Факторы риска, связанные с позитивом и смертностью в связи с инфекциями в крови: Демографическое исследование пациентов с злокачественными заболеваниями B-Cell.

Introduction

Infections remain one of the leading causes of mortality in patients with B‐cell malignancies [ 1 , 2 , 3 , 4 ] and up to one‐third of these patients will suffer from a bloodstream infection (BSI) at some point [ 5 ], with 30‐day mortality rates of 22%–28% [ 6 , 7 ]. Optimal antibiotic treatment requires consideration of local pathogen spectra and current resistance profiles. However, bacterial aetiology and patterns of multidrug resistance differ across regions, even within Europe [ 8 , 9 ]. Hence, surveillance of pathogens causing BSI is crucial.

Overall, the positivity rate of blood cultures is around 10%, and bottles with no growth are usually reported negative after 5–6 days of incubation [ 10 ]. In patients with BSI, reports on Gram stain and species are often available already within the first day, and susceptibility within 1–3 days. BSIs typically warrant intravenous antibiotics, at least for some days until clinical improvement. The uncertainty concerning BSIs in the first days after blood cultures are drawn can potentially result in unnecessary and needlessly long hospital admissions.

The mortality of BSI may differ according to underlying factors. Studies on patients with B‐cell malignancies have identified several factors, such as advanced age, inappropriate antibiotic therapy, prolonged neutropenia, acute renal failure, and use of vasopressors to be associated with short‐term mortality [ 7 , 11 , 12 , 13 ]. However, studies from Northern Europe are lacking, and conclusions do not necessarily apply to a Scandinavian population, partly due to differences in treatment strategies and bacterial epidemiology.

We examined time‐to‐positivity (TTP), and factors associated with blood culture positivity (BCP) and 30‐ and 90‐day mortality of BSI in a Danish population of patients with a B‐cell malignancy. Additionally, we provide insight to the local bacterial spectrum and the changes during a 10‐year period.

Materials and Methods

Study Design

This population‐based cohort study was based on all adult patients (≥ 18 years) diagnosed with lymphoma, multiple myeloma (MM) or chronic lymphocytic leukaemia (CLL) in The North Denmark Region, Denmark between January 1st, 2013, and May 1st, 2022. We restricted the cohort to patients who had at least one blood culture drawn and followed them from the date of cancer diagnosis until death or 31st December 2022; whichever came first.

Data Collection

Patients were identified in The Danish National Lymphoma Registry [ 14 ], The Danish National CLL Registry [ 15 ] and The Danish National MM Registry [ 16 ]. See Supporting Information (SM) for a complete list of included cancer diagnoses (Figure S1 ), and handling of two or more cancer diagnoses of interest. The completeness of these registries is close to 100% [ 17 , 18 ], and the missing patients (< 2%) consist of patients with a B‐cell malignancy who are not receiving treatment at a hospital [ 19 ]. Furthermore, the data quality is high with positive predictive values of 87%–100% [ 20 ]. From these databases, patient‐related baseline data (e.g., date of birth and cancer diagnosis, performance score, and Charlson Comorbidity Index score) were collected. We used the unique Central Person Register (CPR) number to link data from cancer registries to electronic patient health records. From these records, blood tests and vital parameters were collected. See SM for a full list of collected variables (Table S1 ), and handling of outliers (Table S2 ).

We retrieved information on positive blood cultures (species, number of positive blood culture bottles, time‐to‐positivity, and antimicrobial susceptibility testing results) from the regional microbiological database, which is maintained by the Department of Clinical Microbiology, Aalborg University Hospital. See SM for handling of blood cultures.

All data was pseudonymized prior to analysis.

Definitions

Performance status was assessed by the Eastern Cooperative Oncology Group Performance Status (PS) Scale and comorbidities were categorized according to the Charlson Comorbidity Index (CCI) score. Both variables were registered by the attending physician in the cancer registry at the time of cancer diagnosis. Neutropenia was defined as a neutrophil count of < 0.5 cells/μL. Sepsis was defined as a Quick Sequential (Sepsis‐Related) Organ Failure Assessment (qSOFA) Score of ≥ 2 [ 21 ], calculated using respiratory rate, Glascow Coma Scale and systolic blood pressure obtained within the same hour (most abnormal value recorded). Shock was defined as a mean arterial pressure (MAP) of < 65 mmHg, calculated as systolic blood pressure + 2 times diastolic blood pressure divided by 3 [ 22 ]. Based on registry data at the time of diagnosis, medical antineoplastic treatment comprised all intended systemic modalities, including chemotherapy, immunotherapy, targeted therapy, monoclonal antibodies, corticosteroids, and hematopoietic stem cell transplantation, while excluding surgery and radiation therapy.

An episode of BSI was defined as detection of ≥ 1 microorganism(s) in a blood culture set (defined as a positive blood culture). A positive blood culture was re‐registered as a negative blood culture if a clinical microbiologist labelled the microorganism(s) as contamination/irrelevant, based on clinical presentation (e.g., Cutibacterium acnes, Micrococcus luteus , Staphylococcus hominis ). A new unique episode of BSI in the same patient was defined as a blood culture isolate that differed from the previous one or an interval of at least 30 days between two positive blood cultures. Polymicrobial BSI was defined as two or more microorganisms isolated within 48 h from the same patient. These definitions are consistent with the North Denmark Bacteremia Research Database definitions [ 23 ]. Previous episodes of BSIs were defined as any BSI within five years prior to the blood culture of interest.

Statistical Analyses

Baseline characteristics were summarized as counts (%) and medians (interquartile ranges [IQR]). The distribution of causative pathogens, prevalence of multidrug resistant (MDR) organisms, and TTP for all pathogens were described and visualized in bar plots and time‐trend graphs. Follow‐up time was estimated using the reverse Kaplan–Meier (KM) method.

Multiple statistical approaches were used to examine associations between potential risk factors and BCP. Variables were selected based on existing literature and clinical expert knowledge and included baseline characteristics, blood tests, and vital parameters. In case of missing data, multiple imputation by chained equations (MICE; 1000 imputations) was applied, but only for variables with < 10% missingness and ≤ 2 missing values per patient. Imputed datasets were pooled using Rubin's rules and used exclusively for regression analyses. We employed least absolute shrinkage and selection operator (LASSO) regression for variable selection, retaining all variables with non‐zero coefficients, as this method is well suited for identifying relevant predictors in the presence of numerous correlated covariates. Given that LASSO was applied across multiply imputed datasets generated by MICE, variable selection was determined by selection frequency, defined as the proportion of imputed datasets in which a given variable was retained by the LASSO procedure. Variables with a selection frequency exceeding zero were considered eligible for further analysis. These variables were subsequently entered into univariate logistic regression models to quantify the magnitude and direction of individual associations. This two‐step approach avoided unstable multivariable models given the limited number of events and allowed presentation of clinically interpretable odds ratios (ORs) with 95% confidence intervals (CIs). Analyses were restricted to each patient's first blood culture and variables were collected within seven days prior to the blood culture to capture clinically relevant values close to the infection event.

The same analytical strategy (LASSO for variable selection followed by univariate logistic regression) was applied to identify factors associated with 30‐ and 90‐day mortality following a BSI. For mortality analyses, variables were collected from seven days before the positive blood culture until two days after. Only each patient's first BSI episode was included.

Survival probabilities were estimated using the KM estimator, and all‐cause cumulative 30‐ and 90‐day mortality rates were calculated using the reverse KM method and presented with 95% CIs. p ‐values were calculated using the z‐test. KM estimates with 95% CIs were plotted, and log rank‐tests were used to compare mortality between groups (e.g., by age, cancer type, pathogen). Survival analyses were conducted using complete‐case data.

All statistical analyses were performed using R (Version 4.3.2, R Core Team, 2023).

Results

We identified a total of 2778 patients diagnosed with a B‐cell malignancy during the study period. Among these, 1592 (57.3%) had at least one blood culture drawn and were included for further analysis. Of these, lymphoma was the most frequent cancer diagnosis (53.3%), followed by MM (32.2%) and CLL (14.5%). As shown in Table 1 , most patients were 65 years or older (72.5%) and males (59.7%). At the time of cancer diagnosis, 44.9% of the patients were PS 0 and only 9.0% were PS 3–4. During a median follow‐up of approximately 6 years (70.5 months [95% CI: 67.4–74.5]), 341 (21.4%) had at least one episode of BSI in the study period covering 436 BSI episodes.

Patients with and without a BSI were comparable in respect to age, sex and CCI score, but those with BSI had a slightly worse PS (Table 1 ). Median age was slightly higher among patients with lymphoma compared to CLL and MM (74 years vs. 72 and 71 years, respectively). In patients with CLL, 57.8% ( n = 133) were PS 0, whereas 46.3% ( n = 393) with lymphoma and 36.6% ( n = 188) with MM were PS 0. Among patients with CLL, 36.4% ( n = 84) received medical antineoplastic treatment immediately after the cancer diagnosis, whereas it was 71.2% ( n = 365) and 81.8% ( n = 694), respectively, in patients with MM and lymphoma. In patients with lymphoma and MM, the median time from cancer diagnosis to first BSI was approximately 9 and 12 months, respectively, whereas CLL patients had a median time from cancer diagnosis to first BSI of approximately 20 months.

Microbiological Findings

Among 436 BSI episodes, 40 (9.2%) were polymicrobial, totalling 484 positive blood culture isolates. Of these, 250 (52.5%) were Gram‐negative bacteria (GNB), 226 (47.5%) were Gram‐positive bacteria (GPB), and the remaining 8 (1.7%) were yeasts, primarily Candida albicans . In Figure 1 , all isolated pathogens are presented either as species for the most frequent, and in clinically relevant categories for the less frequent. The most frequently isolated pathogens were Escherichia coli (24.8%), Klebsiella spp. (9.3%), and Enterococcus spp. (9.1%). When analysing only on species level, the most frequent were E. coli (24.8%), Staphylococcus aureus (8.1%), Enterococcus faecium (7.0%), and Streptococcus pneumoniae (6.8%). Anaerobic bacteria were detected in 5.6% of the blood cultures.

During the 10‐year study period, the dominance of GPB shifted to GNB from 42.5% (31/73) GNB in 2013–2014 to 58.1% (72/124) in 2021–2022. This was mainly caused by an increase in the frequency of E. coli , as shown in Figure S2 in SM , from 20.5% in 2013–2014 to 31.5% in 2021–2022.

Recurrent BSI, that is, more than one episode of BSI, differed by cancer type: CLL 10.0% ( n = 23), lymphoma 19.9% ( n = 169), and MM 29.0% ( n = 149) (see Table S3 in SM). Patients with CLL had a high frequency of GNB (64.3%), mainly by E. coli , which was detected in 44.8% of the positive blood cultures (see Table S4 in SM). In patients with MM, S. pneumoniae was more frequent than in the general study population (12.8% vs. 6.8%). In patients with lymphoma, enterococci were the causing pathogen in 12.8% of the blood cultures, which was higher than for MM (4.9%).

Extended Spectrum Beta‐Lactamase (ESBL) producing E. coli and Klebsiella spp. accounted for 3.8% (18/476) of all isolated bacteria (prevalence in GNB‐only was 7.2%), whereas the prevalence of carbapenemase‐producing Enterobacterales was low ( n < 5). There were no cases of vancomycin‐resistant enterococci (VRE) or methicillin‐resistant S. aureus (MRSA).

The median TTP across all positive blood cultures was 10.9 h (IQR: 7.2–15.3) and the TTP was < 24 h in 90.1% of the isolates. Only 1.9% of the isolates had a TTP > 48 h. Median TTP for all pathogens is provided in Table S5 in SM.

Risk of Blood Culture Positivity

The LASSO regression identified the following variables as associated with BCP: age, sex, cancer type, PS, CCI, neutrophil count, eGFR‐level, CRP‐level, sepsis, shock, and time from cancer diagnosis to blood culture (all tested variables with corresponding selection frequencies are shown in Tables S6–S8 in SM). When testing the LASSO‐selected variables in an univariate analysis, we found that patients with MM had a higher risk of BCP than patients with CLL (OR: 2.06 [95% CI: 1.08–3.94]), and that having neutropenia carried an increased risk of BCP (OR: 2.41 [95% CI: 1.52–3.80]) (Figure 2 ). Time from cancer diagnosis to blood culture of 6–12 months increased the risk of BCP (OR: 1.68 [95% CI: 0.95–2.98]). The same analyses performed on a complete‐case dataset showed highly similar results (Figure S3 in SM).

Mortality Following a Bloodstream Infection

The LASSO regression identified age, cancer type, PS, neutrophil count, eGFR‐level, sepsis, polymicrobial BSI, MDR bacteria, number of previous BSIs, and time from cancer diagnosis to BC to be associated with 30‐day mortality. Age, cancer type, PS, CCI, eGFR‐level, sepsis, TTP, MDR, and time from cancer diagnosis to BC were identified as associated with 90‐day mortality following a BSI. In the univariate analysis, we found that increasing age and PS were associated with increased risk of both 30‐ and 90‐day mortality (Figures 3 and 4 ). Moreover, patients presenting with sepsis (qSOFA score ≥ 2) had a four‐fold higher OR of dying within 30 days (OR: 4.22 [95% CI: 2.32–7.68]) and more than three times higher OR of dying within 90 days (OR: 3.20 [95% CI: 1.86–5.51]). An eGFR‐level < 60 mL/min/1.73 m 2 increased the risk of both 30‐ and 90‐day mortality, yielding ORs of 2.40 (95% CI: 1.29–4.45) and 1.98 (95% CI: 1.19–3.30), respectively. Having polymicrobial BSI or MDR bacteria indicated an increased risk of 30‐day mortality, with ORs of 1.91 (95% CI: 0.80–4.57) and 2.13 (95% CI: 0.71–6.36), respectively. Neutropenia showed an OR of 0.94 (95% CI: 0.49–1.82) for 30‐day mortality, and TTP an OR of 1.28 (95% CI: 0.63–2.60) for 90‐day mortality, with wide confidence intervals spanning 1.0 in both cases. Analyses performed on complete‐case datasets are shown in Figures S4 and S5 in SM.

Association between the causative pathogen of the BSI and mortality was also assessed (Figure 5 ). Patients with GNB had two‐fold higher odds of 30‐day mortality and 71% higher odds of 90‐day mortality compared to those with GPB (OR: 2.03 [95% CI: 1.14–3.61] and OR: 1.71 [95% CI: 1.05–2.79], respectively). At species level, we found a lower 90‐day mortality among patients with pneumococcal BSI compared to E. coli BSI (OR: 0.26 [95% CI: 0.07–0.92]). Yeast was associated with higher 30‐day mortality compared to E. coli , but the estimate was of poor precision (OR: 10.65 [95% CI: 1.05–108.04]). No other pathogen‐specific mortality differences were detected, likely reflecting low event numbers.

All‐Cause Mortality and Survival Estimates

The cumulative all‐cause mortality following the first BC drawn after time of cancer diagnosis was 9.5% (95% CI: 8.0–11.0) within 30 days and 16.2% (95% CI: 14.3–18.1) within 90 days. 30‐day mortality was significantly higher following a positive BC compared to a negative BC (14.9% [95% CI: 9.3–22.8] vs. 9.0% [95% CI: 7.5–10.7], p = 0.034), as was 90‐day mortality (23.1% [95% CI: 16.2–31.9] vs. 15.5% [95% CI: 13.6–17.6], p = 0.029). Overall, when pooling all first BSI episodes, 30‐day and 90‐day mortality were 17.9% (95% CI: 13.7–22.0) and 27.1% (95% CI: 22.2–31.8), respectively. Figures S6 and S7 in SM show overall survival (OS) following first BSI stratified on the predictive variables. All the plotted survival estimates correlates with the finding of the regression analyses regarding differences between the chosen cut‐offs. The curves split within the first few weeks, indicating that the difference in survival is caused by the BSI itself and not cancer mortality or other causes.

Discussion

In this population‐based cohort study, we analysed the epidemiology of BSIs in patients with B‐cell malignancies. The most frequent causative pathogen was E. coli and we observed a shift from GPB to GNB dominance, likely explained by the rise in BSIs caused by E. coli . Having MM, neutropenia, impaired kidney function, high CRP, sepsis, and shock was found to be associated with BCP and advanced age, poor PS, impaired kidney function and sepsis was associated with mortality following BSI. All‐cause mortality following a BSI was approximately twice as high as that observed following a negative blood culture.

We found that 21.4% of the patients had at least one episode of BSI during the 10‐year study period. This finding corresponds with a previous Danish study on haematological cancers (1992–2002) where 21.0% developed bacteremia within a median follow‐up of 1.1 years [ 24 ]. Overall, our study population had good PS at the time of cancer diagnosis and the follow‐up was long, which strengthens the reliability of the findings. The distribution of the most frequently identified pathogens, together with the observed increase in E. coli , aligns with other comparable reports, further supporting the reliability of our data [ 6 , 10 , 25 ].

The high frequency of E. faecium in our cohort (especially among lymphoma patients) reflects a global trend, where enterococci now are among the most common causes of GPB, including VRE, and are more frequently found in patients with cancer and linked to hospital‐acquired infections, previous exposure to piperacillin/tazobactam, and carbapenems [ 26 , 27 , 28 ]. Patients with B‐cell malignancies are often exposed to broad‐spectrum antibiotics, such as piperacillin/tazobactam and nosocomial pathogens, likely explaining the high frequency of enterococcal infections in this group. The high frequency of S. pneumoniae among patients with multiple myeloma is consistent with previous studies and is likely explained by the hypogammaglobulinemia seen in these patients, which increases susceptibility to infections caused by encapsulated bacteria such as S. pneumoniae [ 24 , 29 ].

In agreement with the low prevalence of MDR bacteria in Denmark, we found an overall prevalence of < 5% of ESBL‐producing E. coli . A recent study of antimicrobial resistance found that among bacteria of interest ( S. aureus , Enterococcus faecalis , E. faecium , E. coli , Klebsiella pneumoniae , Pseudomonas aeruginosa and Acinetobacter spp.), 6.0% (99/1640) of the isolates in a Norwegian cohort were MDR bacteria [ 30 ]. Analysing the same group of bacteria, we found 6.3% (16/253) MDR bacteria in our cohort, underscoring the similarity of MDR rates across Scandinavian countries. In contrast, the English cohort from the same study reported a rate of 16.0% (309/1929), highlighting substantial MDR epidemiology even within Europe and underlining the importance of continuous local surveillance.

MDR bacteremia was identified by LASSO regression as a predictor of 30‐day mortality, with an OR consistent with a clinically relevant effect (OR: 2.13 [95% CI: 0.71–6.36]). Although precision was limited, reflecting the low frequency of MDR in our cohort, our finding is in line with previous studies reporting increased mortality associated with MDR pathogens [ 30 , 31 ].

The median TTP found in our cohort aligns with other studies [ 32 , 33 ]. Although TTP was selected by the LASSO model, univariate analysis did not demonstrate an independent association with 30‐day or 90‐day mortality, suggesting that its predictive value may depend on the presence of other co‐occurring risk factors. These findings support the rationale for antibiotic de‐escalation and early discharge in cases of negative blood cultures after 24–48 h, an approach previously proposed and found to be safe [ 34 , 35 ].

Our results further suggest that it may be possible to predict BCP as early as at time of admission based on routinely obtained clinical parameters. This has not previously been studied specifically in patients with B‐cell malignancies. In non‐cancer cohorts, risk factors such as age, CRP, pulse rate, temperature, neutrophil count, platelet count, and leukocyte count have been associated with BCP [ 36 , 37 ], consistent with our findings. Notably, the ORs for BCP associated with neutropenia, sepsis, renal failure, elevated CRP, and hypotension were similar, ranging from 1.89 to 2.41 (Figure 2 ). This likely reflects collinearity, as these variables share a common biological basis as markers of infection severity. Although CCI was selected by the LASSO model, univariate analysis estimated ORs close to 1.0, suggesting that its contribution may be context‐dependent rather than independently predictive. This should be interpreted with caution, as comorbidity data were recorded at time of diagnosis and may not reflect the patients' comorbidity burden at the time of the BC. The association between time from cancer diagnosis to BC and BCP should likewise be interpreted with caution, as patients presenting later in their disease course represent a selected population who survived to that time point. Survivorship bias may thus have confounded our estimates, as long‐term survivors may differ systematically from earlier presenters with respect to disease trajectory, treatment intensity, and PS.

We found that among patients with BSI, increasing age, kidney failure, and sepsis were associated with both 30‐ and 90‐day mortality. Since BSI mortality is typically driven by organ failure, these associations are biologically plausible and consistent with previous reports [ 11 , 12 , 13 ]. These findings provide candidates for a post‐BSI mortality prediction model to guide targeted interventions and earlier discharge.

An important limitation of this study is the lack of certain clinical information, including drug prescriptions (such as antibiotic use and steroids), source and type of BSI (e.g., pneumonia, mucosal barrier injury‐related, or catheter‐related), disease stage at the time of BSI, and chemotherapy regimen. Unfortunately, those data were not available or incomplete in the registries. Moreover, commonly used prophylactic antimicrobial therapy such as co‐trimoxazole (trimethoprim/sulfamethoxazole) may have suppressed bacterial growth in the blood cultures, potentially leading to an underestimation of the true rate of BCP.

Another important limitation is the external validity, as results may primarily apply to countries with similar health care systems and bacterial epidemiology. Importantly, the low prevalence of MDR pathogens in our cohort differs markedly from other regions of Europe and beyond. Our results may thus not be applicable to settings with a high prevalence of MDR pathogens.

The associations presented in this study must be interpreted with caution, as exploratory retrospective observational studies cannot establish causality. The data‐driven variable selection used in this study, however, reduces the risk of overfitting, is straightforward to replicate, and yields clinically interpretable effect estimates. The design using clinical information (unrelated to study purpose) is well suited for exploratory analyses like this with a potentially large set of predictors. Another strength is the use of Danish nationwide registries, with high completeness quality, making them well suited for real‐world analyses.

Exploratory analyses are both logical and necessary to establish whether any associations even exist. The results of this exploratory study provide a foundation for hypothesis generation and call for future validation in a robust and data‐driven model establishing causality.

In conclusion, this study shows that BSI remain a frequent complication with high mortality in patients with a B‐cell malignancy. Over the 10‐year study period, an increase in E. coli drove a shift in dominance from GPB to GNB. MDR rates are still relatively low in our setting; however, ongoing surveillance is crucial. Further, the results support antibiotic de‐escalation and early discharge. This is guided by the short TTP and possibility of BCP prediction using clinical values available at time of admission. We show that patient‐related factors, laboratory biomarkers, and clinical assessment tools may help identify patients at high risk of BCP and poor outcomes. These findings provide candidates for inclusion in future development of risk‐stratification models to individualize infection management in patients with B‐cell malignancies. Whether such models can improve survival, reduce costs and enhance quality of life warrants prospective studies.

Author Contributions

Performed analyses: M.H.E.; Drafted 1st version of manuscript: M.H.E.; Collected data: M.R.S., K.K.S.; Concept and design: M.H.E., T.C.E.‐G., A.D.F., P.J., L.Ø.P., M.R.S., K.K.S.; Interpreted results: M.H.E., T.C.E.‐G., A.D.F., P.J., L.Ø.P., K.K.S.; Critically revised the manuscript: M.H.E., T.C.E.‐G., A.D.F., P.J., I.E.N., L.Ø.P., M.R.S., K.K.S.; Approved final version: M.H.E., T.C.E.‐G., A.D.F., P.J., I.E.N., L.Ø.P., M.R.S., K.K.S.

Funding

This work was supported by Sygeforsikringen “danmark” (grant number 2022‐0086) and Danish Data Science Academy (grant number 2023‐1210).

Ethics Statement

According to Danish law, ethical approval and informed consent are not required for registry‐based studies. Access to data was granted by The North Denmark Region according to article 30 (project ID: F2023‐002) and §46 (case number 2023‐002807).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: List of all data collected.

Table S2: Limitations for Outliers.

Table S3: Baseline data sorted by cancer type.

Table S4: Frequency of all isolated pathogens in total and by cancer type.

Table S5: Median time to positivity according to group of pathogen.

Table S6: Selection frequencies from LASSO—association with a positive blood culture.

Table S7: Selection frequencies from LASSO—association with 30‐day mortality following a bloodstream infection.

Table S8: Selection frequencies from LASSO—association with 90‐day mortality following a bloodstream infection.

Figure S1: Included cancer diagnoses and corresponding ICD‐10 codes.

Figure S2: Distribution of the five most frequent bacteria during the 10‐year study period.

Figure S3: Forest plot of association with blood culture positivity on complete‐case data.

Figure S4: Forest plot of association with 30‐day mortality on complete‐case data.

Figure S5: Forest plot of association with 90‐day mortality on complete‐case data.

Figure S6: Survival following bloodstream infection stratified by age, kidney function, sepsis, and performance status.

Figure S7: Survival following bloodstream infection stratified by gram stain.

Acknowledgments

During the preparation of this work the author(s) used OpenAI ChatGPT in order to assist with language editing (spelling, grammar, and phrasing). After using this tool/service, the author(s) reviewed and edited the content as needed and take full responsibility for the content of the published article.

Data Availability Statement

It is not permitted under Danish law to freely share the data used in the study. However, researchers can apply for the necessary approvals within the limits of the law and then conduct the necessary studies.

References

1 M. E. Cabrera , M. Civallero , Y. Stepanishyna , et al., “ Cause‐Specific Mortality in a Cohort of 1,435 Patients With Hodgkin Lymphoma Treated Between 1985 and 2014: A Nationwide Chilean Cohort Study ,” JCO Global Oncology 11 ( 2025 ): e2400344 , 10.1200/go-24-00344 . 40324117

2 C. H. Blimark , K. Carlson , C. Day , et al., “ Risk of Infections in Multiple Myeloma. A Population‐Based Study on 8,672 Multiple Myeloma Patients Diagnosed 2008‐2021 From the Swedish Myeloma Registry ,” Haematologica 110 , no. 1 ( 2025 ): 163 – 172 , 10.3324/haematol.2024.285645 . 39021214 PMC11694133

3 L. van der Straten , M. D. Levin , M. A. W. Dinnessen , et al., “ Causes of Death Among Patients Diagnosed With Chronic Lymphocytic Leukemia: A Population‐Based Study in The Netherlands, 1996–2020 ,” Hema 8 , no. 11 ( 2024 ): e70015 , 10.1002/hem3.70015 . PMC11555299 39534384

4 L. Y. Bei , C. Y. Shang , J. Z. Wu , et al., “ Cause‐Specific Mortality in a Population‐Level Cohort of Diffuse Large B‐Cell Lymphoma Following Chemotherapy in the Early 21st Century ,” Annals of Hematology 103 , no. 5 ( 2024 ): 1675 – 1685 , 10.1007/s00277-024-05619-w . 38228775

5 B. W. Teh , S. J. Harrison , M. A. Slavin , and L. J. Worth , “ Epidemiology of Bloodstream Infections in Patients With Myeloma Receiving Current Era Therapy ,” European Journal of Haematology 98 , no. 2 ( 2017 ): 149 – 153 , 10.1111/ejh.12813 . 27717026

6 M. A. Andersen , C. E. Moser , J. Lundgren , and C. U. Niemann , “ Epidemiology of Bloodstream Infections in Patients With Chronic Lymphocytic Leukemia: A Longitudinal Nation‐Wide Cohort Study ,” Leukemia 33 , no. 3 ( 2019 ): 662 – 670 , 10.1038/s41375-018-0316-5 . 30546080

7 B. Islas‐Muñoz , P. Volkow‐Fernández , C. Ibanes‐Gutiérrez , A. Villamar‐Ramírez , D. Vilar‐Compte , and P. Cornejo‐Juárez , “ Bloodstream Infections in Cancer Patients. Risk Factors Associated With Mortality ,” International Journal of Infectious Diseases 1 , no. 71 ( 2018 ): 59 – 64 , 10.1016/j.ijid.2018.03.022 . 29649549

8 European Centre for Disease Prevention and Control (ECDC) , “Surveillance Atlas of Infectious Diseases [Internet]. [cited 2023 Oct 30],” https://atlas.ecdc.europa.eu/public/index.aspx? .

9 F. Baccelli , M. Aguilar‐Guisado , C. G. Vidal , et al., “ Epidemiology of Resistant Bacterial Infections in Patients With Hematological Malignancies or Undergoing Hematopoietic Cell Transplantation in Europe: A Systematic Review by the European Conference on Infections in Leukemia (ECIL) ,” Journal of Infection 91 , no. 3 ( 2025 ): 106571 , 10.1016/j.jinf.2025.106571 . 40812455

10 R. B. Dessau , C. Ø. Andersen , J. Coia , et al., “ The Increasing Incidence and Mortality of Bacteremia in Denmark From 2010 to 2022: A Population‐Based Nationwide Cohort Study ,” Frontiers in Public Health 20 ( 2024 ): 12 , 10.3389/fpubh.2024.1502893 . PMC11695369 39758201

11 S. Chen , K. Lin , Q. Li , et al., “ A Practical Update on the Epidemiology and Risk Factors for the Emergence and Mortality of Bloodstream Infections From Real‐World Data of 3014 Hematological Malignancy Patients Receiving Chemotherapy ,” Journal of Cancer 12 , no. 18 ( 2021 ): 5494 – 5505 , 10.7150/jca.50802 . 34405012 PMC8364636

12 Y. Tang , Q. Cheng , Q. Yang , et al., “ Prognostic Factors and Scoring Model of Hematological Malignancies Patients With Bloodstream Infections ,” Infection 46 , no. 4 ( 2018 ): 513 – 521 , 10.1007/s15010-018-1151-3 . 29767394

13 M. Tumbarello , T. Spanu , M. Caira , et al., “ Factors Associated With Mortality in Bacteremic Patients With Hematologic Malignancies ,” Diagnostic Microbiology and Infectious Disease 64 , no. 3 ( 2009 ): 320 – 326 , 10.1016/j.diagmicrobio.2009.02.008 . 19345033

14 B. Arboe , P. Josefsson , J. Jørgensen , et al., “ Danish National Lymphoma Registry. Clinical ,” Epidemiology 8 ( 2016 ): 577 – 581 , 10.2147/CLEP.S99470 . PMC5094596 27822102

15 C. da Cunha‐Bang , C. Geisler , L. Enggaard , et al., “ The Danish National Chronic Lymphocytic Leukemia Registry ,” Clinical Epidemiology 8 ( 2016 ): 561 – 565 , 10.2147/CLEP.S99486 . 27822100 PMC5094649

16 P. Gimsing , M. Holmstrøm , T. Wirenfelt Klausen , et al., “ The Danish National Multiple Myeloma Registry ,” Clinical Epidemiology 8 ( 2016 ): 583 – 587 , 10.2147/CLEP.S99463 . 27822103 PMC5094522

17 Danish Myeloma Study Group , “Multiple Myeloma Annual Reports 2012‐2022 [Internet]. [cited 2025 Jan 6],” https://myeloma.dk/index.php/aarsrapporter/ .

18 Danish Lymphoma Group , “Malignant Lymphoma and CLL Annual Reports 2004–2023 [Internet]. [cited 2025 Jan 6],” https://www.lymphoma.dk/aarsrapporter/ .

19 P. d. N. Brown , J. M. Nørgaard , P. Gimsing , et al., “ The Haematological Corporate Database ,” Ugeskrift for Laeger 174 , no. 42 ( 2012 ): 2526 . 23079429

20 B. Arboe , T. C. El‐Galaly , M. R. Clausen , et al., “ The Danish National Lymphoma Registry: Coverage and Data Quality ,” PLoS One 11 , no. 6 ( 2016 ): e0157999 , 10.1371/journal.pone.0157999 . 27336800 PMC4919044

21 M. Singer , C. S. Deutschman , C. W. Seymour , et al., “ The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis‐3) ,” Journal of the American Medical Association 315 , no. 8 ( 2016 ): 801 – 810 , 10.1001/jama.2016.0287 . 26903338 PMC4968574

22 O. H. Gauer , Lehrbuch der Physiologie des Menschen , ed. L. Landois and H. U. Rosemann ( Von Urban and Schwarzenberg , 1960 ), 95 – 135 .

23 H. C. Schonheyder , “ Existing Data Sources for Clinical Epidemiology: The North Denmark Bacteremia Research Database ,” Clinical Epidemiology 171 ( 2010 ): 171 , 10.2147/CLEP.S10139 . PMC2943179 20865114

24 M. Nørgaard , H. Larsson , G. Pedersen , H. C. Schønheyder , and H. T. Sørensen , “ Risk of Bacteraemia and Mortality in Patients With Haematological Malignancies ,” Clinical Microbiology and Infection 12 , no. 3 ( 2006 ): 217 – 223 , 10.1111/j.1469-0691.2005.01298.x . 16451407

25 D. J. Diekema , P. R. Hsueh , R. E. Mendes , et al., “ The Microbiology of Bloodstream Infection: 20‐Year Trends From the SENTRY Antimicrobial Surveillance Program ,” Antimicrobial Agents and Chemotherapy 63 , no. 7 ( 2019 ): e00355‐19 , 10.1128/AAC.00355-19 . 31010862 PMC6591610

26 S. M. Jabbari Shiadeh , A. Pormohammad , A. Hashemi , and P. Lak , “ Global Prevalence of Antibiotic Resistance in Blood‐Isolated Enterococcus Faecalis and Enterococcus faecium : A Systematic Review and Meta‐Analysis ,” Infection and Drug Resistance 12 ( 2019 ): 2713 – 2725 , 10.2147/IDR.S206084 . 31564921 PMC6731464

27 E. Sörstedt , G. Ahlbeck , and U. Snygg‐Martin , “ Trends in Enterococcus faecium Bacteremia: Exploring Risk Factors With Emphasis on Prior Antibiotic Exposure ,” Microorganisms 12 , no. 10 ( 2024 ): 1932 , 10.3390/microorganisms12101932 . 39458242 PMC11509189

28 C. Gudiol , J. Ayats , M. Camoez , et al., “ Increase in Bloodstream Infection due to Vancomycin‐Susceptible Enterococcus faecium in Cancer Patients: Risk Factors, Molecular Epidemiology and Outcomes ,” PLoS One 8 , no. 9 ( 2013 ): e74734 , 10.1371/journal.pone.0074734 . 24069339 PMC3778008

29 C. Encinas , M. J. Blanchard , R. A. Fernández , et al., “ Factors Increasing the Risk of Infection in Patients With Multiple Myeloma: From Biology to Prevention ,” Clinical Lymphoma, Myeloma & Leukemia 25 , no. 10 ( 2025 ): e844 – e856.e1 , 10.1016/j.clml.2025.06.012 . 40695637

30 A. S. Danielsen , C. Lim , C. H. Yoon , et al., “ Epidemiology and Excess Mortality of Antimicrobial Resistance in Bacteraemias Among Cancer Patients: A Cohort Study Using Routinely Collected Health Data From Regional Hospital Trusts in Oxford and Oslo, 2008–2018 ,” BMJ Open 15 , no. 6 ( 2025 ): e092740 , 10.1136/bmjopen-2024-092740 . PMC12164652 40514231

31 K. H. Park , Y. J. Jung , H. J. Lee , et al., “ Impact of Multidrug Resistance on Outcomes in Hematologic Cancer Patients With Bacterial Bloodstream Infections ,” Scientific Reports 14 , no. 1 ( 2024 ): 15622 , 10.1038/s41598-024-66524-w . 38972913 PMC11228017

32 M. M. C. Lambregts , A. T. Bernards , M. T. van der Beek , L. G. Visser , and M. G. de Boer , “ Time to Positivity of Blood Cultures Supports Early Re‐Evaluation of Empiric Broad‐Spectrum Antimicrobial Therapy ,” PLoS One 14 , no. 1 ( 2019 ): e0208819 , 10.1371/journal.pone.0208819 . 30601829 PMC6314566

33 F. Edwards , M. Waller , and K. B. Laupland , “ Determinants of Time to Positivity in Bloodstream Infections: An Analysis of a Population‐Based Cohort in Queensland, Australia ,” European Journal of Clinical Microbiology & Infectious Diseases 44 , no. 5 ( 2025 ): 1197 – 1204 , 10.1007/s10096-025-05096-7 . 40056305 PMC12062184

34 J. Bodilsen and H. Nielsen , “ What More Do You Need? Safety of Early Oral Antimicrobials for Low‐Risk Neutropenic Patients With Suspected Infection ,” Clinical Microbiology and Infection 30 , no. 1 ( 2024 ): 15 – 17 , 10.1016/j.cmi.2023.09.026 . 37832936

35 V. Coyle , C. Forde , D. F. McAuley , et al., “ Early Switch to Oral Antibiotic Therapy in Patients With Low‐Risk Neutropenic Sepsis (EASI‐SWITCH): A Randomized Non‐Inferiority Trial ,” Clinical Microbiology and Infection 30 , no. 1 ( 2024 ): 92 – 99 , 10.1016/j.cmi.2023.07.021 . 37517522

36 A. W. Boerman , M. Schinkel , L. Meijerink , et al., “ Using Machine Learning to Predict Blood Culture Outcomes in the Emergency Department: A Single‐Centre, Retrospective, Observational Study ,” BMJ Open 12 , no. 1 ( 2022 ): e053332 , 10.1136/bmjopen-2021-053332 . PMC8728456 34983764

37 R. Z. Marandi , F. B. Hertz , J. Q. Thomassen , et al., “ Prediction of Bloodstream Infection Using Machine Learning Based Primarily on Biochemical Data ,” Scientific Reports 15 , no. 1 ( 2025 ): 17478 , 10.1038/s41598-025-01821-6 . 40394073 PMC12092749