Background

Klebsiella pneumoniae is one of the most important pathogens causing urinary tract infection, pneumonia, intra-abdominal infection, and primary bacteremia. It is also the second most common cause of community- and hospital-acquired gram-negative bacteremia [1–3]. Mortality from Klebsiella pneumoniae bacteremia (KpB) is about 20 to 40% [3–5], and its population-based incidence ranges from 7.1 to 9.7 per 100,000 person-years [5, 6].

There is evidence that the usefulness of routine blood cultures should be reconsidered in acute pyelonephritis [7, 8], cellulitis [9], community-acquired pneumonia [10, 11], and in isolated fever or leukocytosis [12]. Although attempts have been made to reduce unnecessary blood cultures by introducing clinical rules [12–14], clear guidelines have yet to emerge as to when blood cultures should be drawn. And blood cultures are liberally prescribed because of the high mortality due to bacteremia, anxiety about undertreatment, and fear of using inappropriate antimicrobial agents causing limited positivity of 4 to 7% [12, 15, 16].

Although it is widely appreciated that the requirement for follow-up blood cultures in bacteremic patients is questionable, they may still be prescribed too liberally, resulting in low rates of positive outcomes. Although routine follow-up blood culture is recommended in Staphylococcus aureus bacteremia and in infective endocarditis [17], the need for mandatory confirmation of negative conversion in KpB has not been adequately addressed. The objective of the present study was to determine the risk factors for persistent bacteremia and to reevaluate the need for routine follow-up blood cultures in KpB.

Methods

Study setting and patients

Data on all episodes of KpB were collected from the electronic medical record systems between January 2007 and December 2011 at Seoul National University Hospital (a 1600-bed tertiary-care hospital, Seoul, Korea) and Seoul National University Bundang Hospital (a 900-bed tertiary-care hospital, Seongnam, Korea). All adult patients (≥ 18 years) who suffered their first KpB episodes were enrolled. Cases of polymicrobial infection, and repeated episodes of KpB in the same patient, were excluded.

Case- and control-patients were those with persistent KpB and non-persistent KpB, respectively. They were matched 1-to-3 according to age and gender using a greedy algorithm. The data reviewed included primary site of infection, Charlson’s co-morbidity weighted index [18], intensive care unit (ICU) admission during the episode of KpB, severity of illness as measured by the Acute Physiology and Chronic Health Evaluation (APACHE) II score [19] and Pitt bacteremia score [20]; also microbiologic characteristics, both appropriateness of empirical antibiotic regimen and length of time from initial blood culture to antibiotic administration, as well as presence of fever, leukocytosis, and the trend of C-reactive protein (CRP) on the second day after initial blood culture. We also assessed the nosocomial or community onset of KpB, presentation with septic shock, presence of metastatic infection, antibiotic or immunosuppressive therapy within the previous year, history of solid organ transplantation, any invasive procedure undertaken and neutropenia within 72 hours before the onset of KpB, together with in-hospital mortality, and length of hospital stay.

The study was approved by the institutional review board of Seoul National University Hospital and Seoul National University Bundang Hospital.

Definitions

The initial blood culture was defined as the first positive blood culture for K. pneumoniae. Persistent KpB was defined as finding of K. pneumoniae in more than one separate blood cultures for longer than a 2-day period in a single infection episode. When an isolate was resistant to both cefotaxime and ceftazidime, it was suspected of being extended-spectrum β-lactamase (ESBL)-producing K. pneumoniae. A nosocomial infection was defined as an infection that occurred more than 48 hours after admission to hospital [21]. Fever and leukocytosis were defined as body temperature of ≥38°C and white blood cell count of >10,000 /μL. Neutropenia was defined as an absolute neutrophil count below 500/mL. Invasive procedures included insertion of a central venous catheter, endoscopy, endoscopic retrograde cholangiopancreatography, insertion of a nasogastric tube, bronchoscopy, parenteral nutrition, and insertion or manipulation of a percutaneous transhepatic biliary drainage [22].

Antimicrobial therapy was considered inappropriate if the isolate was not susceptible to the chosen antibiotic in vitro. The presence of at least two of the following: fever, leukocytosis, and lack of CRP decrease on the second day after initial blood culture, was pre-defined as an unfavorable treatment response before data collection.

Microbiologic analyses

BacT/ALERT FA and FN (bioMe´rieux, Durham, North Carolina) was used for all blood cultures. Antimicrobial susceptibility was identified with a Microscan WalkAway-96 (Siemens Healthcare Diagnostics, Deerfield, Illinois) and VITEK 2 (bioMe´rieux, Marcy L’etoil, France), using the criteria of the Clinical and Laboratory Standards Institute (CLSI) guidelines. For suspected ESBL-producing K. pneumoniae isolates, ESBL production was confirmed by the double disk synergy test according to the CLSI performance standards [17].

Statistical analyses

Conditional logistic regression was used in univariable and multivariable analyses to manage matched data. Variables with P values of < 0.10 in the univariable analysis or of clinical importance (i.e., intra-abdominal infection) were subjected to multivariable analysis. To evaluate the proposed scoring system, a receiver operating characteristic (ROC) analysis was carried out. P <0.05 was considered statistically significant. PASW for Windows (version 18 software package; SPSS Inc., Chicago, IL, USA) was used in the analyses.

Results

Rate of follow-up blood culture and incidence of persistent KpB

Of a total of 1372 KpB episodes during the study period, 1068 were included in this study, after excluding 173 cases of polymicrobial infection and 131 cases of repeated KpB. Among the 1068 enrolled cases of KpB, follow-up blood cultures were performed in 862 cases (80.7%), and their incidence was highest on the second day after the initial blood culture (Additional file 1: Figure S1). Sixty-two (7.2%) were found to be persistent KpB, and 186 non-persistent KpB cases were matched with the latter as described above.

Risk factors for persistent KpB

Univariable analysis of the data indicated that the factors significantly associated with persistent KpB as opposed to non-persistent KpB included higher Charlson’s comorbidity weighted index score, ICU admission, nosocomial infection, immunosuppressive therapy during the previous year, history of solid organ transplantation, invasive procedure during the previous 72 hours, fever and lack of CRP decrease on the second day after initial blood culture, ESBL-producing KpB, inappropriate empirical antibiotic use, and in-hospital mortality (Table 1).

Table 1 Univariable analysis for risk factors associated with persistent Klebsiella pneumoniae bacteremia (cases) vs. non-persistent Klebsiella pneumoniae bacteremia (controls)
Full size table

Multivariable analysis revealed that the independent risk factors for persistent KpB were intra-abdominal infection (adjusted odds ratio [aOR], 2.99; 95% confidence interval [CI], 1.07-8.31; P = 0.036), higher Charlson’s comorbidity weighted index score (aOR, 1.18 per each point; 95% CI, 1.01-1.36; P = 0.032), history of solid organ transplantation (aOR, 92.23; 95% CI, 1.27-6689.96; P = 0.038), and unfavorable treatment response (aOR, 4.79; 95% CI, 1.89-12.14; P = 0.001). The concept of unfavorable treatment response was introduced and substituted for fever, leukocytosis, and lack of CRP decrease in the multivariable analysis, to avoid the multicolinearity among those three variables. Neither ESBL-producing KpB nor inappropriate empirical antibiotic use was found to be an independent risk factor (Table 2).

Table 2 Independent risk factors for persistent Klebsiella pneumoniae bacteremia
Full size table

Percentage of persistent KpBaccording to clinical score

To estimate the likelihood of persistent KpB, a scoring system was introduced, based on the presence or absence of four risk factors viz. intra-abdominal infection, nosocomial KpB, fever and lack of CRP decrease, the last two being assessed on the second day after initial blood culture. According to their odds ratio in a conditional logistic regression with these four risk factors, the weighted scores given for the risk factors were 1, 2, 3, and 2, respectively (Table 3).

Table 3 Odds ratios and adjusted weighted scores for the four variables used in the scoring system
Full size table

The frequency of persistent KpB exceeded 50% when the score was greater than five; conversely, it was only 4.9% when the score was zero to one with no risk factors, or with intra-abdominal infection as the only risk factor (Figure 1). This indicates that persistent KpB is likely to be rarer among patients with fewer risk factors. A ROC analysis showed that the area under the curve was 0.78 (P <0.0001) and the best cut-off was 2 (sensitivity, 0.81; specificity, 0.64, Additional file 2: Figure S2).

Figure 1
figure 1

Percentage of persistent Klebsiella pneumoniae bacteremia and number of cases according to clinical score. There was a positive relationship between clinical score and percent persistent KpB.

Full size image

Discussion

K. pneumoniae is one of the most important pathogens causing nosocomial and community-acquired infections, and clinicians sometimes encounter persistent KpB in clinical practice. Despite the fact that the CLSI guidelines do not recommend routine follow-up blood cultures except in cases of Staphylococcus aureus bacteremia and infective endocarditis [17], our understanding of the occurrence of persistent KpB and the frequency of follow-up blood cultures is far from complete. The present study showed that, although follow-up blood cultures were performed in 80.7% of the cases, only 7.2% were found to be persistent KpB. Furthermore, more than two consecutive blood cultures were performed in 53.2% of the patients with non-persistent KpB in the present study (Data not shown). These results imply that most of these follow-up blood cultures would have been unwarranted. Studies have shown that only 4 to 7% of blood cultures are positive owing to its liberal prescription [12, 15, 16], and our data with follow-up blood cultures also point to a similar situation. Our multivariable analysis revealed that unfavorable treatment response was an independent risk factor for persistent KpB, along with intra-abdominal infection, high Charlson’s comorbidity weighted index score, and prior solid organ transplantation. All this suggests that, when treating suspected KpB patients, the clinical findings should be considered carefully before opting for follow-up blood culture, to reduce indiscriminate prescription and the ensuing cost and patient inconvenience.

Severe underlying disease and prior organ transplantation are established risk factors for KpB [4, 5, 23, 24], as they were for persistent KpB in this study. An intra-abdominal origin of KpB is a known risk factor for mortality [4, 5], and it was also significantly associated with persistent KpB in our analysis. Although the KpB of respiratory tract origin has been documented to be a risk factor for mortality [4, 5, 23], it was not significantly associated with persistent KpB in this study. It would be interesting to find their association further by taking many factors into account, including the limitations of this study as explained below.

According to Kang et al.[25, 26], ESBL-producing organism and inappropriate empirical antibiotic are not risk factors for increased mortality in bloodstream infections due to K. pneumoniae. We also found that production of ESBL was not an independent risk factor for persistent KpB, which indicates that follow-up blood cultures are not routinely needed in cases of ESBL-producing KpB. However, inappropriate empirical antibiotic regimen may call for further evaluation because of its borderline P value.

Various definitions of persistent bacteremia have been used in Staphylococcus aureus bacteremia [27–29]. Our definition of persistent bacteremia was not as strict as others because both the failure of appropriate antimicrobial therapy and prolonged periods of bacteremia were not essential elements of our consideration. If a stricter definition had been used, the incidence of persistent KpB would have been further reduced, which supports our suggestion that more care should be taken before opting for follow-up blood culture.

Our scoring system, which has fair discriminative ability, could be used as a guide for whether a follow-up blood culture is needed, since there was a positive relationship between clinical score and percent persistent KpB (Figure 1): the percentage of persistent KpB declined sharply as the score decreased, and less than 5% of the patients with zero or one point had persistent bacteremia. On the other hand, follow-up blood culture could be justified in patients with more than five points because more than a half of those would be expected to suffer from persistent bacteremia. According to our data, if follow-up blood cultures were only carried out in patients with scores exceeding 2 points, more than 50% of follow-up blood cultures could be avoided (Figure 1).

There are some limitations to this study. There is a possibility of bias because it was a retrospective study with a limited sample size. Although our scoring system appeared to be useful, as discussed above, it should be further assessed on a larger sample. We were not able to evaluate the possible impact of age and gender on persistent KpB by matching for these factors because of the design of this study. Also, there may have been selection bias in the 19.3% of patients without follow-up blood culture, one third of whom died by the 2nd day after blood culture (data not shown).

Persistent KpB showed significantly worse outcomes, such as in-hospital mortality (P = 0.044) and ICU admission during the KpB episode (P = 0.029) in univariable analyses (Table 1). However, we did not perform multivariable analysis for mortality-related factors, mainly because the study design was focused on persistence of bacteremia, and also the matching was carried out in that context.

Finally, this study was undertaken in university-affiliated tertiary-care hospitals where a large proportion of the patients might have severe underlying diseases including malignancy. However, 64.1% and 25.0% of our cases involved community-onset and community-acquired KpB, respectively (data not shown). The rate of persistent KpB would be expected to be much lower in a community-based hospital since a high Charlson’s comorbidity weighed index score is an independent risk factor for persistent KpB and the comorbidities in patients in community-based hospitals are likely to be less severe.

Conclusions

In conclusion, unfavorable treatment response on the second day after initial blood culture, intra-abdominal infection, high Charlson’s comorbidity weighted index score, and prior solid organ transplantation are independent risk factors for persistent KpB. Since patients with persistent KpB were rare, especially among individuals with few risk factors, routine follow-up blood culture may not be justified. More careful clinical assessment before deciding on follow-up blood culture would reduce costs and inconvenience to patients.