Bloodstream infections (BSIs) are the leading cause of infection-related death and are associated with substantial morbidity and mortality (Lamy et al., 2020), particularly those might induce sepsis/septic shock (Timsit et al., 2020). Although sepsis may develop without bacterial invasion into the blood stream, pathogens isolated from blood are often considered to be the major causative agent of the sepsis and are utilized to guide the antibiotic treatment. Early detection and effective administration of antibiotics are essential to improve clinical outcomes in critical medical situations like sepsis and septic shock. Delaying the administration of effective antibiotics increases patient mortality, lowers treatment success rates, and raises total healthcare costs (Chertoff and Ataya, 2017; Liu et al., 2017). Therefore, the International Guidelines for Management of Sepsis and Septic Shock 2021 advise using antibiotics right away for adult patients who may have septic shock or sepsis, ideally within one hour of diagnosis to get the optimum therapeutic benefit (Evans et al., 2021). A considerable number of suspected septic patients initially enter the health-care system through an emergency department (ED), hence it is vital to establish an early effective strategy to diagnose sepsis and initiate prompt care for those with life-threatening infection in ED. However, the early identification of the pathogen, timely administration of antibiotics and further improvement of clinical outcomes are often challenging in the ED, due to the lack of development of accurate BSI diagnostic tools for rapid and accurate detection of pathogens.

At present, several BSI prediction tools have been proposed to identify patients at high risk of BSI in ED which include single biomarkers such as C-reactive protein (CRP) (Jeong et al., 2012), serum lactate (Lin et al., 2017) and procalcitonin (PCT) (Hoeboer et al., 2015) or a combination of clinical parameters and biomarkers (Lee et al., 2012; Takeshima et al., 2016; Ljungström et al., 2017). Among all the BSI prediction tools, scores based on clinical parameters and/or bedside biomarkers are proved to be the most rapid way of classifying the risk of BSI and/or sepsis which can be used to guide the diagnostic activities. As a well-known prediction model developed to rule out patients with low risk of positive blood culture (BC), the Shapiro score (SS) (Shapiro et al., 2008) has been widely used, verified, and referenced in the ED worldwide (Hodgson et al., 2016; Jessen et al., 2016; Sparks et al., 2022). More recently, a newly investigation had reported that a Modified Shapiro Score (MSS) was able to predict positive BC in the ED in a well-characterized cohort of patients with suspected BSI and in a subset of patients meeting Sepsis-3 criteria (Nestor et al., 2021).

Apart from early screening out the potential candidates for the suspected patients at high risk of BSI, the precise identification of pathogens also remains a big challenge in the ED and the quantity of pathogen in the blood is crucial for the diagnosis of BSIs. Currently, BC combined with antibiotic susceptibility testing (AST) remains the gold standard for BSI diagnosis, while on the other hand, which is hampered by its low sensitivity and high turnaround time. Emerging molecular diagnostic approaches are used to compensate for the deficiencies of BC, which may be divided into pathogenic bacteria detection based on positive BC samples and pathogenic bacteria detection in the whole blood. The former comprises multiplex PCR, fluorescence in situ hybridization, matrix-assisted laser desorption/ionization time-of-flight mass spectrometry, DNA microarray technology, etc. The latter consists of technologies like real-time quantitative PCR (qRT-PCR), next-generation sequencing (NGS), T2 magnetic resonance detection (Timbrook et al., 2017; Sinha et al., 2018). Unfortunately, several above-mentioned molecular tests have a medium sensitivity/specificity, lack of ASTs and therefore have limited diagnostic value in the clinical environment (Marco, 2017; Zhang et al., 2018; Zboromyrska et al., 2019). Droplet digital PCR (ddPCR) is a third generation of qPCR that employs emulsified micro-droplets suspended in oil to generate thousands of droplets that can be counted and used to quantify nucleic acid targets (Wouters et al., 2020). Recently, ddPCR has shown promising potential in resolving polymicrobial infection because it simultaneously achieves unprecedented high sensitivity (able to detect pathogens at low concentrations as low as 10 CFU/ml), high specificity, and absolute quantification without the need for a standard curve (Abram et al., 2020; Wouters et al., 2020). It has been utilized in several medical applications, including liquid biopsy for cancer monitoring, rejection monitoring following organ transplantation and study of prenatal genetic disorders (Caswell et al., 2020; Sorbini et al., 2021; Boldrin et al., 2022). Till now, a few reports had documented that ddPCR had been utilized to detect bacterial infections in septic patients or patients with a highly suspected BSI in the intensive care units (ICU) (Hu et al., 2021; Zheng et al., 2021; Shao et al., 2022; Wu et al., 2022; Lin et al., 2023), which had shown advantages in identifying polymicrobial BSIs and ability to dynamically monitor changes in pathogenic microorganisms in the blood, therefore might be used to evaluate antibiotic efficacy and survival prognosis. As we all know, the intensive care populations mainly with septic shock are to some extent different from a general ED population of patients with varying severity of disease and with relatively lower mortality rates. Therefore, the ratio and species of pathogen identified in the ED might be distinguished from that in the ICU settings. To date, the ddPCR performance specifically for rapid BSI detection in the setting of ED has not been systematically explored yet. Besides, the relationship of the pathogen load detected by ddPCR with the inflammatory and prognostic markers had not been documented and the cost effective of ddPCR had not been studied before. Hence, in the present study, in a prospective cohort of patients with suspected BSIs screened by MSS score in the ED, we comprehensively evaluated the clinical diagnostic application of ddPCR-based methods and made a comparison against the traditional BC as the gold standard. Further, we explored the clinical application value of ddPCR in patients with suspected sepsis.

BSIs are common situations which are associated with poor outcomes, especially in cases of sepsis/septic shock, immune deficiency, and delayed adequate antimicrobial management. Many septic patients are initially treated in the ED, where rapid administration of targeted antibiotics drastically improve prognosis. Unfortunately, traditional BC are too tardive to support early recognition of sepsis and rapid therapy, which suggests that there is room for improvement. In this study, we assessed the diagnostic performance of ddPCR in detecting suspected sepsis patients in ED clinical practice. Based on the current study, the ddPCR assay exhibited high conformance with the BC method in detecting BSIs. Compared with traditional BC, the ddPCR assay could detect more positive cases and pathogen species for guiding clinical treatment, and greatly shorten the turnover time as well as reduce healthcare costs. Besides, we found the pathogen load of ddPCR was positively correlated with inflammation, coagulation and illness severity associated biomarkers and had the potential to predict a poor clinical outcome. In sum, we firstly determined the efficiency and effectiveness of ddPCR in the quick identification of pathogen and initiation of pathogen-oriented therapy in the ED, which might aid the improvement of management of sepsis and septic shock in the ED.

Until now, BC remains the gold standard assay for the detection of pathogen in the blood stream in the ED. However, the positive ratio of BC is quite low, and it usually takes 24–72 hours or more to isolate disease-causing pathogens (Klouche and Schröder, 2008). Recently, as an emerging flexible and universal platform with high sensitivity and excellent accuracy, ddPCR has been reported to be increasingly used in multiple clinical situations including BSIs but this was at small-scale and not routinely used in the urgent ED environment (Caswell et al., 2020; Hu et al., 2021; Sorbini et al., 2021; Boldrin et al., 2022; Wu et al., 2022). Our study focused on patients presenting at ED with suspected of BSI screened by MSS score more than 2 which had been verified to predict positive BC in the ED in previous studies (Nestor et al., 2021). We designed the multiplex ddPCR panel according to the global and local pathogen epidemiological data which covers the common pathogens of bacteria, virus, and fungi in the ED. Our findings were basically in accordance with the previous studies which suggested ddPCR exhibited superior performance in detecting various infections as well as offered faster turnaround time about 3–4 hours compared to BC (Hu et al., 2021; Shao et al., 2022; Wu et al., 2022; Lin et al., 2023). A recent prospective study enrolled 438 plasma samples from 150 ICU patients, documented the total diagnostic sensitivity rose from 9.1% (40/438) by BC alone to 41.1% (180/438) by ddPCR (Wu et al., 2022). Afterwards, in a prospective, observational, single-center investigation with 122 plasma samples from 169 suspected BSIs patients from the department of infectious diseases, the ED, and the ICU, Lin et al. observed that BC and ddPCR positive rate was 11.27% and 30.28%, respectively (Lin et al., 2023). In contrast to those two studies, the superiority in our cohort is significant as the ddPCR positive rate was 48.13%, which was much higher than the rate of 6% for BC. Among the 18 BC+ cases, 88.8% of which were identified by ddPCR with only 2 missed. The positive rate of BC in our study was relatively lower than previous reports, which may be due to the complexity of emergency patients, and the severity of the disease was not so severe as that in ICU. In consideration of the low positive rate of BC in our study, we define positive ddPCR as identifying same pathogen with BC, another microorganism test, the patient’s condition improved after the treatment of ddPCR pathogen or infection indicators correspond to ddPCR. Therefore, the ddPCR testing showed a potential advantage over BC in terms of an overall sensitivity of 88.89% and specificity of 55.61%. We further found that the optimal diagnostic power for quantifying BSI through ddPCR is achieved with a copy cutoff of 155.5 according to Youden index at maximum. On the other hand, in our research, ddPCR exhibited a relatively high sensitivity in suspected sepsis patients in spite of the extremely low levels of microorganisms in the blood or past antibiotic therapy which may aid in prompt identification of the potential candidates. It is worth mentioning our novel finding that we discovered that in patients with liver abscess, ddPCR detection efficiency was highly sensitive and specific. The clinical symptoms and signs of patients with liver abscess are often atypical, which can cause misdiagnosis or missed diagnosis in clinical environment especially in ED (Kaplan et al., 2004). The long cycle of BC and pus culture is not conducive to the early diagnosis of pathogenic bacteria. Our data suggested that ddPCR might have the ability to detect some recessive BSIs of diseases with a low positive rate of BC as liver abscess which need further large-scale research.

It is worth noting that nearly 10–40% of negative BCs were found to be positive using multiplex molecular approaches during previous studies (Wallet et al., 2010; Peker et al., 2018). Consequently, we concentrated on the interpretation of discordant ddPCR+ results and considered whether BC-/ddPCR+ result represents a real BSI under circumstance of additional factors, including epidemiological, clinical, and laboratory results. Importantly, from the perspective of detailed clinical circumstances, the majority of the discordant results were either probable or possible BSIs. The possible reasons might be accounted for the presence of nonviable, nonproliferating, or transient or intermittent bacteremia, intracellular organisms within circulating phagocytic cells, inhibition of bacterial growth by antibiotics, or possible contamination (Wu et al., 2022). Moreover, organisms identified by microbial cfDNA which are judged as possible or unlikely causes of the sepsis alert included reactivated herpesviruses, chronic infections, microorganisms likely to be commensal organisms or subclinical colonization and possible causes of non-sepsis-related acute infection. Taken together, the sensitivity and specificity of ddPCR were higher than BCs, therefore, it’s indispensable to take ddPCR as an add-on complementary assay to the conventional BC method for identifying the possible causative pathogens and related AMR genes for BC-negative septic patients in emergency department practices.

Another strength of applying ddPCR is to detect multiple bacteria and AMR genes, which could guide the anti-inflammatory treatment. Polymicrobial BSI, defined as the presence of at least two distinct bacteria detected from the BCs or ddPCR, has been recorded more frequently, with rates ranging from 5% to 38% of all BSI events (Hochstein et al., 1965; Kiani et al., 1979; Reuben et al., 1989). Hospitalized patients with polymicrobial BSIs had mortality that varied from 21% to 63%, which is almost twice as high as those of patients with monomicrobial infections (Zhang et al., 2019). The ddPCR assay is more likely to detect polymicrobial BSIs because it eliminates the potential bias produced by preferred amplification and allows detection of pathogens at low quantities when compared to BC. Polymicrobial BSIs were discovered in 7% (15/214) of the patients diagnosed by ddPCR in our current investigation, which was similar to earlier findings ( Supplementary Table S9 ) (Lin et al., 2010). Furthermore, our findings illustrated the potential benefit of using ddPCR assays to direct antimicrobial therapy while keeping track of polymicrobial BSIs. The growing number of bacteria with AMR has made clinical anti-infective treatment more difficult. Traditional AST limits timely antibiotics administration and infection control. The current ddPCR in our study, with blaKPC, mecA, blaOXA-48, blaNDM, blaMP, vanA, vanM, seven frequent target AMR genes in one test panel, could directly identify these genes from whole blood in around 3 hours. Notably, prompt and concomitant detection of AMR genes and microbial/polymicrobial infection sources by ddPCR might optimize patient outcomes and allow for better monitoring of resistance mutations (Abram et al., 2020; Hu et al., 2021). In our study, 5 blaKPC, 2 blaNDM, and 10 mecA genes were detected, and blaKPC might be carried by Klebsiella pneumoniae, blaNDM by Klebsiella pneumoniae as well as mecA by Staphylococcus aureus. While no proper causative pathogens were identified for some of blaNDM and mecA genes in several samples. Because the AMR gene identified by ddPCR assay is not from isolated pathogens, which is different from traditional culture-based methods, further research is needed to investigate the causes of mismatches between pathogens and AMR genes. Overall, the data from our investigation and the other clinical trials indicate that the ddPCR test, either alone or in combination with other techniques, can offer a strong platform for early antimicrobial medication start and quick detection of BSIs. However, additional detection panels are required to prevent false-negative findings brought on by the relatively small detection range of ddPCR.

Owing to the clear predominance of bacterial and fungal infections in the context of patients with BSIs, screening for viral infections is rarely part of routine diagnostics. Among all the identified virus by ddPCR, EBV had a substantially higher positive rate than others, which was an interesting finding. EBV, as the first human tumor virus expressing virus cancer genes and immortalizing infected lymphocytes, resides in humans to establish a long-term latent infection and is associated with a variety of human diseases including hematologic malignancies. The levels of viraemia might be considered to be a useful biomarker of immunosuppression, guiding immunotherapy and monitoring disease progression and response to therapy. Our findings were in line with a newly report documented the performance of ddPCR in the children with suspected BSI, with a total of nine viruses including seven EBV out of 44 total identified pathogens by ddPCR (Liu et al., 2023). Our further research discovered that EBV has a link to immunosuppression. 35 (50.72%) were found to have concurrent COVID-19 infection and 21(30.43%) were accompanied with immunosuppression. in our study. In very sick COVID-19 patients, EBV was quite common and was a cause of mortality (Naendrup et al., 2022). Previous research had indicated that viral DNAemia was widespread in severe sepsis, and EBV was the most frequently reactivated virus in plasma in septic patients (Walton et al., 2014; Ong et al., 2017; Mallet et al., 2019). Mallet recently demonstrated that viral DNAemia was highly correlated with the likelihood of secondary infection, a decline in innate immune function over time, and pre-existing immunosuppression (Mallet et al., 2021). Viral DNAemia is expected to be a possible marker of immunosuppression and infection risk in bigger multi-center trials. Herein, ddPCR might be serving as an early warning indicator of the possibility of viraemia in adults with pre-existing or new onset immunosuppressive disorders.

In addition, to our knowledge, we are the first to report that the copies of pathogens in ddPCR were positively correlated with indicators of inflammation severity as well as poor prognosis. One possible explanation is that more serious inflammation in BSIs often exhibits larger loads of causative agent, contributing to a positive ddPCR results. An increasing copy of microorganism facilitates series of inflammatory response, severity and ultimate prognosis. As shown in our study, we found positive correlations between ddPCR pathogen loads and indicators of inflammation (IL-6, CRP, PCT and NLCR levels), severity (Lac) and prognosis (SOFA, APACHII and 28-day mortality). Importantly, we found that the 28-day mortality rate was 2.215 times higher for those with copies greater than 1263. Therefore, the high pathogen load might indicate the development and severity of sepsis. Sepsis-associated dysfunction of coagulation is quite common which usually manifested as abnormal of traditional markers (PLT, PT, APTT and D-dimer) and four new coagulate indicators (TM, TAT, PIC, and t-PAIC). We further found there existed a positive relationship between the pathogen load and these coagulate markers, which suggested the high pathogen load might initiate the disorder of coagulation and thereby aggravate the clinical situations. As for the expectation of prognosis, indeed, we found that clinical variables alone, including the qSOFA score, were unable to accurately estimate outcomes of patients with critical infection at the time of initial evaluation in the ED. Hence, the ddPCR provides us with a complementary tool.

Most recently, a very few clinical trials had suggested that the ddPCR may facilitate precision antimicrobial stewardship by dynamic monitoring pathogens (Shao et al., 2022; Lin et al., 2023). In line with these reports, the enrolled patients’ condition steadily improved following the clinical escalation of antibiotics, and the load of pathogens and drug resistance genes significantly decreased, according to drug resistance gene detection at the same time. In this work, we discovered that variations in pathogen DNA load and number of species detected by ddPCR had the potential to guide antibiotic therapy in critically sick patients with BSIs. With high sensitivity and specificity, the ddPCR detects more pathogens by reducing reporting time, which can drive the optimization of antibiotic management and the rational de-escalation therapy. As a result, ddPCR facilitates precision antimicrobial stewardship and reduces hospitalization costs ultimately. Besides, in view of new diagnostic technology, we not only objectively evaluate the efficiency of its clinical application, but further analyze its value from the perspective of health economics, namely, determine the economic value of the new method. From our data, we found ddPCR assay, with relative low hospitalization/antibiotic cost, high efficiency as well as high specificity and sensitivity, exhibited higher economic efficiency. As a result, the ddPCR assay can help to save medical resources and make it possible to be widely used in clinical practice. Ultimately, patients with suspected sepsis can benefit from the initial first-hand handling in ED in the future.

This study should be considered in light of its limitations. Firstly, this was single-center prospective research, and our findings require further investigation in a multi-center study with a bigger sample size. In addition, the ddPCR system only covered 18 common isolated pathogens and 5 antimicrobial resistance genes, the results of our investigation should be viewed with caution. This revealed the limitation of ddPCR in pathogen detection due to the panel design. Moreover, detections of viruses and AMR genes were not simultaneously compared with other molecular tests, so the results of virus and AMR genes in this study need further investigation. Finally, we only compared the results of ddPCR with conventional BC results, so multiple detection methods should be added to make the results more convincing.

This single-center study firstly demonstrated that ddPCR had an overall superior detection rate of potential pathogens compared to BC in patients with suspected BSI, which suggest that ddPCR can be used for sepsis pathogen diagnosis models to guide antibiotic treatment. In a timely manner, ddPCR delivered precise and quantitative load data on the causal pathogen, which in part reflected the severity of the infection and poor outcome. Consequently, ddPCR offers the ability to monitor the patient’s condition in real-time and to give medics an early warning of sepsis in time-urgent clinical situations as ED.

The original contributions presented in the study are included in the article/ Supplementary Material , further inquiries can be directed to the corresponding authors.

The studies involving humans were approved by Ethics Committee of Shanghai East Hospital. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

SJ: Writing – original draft, Investigation. DZ: Software, Formal analysis, Data curation, Writing – original draft. CW: Writing – review & editing, Validation, Methodology. XL: Writing – original draft, Supervision, Resources, Investigation. QY: Writing – review & editing, Resources, Investigation, Data curation. XB: Writing – review & editing, Visualization, Resources, Investigation. TD: Writing – original draft, Methodology, Investigation. GL: Writing – original draft, Methodology, Investigation. YG: Writing – original draft, Investigation, Formal analysis. YY: Writing – original draft, Methodology. BS: Writing – review & editing, Investigation. SX: Writing – original draft, Investigation. XZ: Writing – review & editing, Visualization, Validation, Supervision. LF: Writing – review & editing, Supervision, Project administration, Methodology, Funding acquisition. LT: Writing – review & editing, Supervision, Project administration, Funding acquisition.