Immunosuppression increases sepsis risk in patients with HMs. Early and accurate pathogen identification is crucial for improving prognosis [11]. While ddPCR shows promise for early pathogen diagnosis [27], further data are needed to support broader clinical use. In our study of 400 patients with HMs and sepsis, ddPCR significantly outperformed traditional BC in pathogen and AMR genes diagnosis.

However, this study found ddPCR’s diagnostic performance in patients with HMs remains improvable. dPCR’s diagnostic performance in infectious diseases, ICU [18] and emergency patients (88.89% sensitivity, 55.61% specificity) [20] surpassed that in patients with HMs. We speculate that the population-specific characteristics of patients with HMs may be one of the reasons for the difference in performance. Additionally, it has been reported in the literature [25] that the sensitivity of ddPCR in patients without empirical antibiotic therapy can reach 100%, while the sensitivity decreases to 55.56% after receiving such therapy. In this study, 75% (300/400) of patients used broad-spectrum antibiotics within 3 days before ddPCR testing, which may be another important reason for the difference in performance.

Further comparative studies revealed that in a retrospective study involving 71 febrile hematological patients [26], ddPCR demonstrated better detection performance than our study, potentially due to differences in pathogen coverage. Unlike that study, our ddPCR did not include five viruses (detection rate: 62.90% in the prior study) and five Gram-negative bacteria (rarely detected, only once). We speculate that the exclusion of virus detection is one of the reasons for the difference in performance. In addition, the detection range of ddPCR in this study did not include Aspergillus and Mucor, which are increasingly prevalent in patients with hematological malignancies, and this further limited its diagnostic efficacy. In summary, key reasons for the potential improvement in ddPCR’s diagnostic performance in this study include the unique characteristics of patients with HMs, prior exposure to broad-spectrum antibiotics, and limited detectable pathogen range. Thus, in subsequent prospective studies, our team will focus on this special population, optimize ddPCR timing, and expand detectable pathogens. This is expected to further clarify and enhance ddPCR’s diagnostic performance in patients with HMs.

Mixed infections are relatively common in patients with HMs. Notably, ddPCR showed a higher detection rate of mixed pathogens (44.59% vs. 26.39% for BC), highlighting its potential in identifying mixed infections in immunosuppressed patients. Additionally, ddPCR detected seven AMR genes, overcoming BC’s inability to provide resistance genes. This enables ddPCR to facilitate more precise use of antimicrobial agents, timely response to drug-resistant bacteria, and rapid control of disease progression.

The timeliness and accuracy of early antibacterial treatment are closely linked to sepsis prognosis. A study [28] showed BC-based sensitive antimicrobial therapy takes 52 h, while the median time to death of carbapenem-resistant Enterobacteriaceae BSI was 4 days. Delayed antibiotic access correlates with higher mortality. A review found 6-h timely antimicrobial therapy reduces death risk by 57.00% (OR = 0.57, 95% CI 0.39–0.82). Each hour delay increases in-hospital mortality by 9.00%, and the absolute mortality rate of patients with septic shock increases by 1.80%/h [10, 29]. Our ddPCR assay, with a TAT 18-fold faster than BC, enabled early pathogen detection in 73.08% of BC-positive cases and 72.00% of clinically suspected infections, facilitating early targeted therapy consistent with prior findings by Li et al. [26]. Thus, the rapidity and accuracy of ddPCR imply its potential to optimize clinical decision-making [22, 30] and possibly decrease the mortality associated with BSIs.

While ddPCR enables rapid pathogen detection, its impact on clinical outcomes and antimicrobial stewardship remains unclear. Prior studies on other rapid diagnostic methods have yielded conflicting results: blaKPC gene identification reduced mortality in CRE infections [31], whereas MALDI-TOF-based detection showed no effect on 30-day mortality [32]. Our study is the first to demonstrate that ddPCR accelerates the decline of inflammatory biomarkers (CRP/PCT) by > 50.00% within 48–72 h in patients with HMs and sepsis, enhances treatment efficacy, and reduces 28-day all-cause mortality.

To further explore the clinical potential of ddPCR, we assessed its value for antimicrobial stewardship. Our study in patients with HMs aligns with prior research in emergency, infectious disease, and ICU settings, confirming ddPCR’s role in real-time antibacterial treatment optimization [18, 23, 25]. Our study first confirms that ddPCR optimizes antimicrobial stewardship via real-time treatment adjustments, reducing AUD and combination therapy using in patients with HMs and sepsis.

Thus, we conclude that ddPCR rapidly detects pathogens and AMR genes, enabling clinicians to timely optimize antimicrobial regimens, avoid treatment delays, and reduce broad-spectrum antibiotic abuse, thereby improving clinical efficacy, reducing mortality, and optimizing antimicrobial stewardship. For patients with HMs and sepsis, it is recommended to simultaneously submit BC and nucleic acid-based rapid pathogen identification tests, along with rapid detection of antimicrobial resistance genes. This study suggests that ddPCR represents a preferable option for these purposes.

Although ddPCR reduced the proportion of antimicrobial agents in total drug expenditure, no significant difference was observed between the two groups. This phenomenon is speculated to be attributed to the higher detection rate of AMR genes in the ddPCR group: while the combined use of antimicrobials was reduced, the utilization of higher-tier and more expensive antimicrobials offset potential cost savings, ultimately leading to no intergroup difference in the proportion of antimicrobial expenditure. Furthermore, numerous factors continue to influence the proportion of antimicrobial agents in total drug expenditure. Consequently, this conclusion should be interpreted with caution and warrants further validation.

In this study, the cost of ddPCR ($203.19) is 5.5 times that of traditional BC ($36.81). Furthermore, as ddPCR remains a self-funded item in China at present, patient acceptance is relatively low, which hinders its application in patients with mild infections to a certain extent. Meanwhile, despite its advantages of high sensitivity and absolute quantification in pathogen detection, ddPCR has inherent limitations: it may produce false positives or false negatives due to factors such as pre-set targets, nucleic acid extraction efficiency, contamination risks, target gene mutations, and matrix interference. In addition, ddPCR still cannot distinguish between viable and non-viable bacteria. When multiple pathogens are detected or the quantified copy number is low, result interpretation remains highly dependent on the clinical experience of physicians.

mNGS exhibits potential in pathogen detection for patients with suspected BSIs due to its advantage of unbiased broad coverage, enabling simultaneous detection of multiple pathogens, including bacteria, fungi, and viruses. ddPCR, characterized by high sensitivity and rapid absolute quantification, also performs prominently. Both methods show promise in pathogen detection for suspected BSIs; however, no studies have clarified differences in their diagnostic performance among sepsis patients with HMs. A study involving 60 ICU critically ill patients with suspected BSIs [24] demonstrated that ddPCR had a higher positive rate (83.30%) than mNGS (68.30%, excluding viruses) and blood culture (16.70%). ddPCR featured faster detection speed and higher detection rate of AMR genes, albeit with a narrower detection range. Consistent with our findings, this study confirms the advantages of ddPCR in terms of high positive rate, high AMR gene detection rate, and rapid detection.

Our study did not compare these two methods, primarily considering factors such as the clinical accessibility of ddPCR, the unique features of infections in patients with HMs, patients’ economic affordability, and the limited sample size of mNGS-tested cases in our institution. As mNGS is increasingly applied in clinical practice, our team will further investigate the diagnostic differences between the two methods in patients with HMs to more accurately define their clinical application value.

This study offers key strengths: (1) It was the first study to systematically evaluate ddPCR in patients with HMs and sepsis. (2) This is the first time that ddPCR has been linked to mortality, the decline rate of inflammatory indicators, and antimicrobial stewardship (e.g., AUD and combined antimicrobial usage). By integrating 10 comprehensive indicators spanning mortality, clinical outcomes, and antimicrobial stewardship, this study represents the most comprehensive exploratory research in this field to date. (3) Multiple analytical approaches were employed to mitigate confounding and strengthen the robustness and reliability of findings, including propensity score matching (PSM), regression analyses, stabilized inverse probability of treatment weighting (sIPTW), and propensity score adjustment.

This study has several limitations: (1) Single-center retrospective design may cause selection bias, needing multicenter prospective validation despite PSM and sensitivity analyses. (2) The detection panel has limited pathogens, missing rare ones (e.g., Cryptococcus), requiring cautious interpretation. (3) We did not compare ddPCR with emerging technologies (e.g., mNGS). This aspect needs to be addressed in future research. (4) The sample sizes of some subgroups were relatively small, limiting the statistical power of the study.