respectively [see Fig. S1, Additional file 1]. Three categories of risk were performed by Kaplan-Meier survival curves using the predetermined cut-off values, which demonstrated a significant difference in 30-day mortality among three strata by log-rank test (P < 0.01) [see Fig. S2, Additional file 1].