Open research questions in Sepsis Diagnosis and Treatment

What the literature leaves open

• Several limitations of this study must be acknowledged. The present study, which was designed to assess admission lactate only, did not allow for evaluation of lactate clearance, representing a key limitation. Future larger, multicenter studies incorporating lactate kinetics and dynamic resuscitation parameters, particularly in low- and middle-income settings, are warranted to better define the prognostic role of lactate in septic shock.

  Prognostic Value of Admission Blood Lactate for Predicting 28-Day Mortality in Patients With Septic Shock: A Prospective Pilot Study · 2026 · DOI

• However, current evidence is limited by predominantly retrospective designs, variable cutoff values, and insufficient external validation of incremental value beyond established tools such as SOFA and APACHE II.

  The CALLY Index in Critical Illness: Risk Stratification and Its Potential Role in Clinical Monitoring · 2026 · DOI

• Purpose Sepsis-associated acute kidney injury (SA-AKI) raises mortality risk, while the independent links of pre-existing diabetes mellitus (DM) and in-hospital glycemic status to SA-AKI remain unclear.

  Impact of diabetes mellitus and glucose level control on early sepsis-associated acute kidney injury: a multicenter retrospective observational study · 2026 · DOI

• However, the prognostic significance of the nadir serum albumin level during hospitalization in patients with spontaneous intracerebral hemorrhage (ICH) admitted to the intensive care unit (ICU) remains unclear.

  Association of nadir serum albumin during hospitalization with mortality in ICU patients with spontaneous intracerebral hemorrhage · 2026 · DOI

• Third, detailed clinical data regarding the type or severity of shock and acute respiratory failure were insufficient; specifically, information on the timing, dose, and dynamic response to vasoactive therapy was limited, precluding assessment of potential differential effects across varying degrees of circulatory or respira- tory compromise.

  Prognostic impact of shock at ICU admission in acute respiratory failure · 2026 · DOI

• Background The Geriatric Nutritional Risk Index (GNRI) is a validated nutritional assessment tool predictive of outcomes in elderly patients, yet its prognostic value in adult cancer patients with sepsis remains unexplored.

  The Geriatric Nutritional Risk Index and its association with all-cause mortality in cancer patients with sepsis: a dual-center retrospective cohort study · 2026 · DOI

• Background and Objectives: Bacteremia is associated with worse outcomes in sepsis, but comparative data between Gram-positive bacteremia (GPB) and Gram-negative bacteremia (GNB) remain scarce.

  Clinical Characteristics, Empirical Antibiotic Appropriateness, and Outcomes of Gram-Positive Versus Gram-Negative Bacteremia in Emergency Department Patients with Sepsis · 2026 · DOI

• The optimal level of care and nursing ratios for risk- stratified cardiovascular diagnoses that do not require critical care therapies remain understudied and the provision of individual therapeutic inventions and hemodynamic monitoring modalities is often dictated by local hospital protocols and unit-level nursing training.

  Cardiac Intensive Care Unit Appropriate Patient Selection and Triage: A Scientific Statement From the American Heart Association · 2026 · DOI

• Future work will focus on prospective multicentre validation to confirm generalisability across different populations, pathogen distributions, and healthcare settings. Sep- sis is dynamic; future work should explore LSTM or Temporal Convolutional Networks to model biomarker velocity over the first 24 h.

  Diagnostic evaluation of Pentraxin-3 and its complementary role to Procalcitonin for respiratory sepsis source identification using explainable machine learning · 2026 · DOI

• Abstract Background International statements suggest using the Global Leadership Initiative on Malnutrition (GLIM) criteria in intensive care units (ICUs); however, the economic impact of GLIM‐defined malnutrition in sepsis remains unclear.

  Association between GLIM‐defined malnutrition and in‐hospital costs in adults with sepsis: A sub analysis of a prospective cohort study · 2026 · DOI

• Future research should focus on personalized hemodynamic management, advanced monitoring technologies, and precision-based therapeutic approaches capable of further improving survival, reducing organ dysfunction, and enhancing the quality of critical care delivered to patients with septic shock throughout the American continent. Nevertheless, further research is required to clarify its long-term effects on survival and organ function. Although additional research is needed to further clarify their long-term impact on survival and organ preservation, these agents represent important therapeutic options that may enhance individualized hemodynamic management. Although evidence regarding mortality benefits remains inconclusive, vasopressin may reduce catecholamine exposure and contribute to improved vascular responsiveness through alternative physiological pathways (Russell et al.

  Hemodynamic Impact of Vasopressor Therapy in Critically Ill Patients with Septic Shock in the Intensive Care Unit · 2026 · DOI

• Nevertheless, we must emphasize that PIIV and other inflammatory parameters, in conjunction with well- established scoring systems, should be validated in large prospective studies to effectively stratify CAP patients upon ICU admission. e inflammation markers—such as CRP—for diseases such as pneumonia, bacteremia, sepsis, and adverse clinical outcomes, despite the presence of some contradictory findings in the literature.

  Assessing the Prognostic Performance of Pan Immune-Inflammation Value and Inflammatory Biomarkers in Critically Ill Patients with Community-Acquired Pneumonia · 2026 · DOI

• Several limitations of this study warrant acknowledgment. Despite the relatively large sample size of the MIMIC-IV 3.0 and validation cohorts, the retrospective design may introduce selection bias and limit generalizability. The comparisons between the included and excluded patients in the MIMIC-IV 3.0 database indicate that our findings are derived from sepsis patients with relatively high severity and mortality risk; their interpretation in milder patients warrants further investigation. Due to substantial missing data for several nutrition/inflammation-based indices, only ten of the 18 indices were direct compared; the remaining should be compared with well-documented indices in further studies. In multivariable analysis, despite adjusting for important clinical parameters, unmeasured or residual confounding may still persist. As the SSIC system was established solely based on limited biochemical parameters, its predictive accuracy for mortality is limited when used alone, and it is therefore recommended to be combined with existing severity scores for better risk assessment. Given the high missing rate of C-reactive protein, this marker was not incorporated into the SSIC system, and further research is therefore warranted. The application of the SSIC system for severity assessment and mortality prediction in sepsis patients requires prospective validation. The specific interventions guided by the SSIC system and their effectiveness in reducing mortality risk also require prospective investigation. Notably, moderate SSIC risk was not associated with increased in-ICU mortality after adjustment in the validation cohort, which should be considered in clinical use, and modifying the SSIC system to fit local populations may be helpful. Finally, dynamic changes in nutrition/inflammation-based indicators may also be related to sepsis mortality and should be investigated in future studies.

  Development and validation of a sepsis-specific systemic inflammation classification system for mortality prediction · 2026 · DOI

• By positioning clini- cal trajectories, immune-cell-defined endotypes, PIICS, multi-omics discordance, and treatment-effect hetero- geneity within one translational framework, the review aims to clarify why the field is moving toward staged decision-making and why current evidence remains insufficient for unrestricted bedside implementation.

  Beyond the sepsis label: heterogeneity, subphenotypes, and the path to precision therapy · 2026 · DOI

• spanning This review has several limitations. The literature is methodologically heterogeneous, clinical cohorts, secondary trial analyses, immune studies, and multi-omics investigations, which limits direct com- parability across proposed subphenotypes and endo- types. Many classification systems remain temporally unstable, incompletely validated across settings, and not yet ready for routine bedside implementation. Much of the evidence for phenotype-guided therapy still comes from retrospective or exploratory analyses rather than prospective phenotype-stratified trials. In addition, this article is a narrative synthesis and does not introduce a fundamentally new taxonomy; its novelty lies in inte- grating existing frameworks, clarifying terminology, highlighting PIICS as a longitudinal gap, and identifying translational barriers that must be addressed before pre- cision approaches can change practice.

  Beyond the sepsis label: heterogeneity, subphenotypes, and the path to precision therapy · 2026 · DOI

• any the dans année le monde.1 Malgré Le sepsis représente un fardeau considérable pour la santé mondiale, contribuant à plus de 5 millions de décès par progrès thérapeutiques, les taux de mortalité demeurent élevés: plus de 25 % des personnes atteintes de sepsis en meurent, et la mortalité hospitalière liée au choc septique avoisine 50 %.2 Des stratégies visant à optimiser le traitement du sepsis ont donc le potentiel d’améliorer substantiellement les soins à l'échelle mondiale. les L’apparition d’une hypoperfusion constitue un facteur clé du sepsis. En effet, le sepsis entraîne une hypoperfusion par vasodilatation et perturbations microcirculatoires, l’ischémie tissulaire résultant d’un déséquilibre systémique et local entre l’apport en oxygène et les besoins tissulaires. Ces processus mènent à une défaillance des organes cibles et, ultimement, au décès. Des efforts considérables ont été déployés pour repérer des marqueurs d’hypoperfusion et mettre au point des stratégies visant à améliorer la perfusion chez les personnes atteintes de sepsis.3–6 L’un des est lactate. Une paramètres lactatémie élevée (> 2 mmol·L−1) reflète de multiples facteurs, notamment un apport insuffisant en oxygène aux tissus, une altération de la respiration aérobie, une accélération de la glycolyse aérobie et une réduction de la clairance hépatique. Une lactatémie plus élevée est associée à une mortalité accrue et a été utilisée pour définir le choc septique en association avec une hypotension persistante.2 Compte tenu de l’association entre la lactatémie et le devenir des personnes soignées, l’utilité du lactate comme l’objet de cible de recherches.7–9 réanimation continue de couramment utilisés faire le les résultats d’une La quête de cibles optimales de réanimation en contexte de choc septique a évolué, passant de protocoles uniformes à une compréhension de plus en plus nuancée selon laquelle l’intensification des soins n’est pas toujours synonyme de meilleures issues. Dans ce numéro du Journal, Liu et coll.10 étude observationnelle rapportent examinant la relation entre le taux de clairance du lactate et la mortalité chez des personnes gravement malades. Cette étude, qui regroupait plus de 13 000 personnes atteintes de sepsis recensées dans les bases de données Medical Information Mart for Intensive Care IV (MIMIC-IV) et Electronic Intensive Care Unit (eICU), a montré que des taux de clairance du lactate croissants étaient associés à une réduction de la mortalité, jusqu'à un certain point. Lorsque les taux de clairance du lactate dépassaient 11 % par heure, une réanimation plus poussée semblait produire des gains décroissants et n'était pas associée à une diminution supplémentaire du risque de décès. Cette observation constitue une avancée importante vers une réanimation personnalisée, permettant d'éviter les deux écueils que sont la sous-réanimation et la sur-réanimation.

  Lactate clearance in patients with septic shock: identifying the point of diminishing returns · 2026 · DOI

• Abnormally elevated cerebral arterial RI is closely associ- ated with adverse neurological outcomes in sepsis. The clini- cal spectrum of SAE ranges from mild cognitive impairment to delirium and coma and represents an independent predic- tor of mortality in sepsis [27, 37, 38]. Higher RI values have been shown to correlate positively with the incidence and severity of SAE as well as with poor mid- and long-term outcomes [39, 40]. From a pathophysiological perspective, an increased RI integrates inflammation-mediated micro- vascular injury and impaired cerebral perfusion, providing hemodynamic evidence for the development of SAE and suggesting that resuscitation strategies based solely on tra- ditional MAP targets may be insufficient to ensure adequate cerebral microcirculatory perfusion. However, several limitations and confounders must be considered when interpreting cerebral RI in clinical practice. First, as noted above, cerebral vessels are extremely sensitive to PaCO₂ [32], mechanical ventilation strategies (e.g., hyper- ventilation or permissive hypercapnia) can substantially alter RI and obscure sepsis-related pathological changes. Second, the widespread use of sedative and analgesic agents in the ICU reduces the cerebral metabolic rate of oxygen (CMRO₂) [26] and, via metabolism–flow coupling, may induce cer- ebral vasoconstriction and an increase in RI. The most criti- cal limitation is the strong blood-pressure dependence of RI when cerebral autoregulation is impaired. In the absence of effective autoregulation, reductions in systemic blood pres- sure directly decrease diastolic cerebral blood flow velocity, resulting in an elevated RI. In this setting, an increased RI primarily reflects systemic hypoperfusion rather than a pri- mary increase in cerebrovascular resistance. In summary, cerebral arterial RI in sepsis is not a static marker of injury but a dynamic window into the interac- tion among autoregulatory capacity, microcirculatory integ- rity, and systemic hemodynamics. Its interpretation must therefore be contextualized within the disease stage, global circulatory status, and ventilation/sedation conditions— further underscoring the necessity of serial, rather than sin- gle-time-point, assessment within the ARISE framework. Recent work on sepsis-associated alterations in cerebral blood flow should also be incorporated here to strengthen the evidence base. Renal resistance index: microcirculatory dysfunction and dissociation from global hemodynamics Among all organ-related resistance indices, the renal artery RI is arguably the best supported by evidence—and one of the most easily misinterpreted. Its main clinical value lies not in reflecting total renal blood flow, but in reveal- ing sepsis-related renal microcirculatory dysfunction and its “dissociation” from global systemic hemodynamics [41, 42].

  ARISE framework: a temporal, multi-organ view of perfusion in sepsis · 2026 · DOI

• The number of patients meeting each definition varied widely (211–47,888 patients), demonstrating that even small differences in thresholds or timing markedly influence patient identification.

  Definitions and Prognoses of Persistent Inflammation, Immunosuppression, and Catabolism Syndrome: A Comparative Analysis via Systematic Review and Observational Study in Japan · 2026 · DOI

• Nebivolol (NEB), a selective β1-blocker with vasodilatory and antioxidant properties, has shown organ-protective effects in sepsis, but its cardiovascular role remains unclear.

  Nebivolol as a potential adjunct therapy for sepsis-induced cardiac dysfunction: evidence from a rat model · 2026 · DOI

• injury allowed independently associated with Third, the results support catecholamine-sparing strategies in patients with increasing doses of norepinephrine. Although vasopressin use was not lower mortality in this analysis, its higher use in the high-exposure group the common practice of adding nonadrenergic vasopressors in refractory shock. Future studies should assess whether earlier introduction of vasopressin reduces catecholaminergic load and improves tissue perfusion. 5.5. Strengths of the study A relevant strength is the integration of clinical, hemodynamic, and metabolic variables. Unlike studies focusing only on mortality or lactate, this lactate included vasopressor dose, analysis clearance, capillary refilling, diuresis, SOFA, and mortality. This approach better reflects the pathophysiological complexity of septic shock. Another strength is the multivariate analysis adjusted for basal gravity. The inclusion of APACHE II, SOFA, initial lactate, and acute kidney to partially reduce confounding by indication. In addition, the comparison of predictive models showed that the incorporation of vasopressor load and tissue perfusion improves mortality discrimination. 5.6.

  Effect of vasopressor catecholamines on tissue perfusion and mortality in patients with septic shock in ICU · 2026 · DOI

• Several limitations should be acknowledged when interpreting the results of this study. First among these is the cross-sectional design, which precludes the determination of causal relationships between nurses’ knowledge, self- confidence, and associated factors. Second, data were obtained via self- reported surveys, which may be influenced by response bias or social desirability, potentially causing participants to overstate their knowledge or confidence. Third, the use of convenience sampling introduces the possibility of selection bias. Nurses who chose to participate may have been more ARTICLE IN PRESS ARTICLE IN PRESS ACCEPTED MANUSCRIPT interested in sepsis, more confident in their clinical knowledge, or more motivated to engage in research than non-participants. As a result, the findings may overestimate the overall level of sepsis knowledge and confidence among the broader nursing workforce. In addition, although participants were recruited from two hospitals and from a range of clinical settings, the sample may not be fully representative of all nurses within these institutions. Fourth, the study was conducted at two hospitals within a single healthcare cluster in eastern China, and the response rate of 44.3% further limits the generalizability of findings to all registered nurses working in other tertiary care or community healthcare settings. Fifth, the study did not assess all contextual and organizational factors that may influence nurses’ sepsis knowledge and confidence, such as institutional policies, workload, access to continuing education, or local implementation practices related to sepsis protocols. Sixth, although the survey incorporated case-based multiple- choice questions to assess applied knowledge, such formats may not fully reflect real-world clinical competence or decision-making abilities under time-sensitive conditions. Objective assessments, such as simulation-based testing or direct observation of clinical practice, could provide a more accurate measure of performance. Finally, while an open-ended question was included to explore organizational support needs, the study lacked in-depth qualitative insights into nurses’ experiences and perceptions of sepsis care, which could have offered richer contextual understanding to guide policy development and educational strategies. Future research should adopt mixed-methods designs, include larger multicenter samples, and integrate qualitative approaches to enhance the validity, applicability, and robustness of findings.

  Assessing nurses’ knowledge and confidence in adult sepsis: evidence from two hospitals in eastern China · 2026 · DOI

• Developing models aimed at preventing poor outcomes in critically ill patients is highly complex and requires individ- ualized assessments to better capture each patient’s evolving condition. Although promising advances have been made, the routine integration of personalized longitudinal pre- diction models into hospital information systems remains limited. The development and implementation of medi- cal services based on longitudinal prediction modelling, 1 3EPMA Journal (2026) 17:427–455 supported by artificial intelligence and machine learning, will therefore represent an important step toward PPPM. However, to fully realize the benefits of these approaches and overcome the barriers to implementing longitudinal prediction models in clinical practice (many of which iden- tified in this review), well-designed studies and standard- ized methodological frameworks are needed in the overall research and medical community. Greater consistency in data collection, modelling strategies, and reporting practices would facilitate comparability across studies and improve the reliability of findings resulting from different patient cohorts. Similar efforts toward methodological standard- ization have already been warranted across other research areas [97–99], in order to support the successful integration of different PPPM strategies into routine clinical practice. In parallel, strengthening the training of healthcare profes- sionals will be essential to ensure the effective and accurate incorporation of these technologies.

  Blood biomarker trajectories in ICU-directed prediction models – A scoping review · 2026 · DOI

• This study has several limitations. Formal quantitative pooling of AUC and ΔAUC values was not feasible due to the lack of variance or uncertainty measures required for robust meta-analysis. In addition, included studies variably reported internally validated and externally validated AUC estimates, introducing methodological heterogeneity that could potentially bias pooled discrimination estimates. Consequently, the comparative assessment of discriminative performance was primarily based on descriptive synthesis. Substantial methodological and statistical heterogeneity was observed across studies included for quantitative analysis, both for specificity and sensitivity estimates. Differences in study design, ICU populations, AI model architectures, validation strategies, outcome definitions, and reporting practices may have influenced pooled estimates and limited generalizability. In addition, several studies did not directly report complete 2×2 contingency data, requiring reconstruction of TP, FP, FN, and TN values from reported performance metrics, which may have introduced estimation error. Calibration performance also could not be synthesized quantitatively because of inconsistent reporting across studies. The PROBAST assessment identified recurrent concerns within the analysis domain, particularly related to incomplete reporting of model development methodologies and external validation practices, which may affect reproducibility and generalizability of reported models. Finally, publication and selective reporting bias cannot be excluded, as studies reporting favorable AI model performance may have been more likely to be published.

  Accuracy of Artificial Intelligence-Based Models Versus Conventional Scoring Systems (APACHE, SOFA, and SAPS) in Predicting Mortality Among ICU Patients: A Systematic Review and Meta-Analysis · 2026 · DOI

• The large, nationwide multicentre cohort of ICU-admitted patients with sepsis gives our study high external validity and generalisability. The well-validated health registries in Sweden and the matched study design further strengthen the internal validity. The study was based on a high-resolution data set linked to national health Scientific Reports | (2026) 16:15705 | https://doi.org/10.1038/s41598-026-53619-9 8 registers including morbidity coding that is considered to have high validity42. Demographic characteristics are much in line with previous studies of ICU patients with sepsis. The proportion of missing data was low (< 2%) and found for education and income variables only. This is less than the 5%, which has been suggested as the maximum upper acceptance limit for large datasets43. All citizens with a Swedish identity number are included in the Cause of Death Register, regardless of whether the death occurs in Sweden or abroad. Thus, the loss to follow-up is considered to be low (< 2%) regarding time to death and cause of death. Limitations include the register-based design. Only patients admitted to ICU were included in the study population, and no comparisons were made with hospitalized septic patients not admitted to ICU. We chose to only have patients admitted from the emergency department to minimise the heterogeneity of the sepsis population, this could also limit generalisability. Potential misclassification cannot be ruled out in registry-based studies. The dataset includes all care episodes from hospitals and outpatient care (not classified as primary care) and conditions significant enough to influence management during hospitalisation are likely to be captured. Milder comorbidities managed exclusively in primary care could be missed. Furthermore, due to limitations in the registry data, we were unable to reliably determine the site of infection and to differentiate between sepsis and septic shock. The Charlson Comorbidity Index was chosen as it is the most widely used comorbidity measure in sepsis research, facilitating comparison with previous studies. Although well-validated and adapted to ICD-10, some of its original weightings, particularly for conditions such as cancer and AIDS, may not fully reflect current prognostic relevance in ICU populations due to advances in treatment. Psychiatric illness and substance abuse were added as separate covariates, as these clinically relevant conditions are not captured by the CCI.

  The impact of community-acquired critical sepsis on long-term mortality and morbidity—a nationwide cohort study · 2026 · DOI

• For decades, no relevant progress has been made in PM treatment, and many drugs investigated alone or in combination regimen have failed to demonstrate efficacy.

  Rethinking the role of procalcitonin in suspected sepsis · 2026 · DOI

Most-cited papers in Sepsis Diagnosis and Treatment

• Global, regional, and national sepsis incidence and mortality, 1990–2017: analysis for the Global Burden of Disease Study · The Lancet · 2020 · 6,763 citations
• Enhancing Recovery From Sepsis · JAMA · 2018 · 918 citations
• Understanding the Inflammatory Cytokine Response in Pneumonia and Sepsis · Archives of Internal Medicine · 2007 · 656 citations
• Rapid Response Teams · Archives of Internal Medicine · 2010 · 653 citations
• International Consensus Criteria for Pediatric Sepsis and Septic Shock · JAMA · 2024 · 471 citations
• The pathophysiology of sepsis and precision-medicine-based immunotherapy · Nature Immunology · 2024 · 315 citations
• Development and Validation of the Phoenix Criteria for Pediatric Sepsis and Septic Shock · JAMA · 2024 · 200 citations
• Association between the stress hyperglycemia ratio and 28-day all-cause mortality in critically ill patients with sepsis: a retrospective cohort study and predictive model establishment based on machine learning · Cardiovascular Diabetology · 2024 · 200 citations
• Methods of determining optimal cut-point of diagnostic biomarkers with application of clinical data in ROC analysis: an update review · BMC Medical Research Methodology · 2024 · 165 citations
• New insights into the role of mitochondrial metabolic dysregulation and immune infiltration in septic cardiomyopathy by integrated bioinformatics analysis and experimental validation · Cellular & Molecular Biology Letters · 2024 · 142 citations

Most recent work

• Association Between Sex and Clinical Outcomes for Critically Ill Patients in India: A Registry-Embedded Cohort Study · Critical Care Medicine · 2026
• Breaking the Stalemate: Advancing Sepsis Therapeutics Beyond Supportive Care · Public Health Reports · 2026
• The bug, the burden, and the biology: beyond host-centric phenotyping in sepsis · Journal of Clinical Investigation · 2026
• Multimodal artificial intelligence for precision critical care: A scoping review · Health Data Science · 2026
• Artificial intelligence in acute and critical care: current challenges and strategic solutions · Frontiers in Public Health · 2026
• Enhancing large language model clinical support information with machine learning risk and explainability: a feasibility study · Intensive Care Medicine Experimental · 2026
• Sepsis in multimorbidity: a domain-based framework for systemic vulnerability and precision care · Frontiers in Medicine · 2026
• End-of-surgery prediction of postoperative infectious complications from intraoperative vital-sign dynamics · npj Digital Medicine · 2026
• Principais e emergentes preditores de mortalidade em pacientes sépticos em Unidades de Terapia Intensiva: síntese crítica das evidências nos últimos cinco anos · Revista JRG de Estudos Acadêmicos · 2026
• ICU command centers and evidence-based artificial intelligence – Letter on Munoz et al. · Intensive and Critical Care Nursing · 2026

Related topics in Medicine