Prediction of 30-Day Mortality Using the Quick Pitt Bacteremia Score in Hospitalized Patients with Klebsiella pneumoniae Infection

Introduction

Klebsiella pneumoniae is a gram-negative, encapsulated bacterium that usually colonizes the human gastrointestinal tract mucosal surface, but may cause extraintestinal opportunistic infections in the urinary tract, lung, brain, or bloodstream, especially in immunocompromised status.^1,2 K. pneumoniae is an important causative pathogen in approximately 10% of nosocomial infections, resulting in poor prognosis owing to its hypervirulence and antibiotic resistance.^3,4 It has been estimated that the mortality rate of patients with K. pneumoniae infection ranged from 5–30% and up to 29–54% in those with bloodstream infection.^5–7 The World Health Organization recently included K. pneumoniae in list of critical microorganisms, which pose a significant threat and major burden to global public health.²

The Pitt bacteremia score (PBS) has been a widely used scoring system for disease severity stratification in patients with bacteremia for decades and has recently been validated for its prognostic role in patients with nonbacteremic infections.^8–10 Recently, a simplified quicker version of the PBS (qPitt) using binary variables for acute illness severity measurement was developed, including body temperature, blood pressure, respiratory rate, cardiac arrest event, and mental status.¹¹ qPitt was investigated as an outcome prediction tool in patients with Staphylococcus aureus bloodstream infection,¹² and bloodstream infection secondary to urinary tract infection in recent studies.¹³ Both studies revealed that qPitt has excellent mortality discrimination ability in these patients.^12,13 As PBS was originally derived from patients with gram-negative bloodstream infection,¹⁴ and qPitt has not been validated in patients with single causative gram-negative pathogen, the goal of this study was to explore the prognostic value of qPitt in patients with K. pneumoniae infection. Furthermore, we tested the mortality discrimination ability of qPitt in subgroups of patients with nonbacteremic infections.

Materials and Methods

Study Design and Patient Enrollment

This retrospective observational cohort study was conducted at a tertiary medical center located in Kaohsiung, Taiwan. There are approximately 1100 licensed bed and 53,000 emergency department (ED) visits annually. The institutional review board waived the requirement for informed consent because of the retrospective and observational nature of the study. The study followed the principles of the Declaration of Helsinki and STROBE guidelines.

Inclusion Criteria

We enrolled all adult patients (aged ≥18 years) who visited the ED between January 1, 2021, and December 31, 2021, and underwent blood or other site-specific cultures which yield K. pneumoniae in this study. If a patient had multiple ED visits during the study period, only the first visit was included.

Exclusion Criteria

To avoid selection bias and heterogeneity, we excluded patients with bloodstream infection from site-specific infection at another body site (ie, secondary bloodstream infection).¹⁵ We also excluded non-hospitalized patients and those with unrelated diagnoses (the cultured bacteria was not related to the cause of patient’s infection. eg, the urine specimen yield K. pneumoniae but the principal diagnosis was biliary tract infection not yield K. pneumoniae).

Data Collection and Study Definitions

All eligible patients and their relevant data, including demographic, clinical, microbiological information and management strategies, were collected from an anonymized electronic medical record system. The patients were divided into survivor and non-survivor groups according to their 30-day survival status for further comparison and analysis. The systemic inflammatory response syndrome (SIRS), quick sequential organ failure assessment (qSOFA), and sequential organ failure assessment (SOFA) scores were calculated based on their original definition using the initial ED physiological and laboratory parameters.¹⁶ A qSOFA score ≥ 2 points was used as the prognostic cutoff value.¹⁷ The PBS, ranging from 0 to 14, was obtained based on the following five variables: temperature, altered mental status, blood pressure, need for mechanical ventilation, and cardiac arrest event.⁸ The qPitt score was calculated based on five scoring items: temperature <36°C, systolic blood pressure ≤90 mmHg or vasopressor administration, respiratory rate ≥25 times/min or need of mechanical ventilation, altered mental status, and cardiac arrest event.¹¹ Each variable was assigned one point, with a maximum score of 5. The sources of K. pneumoniae infection was stratified as primary bacteremia, urinary tract infection, respiratory tract infection, intra-abdominal infection, skin/soft tissue infection, and others (minor infection sites, such as catheter-related infection).¹⁵ The administered antibiotic was considered appropriate if it was active in vitro against the cultured K. pneumoniae (eg, blood, urine, sputum, or bronchoalveolar lavage) by susceptibility testing. The implementation of source control (measures to eliminate the source of infection) was determined by the treating physician’s judgement and recorded according to medical records.

Outcome Measurement and Statistical Analysis

The primary outcome of this study was to compare the 30-day mortality prediction ability of qPitt and other commonly used sepsis severity scoring systems, including qSOFA, SOFA, and PBS, in patients with K. pneumoniae infection. The secondary outcome was the mortality discrimination ability of qPitt in nonbacteremic patients. Quantitative variables were expressed as mean ± standard deviation or median (interquartile range) and were compared using a two-sample t-test or Mann–Whitney U-test, depending on whether they were normally distributed. Qualitative variables were expressed as percentages and compared using the Chi-square or Fisher’s exact test. The mortality discriminative ability of qPitt and other scoring systems was tested using the area under the receiver operating characteristic (ROC) curve. The DeLong method was used to compare the area under the curve (AUC) of these variables.¹⁸ The significance level for statistical testing was defined as two-tailed p < 0.05. All statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS, IL, USA), version 25.0, and MedCalc Statistical Software version 18.2.1 (MedCalc Software bv, Ostend, Belgium). Figures were drafted using GraphPad Prism version 8.4.3 (GraphPad Software, CA, USA).

Results

Initially, 1569 patients with K. pneumoniae infection met the inclusion criteria. After excluding repetitive patients (N = 88), non-hospitalization (N = 378), secondary bacteremia (N = 90), and unrelated diagnoses (N = 146), the remaining 867 patients were included in the final analysis (Figure 1).

Figure 1 Flowchart of patient enrollment.

As shown in Table 1, the total 30-day mortality rate of the enrolled patients was 13.4% (116/867). The mean age of the patients was 66.9±15.8 years, and more than half (57.8%) were male patients. Compared with survivor group, patients in the non-survivor group had significantly higher proportion of chronic liver disease (12.9% vs 5.6%, p<0.01), chronic kidney disease (50.0% vs 30.6%, p<0.01), and malignancy (37.1% vs 22.6%, p<0.01) comorbidities. The non-survivor group also had significantly lower hemoglobin (10.3 [8.7–12.9] g/dL vs 11.8 [9.9–13.6] g/dL, p<0.01) and platelet (158 [97–244] ×10⁹/L vs 218 [159–289] ×10⁹/L, p<0.01) level, longer international normalized ratio (1.24 [1.13–1.46] vs 1.07 [1.02–1.18], p<0.01), higher lactate (3.6 [2.1–7.6] mmol/L vs 1.7 [1.1–2.7] mmol/L, p<0.01), c-reactive protein (108.5 [40–203.9] mg/dL vs 64.4 [20–150.7] mg/dL, p<0.01), and creatinine levels (1.7 [1.0–2.3] mg/dL vs 0.7 [1.0–1.4] mg/dL, p<0.01) than the survivor group (Table 1).

Table 1 Baseline Characteristics of Patients with Klebsiella pneumoniae Infection

Regarding the sepsis severity scoring systems, it was not surprising that patients in the non-survivor group had a significantly higher proportion of SIRS criteria (81.9% vs 66.0%, p<0.01), high qSOFA category (39.7% vs 13.3%, p<0.01), and higher SOFA scores (9 [7–11] vs 2 [0–4], p<0.01). Furthermore, the non-survivor group had significantly higher PBS (5 [3–8] vs 0 [0–2], p<0.01) and qPitt (3 [2–3] vs 0 [0–1], p<0.01) scores than the survivor group (Table 2).

Table 2 Sepsis Severity, Source and Management of Patients with Klebsiella pneumoniae Infection

The most common infection source in the enrolled patients was the urinary tract (32.9%), followed by the bloodstream (26.0%) and skin/soft tissue (17.3%). Approximately three-fourths (71.7%) of the patients in this study received appropriate antibiotic administration (Table 2). The patients in the non-survivor group had a significantly higher proportion of bloodstream infections (33.6% vs 24.8%, p<0.01) and a lower proportion of urinary tract infections (19.0% vs 35.0%, p<0.01) than those in the survivor group. There were comparable proportion of appropriate antibiotic administration between the survivor and non-survivor groups (p=0.85).

The 30-day mortality prediction abilities of the scoring systems were compared. As shown in Figure 2, the AUC of each scoring system was as follows: SOFA, 0.91 (95% confidence interval [CI] = 0.89–0.93), qPitt, 0.87 (95% CI = 0.84–0.89), PBS, 0.87 (95% CI = 0.85–0.89), and qSOFA, 0.73 (95% CI = 0.70–0.76). Moreover, the AUC of qPitt was significantly higher than that of qSOFA (p<0.01) and similar to that of PBS (p=0.65).

Figure 2 The AUC of each scoring system.

We further examined the mortality discriminative performance of the qPitt score, which ranged from 0 to 5. As shown in Figure 3, the 30-day mortality rates in the qPitt group were 0.9% (4/422), 8.4% (16/191), 27.3% (38/139), 45.1% (37/82), 60.7% (17/28), and 80.0% (4/5), respectively. The mortality risk increased significantly in a stepwise manner as qPitt increased (p<0.01). Finally, we tested the ability of qPitt to predict mortality in nonbacteremic patients. As shown in Figure 4, qPitt also demonstrated excellent mortality discrimination ability in these patients (AUC= 0.85, 95% CI = 0.82–0.88, p<0.001).

Figure 3 30-day mortality rates in the qPitt score groups.

Figure 4 The ability of the qPitt to predict mortality in patients without bacteremia.

Discussion

In this single-center, ED-based, retrospective observational study, we evaluated the predictive performance of qPitt for the 30-day mortality risk of patients with K. pneumoniae infection and compared it with other widely used scoring systems, including SOFA, PBS, and qSOFA. We demonstrated that qPitt had a mortality discriminative ability comparable to that of PBS and was superior to that of the qSOFA score. Furthermore, we revealed that qPitt has an excellent mortality prediction ability in nonbacteremic patients with K. pneumoniae infection. To the best of our knowledge, this is the first study to investigate the prognostic role of qPitt in patients with a single causative gram-negative pathogen, including both bloodstream and non-bloodstream infections.^11–13,19

Susceptibility to K. pneumoniae infection is determined by various factors, including pathogen virulence, host defense, and extrinsic stress.^1–3 It was studied that individuals with advanced age or other chronic illness, such as diabetes, malignancy, liver disease, chronic obstructive lung disease, renal failure, or malnutrition, are more vulnerable to K. pneumoniae infections and have a poorer prognosis.^1,3 In patients with K. pneumoniae bloodstream infection, anemia, thrombocytopenia, coagulation disorders, renal failure, hypoalbuminemia, elevated c-reactive protein levels, and hyperlactatemia have been reported to be associated with mortality risk.^20–22 The comorbidities and laboratory results of the non-survivors in our study were consistent with these findings, further strengthening their prognostic significance in patients with K. pneumoniae infections.

The proportion of infection source distribution in our patient groups was also consistent with previous studies.⁵ Patients have bloodstream infections due to their chronic illness and are prone to undergo surgical procedures, immunosuppressant agents, and invasive devices, all of which contribute to their poor physical condition and higher mortality risk.^3,5 In contrast, the urinary tract is a relatively minor infection site, which may slightly influence the patient’s condition, thus providing a better survival chance.⁵ The escalating prevalence of multidrug-resistant bacteria is another global issue, and was estimated to account for one-third of all K. pneumoniae pathogens in a recent meta-analysis.⁴ Fortunately, the relatively low mortality rate of bloodstream infections in our study may be attributed to the low prevalence of antibiotics resistance (ie, a high proportion of appropriate antibiotics administration).

The SOFA score was developed to quantitatively describe the degree of organ dysfunction/failure and complications of critical illness in individuals or groups of patients and is one of the most widely used scoring systems in adult intensive care settings, including sepsis.²³ Although the SOFA score was originally designed to evaluate the extent of organ dysfunction rather than outcome prediction, it has been well validated as a good prognostic discrimination tool in patients with infectious disease.^16,24,25 Therefore, it is reasonable that the SOFA score had the highest prognostic accuracy in our study. Nevertheless, the need for laboratory parameter measurements and a slightly complex calculation process sometimes makes the SOFA score not readily available for risk stratification of these patients.²⁵ Moreover, the iatrogenic confounders, inter-patient variability, and the introduction of new intervention strategies all possibly promote the need for SOFA score update.^23,25

PBS has been used for three decades to stratify acute illness severity and predict mortality in various patient populations with bloodstream infections.^14,26,27 Similar to PBS, qPitt is a more simplified scoring system consisting of five binary variables that retain hypothermia but exclude fever from PBS.¹¹ Hypothermia is deemed a late phase of acute illness and a more severe stage of endotoxic shock in animal experimental studies; thus, it is more strongly correlated with mortality in patients with infectious diseases.^28,29 Another advantage of qPitt variables is the respiratory rate, which is based on the number of breaths/min rather than the oxygen concentration, and the need for mechanical ventilation may allow earlier detection of respiratory failure in these patients.¹² The comparable mortality prediction ability of PBS and qPitt in our study strengthens the value of outcome prediction. Larger prospective studies are required to validate these findings.

Both the qPitt and qSOFA scores were developed to identify high-risk patients based solely based on vital signs and other readily available clinical variables during an initial physical examination, and the absence of laboratory parameters allows making clinical decision more quickly.^12,14 Although they share some similar parameters, the slightly higher threshold in respiratory rate and systolic blood pressure parameters in qPitt may improve its specificity and positive predictive value without the expense of sensitivity or negative predictive value.¹¹ Notably, qPitt outperformed qSOFA in mortality prediction in patients with Staphylococcus aureus bloodstream infection in a previous study, whereas another study in patients with urinary tract infection revealed converse results.^12,19 The significantly superior prognostic value of qPitt in our study encourage future studies to compare both scoring systems in other cohorts, further extending the application potential of qPitt to a wide variety of infections.

Another novel finding of our study is that qPitt predicted mortality in nonbacteremic patients with K. pneumoniae infection, as well as in those with bacteremia. A recent multicenter prospective study investigated the prognostic role of qPitt in patients with nonbacteremic carbapenem-resistant Enterobacteriaceae infections and demonstrated the strong mortality prediction ability of qPitt in these patients.¹⁰ These findings may allow the broader use of this simple score for risk stratification in clinical infectious diseases, even prior to the identification of the microbiological profile or site of infection.^10,12

The main strength of this study was the inclusion of all patients with K. pneumoniae infection, regardless of the infection site, which is more practical for clinical applications. The exclusion of patients with secondary bloodstream infections also avoided misclassification of these patients. Nevertheless, this study also shares the common limitations of a retrospective observational cohort design, which is mostly subject to inherent recall bias. Although we thoroughly reviewed the electronic medical records to obtain detailed information on our enrolled patients, some unknown and unmeasured confounders may still bias the results. Other important elements of sepsis management, including the amount of fluid or vasopressors administered to patients, were not calculated or controlled in this study. The single-center nature of this study warrants future multicenter prospective studies to validate our findings.

Conclusion

In conclusion, in patients with K. pneumoniae infection, qPitt revealed excellent 30-day mortality prediction ability, similar to PBS, and outperformed the qSOFA score. The qPitt also exhibited mortality discrimination ability in nonbacteremic patients with K. pneumoniae infections. Clinicians can use this simplified scoring system to stratify patients earlier and initiate prompt treatment in high-risk patients.