Development and Validation of Deep-Learning-Based Sepsis and Septic Shock Early Prediction System (DeepSEPS) Using Real-World ICU Data

BACKGROUND

Successful sepsis treatment depends on early diagnosis. We aimed to develop and validate a system to predict sepsis and septic shock in real time using deep learning.

METHODS

Clinical data were retrospectively collected from electronic medical records (EMRs). Data from 2010 to 2019 were used as development data, and data from 2020 to 2021 were used as validation data. The collected EMRs consisted of eight vital signs, 13 laboratory data points, and three demographic information items. We validated the deep-learning-based sepsis and septic shock early prediction system (DeepSEPS) using the validation datasets and compared our system with other traditional early warning scoring systems, such as the national early warning score, sequential organ failure assessment (SOFA), and quick sequential organ failure assessment.

RESULTS

DeepSEPS achieved even higher area under receiver operating characteristic curve (AUROC) values (0.7888 and 0.8494 for sepsis and septic shock, respectively) than SOFA. The prediction performance of traditional scoring systems was enhanced because the early prediction time point was close to the onset time of sepsis; however, the DeepSEPS scoring system consistently outperformed all conventional scoring systems at all time points. Furthermore, at the time of onset of sepsis and septic shock, DeepSEPS showed the highest AUROC (0.9346).

CONCLUSIONS

The sepsis and septic shock early warning system developed in this study exhibited a performance that is worth considering when predicting sepsis and septic shock compared to other traditional early warning scoring systems. DeepSEPS showed better performance than existing sepsis prediction programs. This novel real-time system that simultaneously predicts sepsis and septic shock requires further validation.

M1 Cholinergic Signaling Modulates Cytokine Levels and Splenocyte Sub-Phenotypes Following Cecal Ligation and Puncture

Background

The contribution of the central nervous system to sepsis pathobiology is incompletely understood. In previous studies, administration of endotoxin to mice decreased activity of the vagus anti-inflammatory reflex. Treatment with the centrally-acting M1/M4 muscarinic acetylcholine (ACh) receptor (M1/M4AChR) attenuated this endotoxin-mediated change. We hypothesize that decreased M1/M4AChR-mediated activity contributes to inflammation following cecal ligation and puncture (CLP), a mouse model of sepsis.

Methods

Basal forebrain cholinergic activity (immunostaining), serum cytokine/chemokine levels (ELISA) and splenocyte subtypes (flow cytometry) were examined at baseline and following CLP in male C57BL/6 male mice.

Rersults

At 48hrs. post-CLP, activity in basal forebrain cells expressing choline acetyltransferase (ChAT) was half of that observed at baseline. Lower activity was also noted in the hippocampus, which contains projections from ChAT-expressing basal forebrain neurons. Serum levels of TNFα, IL-1β, MIP-1α, IL-6, KC and G-CSF were higher post-CLP than at baseline. Post-CLP numbers of splenic macrophages and inflammatory monocytes, TNFa+ and ILb+ neutrophils and ILb+ monocytes were higher than baseline while numbers of central Dendritic Cells (cDCs), CD4+ and CD8+ T cells were lower. When, following CLP, mice were treated with xanomeline, a central-acting M1AChR agonist, activity in basal forebrain ChAT-expressing neurons and in the hippocampus was significantly higher than in untreated animals. Post-CLP serum concentrations of TNFα, IL-1β, and MIP-1α, but not of IL-6, KC and G-CSF, were significantly lower in xanomline-treated mice than in untreated mice. Post-CLP numbers of splenic neutrophils, macrophages, inflammatory monocytes and TNFα+ neutrophils also were lower in xanomeline-treated mice than in untreated animals. The effects of CLP on percentages of IL-1β+ neutrophils, IL-1β+ monocytes, cDCs, CD4+ T cells and CD8+ T cells were similar in xanomeline - treated and untreated post-CLP mice.

Conclusion

Our findings indicate that M1/M4AChR-mediated responses modulate CLP-induced alterations in the distribution of some, but not all, leukocyte phenotypes and certain cytokines and chemokines.

Presepsin: gelsolin ratio, as a promising marker of sepsis-related organ dysfunction: a prospective observational study

Introduction

We aimed to facilitate the diagnosis and prognosis of sepsis-related organ dysfunction through analyzing presepsin (PSEP) and gelsolin (GSN) levels along with a novel marker, the presepsin:gelsolin (PSEP:GSN) ratio.

Methods

Blood samples were collected from septic patients at the intensive care unit (ICU) at three time points (T1-3): T1: within 12 h after admission; T2: second day morning; T3: third day morning. Sampling points for non-septic ICU patients were T1 and T3. PSEP was measured by a chemiluminescence-based POCT method while GSN was determined by an automated immune turbidimetric assay. Data were compared with routine lab and clinical parameters. Patients were categorized by the Sepsis-3 definitions. PSEP:GSN ratio was evaluated in major sepsis-related organ dysfunctions including hemodynamic instability, respiratory insufficiency and acute kidney injury (AKI).

Results

In our single center prospective observational study, 126 patients were enrolled (23 control, 38 non-septic and 65 septic patients). In contrast to controls, significantly elevated (p < 0.001) admission PSEP:GSN ratios were found in non-septic and septic patients. Regarding 10-day mortality prediction, PSEP:GSN ratios were lower (p < 0.05) in survivors than in non-survivors during follow-up, while the prognostic performance of PSEP:GSN ratio was similar to widely used clinical scores (APACHE II, SAPS II, SOFA). PSEP:GSN ratios were also higher (p < 0.001) in patients with sepsis-related AKI than septic non-AKI patients during follow-up, especially in sepsis-related AKI patients needing renal replacement therapy. Furthermore, increasing PSEP:GSN ratios were in good agreement (p < 0.001) with the dosage and the duration of vasopressor requirement in septic patients. Moreover, PSEP:GSN ratios were markedly greater (p < 0.001) in patients with septic shock than in septic patients without shock. Compared to septic patients requiring oxygen supplementation, substantially elevated (p < 0.001) PSEP:GSN ratios were observed in septic patients with demand for mechanical ventilation, while higher PSEP:GSN ratios (p < 0.001) were also associated with extended periods of mechanical ventilation requirement in septic patients.

Conclusion

PSEP:GSN ratio could be a useful complementary marker besides the routinely used SOFA score regarding the diagnosis and short term mortality prediction of sepsis. Furthermore, the significant increase of this biomarker may also indicate the need for prolonged vasopressor or mechanical ventilation requirement of septic patients. PSEP:GSN ratio could yield valuable information regarding the extent of inflammation and the simultaneous depletion of the patient's scavenger capacity during sepsis.

Clinical trail registration

NIH U.S. National Library of Medicine, ClinicalTrails.gov. Trial identifier: NCT05060679, (https://clinicaltrials.gov/ct2/show/NCT05060679) 23.03.2022, Retrospectively registered.

Effect of Serum Phosphate on the Prognosis of Septic Patients: A Retrospective Study Based on MIMIC-IV Database

Objective

To assess the effect of serum inorganic phosphate (Pi) on the prognosis of patients with sepsis.

Methods

A retrospective analysis of patients with sepsis selected from the Medical Information Mart for Intensive Care (MIMIC)-IV database was performed. Sepsis was diagnosed according to the Third International Consensus Definition for sepsis and septic shock (Sepsis-3). The time-weighted values of the serum Pi measurements within the first 24 h of sepsis were analyzed. The association between serum Pi and in-hospital mortality was evaluated with a generalized linear model (log-binomial model).

Results

The analysis of 11,658 patients from six intensive care units (ICUs) showed a nearly linear correlation between serum Pi and in-hospital mortality in all patients with sepsis, especially in those with acute kidney injury (AKI). The increase of serum Pi was related to a higher risk of AKI, higher norepinephrine doses, ICU mortality, and in-hospital mortality. The generalized linear model showed that serum Pi was an independent predictor for in-hospital mortality in all patients with sepsis even within the normal range. The adjusted risk ratios (RRs) were also significant in subgroup analyses according to kidney function, gender, respiratory infection, vasopressor use, and Sequential Organ Failure Assessment (SOFA) score.

Conclusion

Higher levels of serum Pi, even within the normal range, were significantly associated with a higher risk of in-hospital mortality in patients with sepsis regardless of kidney function, gender, respiratory infection, vasopressor use, and SOFA score.

Thrombocytopenia and platelet count recovery in patients with sepsis-3: a retrospective observational study

Thrombocytopenia is common in critical illness. But there are no studies that focus on thrombocytopenia and platelet recovery in Sepsis-3 patients. We employed a large database to identify sepsis based on Sepsis-3 criteria. Patients were grouped by nadir platelet count during ICU, propensity score matching was used to eliminate covariates imbalance, multivariable cox proportional hazard model was used for evaluating mortality. A total of 9709 patients were enrolled based on Sepsis-3, 1794 (18%) patients developed thrombocytopenia, with 858 (8.8%) exhibiting thrombocytopenia at ICU admission (prevalent), 891 (9.2%) developed thrombocytopenia during ICU stay (incident). In the incident thrombocytopenia group, survivors exhibited higher nadir platelet count, higher rate in platelet count recovery and shorter time to platelet recovery compared to non-survivors. Platelet recovery was not observed until 1 days (IQR, 1-2) after weaning of mechanical ventilation and 1 days (IQR, 1-3) after discontinuation of vasopressor in survivors of incident thrombocytopenia. Furthermore, thrombocytopenia was associated with longer duration of ICU length of stay, longer duration of mechanical ventilation and vasopressor use compared to no thrombocytopenia. Moderate (20-50 × 10⁹/L) and severe (<20 × 10⁹/L) thrombocytopenia group showed increased 28 days mortality compared to no thrombocytopenia, while the mortality rate between mild (51-100 × 10⁹/L) and no thrombocytopenia group (≥100 × 10⁹/L) showed no significant difference. Taken together these data revealed that thrombocytopenia occurred in 18% Sepsis-3 patients; platelet recovery occurred more frequent and earlier in survivors; platelet recovery was not observed until clinical improvement. Thrombocytopenia in Sepsis-3 demonstrated increased disease severity, and patients with platelet count <50 × 10⁹/L showed increased 28 days mortality.

Temporal Differential Expression of Physiomarkers Predicts Sepsis in Critically Ill Adults

BACKGROUND

Sepsis is a life-threatening condition with high mortality rates. Early detection and treatment are critical to improving outcomes. Our primary objective was to develop artificial intelligence capable of predicting sepsis earlier using a minimal set of streaming physiological data in real time.

METHODS AND FINDINGS

A total of 29,552 adult patients were admitted to the intensive care unit across five regional hospitals in Memphis, Tenn, over 18 months from January 2017 to July 2018. From these, 5,958 patients were selected after filtering for continuous (minute-by-minute) physiological data availability. A total of 617 (10.4%) patients were identified as sepsis cases, using the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3) criteria. Physiomarkers, a set of signal processing features, were derived from five physiological data streams including heart rate, respiratory rate, and blood pressure (systolic, diastolic, and mean), captured every minute from the bedside monitors. A support vector machine classifier was used for classification. The model accurately predicted sepsis up to a mean and 95% confidence interval of 17.4 ± 0.22 h before sepsis onset, with an average test accuracy of 83.0% (average sensitivity, specificity, and area under the receiver operating characteristics curve of 0.757, 0.902, and 0.781, respectively).

CONCLUSIONS

This study demonstrates that salient physiomarkers derived from continuous bedside monitoring are temporally and differentially expressed in septic patients. Using this information, minimalistic artificial intelligence models can be developed to predict sepsis earlier in critically ill patients.

Monocyte distribution width (MDW) parameter as a sepsis indicator in intensive care units

OBJECTIVES

Patients in Intensive Care Units (ICU) are a high-risk population for sepsis, recognized as a major cause of admission and death. The aim of the current study was to evaluate the diagnostic accuracy and prognostication of monocyte distribution width (MDW) in sepsis for patients admitted to ICU.

METHODS

Between January and June 2020, we conducted a prospective observational study during the hospitalization of 506 adult patients admitted to the ICU. MDW was evaluated in 2,367 consecutive samples received for routine complete blood counts (CBC) performed once a day and every day during the study. Sepsis was diagnosed according to Sepsis-3 criteria and patients enrolled were classified in the following groups: no sepsis, sepsis and septic shock.

RESULTS

MDW values were significantly higher in patients with sepsis or septic shock in comparison to those within the no sepsis group [median 26.23 (IQR: 23.48-29.83); 28.97 (IQR: 21.27-37.21); 21.99 (IQR: 19.86-24.36) respectively]. ROC analysis demonstrated that AUC is 0.785 with a sensitivity of 66.88% and specificity of 77.79% at a cut-off point of 24.63. In patients that developed an ICU-acquired sepsis MDW showed an increase from 21.33 [median (IQR: 19.47-21.72)] to 29.19 [median (IQR: 27.46-31.47)]. MDW increase is not affected by the aetiology of sepsis, even in patients with COVID-19. In sepsis survivors a decrease of MDW values were found from the first time to the end of their stay [median from 29.14 (IQR: 26.22-32.52) to 25.67 (IQR: 22.93-30.28)].

CONCLUSIONS

In ICU, MDW enhances the sepsis detection and is related to disease severity.

Predictive value of serum gelsolin and Gc globulin in sepsis - a pilot study

BACKGROUND

Simultaneous determination of the two main actin scavenger proteins in sepsis has not been investigated until now. In our pilot study, we elucidated the predictive values of Gc globulin and gelsolin (GSN) in sepsis by comparing them to classic laboratory and clinical parameters.

METHODS

A 5-day follow-up was performed, including 46 septic patients, 28 non-septic patients and 35 outpatients as controls. Serum Gc globulin and GSN levels were determined by automated immune turbidimetric assay on a Cobas 8000/c502 analyzer. Patients were retrospectively categorized according to the sepsis-3 definitions, and 14-day mortality was also investigated.

RESULTS

First-day GSN also differentiated sepsis from non-sepsis (AUC: 0.88) similarly to C-reactive protein (AUC: 0.80) but was slightly inferior to procalcitonin (PCT) (AUC: 0.98) with a cutoff value of GSN at 22.29 mg/L (sensitivity: 83.3%; specificity: 86.2%). Only first-day SOFA scores (0.88) and GSN (0.71) distinguished septic survivors from non-survivors, whereas lactate (0.99), Gc globulin (0.76) and mean arterial pressure (MAP) (0.74) discriminated septic shock from sepsis. Logistic regression analyses revealed SOFA scores and GSN being significant factors regarding 14-day mortality. First-day GSN levels were higher (p<0.05) in septic survivors than in non-survivors. Gc globulin levels remained higher (p<0.01) in sepsis when compared with septic shock during the follow-up period.

CONCLUSIONS

Both serum GSN and Gc globulin may have predictive values in sepsis. Considering the small sample size of our study, further measurements are needed to evaluate our results. Measurement of Gc globulin and GSN maybe useful in assessment of sepsis severity and in therapeutic decision-making.

Accuracy of qSOFA for the diagnosis of sepsis-3: a secondary analysis of a population-based cohort study

Background

We aimed to evaluate the accuracy of quick Sequential (sepsis-related) Organ Failure Assessment (qSOFA) for the diagnosis of sepsis-3, and to analyze the prognosis of infected patients in wards over-diagnosed with qSOFA but missed by sepsis-3, and those missed by qSOFA but in accordance with sepsis-3 criteria. We also intended to validate the performance of qSOFA as one predictor of outcome in patients with suspicion of infection.

Methods

We reviewed the medical records of 1,716 adult patients with infection who were hospitalized from July 1^st, 2012 to June 30^th, 2014 in the Yuetan subdistrict of Beijing, China. Based on the sepsis-3 criteria and qSOFA score proposed by the Third International Consensus Definitions for Sepsis and Septic Shock, these patients were categorized into four groups: qSOFA(-)sepsis(-), qSOFA(+)sepsis(-), qSOFA(-)sepsis(+), and qSOFA(+)sepsis(+). Multivariate logistic regression analysis was used to determine the independent risk factors for in-hospital mortality. The area under the receiver operating characteristic curves (AUROCs) of the qSOFA(+) group were compared with the sepsis(+) group for in-hospital mortality, ICU admission, and invasive ventilation.

Results

Among the 1,716 patients with infection, there were 935 patients (54.5%) with sepsis, and 640 patients (37.3%) with qSOFA ≥2. There were 610 patients in the qSOFA(-)sepsis(-) group, 171 in the qSOFA(+)sepsis(-) group, 466 in the qSOFA(-)sepsis(+) group, and 469 in the qSOFA(+)sepsis(+) group. In the logistic regression analysis, increasing age, bedridden status, and malignancy were all independent risk factors of hospital mortality. Sepsis and qSOFA ≥2 were also independent risk factors of hospital mortality, with an adjusted OR of 3.85 (95% CI: 2.70-5.50) and 13.92 (95% CI: 9.87-16.93) respectively. qSOFA had a sensitivity of 50.2% and a specificity of 78.1% for sepsis-3. The false-positive [qSOFA(+)sepsis(-)] group had 38 patients (22.2%) die during hospitalization, and an adjusted OR of 9.20 (95% CI: 4.86-17.38). In addition, the false-negative [qSOFA(-)sepsis(+)] group had a hospital mortality rate of 7.3% (34/466) and an adjusted OR of 2.59 (95% CI: 1.39-4.83). In comparison, patients meeting neither qSOFA nor sepsis criteria had the lowest hospital mortality [2.6% (16/610)], whereas patients with both qSOFA ≥2 and sepsis had the highest hospital mortality [56.5% (265/469)], with an adjusted OR of 42.02 (95% CI: 24.31-72.64). The discrimination of in-hospital mortality using qSOFA (AUROC, 0.846; 95% CI, 0.824-0.868) was greater compared with sepsis-3 criteria (AUROC, 0.834; 95% CI, 0.805-0.863; P<0.001).

Conclusions

In our analysis, the sensitivity(Se) of qSOFA for the diagnosis of sepsis was lower, and qSOFA score ≥2 might identify a group of patients at a higher risk of mortality, regardless of being septic or not.

Sepsis: evolving concepts and challenges

Sepsis remains a major cause of morbidity and mortality worldwide, with increased burden in low- and middle-resource settings. The role of the inflammatory response in the pathogenesis of the syndrome has supported the modern concept of sepsis. Nevertheless, a definition of sepsis and the criteria for its recognition is a continuous process, which reflects the growing knowledge of its mechanisms and the success and failure of diagnostic and therapeutic interventions. Here we review the evolving concepts of sepsis, from the "systemic inflammatory response syndrome triggered by infection" (Sepsis-1) to "a severe, potentially fatal, organic dysfunction caused by an inadequate or dysregulated host response to infection" (Sepsis-3). We focused in the pathophysiology behind the concept and the criteria for recognition and diagnosis of sepsis. A major challenge in evaluating the host response in sepsis is to characterize what is protective and what is harmful, and we discuss that, at least in part, the apparent dysregulated host response may be an effort to adapt to a hostile environment. The new criteria for recognition and diagnosis of sepsis were derived from robust databases, restricted, however, to developed countries. Since then, the criteria have been supported in different clinical settings and in different economic and epidemiological contexts, but still raise discussion regarding their use for the identification versus the prognostication of the septic patient. Clinicians should not be restricted to definition criteria when evaluating patients with infection and should wisely use the broad array of information obtained by rigorous clinical observation.

The macrophage activation marker sMR as a diagnostic and prognostic marker in patients with acute infectious disease with or without sepsis

Sepsis is a leading cause of mortality. This study aims to assess the utility of the soluble mannose receptor (sMR) as a biomarker of sepsis and mortality in patients hospitalized with suspected infection. Using an in-house ELISA assay the concentration of sMR was analyzed in the serum of patients from three prospective studies. Using Sepsis-3 guidelines, patients were stratified as no infection (NI, n = 68), verified infection without sepsis (NSEP, n = 133) and verified infection with sepsis (SEP, n = 190). Adverse outcome was assessed as death before 28 days. We show that the sensitivity of sMR to predict mortality [area under curve (AUC) = 0.77] exceeded the sensitivity of procalcitonin (PCT, AUC = 0.63), C-reactive protein (CRP, AUC = 0.61) and the macrophage soluble receptor, CD163 (sCD163, AUC = 0.74), while it was less accurate to predict diagnosis of sepsis [AUC(sMR) = 0.69 vs. AUC(PCT) = 0.79, AUC(CRP) = 0.71 and AUC(sCD163) = 0.66]. Median sMR was significantly higher in the group with SEP (0.55 mg/L), compared with the groups without sepsis (NI and NSEP) (0.39 mg/L, p < .0001), and among those who died compared to those who survived (0.89 mg/L vs. 0.44 mg/L, p < .0001). Our results, and the current literature, support further evaluation of sMR as a biomarker of sepsis and mortality among patients hospitalized with suspected infection.

Much Ado About the New Definitions of Sepsis

Following the publication of the new definition of sepsis (Sepsis-3), a plethora of articles have been published in medical journals. Recognizing the epidemiological importance of the previous definitions, first issued in 1992 (Sepsis-1), and subsequently revised in 2001 (Sepsis-2), the most recent opinion emphasizes the failure "to provide adequate groups of patients with homogenous aetiologies, presentations and outcomes", and blamed one of the causes "for the failure of several randomized controlled trials (RCTs), that tested the efficacy of adjuvant sepsis therapies". This review summarizes the recent advances in sepsis definition.