{"paper_id":"4279f67b-2c2c-4687-ba6a-b79b8f25dda9","body_text":"The Efficacy of a Novel Medical Artificial Intelligence Large Model(MedGo)-Guided Identification and Personalized Treatment of Sepsis: Study Protocol for a Single- centered Randomized Controlled Trial | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Study protocol The Efficacy of a Novel Medical Artificial Intelligence Large Model(MedGo)-Guided Identification and Personalized Treatment of Sepsis: Study Protocol for a Single- centered Randomized Controlled Trial Sen Jiang, Tong Liu, Chunxue Wang, Bo An, Haitao Zhang, Lunxian Tang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5873082/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Sepsis, a life-threatening organ dysfunction resulting from a dysregulated host response to infection, remains a major global health challenge with high morbility and mortality. Current diagnostic and management frameworks still lack the satisfied sensitivity and specificity for early detection and personalized treatment. In this study, we aim to evaluate the efficacy of MedGo-Sepsis, a novel large language model developed by our collaborating team and us that integrates clinical and immunological data, to guide the identification and personalized treatment of sepsis in the emergency department. Methods This single-center, randomized controlled trial will enroll adult patients presenting to the emergency department with suspected sepsis. Participants will be randomized 1:1 at the physician team level to either the intervention group receiving standard care augmented by MedGo-Sepsis guidance or the control group receiving standard care alone. MedGo-Sepsis provides real-time risk assessments, detailed immune parameter profiling, stratification based on immune phenotypes, and personalized treatment recommendations. The primary outcome is 28-day all-cause mortality. Secondary outcomes include changes in Sequential Organ Failure Assessment (SOFA) score, diagnostic accuracy, time to appropriate antibiotic administration, resource utilization, patient-reported outcomes, and physician workload. Discussion This trial will assess whether MedGo-Sepsis, through the integration of individualized immune data with large language model technology, improves outcomes for patients with sepsis compared to standard care. The combination of enhanced diagnostics and tailored therapeutic strategies has the potential to advance precision management of sepsis in the critical emergency setting. Trial registration The trial has been prospectively registered in the Chinese Clinical Trial Registry (ChiCTR2400094116) on December 17, 2024. Sepsis Large Language Model Artificial Intelligence Personalized Medicine Emergency Department Figures Figure 1 Background Sepsis, defined as a life-threatening organ dysfunction resulting from a dysregulated host response to infection, continues to pose a significant global health challenge and being a leading contributor to in-hospital mortality [ 1 – 3 ]. Recent estimates indicate that sepsis accounts for nearly 20% of all deaths worldwide, underscoring its extensive impact on healthcare systems [ 4 , 5 ]. The clinical complexity of sepsis, influenced by its heterogeneous pathophysiology and rapid disease progression, often leads to delays in diagnosis and compromises treatment decisions [ 6 , 7 ]. Early identification and prompt intervention are essential to prevent progression to septic shock or multi-organ dysfunction, both of which are associated with markedly increased mortality rates [ 7 , 8 ]. Despite advancements in sepsis awareness and public health initiatives, existing diagnostic and management frameworks still fail to adequately address two critical needs: early and accurate detection, as well as personalized treatment. Clinicians have historically employed scoring systems such as the Sequential Organ Failure Assessment (SOFA), quick SOFA (qSOFA), and National Early Warning Score (NEWS) to identify patients with sepsis and predict clinical outcomes [ 7 ]. Although these tools offer valuable diagnostic and prognostic insights, they have inherent limitations, particularly in the context of early stage of sepsis or atypical presentations [ 9 , 10 ]. These models often lack the necessary sensitivity and specificity for timely intervention, resulting in missed opportunities for achieving optimal patient outcomes [ 9 ].Additionally, their static nature does not account for the dynamic characteristics of sepsis and individual patient variability, thereby hindering their effectiveness in guiding personalized therapeutic strategies [ 11 ]. The biomarkers for identification and management of sepsis has evolved beyond traditional ones such as C-reactive protein (CRP) and procalcitonin (PCT), which are limited by their specificity and sensitivity [ 6 , 12 ].Emerging biomarkers, including soluble triggering receptor expressed on myeloid cells-1 (sTREM-1), monocyte distribution width (MDW), presepsin, high-mobility group box 1 (HMGB1), and interleukin-6 (IL-6), show promise for enhancing diagnostic accuracy and risk stratification in patients with sepsis[ 6 , 13 ].Other extensively studied biomarkers include pancreatic stone protein (PSP) and cluster of differentiation 64 (CD64), along with additional candidates such as resistin, tumor necrosis factor-alpha (TNF-α), interleukin-1 beta (IL-1β), syndecan-1, and zonulin [ 13 , 14 ]. Despite their potential, the clinical applicability of these biomarkers remains inconsistent due to discrepancies in measurement standardization, cost, and performance across various clinical settings [ 13 , 14 ]. Furthermore, these individual biomarkers often fail to capture the complexity, dynamics, and heterogeneity of the immune response in sepsis, which hinders the advancement of effective precision medicine strategies. This highlights the need for innovative approaches that integrate multi-modal data, including comprehensive immune profiling, to provide a more holistic understanding of individual patient responses and facilitate personalized therapeutic interventions. The emergence of artificial intelligence (AI) and machine learning (ML) has opened promising new avenues for the management of sepsis. Numerous predictive models based on electronic health record (EHR) data have been developed to enhance early detection and risk stratification [ 15 ]. Although these models have shown some improvements over traditional scoring systems [ 16 ], their \"black box\" nature often obscures the underlying rationale, which can hinder clinician trust and adoption [ 17 , 18 ]. Furthermore, many existing models primarily rely on structured data, overlooking the valuable insights available in unstructured clinical notes, radiology reports, and immune profiling data [18 4]. Lastly, their dependence on retrospective, single-center datasets raises concerns about the generalization of their findings[ 17 , 5 ]. Immunological profiling has emerged as an essential tool for elucidating the intricate pathophysiology of sepsis, enabling the categorization of patients into distinct immunesubtypes. This approach holds significant promise for advancing personalized therapeutic strategies. Current methodologies for immune profiling, including cytokine assays, flow cytometry, and transcriptomic analysis, have successfully identified critical immune subphenotypes, such as hyper-inflammatory states (and immunosuppressive conditions, thereby providing valuable insights into individual variability in immune responses[ 19 , 20 ]. Nevertheless, substantial challenges persist, including data heterogeneity, a lack of standardization in methodologies, and obstacles in translating research findings into real-time clinical decision-making [ 19 – 21 ].Recently, large language models (LLM) with equipped natural language processing (NLP) capabilities have emerged as transformative tools in medical research. By synthesizing extensive amounts of both structured and unstructured data-such as immune profiling outputs, electronic health records (EHRs), and laboratory results - LLMs have demonstrated considerable potential in deciphering complex immune patterns, correlating these patterns with clinical outcomes, and identifying novel immune subgroups within the context of sepsis [ 22 , 23 ]. Recent research has demonstrated that LLMs capable of integrating immune features, such as cytokine profiles and transcriptomic data, significantly enhance the stratification of sepsis subtypes, potentially revealing previously unrecognized patterns of immune dysfunction [ 24 ]. Despite these advancements, several unresolved challenges impede the widespread adoption of LLM-based approaches in clinical practice. Key issues include the scarcity of large, high-quality, and diverse datasets for model training and validation, the variability of immune system responses across different populations, and the lack of interpretability of LLM outputs, which undermines clinician trust and integration into real-time clinical workflows [ 22 , 25 ]. Therefore, addressing these limitations is essential to fully realize the potential of LLMs in sepsis immune phenotyping and establish a precision medicine framework for this life-threatening condition. In light of the significant potential of LLMs in processing medical data, challenges persist in the application of immune profiling, particularly regarding data heterogeneity, methodological standardization, and the provision of real-time clinical decision support for complex immune responses. To address these challenges, we have developed MedGo, a specialized medical language model designed to accurately interpret intricate clinical and immune data. MedGo integrates domain-specific knowledge derived from clinical guidelines and medical literature, enabling it to analyze both structured and unstructured data while incorporating high-dimensional immune profiling for comprehensive assessments of immune status. Building upon the foundation of MedGo, we have further developed MedGo-Sepsis, which is specifically tailored for the diagnosis and treatment of sepsis. MedGo-Sepsis derived extensive clinical knowledge with individualized immune information, thereby facilitating precise risk assessment and the formulation of personalized treatment strategies for patients with sepsis. It aims to enhance the early detection of sepsis, promoting timely interventions, such as early antibiotic administration, which ultimately improves diagnostic accuracy and clinical outcomes. Recognizing the significance of individual immune profiles in determining treatment effectiveness, MedGo-Sepsis incorporates personalized treatment recommendations based on immune phenotyping, thereby transcending traditional \"one-size-fits-all\" methodologies. To evaluate this innovative approach, we propose a randomized controlled trial comparing MedGo-Sepsis-guided treatment with standard care for patients presenting with suspected sepsis in the emergency department (ED). The primary objective of this trial is to assess the impact of MedGo-Sepsis on 28-day all-cause mortality. Secondary endpoints will include the time to antibiotic administration, physician diagnostic confidence, clinician decision-making efficiency, and physician workload. We hypothesize that the implementation of MedGo-Sepsis will result in improved outcomes, including reduced mortality, by enabling more accurate and timely diagnoses alongside personalized therapeutic interventions. This study aims to advance precision medicine in the management of sepsis, with the potential to transform clinical decision-making and enhance patient care in this critical condition. Methods Study design This study is a single-center, randomized, controlled trial designed to evaluate the efficacy of the MedGo-Sepsis LLM for the early identification and personalized treatment of sepsis in patients within the ED. The study will employ a parallel-group design, with participants randomly assigned in a 1:1 allocation ratio to either the intervention group (MedGo-Sepsis assisted care) or the control group (standard care). This pragmatic superiority trial seeks to demonstrate that the MedGo-Sepsis intervention is more effective than standard care in improving patient outcomes. To minimize the risk of treatment contamination, randomization will occur at the level of physician teams. Physicians in the ED will be organized into two teams, with each team randomized to either the MedGo-Sepsis or standard care arm. Physicians will be assigned to only one arm and will not switch groups at any point during the trial. The trial will be conducted in accordance with the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) 2013 statement ( see Additional file 1). Written informed consent will be obtained from all participants prior to enrollment. The trial is registered at ( http://www.chictr.org.cn ) under the identifier (Trial ID:ChiCTR2400094116 ). A comprehensive schedule detailing enrollment, interventions, assessments, and follow-up visits is presented in the participant timeline (Fig. 1 ). Recruitment and Setting This single-center, randomized controlled trial will be conducted at the ED of Shanghai East Hospital in Shanghai, China. The study aims to enroll adult patients (≥ 18 years of age) who present to the ED with suspected sepsis. Suspected sepsis will be defined by the presence of (1) an infection or suspected infection along with any of the following criteria : (2) a qSOFA score ≥ 2, (3) a SOFA score = 1 or (4) a NEWS score of 4–6[ 8 ]. Patients will be screened for eligibility by the treating emergency physician upon arrival at the ED. Those who meet the inclusion and exclusion criteria will be approached for participation by the study team. The target sample size is 524 participants, with an anticipated enrollment period of 24 months. Recruitment strategies will include distributing flyers in the ED waiting area and informing physicians and nurses about the study to encourage referrals. Potentially eligible patients will be approached by trained research staff who will provide a detailed description of the study and obtain informed consent. Eligibility Criteria Inclusion Criteria: 1)Age ≥ 18 years. 2)Presentation to the ED with suspected sepsis (refer to the definition above). 3)Provision of written informed consent. In cases where patients are unable to provide written informed consent due to the acute severity of their medical condition, or cognitive impairment, consent will be obtained from their legally authorized representative. Exclusion Criteria: 1)Age < 18 years. 2)Pregnancy. 3)Active malignancy. 4)HIV infection. 5)Inability to provide informed consent (e.g., due to cognitive impairment, language barriers, severe illness). 6)Input data that is considered unusable by the AI system (e.g., corrupted, incomplete, or insufficient to generate an output). 7)Active participation in another interventional trial. If a patient declines to participate, they will still receive the same quality of care as those enrolled in this study. Additional Information About Consent: The informed consent process will clearly outline the voluntary nature of participation, the right to withdraw at any time without penalty, and all aspects of the study in straightforward language, presented in, both written and oral formats. For potentially eligible participants with impaired cognitive function, a legally authorized representative or family member will be consulted to obtain informed consent. Withdrawal Criteria Participants will be removed from the study under the following conditions: 1)Participants have the right to withdraw from the study at any time and for any reason without facing repercussions regarding their medical care. 2) The investigator reserves the right to withdraw a participant from the study if the participant develops a condition that makes further participation unsafe or if the participant fails to comply with the study protocol. 3)The Data Monitoring Committee (DMC) may recommend the termination of a participant's involvement or the discontinuation of the study based on safety concerns or other ethical considerations. 4)In the event that a participant cannot be reached for follow-up assessments despite reasonable efforts, they will be classified as lost to follow-up and withdrawn from the study. Blinding Outcome assessors will remain blinded to the treatment allocation. Due to the inherent characteristics of the intervention, it is not feasible to blind participants and treating physicians. The study statistician, responsible for data analysis, will also remain blinded to the treatment allocation until the final analysis is conducted. Unblinding will occur only after the database has been locked and prior to the interpretation of the results.The study statistician will not communicate with the unblinded research team in the hospital and will solely focus on the analysis. Randomization Eligible participants will be randomly assigned in a 1:1 ratio to either the MedGo-Sepsis group or the standard care group. This assignment will utilize a computer-generated random sequence with permuted blocks of size four. Randomization will be stratified by sepsis severity (mild or severe) and the presence of comorbidities (present or absent). Within each stratum, a separate randomization sequence will be created. The randomization sequence will be generated by a statistician who is independent of the clinical care team using SAS software for this purpose. To ensure allocation concealment, a secure, centralized, web-based randomization system will be employed, which investigators can access only after obtaining informed consent from the participant. The allocation assignment for each participant will be disclosed to the treating physician only after the baseline data has been collected and entered into the system. Sample Size Based on a hypothesized absolute risk reduction (ARR) of 5% in 28-day mortality associated with MedGo-Sepsis (decreasing from 20% in the standard care group to 15%)[ 4 , 7 ], a sample size of 472 participants (236 per group) is required to achieve 80% power at a two-sided alpha level of 0.05. This calculation was performed using G*Power 3.1 (version 3.1.9.2, Germany) and employed a two-proportion z-test, assuming a binomial distribution. To account for a 10% dropout rate, we will target the enrollment of 524 participants (262 per group). The study is also adequately powered to detect clinically significant differences in secondary outcomes, such as the time to first antibiotic administration (minimal clinically important difference (MCID) : 2-hour reduction) and length of hospital stay (MCID: 2-day reduction) [ 26 , 27 ]. Interventions MedGo-Sepsis Intervention Group Participants assigned to the MedGo-Sepsis group will receive standard care for suspected sepsis in conjunction with the MedGo-Sepsis intervention. Standard care will be administered in accordance with the guidelines established by the Surviving Sepsis Campaign [ 28 ].This care encompasses, but is not limited to, initial resuscitation with fluids and vasopressors, obtaining cultures prior to the administration of broad-spectrum antibiotics, identifying and managing the source of infection, and providing supportive care for organ dysfunction. Upon presentation to the ED, patient data—including demographics, vital signs, laboratory values, clinical findings, relevant imaging reports, and free-text clinical notes—will be entered into the MedGo-Sepsis system by the treating physician after the established inclusion and exclusion criteria have been met. The system will utilize its natural language processing (NLP) capabilities and sepsis-specific algorithms to provide the following functionalities:(1) Real-time Sepsis Risk Assessment: MedGo-Sepsis will generate a sepsis risk score (probability of sepsis) based on the input data. This score will be prominently displayed within the system interface and integrated into the patient’s electronic health record (EHR). (2) Detailed Measurement of Immune Parameters: The system will utilize immune-related data from hospital assays, including: Cytokine Measurements to assess inflammatory responses; Lymphocyte Phenotyping to analyze CD4+, CD8 + T cells, B cells, and NK cells; and Innate and Adaptive Immunity Monitoring to Evaluate neutrophil functions, monocyte antigen presentation (HLA-DR), NK cell activity, T cell counts, Treg cell proportions, and B cell function through immunoglobulin measurements.(3) Stratification Based on Immune Phenotypes: The MedGo-Sepsis system will categorize patients into distinct immune phenotypes based on the measured immune parameters: Hyperinflammation Subtype: This subtype is characterized by elevated levels of pro-inflammatory cytokines and excessive immune activation, indicated by interleukin-6 (IL-6) levels ≥ 100 pg/mL.Immunosuppressed Subtype: This subtype is defined by diminished T-cell responses and elevated levels of anti-inflammatory cytokines, often indicated by a reduction in lymphocyte counts (lymphocyte count ≤ 900 cells/µL) and a percentage of HLA-DR positive monocytes < 30%[ 29 – 32 ]. (4)Personalized Treatment Recommendations: Utilizing the identified immune phenotype, Medgo-Sepsis will formulate individualized treatment strategies. For patients classified under the Hyperinflammation Subtype, the system will advocate for the initiation of broad-spectrum antibiotics in conjunction with anti-inflammatory agents such as hormone. For patients classified as immunosuppressed, particularly those with lymphopenia, the system will recommend immune-enhancing therapies such as interferon gamma, recombinant interleukin-7, immunoglobulin therapy, or thymosin alpha-1 to restore immune function.(5) Dynamic Monitoring and Continuous Feedback: MedGo-Sepsis will continuously monitor incoming patient data and provide real-time alerts if the patient's condition deteriorates or if deviations from recommended treatment protocols occur. Follow-up assessments of cytokine levels and immune responses in patients post-sepsis treatment will inform adjustments to therapeutic strategies. A review of the literature on cytokine levels and immune responses after treatment will be integrated to enhance the adaptability of treatment plans.Physicians are not required to adhere to the recommendations provided by MedGo-Sepsis and retain full autonomy in making treatment decisions. However, the treating physician must document their reasons for accepting or overruling the system's recommendations for each patient. Training on the use of the MedGo-Sepsis system will be provided to all participating physicians and research staff prior to the start of the study. Standard Care Control Group Participants assigned to the standard care group will receive conventional treatment for suspected sepsis in accordance with the Surviving Sepsis Campaign guidelines [ 32 ], without any intervention from the MedGo-Sepsis system. The physicians responsible for the care of patients in this group will have access to all standard diagnostic tools and treatment options, and their clinical decisions will be guided by their own judgment and experience. Data related the treatment administered in the standard care group, including the timing and type of interventions, will be collected and documented using a standardized data collection form. Outcome Measures Primary Outcome The primary outcome of this study is 28-day all-cause mortality: Specifically, it refers to mortality from any cause within 28 days of enrollment. This outcome provides a definitive assessment of the overall impact of MedGo-Sepsis on patient survival. Secondary Outcomes (1)Organ Dysfunction: Change in Sequential Organ Failure Assessment (SOFA) score from baseline to 72 hours. The SOFA score is a widely used and validated measure of organ dysfunction in sepsis, ranging from 0 (normal function) to 4 (severe dysfunction) for each of the six organ systems [ 1 ].A higher total score indicates greater organ dysfunction and increased severity of sepsis. (2)Diagnostic Performance: 1)Diagnostic Time. 2)Misdiagnosis Rate. 3)Missed Diagnosis Rate. 4)Diagnostic Accuracy. (3) Treatment Efficacy : 1)Time to appropriate antibiotic administration 2)Use of adjunctive therapies (e.g., fluids, vasopressors, corticosteroids, immunomodulatory therapies). 3)SOFA score change at day 7 and day 14. 4)Need for mechanical ventilation. (4)Resource Utilization: 1)Length of hospital stay. 2)Length of ICU stay. 3)Total hospital costs. (5)Patient-Reported Outcomes: Health-Related Quality of Life (HRQoL) was assessed at 28 days and 6 months using the EQ-5D-5L, a standardized measure of health status. This tool generates a single index value that represents overall health, along with scores for five dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression[ 33 ]. (6)Physician-Reported Outcomes: 1)The workload of physicians can be assessed using the NASA Task Load Index (NASA-TLX), a multi-dimensional rating system that evaluates perceived workload across six subscales: mental demand, physical demand, temporal demand, performance, effort, and frustration[ 34 ]. 2) Physician diagnostic confidence was evaluated using a 5-point Likert scale, with 1 indicating at all confident 5 indicating 3)Decision Conflict Scale (DCS): This scale measures decisional conflict related to medical choices and consists of 16 items across five subscales: Informed (knowledge about options), Values Clarity (clearness about personal values), Support (feeling supported in decision-making), Uncertainty (clarity about which option is best), and Effective Decision (feeling the decision made was the best possible) [ 35 ]. 4)Satisfaction With Decision (SWD) scale: This [Number]-item scale measures satisfaction with healthcare decisions. Higher scores on this scale reflect a greater level of satisfaction with the decisions made [ 36 ]. 5)Frequency of MedGo-Sepsis use by physicians: (Proportion of patients for which recommendations were generated by MedGo-Sepsis that were used by the physician). 6)Frequency of overriding MedGo-Sepsis recommendations: (Proportional frequency for which MedGo-Sepsis' recommendations were overridden by the treating physician). (7)Long-Term Mortality:All-cause mortality at 6 months and 1 year. Other Data Collected (1)MedGo-Sepsis Performance:Performance errors (deviations from expert consensus). (2)Safety:All adverse events, including serious adverse events (SAEs). (3)Detailed Resource Use: Specific data on resource utilization (e.g., diagnostic tests, medications) will be collected. Statistical Methods Statistical Analysis The primary analysis will be conducted using the Full Analysis Set (FAS), which will include data from all patients who are randomized, receive at least one dose of treatment and have at least one outcome assessment. Additional supportive analyses will be performed on the Per-Protocol set (PPS) and the Safety Set (SS). PPS will consist of participants who adhere to the study protocol without major violations and who do not have significant missing baseline or outcome data. SS will include all randomized participants who have received at least one dose of treatment and have completed at least one safety assessment. The primary outcome, 28-day all-cause mortality, will be compared between the MedGo-Sepsis and standard care groups using Cox proportional hazards regression. This model will adjust for pre-specified confounding variables, including age, sex, initial SOFA score, presence of comorbidities (e.g., diabetes, chronic lung disease), and site of infection. Hazard ratios (HR) with corresponding 95% confidence intervals (CIs) will be reported. Before interpretation, the proportional hazards assumption will be validated using the Schoenfeld residuals test. Should this assumption be violated, alternative methodologies such as stratified Cox regression or the incorporation of time-dependent covariates will be considered. Secondary outcome analyses will be conducted as follows: (1) Organ Dysfunction: Change in SOFA score at 72 hours will be analyzed using a linear mixed-effects model, which accommodates repeated measures within patients (baseline and 72 hours) and adjusts for the previously mentioned covariates. Changes in individual SOFA subscores will be analyzed using the same methodology. (2)Diagnostic Performance:Diagnostic time will be compared between groups using t-tests or Mann-Whitney U test, depending on the distribution. The misdiagnosis rate, missed diagnosis rate, and diagnostic accuracy, which are proportions, will be compared using Chi-squared tests or Fisher's exact test, as appropriate. (3)Treatment Efficacy:Time to appropriate antibiotic administration will be evaluated using survival analysis, specifically Cox proportional hazards regression or Kaplan-Meier methods, depending on the validity of the proportional hazards assumption. The use of adjunctive therapies, and the need for mechanical ventilation will be analyzed using logistic regression. Changes in the SOFA score at day 7 and day 14 will be assessed using linear mixed-effects models, similar to the analysis conducted at the 72-hour mark. (4)Resource Utilization: The length of hospital stay, ICU length of stay, and total hospital costs will be compared using t-tests, Mann-Whitney U, or generalized linear models, with covariate adjustments as necessary. (5)Patient-Reported Outcomes: HRQoL (EQ-5D-5L) will be assessed using linear regression (analysis, or an appropriate alternative method if the assumption of normality is not satisfied. Additionally, patient satisfaction will be evaluated with ordinal logistic regression (or similar) at each time point, adjusting for baseline scores and other relevant covariates. (6)Physician-Reported Outcomes: Physician workload (NASA-TLX), diagnostic confidence, Decision Conflict Scale (DCS), Satisfaction With Decision (SWD) scores, frequency of MedGo-Sepsis use, and frequency of overriding MedGo-Sepsis recommendations will be analyzed using linear mixed-effects models. This approach will accommodate repeated measures and adjust for relevant covariates such as physician experience and case volume. (7)Long-Term Mortality: Mortality rates at 6 months and 1 year will be analyzed using Cox proportional hazards regression, similar to the analysis of the primary outcome. Statistical Descriptions Baseline characteristics will be presented using descriptive statistics. Continuous variables will be summarized as means with standard deviations (SD) if normally distributed, and as medians with interquartile ranges (IQR) if not. Categorical variables will be reported as frequencies and percentages. Baseline characteristics will be compared between intervention and control groups to assess for balance. Statistical Inference Inferential statistics will be employed to evaluate the impact of MedGo-Sepsis on all outcomes. The primary null hypothesis, which asserts that there is no difference in 28-day mortality between the groups, will be tested using Cox proportional hazards regression. For secondary outcomes, differences will be assessed using the previously mentioned statistical tests. Pre-specified subgroup analyses will be conducted to investigate potential heterogeneity in treatment effects based on the following factors: age, sex, severity of sepsis, and comorbidities. The results of these subgroup analyses will be reported with 95% confidence intervals. Missing Data Missing data will be addressed using multiple imputation with 20 imputed datasets. The imputation model will incorporate all relevant variables. Sensitivity analyses will be performed to evaluate the robustness of the results under different assumptions about the missing data mechanism. Software All statistical analyses will be performed using SPSS software (Version 24.0; IBM Corp., Armonk, NY, USA). Discussion Sepsis is a life-threatening condition characterized by significant heterogeneity in its pathophysiology, which presents ongoing diagnostic and therapeutic challenges in critical care settings [ 4 , 37 ]. Traditional diagnostic tools, such as SOFA, qSOFA and biomarkers like CRP and PCT, often lack the sensitivity and specificity required for timely and accurate diagnosis, particularly during the crucial early stages when intervention is essential [ 6 , 7 , 38 ]. Moreover, variability in immune responses among patients highlights the necessity for innovative, individualized approaches to improve outcomes. This randomized controlled trial (RCT) aims to evaluate the efficacy and safety of MedGo-Sepsis, a LLM specifically designed for sepsis management. Unlike existing LLMs, which primarily assist with general tasks such as clinical documentation and information retrieval [ 25 ], MedGo-Sepsis integrates complex clinical, laboratory, and immunological data to provide real-time, patient-specific diagnostic and therapeutic recommendations. By leveraging advanced NLPcapabilities and immune phenotyping, MedGo-Sepsis represents a novel solution to address the limitations of traditional diagnostic tools. The anticipated outcomes of this trial include improved diagnostic accuracy, reduced time to treatment initiation, and enhanced patient survival. Additionally, the study aims to establish a framework for incorporating LLMs into precision medicine for sepsis care, with potential applications extending to other critical illnesses. By directly addressing current gaps in sepsis diagnosis and management, this trial has the potential to advance both the science and practice of personalized critical care. This trial builds upon our recently finished retrospective study of MedGo (unpublished data), the foundational LLM for MedGo-Sepsis. We demonstrated MedGo's potential in facilitating the early diagnosis of sepsis, thereby providing significant assistance to clinicians, particularly junior physicians, in the time-sensitive context of the ED. However, it also identified a critical limitation: the absence of personalized treatment recommendations. This limitation, combined with findings from an ongoing prospective validation study conducted in our ED, highlights the urgent need to incorporate immune phenotyping data for truly personalized sepsis care. Therefore, we integrate LLM technology with immune phenotyping data, facilitating a comprehensive understanding of the intricate interactions between immune responses and disease progression. This multi-modal strategy aims to address the limitations of existing \"black box\" AI models, which primarily rely on structured EHR data and often fail to capture the dynamic and heterogeneous immune responses associated with sepsis[ 39 , 40 ]. By incorporating both clinical and immunological information, MedGo-Sepsis can generate personalized treatment recommendations, representing a significant advancement in the management of sepsis. Hence, the primary innovation of this study lies in the combination of immune phenotyping with an LLM, enabling more refined patient stratification and targeted interventions. Our central hypothesis posits that this personalized approach will lead to a reduction in 28-day all-cause mortality compared to standard care. Specifically, our prospective study indicates that current tools (qSOFA, SOFA) lack the necessary granularity for immune-based patient stratification. MedGo-Sepsis directly addresses this limitation through its integration of high-dimensional immune profiling data. Besides, our study reveals that variability in real-world data presents significant challenges for implementation, underscoring the necessity for robust and adaptable platforms like MedGo-Sepsis that can handle diverse data sources and varying quality. These challenges motivate the development of platforms like MedGo-Sepsis, which leverage the analytical capabilities of LLMs and comprehensive multi-modal data integration, including immune profiles, to move beyond the limitations of single biomarkers and static clinical scores in the management of sepsis. This trial addresses the limitations of previous AI research by employing a prospective, randomized controlled design (N = 524) that encompasses a comprehensive evaluation of clinician and patient-reported outcomes, resource utilization, and implementation factors. The rigorous methodology utilized, alongside the real-world ED setting, enhances the generalization and clinical relevance of our findings. We acknowledge certain limitations of the trial, including its single-center design, potential issues related to data variability, and the necessity for long-term follow-up to assess the impacts on survivors. Notwithstanding these limitations, this trial is a pivotal step towards rigorously evaluating the potential of LLMs, combined with immune phenotyping, to personalize and improve sepsis care. By addressing these methodological gaps, focusing on clinical effectiveness and real-world implementation, this trial provides valuable evidence to guide the future development and integration of LLM-based tools in critical care. Abbreviations SOFA Sequential Organ Failure Assessment qSOFA quick SOFA NEWS National Early Warning Score CRP C-reactive protein PCT procalcitonin sTREM-1 soluble triggering receptor expressed on myeloid cells-1 MDW monocyte distribution width HMGB1 high-mobility group box 1 IL-6 interleukin-6 PSP pancreatic stone protein CD64 cluster of differentiation 64 TNF-α tumor necrosis factor-alpha IL-1β interleukin-1 beta AI artificial intelligence ML machine learning HER electronic health record LLM large language models NLP natural language processing EHRs electronic health records DMC Data Monitoring Committee ARR absolute risk reduction MCID minimal clinically important difference HLA-DR monocyte antigen presentation HRQoL Health-Related Quality of Life EQ-5D-5L European Quality of Life Five Dimensions Five Levels NASA-TLX NASA Task Load Index DCS Decision Conflict Scale SWD Satisfaction With Decision SAEs serious adverse events FAS Full Analysis Set PPS Per-Protocol set SS Safety Set HR Hazard ratios CIs confidence intervals SD standard deviations IQR interquartile ranges ED Emergency Department. Declarations Ethics approval and consent to participate The Shanghai East Hospital Ethics Committee has approved the study protocol and informed consent procedures (Protocol ID: 2024YS-177). All patients will provide written informed consent. Physicians will obtain face-to-face informed consent during an interview prior to enrollment and will submit the written forms to the Shanghai East Hospital MedGo-Sepsis project team. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Author details 1 Shanghai East Clinical Medical College, Nanjing Medical University, Shanghai, China. 2 Department of Internal Emergency Medicine, Shanghai East Hospital, School of Medicine, Tongji University, Shanghai, China 3 Institute of Ethnology and Anthropology, Chinese Academy of Social Sciences, Beijing, China Funding This work was supported by grants from the the National Natural Science of Foundation (82470074 to L.T. ), the municipal Natural Science Foundation of Shanghai Scientific Committee of China (22ZR1451000 to L.T.), the peak supporting clinical discipline of Shanghai health bureau (2023ZDFC0104 to L.T). the key clinical discipline of Shanghai Pudong health bureau (PWZxk2022-17 to L.T.), the Joint research of Shanghai Pudong health bureau (PW2023-07 to L.T.),the top-notch innovative medical talents of Shanghai Pudong health bureau (2025PDWSYCBJ-03 to L.T.). Author Contribution The study protocol was designed by SJ, TL, CW, and BA. All authors reviewed and revised the protocol and approved the final manuscript. SJ and TL conceived the study, while CW drafted and finalized the paper with inputs from BA. HZ and LT provided critical revisions of the manuscript. Acknowledgements We sincerely thank all the healthcare professionals involved in this study for their dedicated efforts in patient enrollment, data collection, and follow-up. Availability of data and materials This manuscript is a study protocol and does not involve the sharing of data and materials. References Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, et al. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA. 2016;315:801–10. Hotchkiss RS, Monneret G, Payen D. Sepsis-induced immunosuppression: from cellular dysfunctions to immunotherapy. Nat Rev Immunol. 2013;13:862–74. Xie J, Wang H, Kang Y, Zhou L, Liu Z, Qin B, et al. The Epidemiology of Sepsis in Chinese ICUs: A National Cross-Sectional Survey. Crit Care Med. 2020;48:276–84. Rudd KE, Johnson SC, Agesa KM, Shackelford KA, Tsoi D, Kievlan DR, et al. Global, regional, and national sepsis incidence and mortality, 1990–2017: analysis for the Global Burden of Disease Study. Lancet. 2020;395(10219):200–11. Webb T, Holyoak KJ, Lu H. Emergent analogical reasoning in large language models. Nat Hum Behav. 2023;7:1526–41. Barichello T, Generoso JS, Singer M, Dal-Pizzol F. Biomarkers for sepsis: more than just fever and leukocytosis—a narrative review. Crit Care. 2022;26:14. Seymour CW, Gesten F, Prescott HC, Friedrich ME, Iwashyna TJ, Phillips GS, et al. Time to treatment and mortality during mandated emergency care for sepsis. N Engl J Med. 2017;376:2235–44. Emergency Medicine Branch Of Chinese Medical Care International Exchange Promotion Association, Emergency Medical Branch Of Chinese Medical Association. Chinese Medical Doctor Association Emergency Medical Branch, Chinese People's Liberation Army Emergency Medicine Professional Committee. Consensus of Chinese experts on early prevention and blocking of sepsis. Zhonghua Wei Zhong Bing Ji Jiu Yi Xue. 2020;32:518–30. Qiu X, Lei YP, Zhou RX, SIRS, SOFA, qSOFA. NEWS in the diagnosis of sepsis and prediction of adverse outcomes: a systematic review and meta-analysis. Expert Rev Anti Infect Ther. 2023;21:891–900. Wang C, Liang G, Shen J, Kong H, Wu D, Huang J, et al. Long Non-Coding RNAs as Biomarkers and Therapeutic Targets in Sepsis. Front Immunol. 2021;12:722004. Davenport EE, Burnham KL, Radhakrishnan J, Humburg P, Hutton P, Mills TC, et al. Genomic landscape of the individual host response and outcomes in sepsis. Lancet Respir Med. 2016;4:259–71. Saxena J, Das S, Kumar A, Sharma A, Sharma L, Kaushik S, et al. Biomarkers in sepsis. Clin Chim Acta. 2024. 10.1016/j.cca.2024.119891 . He RR, Yue GL, Dong ML, Wang JQ, Cheng C. Sepsis Biomarkers: Advancements and Clinical Applications-A Narrative Review. Int J Mol Sci. 2024;25:9010. Póvoa P, Coelho L, Dal-Pizzol F, Ferrer R, Huttner A, Conway Morris A, et al. How to use biomarkers of infection or sepsis at the bedside: guide to clinicians. Intensive Care Med. 2023;49:142–53. Seymour CW, Kennedy JN, Wang S, Chang CH, Elliott CF, Xu Z, et al. Derivation, validation, and potential treatment implications of novel clinical phenotypes for sepsis. JAMA. 2019;321:2003–17. Rhee C, Dantes R, Epstein L, Murphy DJ, Seymour CW, Iwashyna TJ, et al. Incidence and trends of sepsis in US hospitals using clinical vs claims data, 2009–2014. JAMA. 2017;318:1241–49. Vincent JL, Moreno R, Takala J, Willatts S, De Mendonça A, Bruining H, et al. The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. Intensive Care Med. 1996;22:707–10. Goh KH, Wang L, Yeow AYK, Poh H, Li K, Yeow JJL, et al. Artificial intelligence in sepsis early prediction and diagnosis using unstructured data in healthcare. Nat Commun. 2021;12:711. van der Poll T, Shankar-Hari M, Wiersinga WJ. The immunology of sepsis. Immunity. 2021;54:2450–64. Tang BM, Huang SJ, McLean AS. Genome-wide transcription profiling of human sepsis: a systematic review. Crit Care Med. 2010;38:2374–82. Mathias B, Mira JC, Larson SD. Pediatric sepsis. Curr Opin Pediatr. 2016;28:380–87. Sweeney TE, Azad TD, Donato M, Haynes WA, Perumal TM, Henao R, et al. Unsupervised analysis of transcriptomics in sepsis identifies phenotypes associated with outcomes. Nat Med. 2018;24:634–43. Alanazi A, Aldakhil L, Aldhoayan M, Aldosari B. Machine Learning for Early Prediction of Sepsis in Intensive Care Unit (ICU) Patients. Med (Kaunas). 2023;59:1276. Lalmuanawma S, Hussain J, Chhakchhuak L. Applications of machine learning and artificial intelligence for Covid-19 (SARS-CoV-2) pandemic: A review. Chaos Solitons Fractals. 2020;139:110059. Li J, Dada A, Puladi B, Kleesiek J, Egger J. ChatGPT in healthcare: a taxonomy and systematic review. Comput Methods Programs Biomed. 2024;245:108013. Ferrer R, Martin-Loeches I, Phillips G, Osborn TM, Townsend S, Dellinger RP, et al. Empiric antibiotic treatment reduces mortality in severe sepsis and septic shock from the first hour: results from a guideline-based performance improvement program. Crit Care Med. 2014;42:1749–55. Dellinger RP, Levy MM, Rhodes A, Annane D, Gerlach H, Opal SM, et al. Surviving Sepsis Campaign: International Guidelines for Management of Severe Sepsis and Septic Shock: 2012. Crit Care Med. 2013;41:580–637. Evans L, Rhodes A, Alhazzani W, Antonelli M, Coopersmith CM, French C, et al. Surviving sepsis campaign: international guidelines for management of sepsis and septic shock 2021. Intensive Care Med. 2021;47:1181–47. Giamarellos-Bourboulis EJ, Aschenbrenner AC, Bauer M, Bock C, Calandra T, Gat-Viks I, Kyriazopoulou E, et al. The pathophysiology of sepsis and precision-medicine-based immunotherapy. Nat Immunol. 2024;25:19–28. Rizvi MS, Gallo De Moraes A. New Decade, Old Debate: Blocking the Cytokine Pathways in Infection-Induced Cytokine Cascade. Crit Care Explor. 2021;3:e0364. Kyriazopoulou E, Leventogiannis K, Norrby-Teglund A, Dimopoulos G, Pantazi A, Orfanos SE, et al. Macrophage activation-like syndrome: an immunological entity associated with rapid progression to death in sepsis. BMC Med. 2017;15:172. Francois B, Jeannet R, Daix T, Walton AH, Shotwell MS, Unsinger J, et al. Interleukin-7 restores lymphocytes in septic shock: the IRIS-7 randomized clinical trial. JCI Insight. 2018;3:e98960. Kaplan RM, Hays RD. Health-Related Quality of Life Measurement in Public Health. Annu Rev Public Health. 2022;43:355–73. Hart SG, Staveland LE. Development of NASA-TLX (task load index): results of empirical and theoretical research. Adv Psychol. 1988;52:139–83. O’Connor AM. Validation of a decisional conflict scale. Med Decis Mak. 1995;15:25–30. Holmes-Rovner M, Kroll J, Schmitt N, Rovner DR, Breer ML, Rothert ML, et al. Patient satisfaction with health care decisions: the satisfaction with decision scale. Med Decis Mak. 1996;16:58–64. Weng L, Xu Y, Yin P, Wang Y, Chen Y, Liu W, et al. National incidence and mortality of hospitalized sepsis in China. Crit Care. 2023;27:84. Qiu X, Lei YP, Zhou RX, SIRS, SOFA, qSOFA. NEWS in the diagnosis of sepsis and prediction of adverse outcomes: a systematic review and meta-analysis. Expert Rev Anti Infect Ther. 2023;21:981–900. Komorowski M, Celi LA, Badawi O, Gordon AC, Faisal AA. The Artificial Intelligence Clinician learns optimal treatment strategies for sepsis in intensive care. Nat Med. 2018;24:1716–20. Fleuren LM, Klausch TLT, Zwager CL, Schoonmade LJ, Guo T, Roggeveen LF, et al. Machine learning for the prediction of sepsis: a systematic review and meta-analysis of diagnostic test accuracy. Intensive Care Med. 2020;46:383–400. Additional Declarations No competing interests reported. Supplementary Files SPRITDataSheet.doc Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {\"props\":{\"pageProps\":{\"initialData\":{\"identity\":\"rs-5873082\",\"acceptedTermsAndConditions\":true,\"allowDirectSubmit\":true,\"archivedVersions\":[],\"articleType\":\"Study protocol\",\"associatedPublications\":[],\"authors\":[{\"id\":406253969,\"identity\":\"77f28a19-fc20-4a3d-a538-51d81c048e7d\",\"order_by\":0,\"name\":\"Sen Jiang\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Shanghai East Clinical Medical College, Nanjing Medical University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Sen\",\"middleName\":\"\",\"lastName\":\"Jiang\",\"suffix\":\"\"},{\"id\":406253970,\"identity\":\"ffca7ad3-4ae1-4260-91da-98afea8c6d85\",\"order_by\":1,\"name\":\"Tong Liu\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Tongji University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Tong\",\"middleName\":\"\",\"lastName\":\"Liu\",\"suffix\":\"\"},{\"id\":406253974,\"identity\":\"d1fe6f3c-f469-4015-8be0-6b8bccd613f7\",\"order_by\":2,\"name\":\"Chunxue Wang\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Tongji University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Chunxue\",\"middleName\":\"\",\"lastName\":\"Wang\",\"suffix\":\"\"},{\"id\":406253975,\"identity\":\"108c5040-be1c-4896-97a8-1df24ac5e1d9\",\"order_by\":3,\"name\":\"Bo An\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Chinese Academy of Social Sciences\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Bo\",\"middleName\":\"\",\"lastName\":\"An\",\"suffix\":\"\"},{\"id\":406253976,\"identity\":\"6cf28e56-ef79-4b37-a99f-7c6de9a591cc\",\"order_by\":4,\"name\":\"Haitao Zhang\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Tongji University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Haitao\",\"middleName\":\"\",\"lastName\":\"Zhang\",\"suffix\":\"\"},{\"id\":406253977,\"identity\":\"cdda1a2b-240e-48ea-8e23-4e20381e03c3\",\"order_by\":5,\"name\":\"Lunxian Tang\",\"email\":\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAuUlEQVRIiWNgGAWjYBACxvmHDzA8YLDh4edvIFIL8wy2xIYEhjQZyRkHiNTCPoNHEajlsI0BkCQO8M7uYX+Q2Haex4DhAOOHjzlEaJGcc/ZgQ8KZ2zzmzA3MkjO3EaHFsCEP6JeK2zyWDQfYmHmJ0WJ/IMewIcHgHI/BgQQitTDOAGmpOECKlp5jiTMSziTzSM442EycXxjbmw98+NhmZ8/P33zww0ditCDrbiBN/SgYBaNgFIwC3AAAFj07GTda7NsAAAAASUVORK5CYII=\",\"orcid\":\"\",\"institution\":\"Shanghai East Clinical Medical College, Nanjing Medical University\",\"correspondingAuthor\":true,\"prefix\":\"\",\"firstName\":\"Lunxian\",\"middleName\":\"\",\"lastName\":\"Tang\",\"suffix\":\"\"}],\"badges\":[],\"createdAt\":\"2025-01-21 11:23:38\",\"currentVersionCode\":1,\"declarations\":\"\",\"doi\":\"10.21203/rs.3.rs-5873082/v1\",\"doiUrl\":\"https://doi.org/10.21203/rs.3.rs-5873082/v1\",\"draftVersion\":[],\"editorialEvents\":[],\"editorialNote\":\"\",\"failedWorkflow\":false,\"files\":[{\"id\":75308928,\"identity\":\"ce13d9fd-0e9c-40fb-a9b7-b14af2626b23\",\"added_by\":\"auto\",\"created_at\":\"2025-02-03 08:51:17\",\"extension\":\"jpeg\",\"order_by\":1,\"title\":\"Figure 1\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":91561,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eFlow chart of the clinical trial\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"floatimage1.jpeg\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-5873082/v1/80a0942bff7c4c8aa2e9b814.jpeg\"},{\"id\":91027452,\"identity\":\"f6f2125c-eb75-452a-bd92-6509b16e6bcc\",\"added_by\":\"auto\",\"created_at\":\"2025-09-10 21:16:22\",\"extension\":\"pdf\",\"order_by\":0,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"manuscript-pdf\",\"size\":845201,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"manuscript.pdf\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-5873082/v1/03d1bf97-40f9-4143-a23c-e1d05591254a.pdf\"},{\"id\":75310119,\"identity\":\"82126e18-0452-4d20-94cb-f43ce3538a2f\",\"added_by\":\"auto\",\"created_at\":\"2025-02-03 08:59:17\",\"extension\":\"doc\",\"order_by\":0,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":132608,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"SPRITDataSheet.doc\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-5873082/v1/2731ee1244ac0d544d8fa65a.doc\"}],\"financialInterests\":\"No competing interests reported.\",\"formattedTitle\":\"The Efficacy of a Novel Medical Artificial Intelligence Large Model(MedGo)-Guided Identification and Personalized Treatment of Sepsis: Study Protocol for a Single- centered Randomized Controlled Trial\",\"fulltext\":[{\"header\":\"Background\",\"content\":\"\\u003cp\\u003eSepsis, defined as a life-threatening organ dysfunction resulting from a dysregulated host response to infection, continues to pose a significant global health challenge and being a leading contributor to in-hospital mortality [\\u003cspan additionalcitationids=\\\"CR2\\\" citationid=\\\"CR1\\\" class=\\\"CitationRef\\\"\\u003e1\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR3\\\" class=\\\"CitationRef\\\"\\u003e3\\u003c/span\\u003e]. Recent estimates indicate that sepsis accounts for nearly 20% of all deaths worldwide, underscoring its extensive impact on healthcare systems [\\u003cspan citationid=\\\"CR4\\\" class=\\\"CitationRef\\\"\\u003e4\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR5\\\" class=\\\"CitationRef\\\"\\u003e5\\u003c/span\\u003e]. The clinical complexity of sepsis, influenced by its heterogeneous pathophysiology and rapid disease progression, often leads to delays in diagnosis and compromises treatment decisions [\\u003cspan citationid=\\\"CR6\\\" class=\\\"CitationRef\\\"\\u003e6\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR7\\\" class=\\\"CitationRef\\\"\\u003e7\\u003c/span\\u003e]. Early identification and prompt intervention are essential to prevent progression to septic shock or multi-organ dysfunction, both of which are associated with markedly increased mortality rates [\\u003cspan citationid=\\\"CR7\\\" class=\\\"CitationRef\\\"\\u003e7\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR8\\\" class=\\\"CitationRef\\\"\\u003e8\\u003c/span\\u003e]. Despite advancements in sepsis awareness and public health initiatives, existing diagnostic and management frameworks still fail to adequately address two critical needs: early and accurate detection, as well as personalized treatment.\\u003c/p\\u003e \\u003cp\\u003eClinicians have historically employed scoring systems such as the Sequential Organ Failure Assessment (SOFA), quick SOFA (qSOFA), and National Early Warning Score (NEWS) to identify patients with sepsis and predict clinical outcomes [\\u003cspan citationid=\\\"CR7\\\" class=\\\"CitationRef\\\"\\u003e7\\u003c/span\\u003e]. Although these tools offer valuable diagnostic and prognostic insights, they have inherent limitations, particularly in the context of early stage of sepsis or atypical presentations [\\u003cspan citationid=\\\"CR9\\\" class=\\\"CitationRef\\\"\\u003e9\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR10\\\" class=\\\"CitationRef\\\"\\u003e10\\u003c/span\\u003e]. These models often lack the necessary sensitivity and specificity for timely intervention, resulting in missed opportunities for achieving optimal patient outcomes [\\u003cspan citationid=\\\"CR9\\\" class=\\\"CitationRef\\\"\\u003e9\\u003c/span\\u003e].Additionally, their static nature does not account for the dynamic characteristics of sepsis and individual patient variability, thereby hindering their effectiveness in guiding personalized therapeutic strategies [\\u003cspan citationid=\\\"CR11\\\" class=\\\"CitationRef\\\"\\u003e11\\u003c/span\\u003e].\\u003c/p\\u003e \\u003cp\\u003eThe biomarkers for identification and management of sepsis has evolved beyond traditional ones such as C-reactive protein (CRP) and procalcitonin (PCT), which are limited by their specificity and sensitivity [\\u003cspan citationid=\\\"CR6\\\" class=\\\"CitationRef\\\"\\u003e6\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR12\\\" class=\\\"CitationRef\\\"\\u003e12\\u003c/span\\u003e].Emerging biomarkers, including soluble triggering receptor expressed on myeloid cells-1 (sTREM-1), monocyte distribution width (MDW), presepsin, high-mobility group box 1 (HMGB1), and interleukin-6 (IL-6), show promise for enhancing diagnostic accuracy and risk stratification in patients with sepsis[\\u003cspan citationid=\\\"CR6\\\" class=\\\"CitationRef\\\"\\u003e6\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR13\\\" class=\\\"CitationRef\\\"\\u003e13\\u003c/span\\u003e].Other extensively studied biomarkers include pancreatic stone protein (PSP) and cluster of differentiation 64 (CD64), along with additional candidates such as resistin, tumor necrosis factor-alpha (TNF-α), interleukin-1 beta (IL-1β), syndecan-1, and zonulin [\\u003cspan citationid=\\\"CR13\\\" class=\\\"CitationRef\\\"\\u003e13\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR14\\\" class=\\\"CitationRef\\\"\\u003e14\\u003c/span\\u003e]. Despite their potential, the clinical applicability of these biomarkers remains inconsistent due to discrepancies in measurement standardization, cost, and performance across various clinical settings [\\u003cspan citationid=\\\"CR13\\\" class=\\\"CitationRef\\\"\\u003e13\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR14\\\" class=\\\"CitationRef\\\"\\u003e14\\u003c/span\\u003e]. Furthermore, these individual biomarkers often fail to capture the complexity, dynamics, and heterogeneity of the immune response in sepsis, which hinders the advancement of effective precision medicine strategies. This highlights the need for innovative approaches that integrate multi-modal data, including comprehensive immune profiling, to provide a more holistic understanding of individual patient responses and facilitate personalized therapeutic interventions.\\u003c/p\\u003e \\u003cp\\u003eThe emergence of artificial intelligence (AI) and machine learning (ML) has opened promising new avenues for the management of sepsis. Numerous predictive models based on electronic health record (EHR) data have been developed to enhance early detection and risk stratification [\\u003cspan citationid=\\\"CR15\\\" class=\\\"CitationRef\\\"\\u003e15\\u003c/span\\u003e]. Although these models have shown some improvements over traditional scoring systems [\\u003cspan citationid=\\\"CR16\\\" class=\\\"CitationRef\\\"\\u003e16\\u003c/span\\u003e], their \\\"black box\\\" nature often obscures the underlying rationale, which can hinder clinician trust and adoption [\\u003cspan citationid=\\\"CR17\\\" class=\\\"CitationRef\\\"\\u003e17\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR18\\\" class=\\\"CitationRef\\\"\\u003e18\\u003c/span\\u003e]. Furthermore, many existing models primarily rely on structured data, overlooking the valuable insights available in unstructured clinical notes, radiology reports, and immune profiling data [18 4]. Lastly, their dependence on retrospective, single-center datasets raises concerns about the generalization of their findings[\\u003cspan citationid=\\\"CR17\\\" class=\\\"CitationRef\\\"\\u003e17\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR5\\\" class=\\\"CitationRef\\\"\\u003e5\\u003c/span\\u003e].\\u003c/p\\u003e \\u003cp\\u003eImmunological profiling has emerged as an essential tool for elucidating the intricate pathophysiology of sepsis, enabling the categorization of patients into distinct immunesubtypes. This approach holds significant promise for advancing personalized therapeutic strategies. Current methodologies for immune profiling, including cytokine assays, flow cytometry, and transcriptomic analysis, have successfully identified critical immune subphenotypes, such as hyper-inflammatory states (and immunosuppressive conditions, thereby providing valuable insights into individual variability in immune responses[\\u003cspan citationid=\\\"CR19\\\" class=\\\"CitationRef\\\"\\u003e19\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR20\\\" class=\\\"CitationRef\\\"\\u003e20\\u003c/span\\u003e]. Nevertheless, substantial challenges persist, including data heterogeneity, a lack of standardization in methodologies, and obstacles in translating research findings into real-time clinical decision-making [\\u003cspan additionalcitationids=\\\"CR20\\\" citationid=\\\"CR19\\\" class=\\\"CitationRef\\\"\\u003e19\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR21\\\" class=\\\"CitationRef\\\"\\u003e21\\u003c/span\\u003e].Recently, large language models (LLM) with equipped natural language processing (NLP) capabilities have emerged as transformative tools in medical research. By synthesizing extensive amounts of both structured and unstructured data-such as immune profiling outputs, electronic health records (EHRs), and laboratory results - LLMs have demonstrated considerable potential in deciphering complex immune patterns, correlating these patterns with clinical outcomes, and identifying novel immune subgroups within the context of sepsis [\\u003cspan citationid=\\\"CR22\\\" class=\\\"CitationRef\\\"\\u003e22\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR23\\\" class=\\\"CitationRef\\\"\\u003e23\\u003c/span\\u003e]. Recent research has demonstrated that LLMs capable of integrating immune features, such as cytokine profiles and transcriptomic data, significantly enhance the stratification of sepsis subtypes, potentially revealing previously unrecognized patterns of immune dysfunction [\\u003cspan citationid=\\\"CR24\\\" class=\\\"CitationRef\\\"\\u003e24\\u003c/span\\u003e]. Despite these advancements, several unresolved challenges impede the widespread adoption of LLM-based approaches in clinical practice. Key issues include the scarcity of large, high-quality, and diverse datasets for model training and validation, the variability of immune system responses across different populations, and the lack of interpretability of LLM outputs, which undermines clinician trust and integration into real-time clinical workflows [\\u003cspan citationid=\\\"CR22\\\" class=\\\"CitationRef\\\"\\u003e22\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR25\\\" class=\\\"CitationRef\\\"\\u003e25\\u003c/span\\u003e]. Therefore, addressing these limitations is essential to fully realize the potential of LLMs in sepsis immune phenotyping and establish a precision medicine framework for this life-threatening condition.\\u003c/p\\u003e \\u003cp\\u003eIn light of the significant potential of LLMs in processing medical data, challenges persist in the application of immune profiling, particularly regarding data heterogeneity, methodological standardization, and the provision of real-time clinical decision support for complex immune responses. To address these challenges, we have developed MedGo, a specialized medical language model designed to accurately interpret intricate clinical and immune data. MedGo integrates domain-specific knowledge derived from clinical guidelines and medical literature, enabling it to analyze both structured and unstructured data while incorporating high-dimensional immune profiling for comprehensive assessments of immune status. Building upon the foundation of MedGo, we have further developed MedGo-Sepsis, which is specifically tailored for the diagnosis and treatment of sepsis. MedGo-Sepsis derived extensive clinical knowledge with individualized immune information, thereby facilitating precise risk assessment and the formulation of personalized treatment strategies for patients with sepsis. It aims to enhance the early detection of sepsis, promoting timely interventions, such as early antibiotic administration, which ultimately improves diagnostic accuracy and clinical outcomes. Recognizing the significance of individual immune profiles in determining treatment effectiveness, MedGo-Sepsis incorporates personalized treatment recommendations based on immune phenotyping, thereby transcending traditional \\\"one-size-fits-all\\\" methodologies.\\u003c/p\\u003e \\u003cp\\u003eTo evaluate this innovative approach, we propose a randomized controlled trial comparing MedGo-Sepsis-guided treatment with standard care for patients presenting with suspected sepsis in the emergency department (ED). The primary objective of this trial is to assess the impact of MedGo-Sepsis on 28-day all-cause mortality. Secondary endpoints will include the time to antibiotic administration, physician diagnostic confidence, clinician decision-making efficiency, and physician workload. We hypothesize that the implementation of MedGo-Sepsis will result in improved outcomes, including reduced mortality, by enabling more accurate and timely diagnoses alongside personalized therapeutic interventions. This study aims to advance precision medicine in the management of sepsis, with the potential to transform clinical decision-making and enhance patient care in this critical condition.\\u003c/p\\u003e\"},{\"header\":\"Methods\",\"content\":\"\\u003cdiv id=\\\"Sec3\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003eStudy design\\u003c/h2\\u003e \\u003cp\\u003eThis study is a single-center, randomized, controlled trial designed to evaluate the efficacy of the MedGo-Sepsis LLM for the early identification and personalized treatment of sepsis in patients within the ED. The study will employ a parallel-group design, with participants randomly assigned in a 1:1 allocation ratio to either the intervention group (MedGo-Sepsis assisted care) or the control group (standard care). This pragmatic superiority trial seeks to demonstrate that the MedGo-Sepsis intervention is more effective than standard care in improving patient outcomes. To minimize the risk of treatment contamination, randomization will occur at the level of physician teams. Physicians in the ED will be organized into two teams, with each team randomized to either the MedGo-Sepsis or standard care arm. Physicians will be assigned to only one arm and will not switch groups at any point during the trial. The trial will be conducted in accordance with the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) 2013 statement ( see Additional file 1). Written informed consent will be obtained from all participants prior to enrollment. The trial is registered at (\\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003ehttp://www.chictr.org.cn\\u003c/span\\u003e\\u003cspan address=\\\"http://www.chictr.org.cn\\\" targettype=\\\"URL\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e) under the identifier (Trial ID:ChiCTR2400094116 ). A comprehensive schedule detailing enrollment, interventions, assessments, and follow-up visits is presented in the participant timeline (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig1\\\" class=\\\"InternalRef\\\"\\u003e1\\u003c/span\\u003e ).\\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003c/div\\u003e\\n\\u003ch3\\u003eRecruitment and Setting\\u003c/h3\\u003e\\n\\u003cp\\u003eThis single-center, randomized controlled trial will be conducted at the ED of Shanghai East Hospital in Shanghai, China. The study aims to enroll adult patients (\\u0026ge;\\u0026thinsp;18 years of age) who present to the ED with suspected sepsis. Suspected sepsis will be defined by the presence of (1) an infection or suspected infection along with any of the following criteria : (2) a qSOFA score\\u0026thinsp;\\u0026ge;\\u0026thinsp;2, (3) a SOFA score\\u0026thinsp;=\\u0026thinsp;1 or (4) a NEWS score of 4\\u0026ndash;6[\\u003cspan citationid=\\\"CR8\\\" class=\\\"CitationRef\\\"\\u003e8\\u003c/span\\u003e]. Patients will be screened for eligibility by the treating emergency physician upon arrival at the ED. Those who meet the inclusion and exclusion criteria will be approached for participation by the study team. The target sample size is 524 participants, with an anticipated enrollment period of 24 months. Recruitment strategies will include distributing flyers in the ED waiting area and informing physicians and nurses about the study to encourage referrals. Potentially eligible patients will be approached by trained research staff who will provide a detailed description of the study and obtain informed consent.\\u003c/p\\u003e\\n\\u003ch3\\u003eEligibility Criteria\\u003c/h3\\u003e\\n\\u003cdiv id=\\\"Sec6\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003eInclusion Criteria:\\u003c/h2\\u003e \\u003cp\\u003e1)Age\\u0026thinsp;\\u0026ge;\\u0026thinsp;18 years.\\u003c/p\\u003e \\u003cp\\u003e2)Presentation to the ED with suspected sepsis (refer to the definition above).\\u003c/p\\u003e \\u003cp\\u003e3)Provision of written informed consent. In cases where patients are unable to provide written informed consent due to the acute severity of their medical condition, or cognitive impairment, consent will be obtained from their legally authorized representative.\\u003c/p\\u003e \\u003c/div\\u003e\\n\\u003ch3\\u003eExclusion Criteria:\\u003c/h3\\u003e\\n\\u003cp\\u003e1)Age\\u0026thinsp;\\u0026lt;\\u0026thinsp;18 years.\\u003c/p\\u003e \\u003cp\\u003e2)Pregnancy.\\u003c/p\\u003e \\u003cp\\u003e3)Active malignancy.\\u003c/p\\u003e \\u003cp\\u003e4)HIV infection.\\u003c/p\\u003e \\u003cp\\u003e5)Inability to provide informed consent (e.g., due to cognitive impairment, language barriers, severe illness).\\u003c/p\\u003e \\u003cp\\u003e6)Input data that is considered unusable by the AI system (e.g., corrupted, incomplete, or insufficient to generate an output).\\u003c/p\\u003e \\u003cp\\u003e7)Active participation in another interventional trial. If a patient declines to participate, they will still receive the same quality of care as those enrolled in this study.\\u003c/p\\u003e \\u003cp\\u003eAdditional Information About Consent: The informed consent process will clearly outline the voluntary nature of participation, the right to withdraw at any time without penalty, and all aspects of the study in straightforward language, presented in, both written and oral formats. For potentially eligible participants with impaired cognitive function, a legally authorized representative or family member will be consulted to obtain informed consent.\\u003c/p\\u003e \\u003cdiv id=\\\"Sec8\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003eWithdrawal Criteria\\u003c/h2\\u003e \\u003cp\\u003eParticipants will be removed from the study under the following conditions:\\u003c/p\\u003e \\u003cp\\u003e1)Participants have the right to withdraw from the study at any time and for any reason without facing repercussions regarding their medical care.\\u003c/p\\u003e \\u003cp\\u003e2) The investigator reserves the right to withdraw a participant from the study if the participant develops a condition that makes further participation unsafe or if the participant fails to comply with the study protocol.\\u003c/p\\u003e \\u003cp\\u003e3)The Data Monitoring Committee (DMC) may recommend the termination of a participant's involvement or the discontinuation of the study based on safety concerns or other ethical considerations.\\u003c/p\\u003e \\u003cp\\u003e4)In the event that a participant cannot be reached for follow-up assessments despite reasonable efforts, they will be classified as lost to follow-up and withdrawn from the study.\\u003c/p\\u003e \\u003c/div\\u003e\\n\\u003ch3\\u003eBlinding\\u003c/h3\\u003e\\n\\u003cp\\u003eOutcome assessors will remain blinded to the treatment allocation. Due to the inherent characteristics of the intervention, it is not feasible to blind participants and treating physicians. The study statistician, responsible for data analysis, will also remain blinded to the treatment allocation until the final analysis is conducted. Unblinding will occur only after the database has been locked and prior to the interpretation of the results.The study statistician will not communicate with the unblinded research team in the hospital and will solely focus on the analysis.\\u003c/p\\u003e\\n\\u003ch3\\u003eRandomization\\u003c/h3\\u003e\\n\\u003cp\\u003eEligible participants will be randomly assigned in a 1:1 ratio to either the MedGo-Sepsis group or the standard care group. This assignment will utilize a computer-generated random sequence with permuted blocks of size four. Randomization will be stratified by sepsis severity (mild or severe) and the presence of comorbidities (present or absent). Within each stratum, a separate randomization sequence will be created. The randomization sequence will be generated by a statistician who is independent of the clinical care team using SAS software for this purpose. To ensure allocation concealment, a secure, centralized, web-based randomization system will be employed, which investigators can access only after obtaining informed consent from the participant. The allocation assignment for each participant will be disclosed to the treating physician only after the baseline data has been collected and entered into the system.\\u003c/p\\u003e \\u003cdiv id=\\\"Sec11\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003eSample Size\\u003c/h2\\u003e \\u003cp\\u003eBased on a hypothesized absolute risk reduction (ARR) of 5% in 28-day mortality associated with MedGo-Sepsis (decreasing from 20% in the standard care group to 15%)[\\u003cspan citationid=\\\"CR4\\\" class=\\\"CitationRef\\\"\\u003e4\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR7\\\" class=\\\"CitationRef\\\"\\u003e7\\u003c/span\\u003e], a sample size of 472 participants (236 per group) is required to achieve 80% power at a two-sided alpha level of 0.05. This calculation was performed using G*Power 3.1 (version 3.1.9.2, Germany) and employed a two-proportion z-test, assuming a binomial distribution. To account for a 10% dropout rate, we will target the enrollment of 524 participants (262 per group). The study is also adequately powered to detect clinically significant differences in secondary outcomes, such as the time to first antibiotic administration (minimal clinically important difference (MCID) : 2-hour reduction) and length of hospital stay (MCID: 2-day reduction) [\\u003cspan citationid=\\\"CR26\\\" class=\\\"CitationRef\\\"\\u003e26\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR27\\\" class=\\\"CitationRef\\\"\\u003e27\\u003c/span\\u003e].\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec12\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003eInterventions\\u003c/h2\\u003e \\u003cdiv id=\\\"Sec13\\\" class=\\\"Section3\\\"\\u003e \\u003ch2\\u003eMedGo-Sepsis Intervention Group\\u003c/h2\\u003e \\u003cp\\u003eParticipants assigned to the MedGo-Sepsis group will receive standard care for suspected sepsis in conjunction with the MedGo-Sepsis intervention. Standard care will be administered in accordance with the guidelines established by the Surviving Sepsis Campaign [\\u003cspan citationid=\\\"CR28\\\" class=\\\"CitationRef\\\"\\u003e28\\u003c/span\\u003e].This care encompasses, but is not limited to, initial resuscitation with fluids and vasopressors, obtaining cultures prior to the administration of broad-spectrum antibiotics, identifying and managing the source of infection, and providing supportive care for organ dysfunction.\\u003c/p\\u003e \\u003cp\\u003eUpon presentation to the ED, patient data\\u0026mdash;including demographics, vital signs, laboratory values, clinical findings, relevant imaging reports, and free-text clinical notes\\u0026mdash;will be entered into the MedGo-Sepsis system by the treating physician after the established inclusion and exclusion criteria have been met. The system will utilize its natural language processing (NLP) capabilities and sepsis-specific algorithms to provide the following functionalities:(1) Real-time Sepsis Risk Assessment: MedGo-Sepsis will generate a sepsis risk score (probability of sepsis) based on the input data. This score will be prominently displayed within the system interface and integrated into the patient\\u0026rsquo;s electronic health record (EHR). (2) Detailed Measurement of Immune Parameters: The system will utilize immune-related data from hospital assays, including: Cytokine Measurements to assess inflammatory responses; Lymphocyte Phenotyping to analyze CD4+, CD8\\u0026thinsp;+\\u0026thinsp;T cells, B cells, and NK cells; and Innate and Adaptive Immunity Monitoring to Evaluate neutrophil functions, monocyte antigen presentation (HLA-DR), NK cell activity, T cell counts, Treg cell proportions, and B cell function through immunoglobulin measurements.(3) Stratification Based on Immune Phenotypes: The MedGo-Sepsis system will categorize patients into distinct immune phenotypes based on the measured immune parameters: Hyperinflammation Subtype: This subtype is characterized by elevated levels of pro-inflammatory cytokines and excessive immune activation, indicated by interleukin-6 (IL-6) levels\\u0026thinsp;\\u0026ge;\\u0026thinsp;100 pg/mL.Immunosuppressed Subtype: This subtype is defined by diminished T-cell responses and elevated levels of anti-inflammatory cytokines, often indicated by a reduction in lymphocyte counts (lymphocyte count\\u0026thinsp;\\u0026le;\\u0026thinsp;900 cells/\\u0026micro;L) and a percentage of HLA-DR positive monocytes\\u0026thinsp;\\u0026lt;\\u0026thinsp;30%[\\u003cspan additionalcitationids=\\\"CR30 CR31\\\" citationid=\\\"CR29\\\" class=\\\"CitationRef\\\"\\u003e29\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR32\\\" class=\\\"CitationRef\\\"\\u003e32\\u003c/span\\u003e]. (4)Personalized Treatment Recommendations: Utilizing the identified immune phenotype, Medgo-Sepsis will formulate individualized treatment strategies. For patients classified under the Hyperinflammation Subtype, the system will advocate for the initiation of broad-spectrum antibiotics in conjunction with anti-inflammatory agents such as hormone. For patients classified as immunosuppressed, particularly those with lymphopenia, the system will recommend immune-enhancing therapies such as interferon gamma, recombinant interleukin-7, immunoglobulin therapy, or thymosin alpha-1 to restore immune function.(5) Dynamic Monitoring and Continuous Feedback: MedGo-Sepsis will continuously monitor incoming patient data and provide real-time alerts if the patient's condition deteriorates or if deviations from recommended treatment protocols occur. Follow-up assessments of cytokine levels and immune responses in patients post-sepsis treatment will inform adjustments to therapeutic strategies. A review of the literature on cytokine levels and immune responses after treatment will be integrated to enhance the adaptability of treatment plans.Physicians are not required to adhere to the recommendations provided by MedGo-Sepsis and retain full autonomy in making treatment decisions. However, the treating physician must document their reasons for accepting or overruling the system's recommendations for each patient. Training on the use of the MedGo-Sepsis system will be provided to all participating physicians and research staff prior to the start of the study.\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec14\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003eStandard Care Control Group\\u003c/h2\\u003e \\u003cp\\u003eParticipants assigned to the standard care group will receive conventional treatment for suspected sepsis in accordance with the Surviving Sepsis Campaign guidelines [\\u003cspan citationid=\\\"CR32\\\" class=\\\"CitationRef\\\"\\u003e32\\u003c/span\\u003e], without any intervention from the MedGo-Sepsis system. The physicians responsible for the care of patients in this group will have access to all standard diagnostic tools and treatment options, and their clinical decisions will be guided by their own judgment and experience. Data related the treatment administered in the standard care group, including the timing and type of interventions, will be collected and documented using a standardized data collection form.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec15\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003eOutcome Measures\\u003c/h2\\u003e \\u003cdiv id=\\\"Sec16\\\" class=\\\"Section3\\\"\\u003e \\u003ch2\\u003ePrimary Outcome\\u003c/h2\\u003e \\u003cp\\u003eThe primary outcome of this study is 28-day all-cause mortality: Specifically, it refers to mortality from any cause within 28 days of enrollment. This outcome provides a definitive assessment of the overall impact of MedGo-Sepsis on patient survival.\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec17\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003eSecondary Outcomes\\u003c/h2\\u003e \\u003cp\\u003e(1)Organ Dysfunction:\\u003c/p\\u003e \\u003cp\\u003eChange in Sequential Organ Failure Assessment (SOFA) score from baseline to 72 hours. The SOFA score is a widely used and validated measure of organ dysfunction in sepsis, ranging from 0 (normal function) to 4 (severe dysfunction) for each of the six organ systems [\\u003cspan citationid=\\\"CR1\\\" class=\\\"CitationRef\\\"\\u003e1\\u003c/span\\u003e].A higher total score indicates greater organ dysfunction and increased severity of sepsis.\\u003c/p\\u003e \\u003cp\\u003e(2)Diagnostic Performance:\\u003c/p\\u003e \\u003cp\\u003e1)Diagnostic Time.\\u003c/p\\u003e \\u003cp\\u003e2)Misdiagnosis Rate.\\u003c/p\\u003e \\u003cp\\u003e3)Missed Diagnosis Rate.\\u003c/p\\u003e \\u003cp\\u003e4)Diagnostic Accuracy.\\u003c/p\\u003e \\u003cp\\u003e(3) \\u003cb\\u003eTreatment Efficacy\\u003c/b\\u003e:\\u003c/p\\u003e \\u003cp\\u003e1)Time to appropriate antibiotic administration\\u003c/p\\u003e \\u003cp\\u003e2)Use of adjunctive therapies (e.g., fluids, vasopressors, corticosteroids, immunomodulatory therapies).\\u003c/p\\u003e \\u003cp\\u003e3)SOFA score change at day 7 and day 14.\\u003c/p\\u003e \\u003cp\\u003e4)Need for mechanical ventilation.\\u003c/p\\u003e \\u003cp\\u003e(4)Resource Utilization:\\u003c/p\\u003e \\u003cp\\u003e1)Length of hospital stay.\\u003c/p\\u003e \\u003cp\\u003e2)Length of ICU stay.\\u003c/p\\u003e \\u003cp\\u003e3)Total hospital costs.\\u003c/p\\u003e \\u003cp\\u003e(5)Patient-Reported Outcomes:\\u003c/p\\u003e \\u003cp\\u003eHealth-Related Quality of Life (HRQoL) was assessed at 28 days and 6 months using the EQ-5D-5L, a standardized measure of health status. This tool generates a single index value that represents overall health, along with scores for five dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression[\\u003cspan citationid=\\\"CR33\\\" class=\\\"CitationRef\\\"\\u003e33\\u003c/span\\u003e].\\u003c/p\\u003e \\u003cp\\u003e(6)Physician-Reported Outcomes:\\u003c/p\\u003e \\u003cp\\u003e1)The workload of physicians can be assessed using the NASA Task Load Index (NASA-TLX), a multi-dimensional rating system that evaluates perceived workload across six subscales: mental demand, physical demand, temporal demand, performance, effort, and frustration[\\u003cspan citationid=\\\"CR34\\\" class=\\\"CitationRef\\\"\\u003e34\\u003c/span\\u003e].\\u003c/p\\u003e \\u003cp\\u003e2) Physician diagnostic confidence was evaluated using a 5-point Likert scale, with 1 indicating at all confident 5 indicating\\u003c/p\\u003e \\u003cp\\u003e3)Decision Conflict Scale (DCS): This scale measures decisional conflict related to medical choices and consists of 16 items across five subscales: Informed (knowledge about options), Values Clarity (clearness about personal values), Support (feeling supported in decision-making), Uncertainty (clarity about which option is best), and Effective Decision (feeling the decision made was the best possible) [\\u003cspan citationid=\\\"CR35\\\" class=\\\"CitationRef\\\"\\u003e35\\u003c/span\\u003e].\\u003c/p\\u003e \\u003cp\\u003e4)Satisfaction With Decision (SWD) scale: This [Number]-item scale measures satisfaction with healthcare decisions. Higher scores on this scale reflect a greater level of satisfaction with the decisions made [\\u003cspan citationid=\\\"CR36\\\" class=\\\"CitationRef\\\"\\u003e36\\u003c/span\\u003e].\\u003c/p\\u003e \\u003cp\\u003e5)Frequency of MedGo-Sepsis use by physicians: (Proportion of patients for which recommendations were generated by MedGo-Sepsis that were used by the physician).\\u003c/p\\u003e \\u003cp\\u003e6)Frequency of overriding MedGo-Sepsis recommendations: (Proportional frequency for which MedGo-Sepsis' recommendations were overridden by the treating physician).\\u003c/p\\u003e \\u003cp\\u003e(7)Long-Term Mortality:All-cause mortality at 6 months and 1 year.\\u003c/p\\u003e \\u003cp\\u003eOther Data Collected\\u003c/p\\u003e \\u003cp\\u003e(1)MedGo-Sepsis Performance:Performance errors (deviations from expert consensus).\\u003c/p\\u003e \\u003cp\\u003e(2)Safety:All adverse events, including serious adverse events (SAEs).\\u003c/p\\u003e \\u003cp\\u003e(3)Detailed Resource Use: Specific data on resource utilization (e.g., diagnostic tests, medications) will be collected.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec18\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003eStatistical Methods\\u003c/h2\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec19\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003eStatistical Analysis\\u003c/h2\\u003e \\u003cp\\u003eThe primary analysis will be conducted using the Full Analysis Set (FAS), which will include data from all patients who are randomized, receive at least one dose of treatment and have at least one outcome assessment. Additional supportive analyses will be performed on the Per-Protocol set (PPS) and the Safety Set (SS). PPS will consist of participants who adhere to the study protocol without major violations and who do not have significant missing baseline or outcome data. SS will include all randomized participants who have received at least one dose of treatment and have completed at least one safety assessment. The primary outcome, 28-day all-cause mortality, will be compared between the MedGo-Sepsis and standard care groups using Cox proportional hazards regression. This model will adjust for pre-specified confounding variables, including age, sex, initial SOFA score, presence of comorbidities (e.g., diabetes, chronic lung disease), and site of infection. Hazard ratios (HR) with corresponding 95% confidence intervals (CIs) will be reported. Before interpretation, the proportional hazards assumption will be validated using the Schoenfeld residuals test. Should this assumption be violated, alternative methodologies such as stratified Cox regression or the incorporation of time-dependent covariates will be considered.\\u003c/p\\u003e \\u003cp\\u003eSecondary outcome analyses will be conducted as follows:\\u003c/p\\u003e \\u003cp\\u003e(1) Organ Dysfunction: Change in SOFA score at 72 hours will be analyzed using a linear mixed-effects model, which accommodates repeated measures within patients (baseline and 72 hours) and adjusts for the previously mentioned covariates. Changes in individual SOFA subscores will be analyzed using the same methodology.\\u003c/p\\u003e \\u003cp\\u003e(2)Diagnostic Performance:Diagnostic time will be compared between groups using t-tests or Mann-Whitney U test, depending on the distribution. The misdiagnosis rate, missed diagnosis rate, and diagnostic accuracy, which are proportions, will be compared using Chi-squared tests or Fisher's exact test, as appropriate.\\u003c/p\\u003e \\u003cp\\u003e(3)Treatment Efficacy:Time to appropriate antibiotic administration will be evaluated using survival analysis, specifically Cox proportional hazards regression or Kaplan-Meier methods, depending on the validity of the proportional hazards assumption. The use of adjunctive therapies, and the need for mechanical ventilation will be analyzed using logistic regression. Changes in the SOFA score at day 7 and day 14 will be assessed using linear mixed-effects models, similar to the analysis conducted at the 72-hour mark.\\u003c/p\\u003e \\u003cp\\u003e(4)Resource Utilization: The length of hospital stay, ICU length of stay, and total hospital costs will be compared using t-tests, Mann-Whitney U, or generalized linear models, with covariate adjustments as necessary.\\u003c/p\\u003e \\u003cp\\u003e(5)Patient-Reported Outcomes: HRQoL (EQ-5D-5L) will be assessed using linear regression (analysis, or an appropriate alternative method if the assumption of normality is not satisfied. Additionally, patient satisfaction will be evaluated with ordinal logistic regression (or similar) at each time point, adjusting for baseline scores and other relevant covariates.\\u003c/p\\u003e \\u003cp\\u003e(6)Physician-Reported Outcomes: Physician workload (NASA-TLX), diagnostic confidence, Decision Conflict Scale (DCS), Satisfaction With Decision (SWD) scores, frequency of MedGo-Sepsis use, and frequency of overriding MedGo-Sepsis recommendations will be analyzed using linear mixed-effects models. This approach will accommodate repeated measures and adjust for relevant covariates such as physician experience and case volume.\\u003c/p\\u003e \\u003cp\\u003e(7)Long-Term Mortality: Mortality rates at 6 months and 1 year will be analyzed using Cox proportional hazards regression, similar to the analysis of the primary outcome.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec20\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003eStatistical Descriptions\\u003c/h2\\u003e \\u003cp\\u003eBaseline characteristics will be presented using descriptive statistics. Continuous variables will be summarized as means with standard deviations (SD) if normally distributed, and as medians with interquartile ranges (IQR) if not. Categorical variables will be reported as frequencies and percentages. Baseline characteristics will be compared between intervention and control groups to assess for balance.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec21\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003eStatistical Inference\\u003c/h2\\u003e \\u003cp\\u003eInferential statistics will be employed to evaluate the impact of MedGo-Sepsis on all outcomes. The primary null hypothesis, which asserts that there is no difference in 28-day mortality between the groups, will be tested using Cox proportional hazards regression. For secondary outcomes, differences will be assessed using the previously mentioned statistical tests. Pre-specified subgroup analyses will be conducted to investigate potential heterogeneity in treatment effects based on the following factors: age, sex, severity of sepsis, and comorbidities. The results of these subgroup analyses will be reported with 95% confidence intervals.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec22\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003eMissing Data\\u003c/h2\\u003e \\u003cp\\u003eMissing data will be addressed using multiple imputation with 20 imputed datasets. The imputation model will incorporate all relevant variables. Sensitivity analyses will be performed to evaluate the robustness of the results under different assumptions about the missing data mechanism.\\u003c/p\\u003e \\u003cdiv id=\\\"Sec23\\\" class=\\\"Section3\\\"\\u003e \\u003ch2\\u003eSoftware\\u003c/h2\\u003e \\u003cp\\u003eAll statistical analyses will be performed using SPSS software (Version 24.0; IBM Corp., Armonk, NY, USA).\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e\"},{\"header\":\"Discussion\",\"content\":\"\\u003cp\\u003eSepsis is a life-threatening condition characterized by significant heterogeneity in its pathophysiology, which presents ongoing diagnostic and therapeutic challenges in critical care settings [\\u003cspan citationid=\\\"CR4\\\" class=\\\"CitationRef\\\"\\u003e4\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR37\\\" class=\\\"CitationRef\\\"\\u003e37\\u003c/span\\u003e]. Traditional diagnostic tools, such as SOFA, qSOFA and biomarkers like CRP and PCT, often lack the sensitivity and specificity required for timely and accurate diagnosis, particularly during the crucial early stages when intervention is essential [\\u003cspan citationid=\\\"CR6\\\" class=\\\"CitationRef\\\"\\u003e6\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR7\\\" class=\\\"CitationRef\\\"\\u003e7\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR38\\\" class=\\\"CitationRef\\\"\\u003e38\\u003c/span\\u003e]. Moreover, variability in immune responses among patients highlights the necessity for innovative, individualized approaches to improve outcomes. This randomized controlled trial (RCT) aims to evaluate the efficacy and safety of MedGo-Sepsis, a LLM specifically designed for sepsis management.\\u003c/p\\u003e \\u003cp\\u003eUnlike existing LLMs, which primarily assist with general tasks such as clinical documentation and information retrieval [\\u003cspan citationid=\\\"CR25\\\" class=\\\"CitationRef\\\"\\u003e25\\u003c/span\\u003e], MedGo-Sepsis integrates complex clinical, laboratory, and immunological data to provide real-time, patient-specific diagnostic and therapeutic recommendations. By leveraging advanced NLPcapabilities and immune phenotyping, MedGo-Sepsis represents a novel solution to address the limitations of traditional diagnostic tools. The anticipated outcomes of this trial include improved diagnostic accuracy, reduced time to treatment initiation, and enhanced patient survival. Additionally, the study aims to establish a framework for incorporating LLMs into precision medicine for sepsis care, with potential applications extending to other critical illnesses. By directly addressing current gaps in sepsis diagnosis and management, this trial has the potential to advance both the science and practice of personalized critical care.\\u003c/p\\u003e \\u003cp\\u003eThis trial builds upon our recently finished retrospective study of MedGo (unpublished data), the foundational LLM for MedGo-Sepsis. We demonstrated MedGo's potential in facilitating the early diagnosis of sepsis, thereby providing significant assistance to clinicians, particularly junior physicians, in the time-sensitive context of the ED. However, it also identified a critical limitation: the absence of personalized treatment recommendations. This limitation, combined with findings from an ongoing prospective validation study conducted in our ED, highlights the urgent need to incorporate immune phenotyping data for truly personalized sepsis care. Therefore, we integrate LLM technology with immune phenotyping data, facilitating a comprehensive understanding of the intricate interactions between immune responses and disease progression. This multi-modal strategy aims to address the limitations of existing \\\"black box\\\" AI models, which primarily rely on structured EHR data and often fail to capture the dynamic and heterogeneous immune responses associated with sepsis[\\u003cspan citationid=\\\"CR39\\\" class=\\\"CitationRef\\\"\\u003e39\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR40\\\" class=\\\"CitationRef\\\"\\u003e40\\u003c/span\\u003e]. By incorporating both clinical and immunological information, MedGo-Sepsis can generate personalized treatment recommendations, representing a significant advancement in the management of sepsis. Hence, the primary innovation of this study lies in the combination of immune phenotyping with an LLM, enabling more refined patient stratification and targeted interventions. Our central hypothesis posits that this personalized approach will lead to a reduction in 28-day all-cause mortality compared to standard care. Specifically, our prospective study indicates that current tools (qSOFA, SOFA) lack the necessary granularity for immune-based patient stratification. MedGo-Sepsis directly addresses this limitation through its integration of high-dimensional immune profiling data.\\u003c/p\\u003e \\u003cp\\u003eBesides, our study reveals that variability in real-world data presents significant challenges for implementation, underscoring the necessity for robust and adaptable platforms like MedGo-Sepsis that can handle diverse data sources and varying quality. These challenges motivate the development of platforms like MedGo-Sepsis, which leverage the analytical capabilities of LLMs and comprehensive multi-modal data integration, including immune profiles, to move beyond the limitations of single biomarkers and static clinical scores in the management of sepsis. This trial addresses the limitations of previous AI research by employing a prospective, randomized controlled design (N\\u0026thinsp;=\\u0026thinsp;524) that encompasses a comprehensive evaluation of clinician and patient-reported outcomes, resource utilization, and implementation factors. The rigorous methodology utilized, alongside the real-world ED setting, enhances the generalization and clinical relevance of our findings.\\u003c/p\\u003e \\u003cp\\u003eWe acknowledge certain limitations of the trial, including its single-center design, potential issues related to data variability, and the necessity for long-term follow-up to assess the impacts on survivors. Notwithstanding these limitations, this trial is a pivotal step towards rigorously evaluating the potential of LLMs, combined with immune phenotyping, to personalize and improve sepsis care. By addressing these methodological gaps, focusing on clinical effectiveness and real-world implementation, this trial provides valuable evidence to guide the future development and integration of LLM-based tools in critical care.\\u003c/p\\u003e\"},{\"header\":\"Abbreviations\",\"content\":\"\\u003cdiv class=\\\"DefinitionList\\\"\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eSOFA\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eSequential Organ Failure Assessment\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eqSOFA\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003equick SOFA\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eNEWS\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eNational Early Warning Score\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eCRP\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eC-reactive protein\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003ePCT\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eprocalcitonin\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003esTREM-1\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003esoluble triggering receptor expressed on myeloid cells-1\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eMDW\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003emonocyte distribution width\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eHMGB1\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003ehigh-mobility group box 1\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eIL-6\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003einterleukin-6\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003ePSP\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003epancreatic stone protein\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eCD64\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003ecluster of differentiation 64\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eTNF-α\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003etumor necrosis factor-alpha\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eIL-1β\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003einterleukin-1 beta\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eAI\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eartificial intelligence\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eML\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003emachine learning\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eHER\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eelectronic health record\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eLLM\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003elarge language models\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eNLP\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003enatural language processing\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eEHRs\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eelectronic health records\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eDMC\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eData Monitoring Committee\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eARR\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eabsolute risk reduction\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eMCID\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eminimal clinically important difference\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eHLA-DR\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003emonocyte antigen presentation\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eHRQoL\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eHealth-Related Quality of Life\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eEQ-5D-5L\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eEuropean Quality of Life Five Dimensions Five Levels\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eNASA-TLX\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eNASA Task Load Index\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eDCS\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eDecision Conflict Scale\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eSWD\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eSatisfaction With Decision\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eSAEs\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eserious adverse events\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eFAS\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eFull Analysis Set\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003ePPS\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003ePer-Protocol set\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eSS\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eSafety Set\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eHR\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eHazard ratios\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eCIs\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003econfidence intervals\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eSD\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003estandard deviations\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eIQR\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003einterquartile ranges\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eED\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eEmergency Department.\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003c/div\\u003e\"},{\"header\":\"Declarations\",\"content\":\" \\u003cp\\u003e \\u003cstrong\\u003eEthics approval and consent to participate\\u003c/strong\\u003e \\u003cp\\u003e The Shanghai East Hospital Ethics Committee has approved the study protocol and informed consent procedures (Protocol ID: 2024YS-177). All patients will provide written informed consent. Physicians will obtain face-to-face informed consent during an interview prior to enrollment and will submit the written forms to the Shanghai East Hospital MedGo-Sepsis project team.\\u003c/p\\u003e \\u003c/p\\u003e \\u003cp\\u003e \\u003cstrong\\u003eConsent for publication\\u003c/strong\\u003e \\u003cp\\u003eNot applicable.\\u003c/p\\u003e \\u003c/p\\u003e \\u003cp\\u003e \\u003cstrong\\u003eCompeting interests\\u003c/strong\\u003e \\u003cp\\u003eThe authors declare that they have no competing interests.\\u003c/p\\u003e \\u003c/p\\u003e\\u003cp\\u003e \\u003ch2\\u003eAuthor details\\u003c/h2\\u003e \\u003cp\\u003e \\u003csup\\u003e1\\u003c/sup\\u003eShanghai East Clinical Medical College, Nanjing Medical University, Shanghai, China.\\u003c/p\\u003e \\u003cp\\u003e \\u003csup\\u003e2\\u003c/sup\\u003e Department of Internal Emergency Medicine, Shanghai East Hospital, School of Medicine, Tongji University, Shanghai, China\\u003c/p\\u003e \\u003cp\\u003e \\u003csup\\u003e3\\u003c/sup\\u003e Institute of Ethnology and Anthropology, Chinese Academy of Social Sciences, Beijing, China\\u003c/p\\u003e \\u003c/p\\u003e\\u003ch2\\u003eFunding\\u003c/h2\\u003e \\u003cp\\u003eThis work was supported by grants from the the National Natural Science of Foundation (82470074 to L.T. ), the municipal Natural Science Foundation of Shanghai Scientific Committee of China (22ZR1451000 to L.T.), the peak supporting clinical discipline of Shanghai health bureau (2023ZDFC0104 to L.T). the key clinical discipline of Shanghai Pudong health bureau (PWZxk2022-17 to L.T.), the Joint research of Shanghai Pudong health bureau (PW2023-07 to L.T.),the top-notch innovative medical talents of Shanghai Pudong health bureau (2025PDWSYCBJ-03 to L.T.).\\u003c/p\\u003e\\u003ch2\\u003eAuthor Contribution\\u003c/h2\\u003e\\u003cp\\u003eThe study protocol was designed by SJ, TL, CW, and BA. All authors reviewed and revised the protocol and approved the final manuscript. SJ and TL conceived the study, while CW drafted and finalized the paper with inputs from BA. HZ and LT provided critical revisions of the manuscript.\\u003c/p\\u003e\\u003ch2\\u003eAcknowledgements\\u003c/h2\\u003e \\u003cp\\u003eWe sincerely thank all the healthcare professionals involved in this study for their dedicated efforts in patient enrollment, data collection, and follow-up.\\u003c/p\\u003e\\u003ch2\\u003eAvailability of data and materials\\u003c/h2\\u003e \\u003cp\\u003eThis manuscript is a study protocol and does not involve the sharing of data and materials.\\u003c/p\\u003e\"},{\"header\":\"References\",\"content\":\"\\u003col\\u003e\\u003cli\\u003e\\u003cspan\\u003eSinger M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, et al. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA. 2016;315:801\\u0026ndash;10.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eHotchkiss RS, Monneret G, Payen D. Sepsis-induced immunosuppression: from cellular dysfunctions to immunotherapy. Nat Rev Immunol. 2013;13:862\\u0026ndash;74.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eXie J, Wang H, Kang Y, Zhou L, Liu Z, Qin B, et al. The Epidemiology of Sepsis in Chinese ICUs: A National Cross-Sectional Survey. Crit Care Med. 2020;48:276\\u0026ndash;84.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eRudd KE, Johnson SC, Agesa KM, Shackelford KA, Tsoi D, Kievlan DR, et al. Global, regional, and national sepsis incidence and mortality, 1990\\u0026ndash;2017: analysis for the Global Burden of Disease Study. Lancet. 2020;395(10219):200\\u0026ndash;11.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eWebb T, Holyoak KJ, Lu H. Emergent analogical reasoning in large language models. Nat Hum Behav. 2023;7:1526\\u0026ndash;41.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eBarichello T, Generoso JS, Singer M, Dal-Pizzol F. Biomarkers for sepsis: more than just fever and leukocytosis\\u0026mdash;a narrative review. Crit Care. 2022;26:14.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eSeymour CW, Gesten F, Prescott HC, Friedrich ME, Iwashyna TJ, Phillips GS, et al. Time to treatment and mortality during mandated emergency care for sepsis. N Engl J Med. 2017;376:2235\\u0026ndash;44.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eEmergency Medicine Branch Of Chinese Medical Care International Exchange Promotion Association, Emergency Medical Branch Of Chinese Medical Association. Chinese Medical Doctor Association Emergency Medical Branch, Chinese People's Liberation Army Emergency Medicine Professional Committee. Consensus of Chinese experts on early prevention and blocking of sepsis. Zhonghua Wei Zhong Bing Ji Jiu Yi Xue. 2020;32:518\\u0026ndash;30.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eQiu X, Lei YP, Zhou RX, SIRS, SOFA, qSOFA. NEWS in the diagnosis of sepsis and prediction of adverse outcomes: a systematic review and meta-analysis. Expert Rev Anti Infect Ther. 2023;21:891\\u0026ndash;900.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eWang C, Liang G, Shen J, Kong H, Wu D, Huang J, et al. Long Non-Coding RNAs as Biomarkers and Therapeutic Targets in Sepsis. Front Immunol. 2021;12:722004.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eDavenport EE, Burnham KL, Radhakrishnan J, Humburg P, Hutton P, Mills TC, et al. Genomic landscape of the individual host response and outcomes in sepsis. Lancet Respir Med. 2016;4:259\\u0026ndash;71.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eSaxena J, Das S, Kumar A, Sharma A, Sharma L, Kaushik S, et al. Biomarkers in sepsis. Clin Chim Acta. 2024. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.1016/j.cca.2024.119891\\u003c/span\\u003e\\u003cspan address=\\\"10.1016/j.cca.2024.119891\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eHe RR, Yue GL, Dong ML, Wang JQ, Cheng C. Sepsis Biomarkers: Advancements and Clinical Applications-A Narrative Review. Int J Mol Sci. 2024;25:9010.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eP\\u0026oacute;voa P, Coelho L, Dal-Pizzol F, Ferrer R, Huttner A, Conway Morris A, et al. How to use biomarkers of infection or sepsis at the bedside: guide to clinicians. Intensive Care Med. 2023;49:142\\u0026ndash;53.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eSeymour CW, Kennedy JN, Wang S, Chang CH, Elliott CF, Xu Z, et al. Derivation, validation, and potential treatment implications of novel clinical phenotypes for sepsis. JAMA. 2019;321:2003\\u0026ndash;17.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eRhee C, Dantes R, Epstein L, Murphy DJ, Seymour CW, Iwashyna TJ, et al. Incidence and trends of sepsis in US hospitals using clinical vs claims data, 2009\\u0026ndash;2014. JAMA. 2017;318:1241\\u0026ndash;49.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eVincent JL, Moreno R, Takala J, Willatts S, De Mendon\\u0026ccedil;a A, Bruining H, et al. The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. Intensive Care Med. 1996;22:707\\u0026ndash;10.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eGoh KH, Wang L, Yeow AYK, Poh H, Li K, Yeow JJL, et al. Artificial intelligence in sepsis early prediction and diagnosis using unstructured data in healthcare. Nat Commun. 2021;12:711.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003evan der Poll T, Shankar-Hari M, Wiersinga WJ. The immunology of sepsis. Immunity. 2021;54:2450\\u0026ndash;64.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eTang BM, Huang SJ, McLean AS. Genome-wide transcription profiling of human sepsis: a systematic review. Crit Care Med. 2010;38:2374\\u0026ndash;82.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eMathias B, Mira JC, Larson SD. Pediatric sepsis. Curr Opin Pediatr. 2016;28:380\\u0026ndash;87.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eSweeney TE, Azad TD, Donato M, Haynes WA, Perumal TM, Henao R, et al. Unsupervised analysis of transcriptomics in sepsis identifies phenotypes associated with outcomes. Nat Med. 2018;24:634\\u0026ndash;43.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eAlanazi A, Aldakhil L, Aldhoayan M, Aldosari B. Machine Learning for Early Prediction of Sepsis in Intensive Care Unit (ICU) Patients. Med (Kaunas). 2023;59:1276.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eLalmuanawma S, Hussain J, Chhakchhuak L. Applications of machine learning and artificial intelligence for Covid-19 (SARS-CoV-2) pandemic: A review. Chaos Solitons Fractals. 2020;139:110059.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eLi J, Dada A, Puladi B, Kleesiek J, Egger J. ChatGPT in healthcare: a taxonomy and systematic review. Comput Methods Programs Biomed. 2024;245:108013.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eFerrer R, Martin-Loeches I, Phillips G, Osborn TM, Townsend S, Dellinger RP, et al. Empiric antibiotic treatment reduces mortality in severe sepsis and septic shock from the first hour: results from a guideline-based performance improvement program. Crit Care Med. 2014;42:1749\\u0026ndash;55.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eDellinger RP, Levy MM, Rhodes A, Annane D, Gerlach H, Opal SM, et al. Surviving Sepsis Campaign: International Guidelines for Management of Severe Sepsis and Septic Shock: 2012. Crit Care Med. 2013;41:580\\u0026ndash;637.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eEvans L, Rhodes A, Alhazzani W, Antonelli M, Coopersmith CM, French C, et al. Surviving sepsis campaign: international guidelines for management of sepsis and septic shock 2021. Intensive Care Med. 2021;47:1181\\u0026ndash;47.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eGiamarellos-Bourboulis EJ, Aschenbrenner AC, Bauer M, Bock C, Calandra T, Gat-Viks I, Kyriazopoulou E, et al. The pathophysiology of sepsis and precision-medicine-based immunotherapy. Nat Immunol. 2024;25:19\\u0026ndash;28.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eRizvi MS, Gallo De Moraes A. New Decade, Old Debate: Blocking the Cytokine Pathways in Infection-Induced Cytokine Cascade. Crit Care Explor. 2021;3:e0364.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eKyriazopoulou E, Leventogiannis K, Norrby-Teglund A, Dimopoulos G, Pantazi A, Orfanos SE, et al. Macrophage activation-like syndrome: an immunological entity associated with rapid progression to death in sepsis. BMC Med. 2017;15:172.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eFrancois B, Jeannet R, Daix T, Walton AH, Shotwell MS, Unsinger J, et al. Interleukin-7 restores lymphocytes in septic shock: the IRIS-7 randomized clinical trial. JCI Insight. 2018;3:e98960.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eKaplan RM, Hays RD. Health-Related Quality of Life Measurement in Public Health. Annu Rev Public Health. 2022;43:355\\u0026ndash;73.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eHart SG, Staveland LE. Development of NASA-TLX (task load index): results of empirical and theoretical research. Adv Psychol. 1988;52:139\\u0026ndash;83.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eO\\u0026rsquo;Connor AM. Validation of a decisional conflict scale. Med Decis Mak. 1995;15:25\\u0026ndash;30.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eHolmes-Rovner M, Kroll J, Schmitt N, Rovner DR, Breer ML, Rothert ML, et al. Patient satisfaction with health care decisions: the satisfaction with decision scale. Med Decis Mak. 1996;16:58\\u0026ndash;64.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eWeng L, Xu Y, Yin P, Wang Y, Chen Y, Liu W, et al. National incidence and mortality of hospitalized sepsis in China. Crit Care. 2023;27:84.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eQiu X, Lei YP, Zhou RX, SIRS, SOFA, qSOFA. NEWS in the diagnosis of sepsis and prediction of adverse outcomes: a systematic review and meta-analysis. Expert Rev Anti Infect Ther. 2023;21:981\\u0026ndash;900.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eKomorowski M, Celi LA, Badawi O, Gordon AC, Faisal AA. The Artificial Intelligence Clinician learns optimal treatment strategies for sepsis in intensive care. Nat Med. 2018;24:1716\\u0026ndash;20.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eFleuren LM, Klausch TLT, Zwager CL, Schoonmade LJ, Guo T, Roggeveen LF, et al. Machine learning for the prediction of sepsis: a systematic review and meta-analysis of diagnostic test accuracy. Intensive Care Med. 2020;46:383\\u0026ndash;400.\\u003c/span\\u003e\\u003c/li\\u003e\\u003c/ol\\u003e\"}],\"fulltextSource\":\"\",\"fullText\":\"\",\"funders\":[],\"hasAdminPriorityOnWorkflow\":false,\"hasManuscriptDocX\":true,\"hasOptedInToPreprint\":true,\"hasPassedJournalQc\":\"\",\"hasAnyPriority\":false,\"hideJournal\":true,\"highlight\":\"\",\"institution\":\"\",\"isAcceptedByJournal\":false,\"isAuthorSuppliedPdf\":false,\"isDeskRejected\":\"\",\"isHiddenFromSearch\":false,\"isInQc\":false,\"isInWorkflow\":false,\"isPdf\":false,\"isPdfUpToDate\":true,\"isWithdrawnOrRetracted\":false,\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"identity\":\"researchsquare\",\"isNatureJournal\":false,\"hasQc\":true,\"allowDirectSubmit\":true,\"externalIdentity\":\"\",\"sideBox\":\"\",\"snPcode\":\"\",\"submissionUrl\":\"/submission\",\"title\":\"Research Square\",\"twitterHandle\":\"researchsquare\",\"acdcEnabled\":true,\"dfaEnabled\":false,\"editorialSystem\":\"\",\"reportingPortfolio\":\"\",\"inReviewEnabled\":false,\"inReviewRevisionsEnabled\":true},\"keywords\":\"Sepsis, Large Language Model, Artificial Intelligence, Personalized Medicine, Emergency Department\",\"lastPublishedDoi\":\"10.21203/rs.3.rs-5873082/v1\",\"lastPublishedDoiUrl\":\"https://doi.org/10.21203/rs.3.rs-5873082/v1\",\"license\":{\"name\":\"CC BY 4.0\",\"url\":\"https://creativecommons.org/licenses/by/4.0/\"},\"manuscriptAbstract\":\"\\u003cp\\u003e\\u003cstrong\\u003eBackground\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eSepsis, a life-threatening organ dysfunction resulting from a dysregulated host response to infection, remains a major global health challenge with high morbility and mortality. Current diagnostic and management frameworks still lack the satisfied sensitivity and specificity for early detection and personalized treatment. In this study, we aim to evaluate the efficacy of MedGo-Sepsis, a novel large language model developed by our collaborating team and us that integrates clinical and immunological data, to guide the identification and personalized treatment of sepsis in the emergency department.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eMethods\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThis single-center, randomized controlled trial will enroll adult patients presenting to the emergency department with suspected sepsis. Participants will be randomized 1:1 at the physician team level to either the intervention group receiving standard care augmented by MedGo-Sepsis guidance or the control group receiving standard care alone. MedGo-Sepsis provides real-time risk assessments, detailed immune parameter profiling, stratification based on immune phenotypes, and personalized treatment recommendations. The primary outcome is 28-day all-cause mortality. Secondary outcomes include changes in Sequential Organ Failure Assessment (SOFA) score, diagnostic accuracy, time to appropriate antibiotic administration, resource utilization, patient-reported outcomes, and physician workload.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eDiscussion\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThis trial will assess whether MedGo-Sepsis, through the integration of individualized immune data with large language model technology, improves outcomes for patients with sepsis compared to standard care. The combination of enhanced diagnostics and tailored therapeutic strategies has the potential to advance precision management of sepsis in the critical emergency setting.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eTrial registration\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThe trial has been prospectively registered in the Chinese Clinical Trial Registry (ChiCTR2400094116) on December 17, 2024.\\u003c/p\\u003e\",\"manuscriptTitle\":\"The Efficacy of a Novel Medical Artificial Intelligence Large Model(MedGo)-Guided Identification and Personalized Treatment of Sepsis: Study Protocol for a Single- centered Randomized Controlled Trial\",\"msid\":\"\",\"msnumber\":\"\",\"nonDraftVersions\":[{\"code\":1,\"date\":\"2025-02-03 08:51:12\",\"doi\":\"10.21203/rs.3.rs-5873082/v1\",\"editorialEvents\":[{\"type\":\"communityComments\",\"content\":0}],\"status\":\"published\",\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"identity\":\"researchsquare\",\"isNatureJournal\":false,\"hasQc\":true,\"allowDirectSubmit\":true,\"externalIdentity\":\"\",\"sideBox\":\"\",\"snPcode\":\"\",\"submissionUrl\":\"/submission\",\"title\":\"Research Square\",\"twitterHandle\":\"researchsquare\",\"acdcEnabled\":true,\"dfaEnabled\":false,\"editorialSystem\":\"\",\"reportingPortfolio\":\"\",\"inReviewEnabled\":false,\"inReviewRevisionsEnabled\":true}}],\"origin\":\"\",\"ownerIdentity\":\"a5fcb30e-41a6-4cc4-aa4a-4a640579358c\",\"owner\":[],\"postedDate\":\"February 3rd, 2025\",\"published\":true,\"recentEditorialEvents\":[],\"rejectedJournal\":[],\"revision\":\"\",\"amendment\":\"\",\"status\":\"posted\",\"subjectAreas\":[],\"tags\":[],\"updatedAt\":\"2025-09-10T21:08:15+00:00\",\"versionOfRecord\":[],\"versionCreatedAt\":\"2025-02-03 08:51:12\",\"video\":\"\",\"vorDoi\":\"\",\"vorDoiUrl\":\"\",\"workflowStages\":[]},\"version\":\"v1\",\"identity\":\"rs-5873082\",\"journalConfig\":\"researchsquare\"},\"__N_SSP\":true},\"page\":\"/article/[identity]/[[...version]]\",\"query\":{\"redirect\":\"/article/rs-5873082\",\"identity\":\"rs-5873082\",\"version\":[\"v1\"]},\"buildId\":\"8U1c8b4HqxoKbykW_rLl7\",\"isFallback\":false,\"isExperimentalCompile\":false,\"dynamicIds\":[84888],\"gssp\":true,\"scriptLoader\":[]}","source_license":"CC-BY-4.0","license_restricted":false}