Clinical subphenotypes of sepsis based on mixed data and differences in treatment effects: a cluster analysis of multicentre observational studies

Clinical subphenotypes of sepsis based on mixed data and differences in treatment effects: a cluster analysis of multicentre observational studies | 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 Research Article Clinical subphenotypes of sepsis based on mixed data and differences in treatment effects: a cluster analysis of multicentre observational studies Yuta Yokokawa, Rieko Sakurai, Daisuke Kudo, Gen Tamiya, Shigeki Kushimoto This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7716199/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 15 Jan, 2026 Read the published version in Critical Care → Version 1 posted 9 You are reading this latest preprint version Abstract Background Sepsis is a heterogeneous syndrome, and the treatment effects may vary depending on its subphenotype. Previous studies have not fully used mixed clinical data, nor have investigated the effects of these multiple treatments across subphenotypes. In this study, we aimed to classify patients with sepsis into subphenotypes based on mixed clinical data and examined the differences in treatment effectiveness by subphenotype. Methods This study was a secondary analysis of multicenter registries that enrolled patients with sepsis admitted to intensive care units in Japan. The patients aged 16 years or older admitted to the ICU due to a diagnosis of sepsis were included and fifty-two variables at admission were used in the cluster analysis. We applied the state-of-the-art clustering method named k-UMAP, which uses uniform manifold approximation and projection for dimensionality reduction, followed by clustering using k-means and the previous clustering methods k-prototype and KAMILA. To examine differences in the effectiveness of the six treatments by subphenotype, a logistic regression model was used for propensity score-based weighted data to calculate the odds ratios for the interaction of the treatments and subphenotype in each clustering method. The primary outcome was the in-hospital mortality rate. Results The analysis included 1,756 patients. The results of three clustering methods using mixed clinical data led to the three conditions with high robustness being selected: k-UMAP [number of clusters (k) = 3], k-UMAP [k = 5], and KAMILA [k = 3]. Hospital mortality, patient characteristics, and treatment effectiveness varied by subphenotype. Recombinant thrombomodulin was significantly effective for subphenotype 1 in the k-UMAP [k = 3] and in the k-UMAP [k = 5]. Antithrombin III was significantly effective for subphenotype 2 in the k-UMAP [k = 5]. On the other hand, in KAMILA [k = 3], no significant treatment differences were observed between subphenotypes. Conclusions Using state-of-the-art clustering methods for mixed data were identified three to five subphenotypes which were associated with clinical information and outcomes. The effects of treatments differed for each subphenotype, and k-UMAP may reveal appropriate treatment targets that have not been proven effective. Sepsis subphenotype mixed data cluster analysis treatment effects Figures Figure 1 Figure 2 Figure 3 Introduction Sepsis affects millions of people worldwide, and its mortality rate remains high [ 1 ]. The Guidelines for Management of Sepsis and Septic Shock [ 2 , 3 ] state that the initial fluid resuscitation [ 4 ], early administration of appropriate antimicrobials [ 5 , 6 ], and administration of vasopressors for septic shock [ 7 ] are the bundle to be administered within 1 h of recognition. However, no specific treatment has been shown to improve clinical outcomes in patients with sepsis. Several therapies have been investigated in clinical trials; however, none have adequately demonstrated effectiveness [ 8 – 14 ]. One reason treatment effectiveness has not been demonstrated is the heterogeneity of patients with sepsis [ 15 – 18 ]. Patients with sepsis have various clinical conditions; however, past studies have evaluated "sepsis" as a single population. Thus, effective treatment for some subphenotypes of sepsis may not improve outcomes for all patients with sepsis. Effective treatments are assumed to exist for appropriately classified subphenotypes [ 19 ]. Based on these hypotheses, several studies have been conducted to classify patients with sepsis into subphenotypes by applying unsupervised approaches such as cluster analysis to clinical and biomarker data [ 20 – 24 ]. However, these studies did not have enough impact to change clinical practice. One of the major challenges in the cluster analysis of clinical data is the use of clinical data, including categorical variables (e.g., sex, comorbidities, and medications) and continuous variables (e.g., vital signs and blood test data). The dataset contains numerical and categorical data, called “heterogeneous” or “mixed” data. Mixed data are challenging to use in cluster analysis because it is difficult to directly apply mathematical operations to both types of variables [ 25 ], which is a barrier to fully using clinical data. Additionally, simply dividing patients with sepsis into subphenotypes has little clinical value. The subphenotype classification should reflect the pathophysiological status and impact on clinical practice by recognizing differential treatment responses [ 15 ]. Patients with sepsis are treated with multiple adjunct therapies depending on their medical condition; however, no study has investigated the effects of these multiple treatments across subphenotypes. Identifying subphenotypes with different treatment effects could lead to developing precision medicine for sepsis. Therefore, this study aimed to (1) apply a state-of-the-art clustering method for mixed data, (2) classify patients with sepsis into subphenotypes using clinical information, and (3) examine differences in multiple treatment effects by subphenotype. Materials and methods Study population and setting In our study, we used a dataset from multicenter registries that involved patients (age ≥ 16 years) admitted to intensive care units (ICUs) with severe sepsis or septic shock according to the International Sepsis Definitions Conference criteria [ 26 , 27 ]. We used two registries for the derivation. The Tohoku Sepsis Registry (UMIN000010297) was a prospective observational study involving 616 patients admitted to 10 institutions (three university hospitals and seven community hospitals) in the Tohoku District, Northeastern Japan, between January 2015 and December 2015 [ 28 ]. The Focused Outcomes Research in Emergency Care for Acute Respiratory Distress Syndrome, Sepsis, and Trauma (FORECAST) sepsis study (UMIN000019742) is a multicenter, prospective study that enrolled 1,184 patients admitted to 59 ICUs in Japan between January 2016 and March 2017 [ 29 ]. We also used two other multicenter registries for validation: the Japanese Association for Acute Medicine Sepsis Prognostication in the ICU and Emergency Room (JAAM SPICE-ICU) [ 30 ] and the Japanese Association for Acute Medicine Multicenter Assessment for Sepsis Treatment and Outcome (JAAM MAESTRO) [ 31 ] (Supplemental document). A detailed description of the methods and analytical processes used in these studies is provided in Additional file 1. Each institution’s Institutional Review Board approved the studies, and the need for informed consent was waived. Clinical variables and outcomes We used 52 variables from patient information at ICU admission for cluster analysis based on their availability across registries and their clinical importance: demographic variables (including age, sex, body-mass index, comorbidities, medication, and admission route), vital signs (including respiratory rate, heart rate, systolic and diastolic blood pressure, and body temperature), Glasgow Coma Scale score, parameters of inflammation (including white blood cell count and C-reactive protein), index parameters of organ failure (including bilirubin, creatinine, partial pressure of arterial oxygen, and partial pressure of arterial carbon dioxide), markers of coagulation (including platelet counts, international normalized ratio, activated partial thromboplastin time, and fibrinogen), sub-score of sepsis-related organ failure assessment (SOFA) score [ 32 ], and serum levels of glucose, sodium, potassium, hematocrit, lactate, and potential hydrogen. A detailed description of variables is provided in Additional file 1: Tabel S1. The primary outcome was in-hospital mortality. Interventions We investigated the effects of six treatments that were measured in the registries and which divided opinions on whether to do so according to the guidelines and previous studies [ 8 – 10 , 13 , 14 ]. These treatments include corticosteroids, polymyxin B-immobilized fiber column direct hemoperfusion, recombinant thrombomodulin (rTM), antithrombin III (AT III), immunoglobulin G, and vasopressin. Statistical methods To explore the subphenotypes of patients with sepsis, cluster analysis was performed using the variables of patient information at ICU admission. Among these variables, we eliminated one with a missing rate above 0.5 and one correlated with others with coefficients above 0.8. Serum levels of procalcitonin, fibrinin/fibrinogen degradation products, and antithrombin III were excluded from the analysis. Multiple imputation with chained equation was used to account for missing data with the mice package (Additional file 1: Table S3). Log transformation and normalization were used for nonnormal data. Clustering methods We applied different methods for subphenotype identification. We considered the k-prototype [ 33 ] and KAymeans for MIxed LArge data (KAMILA) [ 34 ] which were recommended in a previous study that compared the performance of several clustering methods for mixed data through simulation and real data analysis [ 35 ]. We also considered another approach, k-means [ 36 ] on the Uniform Manifold Approximation and Projection for Dimension Reduction (UMAP) [ 37 ]: k-UMAP, which is analogous to principal component analysis (PCA) [ 38 ]. PCA is used for dimension reduction to reduce noise but cannot be applied to mixed data. In contrast, UMAP is also used for dimension reduction and can select appropriate metrics depending on the data. Therefore, we adapted UMAP to handle mixed data as numeric data by transforming the embeddings. We examined the performance of k-UMAP using the same simulation as in a previous study [ 35 ] and confirmed its superiority for clustering datasets with many categorical variables (Additional file 1: Figure S1 ). (Refer to Additional file 1: Clustering methods and Table S2 for further details on each clustering method). To compare the results of each clustering method, we use the adjusted rand indexes (ARIs) [ 39 ], which measure the similarity between the clusters obtained from two different clustering methods. Consensus clustering We evaluated the performance of the clustering methods using consensus clustering [ 40 ] to determine the most stable method. In consensus clustering, stability or consensus is measured by counting and normalizing the number of times each sample pair is classified into the same cluster after each iteration. The consensus values can be interpreted as the degree of stability; if all consensus values are 1 or 0, clustering is perfectly stable. The consensus for each cluster was weighted and averaged the consensus for each pair and called the cluster consensus. The weights represent the ratios of the number of samples in a cluster. We adopted clusters based on a combination of consensus clustering outputs, a clear separation of the consensus matrix heat maps, characteristics of the consensus cumulative distribution function plots, and adequate pairwise consensus values between cluster members. These values were used to determine the number of clusters. Cluster validation The same clustering method was applied to the validation dataset. For cluster analysis, we used 44 variables from patient information at ICU admission that were available in the validation dataset and common to the derivation datasets, as well as their clinical importance. Evaluation of treatment effects in sepsis subphenotypes To align the cohort studies in our dataset with the randomized clinical trial situation, which is the gold standard for estimating causal relationships and treatment effects, we estimated the propensity score (PS) [ 41 , 42 ]. We corrected the imbalance of background factors between the treated and untreated groups using inverse probability weighting of the PS [ 41 ]. Most patients in our dataset received multiple treatments, making it difficult to determine the effects of any single treatment. Thus, we considered the combined treatments as confounding factors and included them in estimating PS weights, as well as the variables used in clustering. Logistic regression was used to estimate PS weights. After adjusting for PS, we used a mixed model to estimate the treatment effects by subphenotype. Among the datasets, we conducted a sub-analysis of patients with septic shock, defined as those who received vasopressors. We assessed the differences in characteristics between clusters using analysis of variance for continuous variables and chi-square tests for categorical variables. Statistical analyses were performed using R software (version 4.1.1; R Foundation, Vienna, Austria) (Additional file 1: Table S3). All reported p-values were two-sided and considered significant if < 0.05. Results Patients in the dataset From the two multicenter registries, we excluded 43 patients who declined active treatment within 3 days of admission and only one patient with a history of acquired immunodeficiency syndrome; 1,756 patients were eligible for analysis. The mean age was 71.0 years, and 61.1% of patients were men. The in-hospital mortality rate was 22.1% (Table 1 ). Variable distributions, missingness, and correlations are shown in Additional file 1 (Additional file 1: Table S4, Figure S2 , and S3). Table 1 Characteristics of patients clustered using k-UMAP [nn = 100, k = 3] ALL (n = 1,756) Subphenotypes P value subphenotype 1 (n = 419) subphenotype 2 (n = 736) subphenotype 3 (n = 601) Age, median (IQR), yr 73 (64–82) 78 (68–85) 74 (64–82) 71 (62–79) < 0.001 Male sex, n (%) 1,073 (61.1) 8 (1.9) 719 (97.7) 346 (57.6) < 0.001 Body mass index, median (IQR), kg/m 2 21.9 (19.1–24.7) 21.4 (18.5–24.7) 21.9 (19.1–24.7) 22.0 (19.4–24.7) 0.09 Admission from ER, n (%) 1,125 (64.1) 394 (94.3) 645 (87.6) 86 (14.3) < 0.001 Past Medical history, n (%) Acute myocardial infarction 83 (4.7) 8 (1.9) 51 (6.9) 24 (4.0) < 0.001 Acute Heart Failure 189 (10.8) 38 (9.1) 71 (9.7) 80 (13.3) 0.04 Peripheral Arterial Disease 53 (3.0) 7 (1.7) 28 (3.8) 18 (3.0) 0.13 Stroke 219 (12.5) 52 (12.4) 118 (16.0) 49 (8.2) < 0.001 Chronic Lung Disease 101 (5.7) 19 (4.5) 58 (7.9) 24 (4.0) 0.005 Collagen Disease 113 (6.4) 9 (2.2) 16 (2.2) 88 (14.6) < 0.001 Peptic Ulcer 57 (3.2) 10 (2.4) 36 (4.9) 11 (1.8) 0.004 Liver Disease 90 (5.1) 20 (4.8) 33 (4.5) 37 (6.2) 0.36 Diabetes 442 (25.2) 102 (24.3) 190 (25.8) 150 (25.0) 0.85 Chronic kidney disease 142 (8.1) 30 (7.2) 56 (7.6) 56 (9.3) 0.38 Malignancy 242 (13.8) 47 (11.2) 98 (13.3) 97 (16.1) 0.07 Metastatic Neoplasm 40 (2.3) 1 (0.2) 17 (2.3) 22 (3.7) 0.002 Medication, n (%) Steroid 213 (12.1) 28 (6.7) 44 (6.0) 141 (23.5) < 0.001 Immunosuppressant 66 (3.8) 5 (1.2) 8 (1.1) 53 (8.8) < 0.001 Statin 187 (10.7) 43 (10.3) 90 (12.3) 54 (9.0) 0.15 Antiplatelet 273 (15.6) 43 (10.3) 138 (18.8) 92 (15.3) 0.0006 Beta Blocker 161 (9.2) 27 (6.4) 68 (9.3) 66 (11.0) 0.047 Radiotherapy, n (%) 7 (0.4) 0 (0.0) 3 (0.4) 4 (0.7) 0.25 Vital signs, median (IQR) Glasgow Coma Scale 14 (10–15) 14 (9–15) 14 (9–15) 14 (11–15) 0.0004 Respiratory rate, breath/min 25 (20–30) 25 (21–30) 25 (21–30) 24 (20–30) 0.009 Heart Rate, beats/min 109 (94–124) 107 (92–124) 110 (95–126) 108 (93–124) 0.25 Systolic blood pressure, mmHg 101 (82–127) 105 (84–130) 103 (83–130) 97 (80–123) 0.002 Diastolic blood pressure, mmHg 59 (47–74) 60 (46–74) 60 (49–76) 57 (45–72) 0.007 Body Temperature, ℃ 37.6 (36.7–38.7) 37.7 (36.5–38.6) 37.8 (36.8–38.9) 37.3(36.6–38.4) < 0.001 Blood test Data, median (IQR) White cell count, cells 10 3 /µL 11.4 (6.0-17.8) 11.4 (6.418.1) 11.5 (6.0-17.3) 11.4 (5.9–18.1) 0.77 Hematocrit, % 34.0 (28.8–39.4) 34.1 (29.4–39.2) 35.8 (30.3–40.7) 32.4 (27.2–37.1) < 0.001 Platelets count, platelets 10 4 /µL 14.9 (9.4–22.3) 15.9 (9.8–23.9) 15.4 (10.2–22.4) 13.4 (7.9–21.4) < 0.001 Creatinine, mg/dL 1.4 (0.9–2.5) 1.3 (0.7–2.2) 1.4 (0.9–2.5) 1.5 (0.9–2.6) < 0.001 Bilirubin, mg/dL 0.9 (0.6–1.5) 0.8 (0.5–1.4) 0.9 (0.6–1.6) 0.9 (0.6–1.6) 0.01 Glucose, mg/dL 141 (109–191) 143 (110–189) 147 (116–200) 132 (103–178) < 0.001 Sodium, mEq/L 137 (133–140) 137 (134–141) 137 (133–140) 136 (133–140) < 0.001 Potassium, mEq/L 4.0 (3.6–4.6) 3.9 (3.5–4.6) 4.1 (3.7–4.7) 4.0 (3.6–4.6) 0.004 C-reactive protein, mg/dL 14.8 (6.2–24.2) 12.8 (4.5–23.7) 14.1 (5.5–24.1) 16.8 (9.0-24.9) < 0.001 PT-INR 1.2 (1.1–1.4) 1.20 (1.07–1.40) 1.20 (1.10–1.40) 1.30 (1.11–1.50) < 0.001 APTT, sec 35.7 (30.3–44.3) 34.2 (29.1–42.0) 34.6 (29.6–41.6) 38.6 (32.8–48.8) < 0.001 D-dimer, µg/mL 6.7 (3.0-15.7) 7.0 (2.9–16.9) 5.5 (2.6–13.0) 8.4 (3.9–18.6) < 0.001 Fibrinogen, mg/mL 439 (313–578) 443 (313–577) 456 (317–597) 424 (300–557) 0.02 pH 7.39 (7.31–7.45) 7.40 (7.32–7.46) 7.40 (7.31–7.45) 7.39 (7.30–7.45) 0.09 PaCO2, mmHg 34.0 (28.5–41.2) 33.0 (28.0-39.6) 33.9 (28.8–41.4) 35.0 (29.0–42.0) 0.03 PaO2, mmHg 87.5 (69.8–123) 87.5 (68.8-128.5) 84.1 (68.0-119.0) 90.9 (73.0-126.4) 0.03 Base excess, mEq/L -3.2 (-7.0-0.1) -3.0 (-6.6-0.3) -3.0 (-7.1- -0.1) -3.5 (-7.5-0.3) 0.55 Lactate, mmol/L 2.9 (1.9-5.0) 3.1 (2.0-5.1) 3.1 (2.0-5.4) 2.6 (1.6–4.6) < 0.001 SOFA score, median (IQR) 8 (5–11) 8 (5–11) 8 (5–11) 9 (6–12) 0.02 Respiration 2 (1–3) 2 (1–2) 2 (1–3) 2 (1–3) 0.02 Coagulation 1 (0–2) 0 (0–1) 0 (0–1) 1 (0–2) < 0.001 Liver 0 (0–1) 0 (0–1) 0 (0–1) 0 (0–1) 0.15 Cardiovascular 3 (0–4) 1 (0–4) 3 (0–4) 3 (0–4) 0.001 Central Nervous System 1 (0–3) 1 (1–3) 1 (1–3) 1 (0–2) < 0.001 Renal 1 (0–3) 1 (0–3) 1 (0–3) 1 (0–3) 0.03 APACHE II score, median (IQR) 22 (16–28) 21 (16–27) 21 (16–29) 22 (17–29) 0.05 SIRS, median (IQR) 3 (2–4) 3 (2–4) 3 (2–4) 3 (2–3) 0.05 JAAM DIC score, median (IQR) 3 (2–5) 3 (2–5) 3 (2–5) 4 (2–6) < 0.001 Management, n (%) Noradrenaline 1,000 (57.0) 220 (52.6) 387 (52.7) 393 (65.4) < 0.001 Vasopressin 209 (11.9) 46 (11.0) 83 (11.3) 80 (13.3) 0.42 Corticosteroids 464 (27.1) 86 (21.0) 182 (25.5) 196 (33.3) < 0.001 Thrombomodulin 360 (21.1) 73 (17.9) 127 (17.8) 160 (27.2) < 0.001 Antithrombin III 333 (19.5) 62 (15.2) 107 (15.0) 164 (27.9) < 0.001 Immunoglobulin 332 (19.5) 60 (14.7) 103 (14.5) 169 (28.8) < 0.001 PMX-DHP 141 (8.2) 25 (6.1) 59 (8.3) 57 (9.7) 0.12 Outcomes In-hospital death, n (%) 379 (22.1) 158 (18.2) 158 (22.0) 146 (24.8) 0.04 Abbreviation : nn number of neighbors, SD standard deviation, IQR interquartile range, ER emergency room, APTT Activated partial thromboplastin time, SOFA score Sequential Organ Failure Assessment score, APACHE II score Acute Physiology and Chronic Health Evaluation II score, SIRS Systemic Inflammatory Response Syndrome, JAAM DIC score Japanese Association for Acute Medicine Disseminated Intravascular Coagulation score, PMX-DHP polymyxin B-immobilized fiber column hemoperfusion Cluster analysis Figure 1 illustrates the ARI. In any number of subphenotypes, the ARIs of k-UMAP and k-prototype with high lambda tended to be high, and those of the KAMILA and k-prototype with low lambda tended to be high. A detailed description of each number of subphenotypes is provided in Additional file 1: Figure S4-9. Figure 2 shows the cluster consensus. We observed that the ranking of candidate clusters by cluster consensus was reversed when k was greater than 3. Figure 2 shows that fewer clusters tended to be more robust for all clustering methods, particularly for KAMILA and the k-prototype with lambda = 1 and 0.01; however, we intended to explore as many characteristics as possible. Therefore, we selected the following three results of clustering conditions considering consensus and redundancy: two results of k-UMAP with more robustness (k-UMAP [number of neighbors (nn) = 100, k = 3], and k-UMAP [nn = 200, k = 5]) and one result of KAMILA (KAMILA [k = 3]) because the KAMILA and k-prototype with low lambda had almost the same results of clustering as those of ARIs. Additional file 1: Figure S13 illustrates the transitions between k-UMAP [nn = 100, k = 3] and k-UMAP [nn = 200, k = 5]. This figure indicates that k-UMAP [nn = 200, k = 5] is a divided population of k-UMAP [nn = 100, k = 3]. Characteristics of sepsis subphenotypes in each cluster analysis In the k-UMAP [nn = 100 k = 3], patients with subphenotype 3 were likely to have high in-hospital mortality (Table 1 ), but the severity of organ dysfunction was not significantly different among each subphenotype. The percentages of sex, admission route, steroid medication, and immunosuppressant use were particularly distinctive for each subphenotype. Subphenotype 1 included a lower percentage of men and a higher percentage of admission routes from the emergency room. Subphenotype 3 had a lower percentage of admission routes from the emergency room, a higher percentage of a medical history of collagen disease, and medication histories of steroids and immunosuppressants. In the k-UMAP [nn = 200, k = 5], patients with subphenotypes 3–5 likely had high in-hospital mortality (Additional file 1: Table S5). However, the severity of organ dysfunction differed less among each subphenotype, similar to the k-UMAP [nn = 100 k = 3]. In KAMILA [k = 3], patients with subphenotypes 2 and 3 likely had a high SOFA score, lactate levels, and in-hospital mortality (Additional file 1: Table S6). Cluster analysis in the validation dataset Additional file 1: Figure S8 and Table S7 show the clustering results and characteristics of the sepsis subphenotypes for k-UMAP [nn = 100, k = 3] in the validation dataset. The in-hospital mortality, SOFA, and Acute Physiology and Chronic Health Evaluation II scores, characteristics, admission route, medical histories, and blood test data for each subphenotype were similar to those in the derivation dataset. The clustering results of the derivation and validation data in k-UMAP are plotted in UMAP (Additional file 1: Figure S9), and a similar distribution was obtained. Treatment effects in sepsis subphenotypes Figure 3 shows the odds ratios (ORs) for the interaction between treatment and subphenotypes for in-hospital mortality in each clustering method. In the k-UMAP [nn = 100 k = 3], rTM was significantly effective for sub phenotype 1 (OR 0.37, 95% Confidence interval [CI] 0.13–0.80). In the k-UMAP [nn = 200 k = 5], rTM was significantly effective for subphenotype 1 (OR 0.30, 95% CI 0.09–0.70) and AT III was significantly effective for subphenotype 2 (OR 0.49, 95% CI 0.20–0.92). In contrast, no effective treatment was available for the subphenotypes in KAMILA [k = 3]. Additional file 1: Figure S10 shows the same odds for patients with septic shock and the overall trend was similar to all patients. AT Ⅲ was significantly effective for subphenotype 2 in the k-UMAP [nn = 100, k = 3] (OR 0.43, 95% CI 0.19–0.80), subphenotype 2 in the k-UMAP [nn = 200, k = 5] (OR 0.41, 95% CI 0.16–0.84) and subphenotype 1 in the KAMILA [k = 3] (OR 0.31, 95% CI 0.09–0.73). rTM was significantly effective for subphenotype1 in the k-UMAP [nn = 200, k = 5] (OR 0.42, 95% CI 0.13–0.99). Discussion In this study, we explored the subphenotypes of Japanese patients with sepsis enrolled in cohort studies. For this purpose, we applied the following state-of-the-art clustering methods to the mixed data: k-prototype, KAMILA, and k-UMAP. We identified three or five subphenotypes with various features, and only k-UMAP could detect these subphenotypes in Japanese patients with sepsis who benefited from some treatments. As conventional clustering methods, such as the original k-means method, have not been developed for mixed data, previous studies have only used continuous variables in their analyses [ 21 – 23 ]. In this study, we adopted state-of-the-art clustering methods for mixed data to consider categorical data as features. In the case of mixed data, one of the problems is the lack of consensus on determining the weight between the variables measured using different metrics, such as Euclidean or Hamming. Therefore, we indirectly investigated the clustering method for the weight balance between different metrics by comparing the k-prototype with several values of the lambda parameter using other clustering methods. According to the ARI results, patients were classified into close subphenotypes when using k-UMAP and k-prototype with high lambda. As mentioned in Additional file 1, we used the same metrics for the k-prototype and k-UMAP. This metric indicates that k-UMAP assigned a larger weight to the categorical data than the k-prototype with a low lambda. In contrast, KAMILA and the k-prototype with low lambda values were classified as close subphenotypes. Thus, KAMILA tends to assign larger weights to numerical data than k-UMAP. UMAP has been popular as an effective visualization method in machine learning and medicine and natural language processing, where improved clustering methods have been reported [ 43 ]. However, to our knowledge, the performance of k-UMAP has not been investigated for mixed datasets. Therefore, we examined this using the same simulation as that used in a previous study [ 35 ]. The simulation indicated that the characteristics of k-UMAP were stable for a dataset heavily containing categorical data. Thus, k-UMAP is efficient when categorical data is clinically important. Otherwise, the k-UMAP algorithm differs from other partitioning clustering methods owing to the UMAP; therefore, k-UMAP can be used with other clustering methods to explore a wide range of patient characteristics. By comparing the subphenotypes in previous studies with those of k-UMAP, subphenotypes characterized by high mortality, cardiovascular failure, and coagulopathy were identified (Additional file 1: Table S8) [ 21 , 23 , 24 , 44 ]. Moreover, k-UMAP could be characterized better than the subphenotypes in previous studies because it could reflect differences in elements of categorical variables, such as sex, background disease, medication history, and admission route. The three subphenotypes of k-UMAP were characterized as follows: subphenotype 1 was the older, female, emergency room admission type with the lowest in-hospital mortality; subphenotype 2 was the male, lifestyle-related disease type, such as cardiovascular disease, chronic lung disease, and peptic ulcer; and subphenotype 3 was the younger, immunocompromised type with the highest in-hospital mortality. Another contribution of this study is the evaluation of treatment effects in patients who received several treatments for sepsis. For subphenotype classification to be used in clinical practice, it should be impactful, such as recognizing differential treatment responses [ 15 ]. Generally, patients with sepsis are treated with multiple adjunct therapies depending on their medical condition. However, no studies have investigated the effects of these multiple therapies across subphenotypes. This study showed that the effects of multiple treatments varied based on the subphenotype. By examining treatment effects using the same subphenotype, the effects of each treatment can be compared among subphenotypes. A predictive model for subphenotypes could contribute to identify specific sepsis patients, and a novel treatment flow may be provided. In addition, this approach will be an important progress to individualized treatment strategies for sepsis. This study has several strengths. We used a reliable sepsis dataset with many variables and a few missing values. We investigated various clustering methods and compared their results. We adjusted covariate effects to exclude reverse causality. This study has some limitations. First, the statistical power was insufficient because of the small sample size for each subphenotype. Second, cluster analysis results may depend on a combination of variables. Third, the dataset excluded the timing of treatments; therefore, we could not precisely model the interactions between treatments. Therefore, future studies should use covariates as time-series data to build models that reflect multiple treatment effects and progress. Finally, because the datasets and patients were all Japanese, this study may have been influenced by the ethnic characteristics of the Japanese population and the trend toward more treatment in Japanese hospitals. Future studies in other populations and cultures are required to identify the global subphenotypes associated with treatment effects. Conclusion In this retrospective analysis of datasets from patients with sepsis, three state-of-the-art methods were applied to cluster analysis for mixed data, and three to five subphenotypes were identified to be associated with clinical information and outcomes. The effects of treatments differed for each subphenotype, and k-UMAP detected the subphenotypes of Japanese patients with sepsis who benefited from some treatments. Identifying patients for whom treatment is more effective can lead to precision medicine in critical care, and k-UMAP may be a good option for clustering methods using mixed clinical data. Abbreviations intensive care units (ICUs) , The Focused Outcomes Research in Emergency Care for Acute Respiratory Distress Syndrome, Sepsis, and Trauma (FORECAST), Japanese Association for Acute Medicine Sepsis Prognostication in the ICU and Emergency Room (JAAM SPICE-ICU), Japanese Association for Acute Medicine Multicenter Assessment for Sepsis Treatment and Outcome (JAAM MAESTRO), sepsis-related organ failure assessment (SOFA), recombinant thrombomodulin (rTM), antithrombin III (AT III), KAymeans for MIxed LArge data (KAMILA), Uniform Manifold Approximation and Projection for Dimension Reduction (UMAP), principal component analysis (PCA), adjusted rand indexes (ARIs), propensity score (PS), number of neighbors (nn), odds ratios (ORs) Declarations Ethics approval and consent to participate Two original studies were approved, and the need for informed consent was waived by the Institutional Review Boards of the participating hospitals. Consent for publication Not applicable. Availability of data and materials The datasets generated and/or analyzed during the original studies are available at Mendeley Data, https://data.mendeley.com/datasets/vvv89kw3k5/1 (Tohoku Sepsis Registry). The datasets of the FORECAST sepsis study, JAAM SPICE-ICU, and JAAM MAESTRO are restrictions apply to the availability of the data, which were used under license for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission of the Japanese Association for Acute Medicine. Competing interests The authors declare no conflicts of interest with respect to this study. Funding This work (writing of the manuscript) was supported by departmental funds. Author contributions YY, RS, GT, and SK designed the study. YY, RS, and GT wrote the analysis plan, and YY and RS analyzed the data. YY, RS, DK, and SK drafted the initial manuscript. All authors critically revised the manuscript. The corresponding author attests that all listed authors meet the authorship criteria and that no others meeting the criteria have been omitted. Acknowledgments We thank all investigators involved in the Tohoku Sepsis Registry, FORECAST Sepsis Study, JAAM SPICE-ICU, and JAAM MAESTRO for contributing to data collection and assessment. We thank Hikaru Matsuoka for technical advice. This work was supported by the MEXT/JSPS WISE Program and the Advanced Graduate Program for Future Medicine and Health Care at Tohoku University. References Fleischmann-Struzek C, Mellhammar L, Rose N, Cassini A, Rudd KE, Schlattmann P, Allegranzi B, Reinhart K: Incidence and mortality of hospital- and ICU-treated sepsis: results from an updated and expanded systematic review and meta-analysis . Intensive Care Med 2020, 46 :1552-1562. Egi M, Ogura H, Yatabe T, Atagi K, Inoue S, Iba T, Kakihana Y, Kawasaki T, Kushimoto S, Kuroda Y et al : The Japanese Clinical Practice Guidelines for Management of Sepsis and Septic Shock 2020 (J-SSCG 2020) . J Intensive Care 2021, 9 :53. Evans L, Rhodes A, Alhazzani W, Antonelli M, Coopersmith CM, French C, Machado FR, McIntyre L, Ostermann M, Prescott HC et al : Surviving sepsis campaign: international guidelines for management of sepsis and septic shock 2021 . Intensive Care Med 2021, 47 :1181-1247. Investigators P, Rowan KM, Angus DC, Bailey M, Barnato AE, Bellomo R, Canter RR, Coats TJ, Delaney A, Gimbel E et al : Early, Goal-Directed Therapy for Septic Shock - A Patient-Level Meta-Analysis . N Engl J Med 2017, 376 :2223-2234. Seymour CW, Gesten F, Prescott HC, Friedrich ME, Iwashyna TJ, Phillips GS, Lemeshow S, Osborn T, Terry KM, Levy MM: Time to Treatment and Mortality during Mandated Emergency Care for Sepsis . N Engl J Med 2017, 376 :2235-2244. Peltan ID, Brown SM, Bledsoe JR, Sorensen J, Samore MH, Allen TL, Hough CL: ED Door-to-Antibiotic Time and Long-term Mortality in Sepsis . Chest 2019, 155 :938-946. Avni T, Lador A, Lev S, Leibovici L, Paul M, Grossman A: Vasopressors for the Treatment of Septic Shock: Systematic Review and Meta-Analysis . PLoS One 2015, 10 :e0129305. Iizuka Y, Sanui M, Sasabuchi Y, Lefor AK, Hayakawa M, Saito S, Uchino S, Yamakawa K, Kudo D, Takimoto K et al : Low-dose immunoglobulin G is not associated with mortality in patients with sepsis and septic shock . Crit Care 2017, 21 :181. Rygard SL, Butler E, Granholm A, Moller MH, Cohen J, Finfer S, Perner A, Myburgh J, Venkatesh B, Delaney A: Low-dose corticosteroids for adult patients with septic shock: a systematic review with meta-analysis and trial sequential analysis . Intensive Care Med 2018, 44 :1003-1016. Vincent JL, Francois B, Zabolotskikh I, Daga MK, Lascarrou JB, Kirov MY, Pettila V, Wittebole X, Meziani F, Mercier E et al : Effect of a Recombinant Human Soluble Thrombomodulin on Mortality in Patients With Sepsis-Associated Coagulopathy: The SCARLET Randomized Clinical Trial . JAMA 2019, 321 :1993-2002. Santacruz CA, Pereira AJ, Celis E, Vincent JL: Which Multicenter Randomized Controlled Trials in Critical Care Medicine Have Shown Reduced Mortality? A Systematic Review . Crit Care Med 2019, 47 :1680-1691. Fujii T, Luethi N, Young PJ, Frei DR, Eastwood GM, French CJ, Deane AM, Shehabi Y, Hajjar LA, Oliveira G et al : Effect of Vitamin C, Hydrocortisone, and Thiamine vs Hydrocortisone Alone on Time Alive and Free of Vasopressor Support Among Patients With Septic Shock: The VITAMINS Randomized Clinical Trial . JAMA 2020, 323 :423-431. Dellinger RP, Bagshaw SM, Antonelli M, Foster DM, Klein DJ, Marshall JC, Palevsky PM, Weisberg LS, Schorr CA, Trzeciak S et al : Effect of Targeted Polymyxin B Hemoperfusion on 28-Day Mortality in Patients With Septic Shock and Elevated Endotoxin Level: The EUPHRATES Randomized Clinical Trial . JAMA 2018, 320 :1455-1463. James A. Russell MD, Keith R. Walley, M.D., Joel Singer, Ph.D., Anthony C. Gordon, M.B., B.S., M.D.,, Paul C. Hébert MD, D. James Cooper, B.M., B.S., M.D., Cheryl L. Holmes, M.D., Sangeeta Mehta, M.D.,, John T. Granton MD, Michelle M. Storms, B.Sc.N., Deborah J. Cook, M.D., Jeffrey J. Presneill, M.B., B.S., Ph.D.,, and Dieter Ayers MS, for the VASST Investigator: Vasopressin versus Norepinephrine Infusion in Patients with Septic Shock.pdf . N Engl J Med 2008, 358 :877-887. Reddy K, Sinha P, O'Kane CM, Gordon AC, Calfee CS, McAuley DF: Subphenotypes in critical care: translation into clinical practice . Lancet Respir Med 2020, 8 :631-643. Maslove DM, Tang B, Shankar-Hari M, Lawler PR, Angus DC, Baillie JK, Baron RM, Bauer M, Buchman TG, Calfee CS et al : Redefining critical illness . Nat Med 2022, 28 :1141-1148. Wang W, Liu CF: Sepsis heterogeneity . World J Pediatr 2023, 19 :919-927. Harhay MO, Casey JD, Clement M, Collins SP, Gayat E, Gong MN, Jaber S, Laterre PF, Marshall JC, Matthay MA et al : Contemporary strategies to improve clinical trial design for critical care research: insights from the First Critical Care Clinical Trialists Workshop . Intensive Care Med 2020, 46 :930-942. Anthony C. Gordon NA-L, Lieuwe D. Bos, Jose Dianti, Janet Victoria Diaz, Tomoko Fujii, Evangelos J. Giamarellos-Bourboulis, Jean-Louis Vincent, Carolyn S. Calfee: From ICU Syndromes to ICU Subphenotypes: Consensus Report and Recommendations For Developing Precision Medicine in ICU . Am J Respir Crit Care Med 2024, Online ahead of print. Knox DB, Lanspa MJ, Kuttler KG, Brewer SC, Brown SM: Phenotypic clusters within sepsis-associated multiple organ dysfunction syndrome . Intensive Care Med 2015, 41 :814-822. Seymour CW, Kennedy JN, Wang S, Chang CH, Elliott CF, Xu Z, Berry S, Clermont G, Cooper G, Gomez H et al : Derivation, Validation, and Potential Treatment Implications of Novel Clinical Phenotypes for Sepsis . JAMA 2019, 321 :2003-2017. Kudo D, Goto T, Uchimido R, Hayakawa M, Yamakawa K, Abe T, Shiraishi A, Kushimoto S: Coagulation phenotypes in sepsis and effects of recombinant human thrombomodulin: an analysis of three multicentre observational studies . Crit Care 2021, 25 :114. Bhavani SV, Semler M, Qian ET, Verhoef PA, Robichaux C, Churpek MM, Coopersmith CM: Development and validation of novel sepsis subphenotypes using trajectories of vital signs . Intensive Care Med 2022, 48 :1582-1592. Zhang Z, Chen L, Liu X, Yang J, Huang J, Yang Q, Hu Q, Jin K, Celi LA, Hong Y: Exploring disease axes as an alternative to distinct clusters for characterizing sepsis heterogeneity . Intensive Care Med 2023, 49 :1349-1359. Ahmad A, Khan SS: Survey of State-of-the-Art Mixed Data Clustering Algorithms . IEEE Access 2019, 7 :31883-31902. Dellinger RP, Levy MM, Rhodes A, Annane D, Gerlach H, Opal SM, Sevransky JE, Sprung CL, Douglas IS, Jaeschke R et al : Surviving sepsis campaign: international guidelines for management of severe sepsis and septic shock: 2012 . Crit Care Med 2013, 41 :580-637. Manu Shankar-Hari M, MSc; Gary S. Phillips, MAS; Mitchell L. Levy, MD; ChristopherW. Seymour, MD, MSc; Vincent X. Liu, MD, MSc; Clifford S. Deutschman, MD; Derek C. Angus, MD, MPh; Gordon D. Rubenfeld, MD, MSc; Mervyn Singer, MD, FRCP; for the Sepsis Definitions Task Force: Developing a NewDefinition and Assessing New Clinical Criteria for Septic Shock For the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3) . JAMA 2016, 315 :775-787. Kudo D, Kushimoto S, Miyagawa N, Sato T, Hasegawa M, Ito F, Yamanouchi S, Honda H, Andoh K, Furukawa H et al : The impact of organ dysfunctions on mortality in patients with severe sepsis: A multicenter prospective observational study . J Crit Care 2018, 45 :178-183. Abe T, Ogura H, Shiraishi A, Kushimoto S, Saitoh D, Fujishima S, Mayumi T, Shiino Y, Nakada T-a, Tarui T et al : Characteristics, management, and in-hospital mortality among patients with severe sepsis in intensive care units in Japan: the FORECAST study . Critical Care 2018, 22:322. Abe T, Yamakawa K, Ogura H, Kushimoto S, Saitoh D, Fujishima S, Otomo Y, Kotani J, Umemura Y, Sakamoto Y et al : Epidemiology of sepsis and septic shock in intensive care units between sepsis-2 and sepsis-3 populations: sepsis prognostication in intensive care unit and emergency room (SPICE-ICU) . J Intensive Care 2020, 8:44. Abe T, Umemura Y, Ogura H, Kushimoto S, Fujishima S, Saitoh D, Gando S: Early versus delayed vasopressor administration in patients with septic shock . Acute Med Surg 2023, 10 :e852. J L Vincent RM, J Takala, S Willatts, A De Mendonça, H Bruining, C K Reinhart, P M Suter, L G Thijs: The SOFA (Sepsis.related Organ Failure Assessment) score to describe organ dysfunction/failure . Intensive Care Med 1996, 22 :707-710. HUANG Z: Extensions to the k-Means Algorithm for Clustering Large DataSets with Categorical Values.pdf> . Data Mining and Knowledge Discovery 1998, 2 :283–304. Foss A, Markatou M, Ray B, Heching A: A semiparametric method for clustering mixed data . Machine Learning 2016, 105 :419-458. Preud'homme G, Duarte K, Dalleau K, Lacomblez C, Bresso E, Smail-Tabbone M, Couceiro M, Devignes MD, Kobayashi M, Huttin O et al : Head-to-head comparison of clustering methods for heterogeneous data: a simulation-driven benchmark . Sci Rep 2021, 11 :4202. MACQUEEN J: SOME METHODS FOR CLASSIFICATION AND ANALYSIS OF MULTIVARIATE OBSERVATIONS . Berkeley Symp on Math Statist and Prob 1967:281-297. Leland McInnes JH, James Melville: UMAP Uniform Manifold Approximation and Projection for Dimension Reduction . arXiv 2020(1802). Chris Ding XH: Advances in Knowledge Discovery and Data Mining , vol. 3056. Sydney, Australia: 8th Pacific-Asia Conference; 2004. Hubert L: Comparing Partitions . Journal of Classification 1985, 2 :193-218. STEFANO MONTI PT, JILL MESIROV, TODD GOLUB: Consensus Clustering A Resampling-Based Method for Class Discovery and Visualization of Gene Expression Microarray Data.pdf . Machine Learning 2003, 52 :91–118. PAUL R. ROSENBAUM DBR: The central role of the propensity score in observational studies for causal effects . Biometrika 1983, 70 : 41–55. Austin PC: An Introduction to Propensity Score Methods for Reducing the Effects of Confounding in Observational Studies . Multivariate Behav Res 2011, 46 :399-424. Allaoui M, Kherfi ML, Cheriet A: Considerably Improving Clustering Algorithms Using UMAP Dimensionality Reduction Technique: A Comparative Study . In: Image and Signal Processing. edn.; 2020: 317-325. Soussi S, Sharma D, Juni P, Lebovic G, Brochard L, Marshall JC, Lawler PR, Herridge M, Ferguson N, Del Sorbo L et al : Identifying clinical subtypes in sepsis-survivors with different one-year outcomes: a secondary latent class analysis of the FROG-ICU cohort . Crit Care 2022, 26 :114. Additional Declarations No competing interests reported. Supplementary Files Additionalfile1yutayokokawa.docx Graphicalabstractimage.pdf Cite Share Download PDF Status: Published Journal Publication published 15 Jan, 2026 Read the published version in Critical Care → Version 1 posted Editorial decision: Revision requested 18 Oct, 2025 Reviews received at journal 13 Oct, 2025 Reviews received at journal 11 Oct, 2025 Reviewers agreed at journal 01 Oct, 2025 Reviewers agreed at journal 29 Sep, 2025 Reviewers invited by journal 29 Sep, 2025 Editor assigned by journal 28 Sep, 2025 Submission checks completed at journal 28 Sep, 2025 First submitted to journal 25 Sep, 2025 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-7716199","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":528150988,"identity":"89c2bd1c-3ab0-4516-81a3-7f6eee973463","order_by":0,"name":"Yuta Yokokawa","email":"","orcid":"","institution":"Tohoku University Graduate School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Yuta","middleName":"","lastName":"Yokokawa","suffix":""},{"id":528150989,"identity":"b3afbe79-265d-487c-8360-e38cc86f57aa","order_by":1,"name":"Rieko Sakurai","email":"","orcid":"","institution":"RIKEN Ctr Adv Intelligence Project","correspondingAuthor":false,"prefix":"","firstName":"Rieko","middleName":"","lastName":"Sakurai","suffix":""},{"id":528150990,"identity":"be581f59-0c9c-411c-829f-e652d9545dbe","order_by":2,"name":"Daisuke Kudo","email":"","orcid":"","institution":"Tohoku University Graduate School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Daisuke","middleName":"","lastName":"Kudo","suffix":""},{"id":528150991,"identity":"537f5ea4-2666-4947-807b-cff1a602b2ab","order_by":3,"name":"Gen Tamiya","email":"","orcid":"","institution":"RIKEN Ctr Adv Intelligence Project","correspondingAuthor":false,"prefix":"","firstName":"Gen","middleName":"","lastName":"Tamiya","suffix":""},{"id":528150992,"identity":"573f1838-de5f-4177-9e86-9971da01b94b","order_by":4,"name":"Shigeki Kushimoto","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA50lEQVRIiWNgGAWjYBACCRCRwCAhx8bMfADElyFWi4UxH3tbAojPQ5wWBoaKxHk8ZwxALMJaJGekP/7woEKCsU0i5/OrGzUWPAzsh49uwKdFWiLHTCLhjAQzm0TuNuucY0CH8aSl3cCnRU4ih40hsU2CDaTFOIcNqEWCx4yAFqDDEv9J8LBJ5DwzzvlHhBZpiQQDicQGCQk2njPMj3PbiNAi2fMG6JdjEgZs7G1mzLl9QOsI+UXiePrjjz9q6urnNzM//pzzrU6On/3wMbxakAEbOJLYiFUOAswfSFE9CkbBKBgFIwcAAPbbQFxht4PrAAAAAElFTkSuQmCC","orcid":"","institution":"Tohoku University Graduate School of Medicine","correspondingAuthor":true,"prefix":"","firstName":"Shigeki","middleName":"","lastName":"Kushimoto","suffix":""}],"badges":[],"createdAt":"2025-09-25 21:53:12","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7716199/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7716199/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s13054-025-05819-4","type":"published","date":"2026-01-15T16:30:14+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":93538508,"identity":"1e3f287f-b04b-429f-89d3-59d4f5749c27","added_by":"auto","created_at":"2025-10-15 02:09:37","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":92909,"visible":true,"origin":"","legend":"","description":"","filename":"Manuscriptyutayokokawa.docx","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/f60c70c4999dfa327e5b8136.docx"},{"id":93538541,"identity":"4d0d35d8-891e-4b32-bdb1-0cc520846f8e","added_by":"auto","created_at":"2025-10-15 02:09:39","extension":"jpg","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":172809,"visible":true,"origin":"","legend":"","description":"","filename":"Figure1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/6d905bbf5ceeeae018f10b25.jpg"},{"id":93538516,"identity":"76b0c342-5c40-4f86-9fe5-76bd6defcdb9","added_by":"auto","created_at":"2025-10-15 02:09:38","extension":"jpg","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":85987,"visible":true,"origin":"","legend":"","description":"","filename":"Figure2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/cc1147169edea7d474197b80.jpg"},{"id":93538546,"identity":"0b1278a0-8cbb-42c8-9a60-a7631f7f3d4d","added_by":"auto","created_at":"2025-10-15 02:09:39","extension":"jpg","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":97726,"visible":true,"origin":"","legend":"","description":"","filename":"Figure3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/1272708c93b977b04604959d.jpg"},{"id":93538559,"identity":"94689be1-42aa-465b-9993-297a4275c7cb","added_by":"auto","created_at":"2025-10-15 02:09:39","extension":"json","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8295,"visible":true,"origin":"","legend":"","description":"","filename":"512809b915aa4c66bdb092ba83b79360.json","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/a3a85219d07982e359ec8c6e.json"},{"id":93538547,"identity":"8b41dd9b-a0ff-47a6-b813-067dd5561b61","added_by":"auto","created_at":"2025-10-15 02:09:39","extension":"docx","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4300749,"visible":true,"origin":"","legend":"","description":"","filename":"Additionalfile1yutayokokawa.docx","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/61db771cb4dd3e28576bb4b4.docx"},{"id":93538537,"identity":"3fee7f55-3d20-48b3-b252-a6dacc654ae2","added_by":"auto","created_at":"2025-10-15 02:09:39","extension":"pdf","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2213276,"visible":true,"origin":"","legend":"","description":"","filename":"Graphicalabstractimage.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/c9de62f5af65b7124fac329a.pdf"},{"id":93538535,"identity":"5b7e96dc-4726-4141-8466-683298eebc10","added_by":"auto","created_at":"2025-10-15 02:09:39","extension":"xml","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":155387,"visible":true,"origin":"","legend":"","description":"","filename":"512809b915aa4c66bdb092ba83b793601enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/770066c1fac0ba00cfc1e70b.xml"},{"id":93538613,"identity":"6cb0bb12-5342-48ff-879c-972ea34841b2","added_by":"auto","created_at":"2025-10-15 02:09:42","extension":"jpg","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":172809,"visible":true,"origin":"","legend":"","description":"","filename":"Figure1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/f9076ec19f603a4dadcbdd35.jpg"},{"id":93538525,"identity":"46f4d64b-39cb-4d6d-8d2a-64654d251d20","added_by":"auto","created_at":"2025-10-15 02:09:38","extension":"jpg","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":85987,"visible":true,"origin":"","legend":"","description":"","filename":"Figure2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/fdb98cdd54b230faff94e8a6.jpg"},{"id":93538583,"identity":"ffaaca46-5d83-4be0-bb17-a04d95256f58","added_by":"auto","created_at":"2025-10-15 02:09:40","extension":"jpg","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":97726,"visible":true,"origin":"","legend":"","description":"","filename":"Figure3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/ffbad3b7cc88de30496e6772.jpg"},{"id":93538549,"identity":"5c4529b5-6c57-489d-8250-6969074d6db8","added_by":"auto","created_at":"2025-10-15 02:09:39","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":62572,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure1.png","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/3c48a5bf30a1d5f05b7ee090.png"},{"id":93539398,"identity":"ac1fc9e2-9ef2-4c51-8aca-07f341b55e50","added_by":"auto","created_at":"2025-10-15 02:17:42","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":27126,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure2.png","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/3e9d12e55abceaf8adf6bc20.png"},{"id":93539399,"identity":"2a34f217-eb8e-47d0-ac9a-b3d5a06cfc56","added_by":"auto","created_at":"2025-10-15 02:17:42","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":20290,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure3.png","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/45688157dd3659b70391c421.png"},{"id":93538515,"identity":"3d76d1e9-b6a4-4348-96ca-84efea773d2a","added_by":"auto","created_at":"2025-10-15 02:09:38","extension":"xml","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":154785,"visible":true,"origin":"","legend":"","description":"","filename":"512809b915aa4c66bdb092ba83b793601structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/60e8ada95b46c9e9076b64e6.xml"},{"id":93538532,"identity":"118390eb-5259-4f77-a836-c10cea4f4a6b","added_by":"auto","created_at":"2025-10-15 02:09:39","extension":"html","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":163900,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/609a92f04e1b357cf5f2520e.html"},{"id":93538557,"identity":"4eea9626-9812-43b2-bc02-c36b8d01a43a","added_by":"auto","created_at":"2025-10-15 02:09:39","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":172809,"visible":true,"origin":"","legend":"\u003cp\u003eAdjusted Rand Indexes between clustering methods at each number of clusters.\u003c/p\u003e\n\u003cp\u003eThe Adjusted Rand Index (ARI) is a measure of the similarity between two data of clustering methods. Figure a to f show the ARI between clustering methods (k-UMAP, k-prototype and KAMILA) at cluster number 2 to 7. The ARIs of the KAMILA and k-prototype with low lambda tended to be higher. The ARIs of k-UMAP and k-prototype with high lambda also tended to be higher. The pitem is a resampling rate.\u003c/p\u003e\n\u003cp\u003e*nn: number of neighbors is a parameter which controls how UMAP balances local versus global structure in the data\u003c/p\u003e\n\u003cp\u003eAbbreviations: \u003cem\u003ek\u003c/em\u003enumber of clusters, \u003cem\u003eKAMILA\u003c/em\u003e KAymeans for MIxed LArge data\u003c/p\u003e","description":"","filename":"Figure1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/01623c7820d5f42e4e867183.jpg"},{"id":93538550,"identity":"03c23015-0094-4fed-a1f6-d6413f0ea4fa","added_by":"auto","created_at":"2025-10-15 02:09:39","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":85987,"visible":true,"origin":"","legend":"\u003cp\u003eCluster consistencies of clustering methods at each number of clusters\u003c/p\u003e\n\u003cp\u003eFigure a-f shows the cluster consistencies for each clustering method with subphenotypes 2-7. The pitem is rate of item resampling. When the number of clusters were 2 and 3, cluster consistencies were high for the KAMILA and k-prototype with lambda=1 and 0.01. On the other hands, ranking by cluster consistency reverses when k is greater than 3. The KAMILA and k-prototype were less likely to decrease the cluster consistency by resampling\u003c/p\u003e\n\u003cp\u003e*nn: number of neighbors is a parameter which controls how UMAP balances local versus global structure in the data\u003c/p\u003e\n\u003cp\u003e*lambda is a real valued parameter that controls the trade off between Euclidean distance for numeric variables and simple matching distance for factor variables for cluster assignment\u003c/p\u003e\n\u003cp\u003eAbbreviations: \u003cem\u003ek\u003c/em\u003enumber of clusters, \u003cem\u003eKAMILA\u003c/em\u003e KAymeans for MIxed LArge data\u003c/p\u003e","description":"","filename":"Figure2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/6de2b964c73d948b77f8b774.jpg"},{"id":93539379,"identity":"c9818850-5e9c-43ea-ad5e-a094188aad43","added_by":"auto","created_at":"2025-10-15 02:17:37","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":97726,"visible":true,"origin":"","legend":"\u003cp\u003eOdds ratios for interaction of the treatment and subphenotypes for hospital mortality in each clustering methods. a: k-UMAP [nn=100, k=3], b: k-UMAP [nn=200, k=5], c: KAMILA [k=3]. The odds ratios (ORs) for the interaction between treatment and subphenotypes for in-hospital mortality in each clustering method. In the k-UMAP [nn=100 k=3], rTM was significantly effective for subphenotype 1 (OR 0.37, 95% CI 0.13–0.80). In the k-UMAP [nn=200 k=5], rTM was significantly effective for subphenotype 1 (OR 0.30, 95% CI 0.09–0.70) and AT III was significantly effective for subphenotype 2 (OR 0.49, 95% CI 0.20–0.92). In contrast, no effective treatment was available for the subphenotypes in KAMILA [k=3].\u003c/p\u003e\n\u003cp\u003eAbbreviations: \u003cem\u003eATⅢ\u003c/em\u003eantithrombin III, \u003cem\u003eIVIg\u003c/em\u003e immunoglobulin, \u003cem\u003ePMX\u003c/em\u003e polymyxin B-immobilized fiber column hemoperfusion, \u003cem\u003erTM\u003c/em\u003e recombinant thrombomodulin, \u003cem\u003eVAS\u003c/em\u003e vasopressin, \u003cem\u003enn\u003c/em\u003e number of neighbors, \u003cem\u003ek \u003c/em\u003enumber of clusters, \u003cem\u003eCI \u003c/em\u003eConfidence interval\u003c/p\u003e","description":"","filename":"Figure3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/7b2ead58f8ccbaa37b97077d.jpg"},{"id":100614773,"identity":"66228d25-ca77-4a81-aa91-6858bac09563","added_by":"auto","created_at":"2026-01-19 17:24:39","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3358427,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/956e4d41-4c23-40c9-9fb5-c55625d563e4.pdf"},{"id":93538512,"identity":"99405bd1-4958-40be-a3d6-c8e0eebd973a","added_by":"auto","created_at":"2025-10-15 02:09:37","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":4300749,"visible":true,"origin":"","legend":"","description":"","filename":"Additionalfile1yutayokokawa.docx","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/fcc3fbaf7f60098bf15ca095.docx"},{"id":93538527,"identity":"ad10da30-3e7a-4240-94b8-19e74e01ee65","added_by":"auto","created_at":"2025-10-15 02:09:38","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":2213276,"visible":true,"origin":"","legend":"","description":"","filename":"Graphicalabstractimage.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7716199/v1/0de675cc5f08da09c0e7740c.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Clinical subphenotypes of sepsis based on mixed data and differences in treatment effects: a cluster analysis of multicentre observational studies","fulltext":[{"header":"Introduction","content":"\u003cp\u003eSepsis affects millions of people worldwide, and its mortality rate remains high [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. The Guidelines for Management of Sepsis and Septic Shock [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e] state that the initial fluid resuscitation [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e], early administration of appropriate antimicrobials [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e], and administration of vasopressors for septic shock [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e] are the bundle to be administered within 1 h of recognition. However, no specific treatment has been shown to improve clinical outcomes in patients with sepsis. Several therapies have been investigated in clinical trials; however, none have adequately demonstrated effectiveness [\u003cspan additionalcitationids=\"CR9 CR10 CR11 CR12 CR13\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eOne reason treatment effectiveness has not been demonstrated is the heterogeneity of patients with sepsis [\u003cspan additionalcitationids=\"CR16 CR17\" citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Patients with sepsis have various clinical conditions; however, past studies have evaluated \"sepsis\" as a single population. Thus, effective treatment for some subphenotypes of sepsis may not improve outcomes for all patients with sepsis. Effective treatments are assumed to exist for appropriately classified subphenotypes [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eBased on these hypotheses, several studies have been conducted to classify patients with sepsis into subphenotypes by applying unsupervised approaches such as cluster analysis to clinical and biomarker data [\u003cspan additionalcitationids=\"CR21 CR22 CR23\" citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. However, these studies did not have enough impact to change clinical practice. One of the major challenges in the cluster analysis of clinical data is the use of clinical data, including categorical variables (e.g., sex, comorbidities, and medications) and continuous variables (e.g., vital signs and blood test data). The dataset contains numerical and categorical data, called \u0026ldquo;heterogeneous\u0026rdquo; or \u0026ldquo;mixed\u0026rdquo; data. Mixed data are challenging to use in cluster analysis because it is difficult to directly apply mathematical operations to both types of variables [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e], which is a barrier to fully using clinical data.\u003c/p\u003e\u003cp\u003eAdditionally, simply dividing patients with sepsis into subphenotypes has little clinical value. The subphenotype classification should reflect the pathophysiological status and impact on clinical practice by recognizing differential treatment responses [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Patients with sepsis are treated with multiple adjunct therapies depending on their medical condition; however, no study has investigated the effects of these multiple treatments across subphenotypes. Identifying subphenotypes with different treatment effects could lead to developing precision medicine for sepsis.\u003c/p\u003e\u003cp\u003eTherefore, this study aimed to (1) apply a state-of-the-art clustering method for mixed data, (2) classify patients with sepsis into subphenotypes using clinical information, and (3) examine differences in multiple treatment effects by subphenotype.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cp\u003eStudy population and setting\u003c/p\u003e\u003cp\u003eIn our study, we used a dataset from multicenter registries that involved patients (age\u0026thinsp;\u0026ge;\u0026thinsp;16 years) admitted to intensive care units (ICUs) with severe sepsis or septic shock according to the International Sepsis Definitions Conference criteria [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. We used two registries for the derivation. The Tohoku Sepsis Registry (UMIN000010297) was a prospective observational study involving 616 patients admitted to 10 institutions (three university hospitals and seven community hospitals) in the Tohoku District, Northeastern Japan, between January 2015 and December 2015 [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. The Focused Outcomes Research in Emergency Care for Acute Respiratory Distress Syndrome, Sepsis, and Trauma (FORECAST) sepsis study (UMIN000019742) is a multicenter, prospective study that enrolled 1,184 patients admitted to 59 ICUs in Japan between January 2016 and March 2017 [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. We also used two other multicenter registries for validation: the Japanese Association for Acute Medicine Sepsis Prognostication in the ICU and Emergency Room (JAAM SPICE-ICU) [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e] and the Japanese Association for Acute Medicine Multicenter Assessment for Sepsis Treatment and Outcome (JAAM MAESTRO) [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e] (Supplemental document). A detailed description of the methods and analytical processes used in these studies is provided in Additional file 1.\u003c/p\u003e\u003cp\u003e Each institution\u0026rsquo;s Institutional Review Board approved the studies, and the need for informed consent was waived.\u003c/p\u003e\u003cp\u003eClinical variables and outcomes\u003c/p\u003e\u003cp\u003eWe used 52 variables from patient information at ICU admission for cluster analysis based on their availability across registries and their clinical importance: demographic variables (including age, sex, body-mass index, comorbidities, medication, and admission route), vital signs (including respiratory rate, heart rate, systolic and diastolic blood pressure, and body temperature), Glasgow Coma Scale score, parameters of inflammation (including white blood cell count and C-reactive protein), index parameters of organ failure (including bilirubin, creatinine, partial pressure of arterial oxygen, and partial pressure of arterial carbon dioxide), markers of coagulation (including platelet counts, international normalized ratio, activated partial thromboplastin time, and fibrinogen), sub-score of sepsis-related organ failure assessment (SOFA) score [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e], and serum levels of glucose, sodium, potassium, hematocrit, lactate, and potential hydrogen. A detailed description of variables is provided in Additional file 1: Tabel S1. The primary outcome was in-hospital mortality.\u003c/p\u003e\u003cp\u003eInterventions\u003c/p\u003e\u003cp\u003eWe investigated the effects of six treatments that were measured in the registries and which divided opinions on whether to do so according to the guidelines and previous studies [\u003cspan additionalcitationids=\"CR9\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. These treatments include corticosteroids, polymyxin B-immobilized fiber column direct hemoperfusion, recombinant thrombomodulin (rTM), antithrombin III (AT III), immunoglobulin G, and vasopressin.\u003c/p\u003e\u003cp\u003eStatistical methods\u003c/p\u003e\u003cp\u003eTo explore the subphenotypes of patients with sepsis, cluster analysis was performed using the variables of patient information at ICU admission. Among these variables, we eliminated one with a missing rate above 0.5 and one correlated with others with coefficients above 0.8. Serum levels of procalcitonin, fibrinin/fibrinogen degradation products, and antithrombin III were excluded from the analysis. Multiple imputation with chained equation was used to account for missing data with the mice package (Additional file 1: Table S3). Log transformation and normalization were used for nonnormal data.\u003c/p\u003e\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eClustering methods\u003c/h2\u003e\u003cp\u003eWe applied different methods for subphenotype identification. We considered the k-prototype [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e] and KAymeans for MIxed LArge data (KAMILA) [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e] which were recommended in a previous study that compared the performance of several clustering methods for mixed data through simulation and real data analysis [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eWe also considered another approach, k-means [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e] on the Uniform Manifold Approximation and Projection for Dimension Reduction (UMAP) [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]: k-UMAP, which is analogous to principal component analysis (PCA) [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]. PCA is used for dimension reduction to reduce noise but cannot be applied to mixed data. In contrast, UMAP is also used for dimension reduction and can select appropriate metrics depending on the data. Therefore, we adapted UMAP to handle mixed data as numeric data by transforming the embeddings.\u003c/p\u003e\u003cp\u003eWe examined the performance of k-UMAP using the same simulation as in a previous study [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e] and confirmed its superiority for clustering datasets with many categorical variables (Additional file 1: Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). (Refer to Additional file 1: Clustering methods and Table \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e for further details on each clustering method).\u003c/p\u003e\u003cp\u003eTo compare the results of each clustering method, we use the adjusted rand indexes (ARIs) [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e], which measure the similarity between the clusters obtained from two different clustering methods.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eConsensus clustering\u003c/h3\u003e\n\u003cp\u003eWe evaluated the performance of the clustering methods using consensus clustering [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e] to determine the most stable method. In consensus clustering, stability or consensus is measured by counting and normalizing the number of times each sample pair is classified into the same cluster after each iteration. The consensus values can be interpreted as the degree of stability; if all consensus values are 1 or 0, clustering is perfectly stable. The consensus for each cluster was weighted and averaged the consensus for each pair and called the cluster consensus. The weights represent the ratios of the number of samples in a cluster. We adopted clusters based on a combination of consensus clustering outputs, a clear separation of the consensus matrix heat maps, characteristics of the consensus cumulative distribution function plots, and adequate pairwise consensus values between cluster members. These values were used to determine the number of clusters.\u003c/p\u003e\n\u003ch3\u003eCluster validation\u003c/h3\u003e\n\u003cp\u003eThe same clustering method was applied to the validation dataset. For cluster analysis, we used 44 variables from patient information at ICU admission that were available in the validation dataset and common to the derivation datasets, as well as their clinical importance.\u003c/p\u003e\n\u003ch3\u003eEvaluation of treatment effects in sepsis subphenotypes\u003c/h3\u003e\n\u003cp\u003eTo align the cohort studies in our dataset with the randomized clinical trial situation, which is the gold standard for estimating causal relationships and treatment effects, we estimated the propensity score (PS) [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e]. We corrected the imbalance of background factors between the treated and untreated groups using inverse probability weighting of the PS [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eMost patients in our dataset received multiple treatments, making it difficult to determine the effects of any single treatment. Thus, we considered the combined treatments as confounding factors and included them in estimating PS weights, as well as the variables used in clustering. Logistic regression was used to estimate PS weights. After adjusting for PS, we used a mixed model to estimate the treatment effects by subphenotype.\u003c/p\u003e\u003cp\u003eAmong the datasets, we conducted a sub-analysis of patients with septic shock, defined as those who received vasopressors. We assessed the differences in characteristics between clusters using analysis of variance for continuous variables and chi-square tests for categorical variables. Statistical analyses were performed using R software (version 4.1.1; R Foundation, Vienna, Austria) (Additional file 1: Table S3). All reported p-values were two-sided and considered significant if\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003ePatients in the dataset\u003c/h2\u003e\u003cp\u003eFrom the two multicenter registries, we excluded 43 patients who declined active treatment within 3 days of admission and only one patient with a history of acquired immunodeficiency syndrome; 1,756 patients were eligible for analysis. The mean age was 71.0 years, and 61.1% of patients were men. The in-hospital mortality rate was 22.1% (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Variable distributions, missingness, and correlations are shown in Additional file 1 (Additional file 1: Table S4, Figure \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e, and S3).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eCharacteristics of patients clustered using k-UMAP [nn\u0026thinsp;=\u0026thinsp;100, k\u0026thinsp;=\u0026thinsp;3]\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eALL (n\u0026thinsp;=\u0026thinsp;1,756)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colspan=\"3\" nameend=\"c5\" namest=\"c3\"\u003e\u003cp\u003eSubphenotypes\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e\u003cem\u003eP\u003c/em\u003e value\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003esubphenotype 1\u003c/p\u003e\u003cp\u003e(n\u0026thinsp;=\u0026thinsp;419)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003esubphenotype 2\u003c/p\u003e\u003cp\u003e(n\u0026thinsp;=\u0026thinsp;736)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003esubphenotype 3\u003c/p\u003e\u003cp\u003e(n\u0026thinsp;=\u0026thinsp;601)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge, median (IQR), yr\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e73 (64\u0026ndash;82)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e78 (68\u0026ndash;85)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e74 (64\u0026ndash;82)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e71 (62\u0026ndash;79)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMale sex, \u003cem\u003en\u003c/em\u003e (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1,073 (61.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8 (1.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e719 (97.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e346 (57.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBody mass index, median (IQR), kg/m\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e21.9 (19.1\u0026ndash;24.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e21.4 (18.5\u0026ndash;24.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e21.9 (19.1\u0026ndash;24.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e22.0 (19.4\u0026ndash;24.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.09\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAdmission from ER, \u003cem\u003en\u003c/em\u003e (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1,125 (64.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e394 (94.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e645 (87.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e86 (14.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePast Medical history, \u003cem\u003en\u003c/em\u003e (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAcute myocardial infarction\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e83 (4.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8 (1.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e51 (6.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e24 (4.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAcute Heart Failure\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e189 (10.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e38 (9.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e71 (9.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e80 (13.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.04\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePeripheral Arterial Disease\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e53 (3.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7 (1.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e28 (3.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e18 (3.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.13\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eStroke\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e219 (12.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e52 (12.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e118 (16.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e49 (8.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eChronic Lung Disease\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e101 (5.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e19 (4.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e58 (7.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e24 (4.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.005\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCollagen Disease\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e113 (6.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e9 (2.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e16 (2.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e88 (14.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePeptic Ulcer\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e57 (3.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e10 (2.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e36 (4.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e11 (1.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.004\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLiver Disease\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e90 (5.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e20 (4.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e33 (4.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e37 (6.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.36\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDiabetes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e442 (25.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e102 (24.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e190 (25.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e150 (25.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.85\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eChronic kidney disease\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e142 (8.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e30 (7.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e56 (7.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e56 (9.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.38\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMalignancy\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e242 (13.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e47 (11.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e98 (13.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e97 (16.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.07\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMetastatic Neoplasm\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e40 (2.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1 (0.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e17 (2.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e22 (3.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.002\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMedication, \u003cem\u003en\u003c/em\u003e (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSteroid\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e213 (12.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e28 (6.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e44 (6.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e141 (23.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eImmunosuppressant\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e66 (3.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5 (1.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e8 (1.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e53 (8.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eStatin\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e187 (10.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e43 (10.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e90 (12.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e54 (9.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.15\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAntiplatelet\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e273 (15.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e43 (10.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e138 (18.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e92 (15.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.0006\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBeta Blocker\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e161 (9.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e27 (6.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e68 (9.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e66 (11.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.047\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRadiotherapy, \u003cem\u003en\u003c/em\u003e (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e7 (0.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0 (0.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e3 (0.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e4 (0.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.25\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eVital signs, median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGlasgow Coma Scale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e14 (10\u0026ndash;15)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e14 (9\u0026ndash;15)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e14 (9\u0026ndash;15)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e14 (11\u0026ndash;15)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.0004\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRespiratory rate, breath/min\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e25 (20\u0026ndash;30)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e25 (21\u0026ndash;30)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e25 (21\u0026ndash;30)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e24 (20\u0026ndash;30)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.009\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHeart Rate, beats/min\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e109 (94\u0026ndash;124)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e107 (92\u0026ndash;124)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e110 (95\u0026ndash;126)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e108 (93\u0026ndash;124)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.25\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSystolic blood pressure, mmHg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e101 (82\u0026ndash;127)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e105 (84\u0026ndash;130)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e103 (83\u0026ndash;130)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e97 (80\u0026ndash;123)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.002\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDiastolic blood pressure, mmHg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e59 (47\u0026ndash;74)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e60 (46\u0026ndash;74)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e60 (49\u0026ndash;76)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e57 (45\u0026ndash;72)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.007\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBody Temperature, ℃\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e37.6 (36.7\u0026ndash;38.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e37.7 (36.5\u0026ndash;38.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e37.8 (36.8\u0026ndash;38.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e37.3(36.6\u0026ndash;38.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBlood test Data, median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWhite cell count, cells 10\u003csup\u003e3\u003c/sup\u003e/\u0026micro;L\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e11.4 (6.0-17.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e11.4 (6.418.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e11.5 (6.0-17.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e11.4 (5.9\u0026ndash;18.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.77\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHematocrit, %\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e34.0 (28.8\u0026ndash;39.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e34.1 (29.4\u0026ndash;39.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e35.8 (30.3\u0026ndash;40.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e32.4 (27.2\u0026ndash;37.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePlatelets count, platelets 10\u003csup\u003e4\u003c/sup\u003e/\u0026micro;L\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e14.9 (9.4\u0026ndash;22.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e15.9 (9.8\u0026ndash;23.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e15.4 (10.2\u0026ndash;22.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e13.4 (7.9\u0026ndash;21.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCreatinine, mg/dL\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.4 (0.9\u0026ndash;2.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.3 (0.7\u0026ndash;2.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1.4 (0.9\u0026ndash;2.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e1.5 (0.9\u0026ndash;2.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBilirubin, mg/dL\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.9 (0.6\u0026ndash;1.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.8 (0.5\u0026ndash;1.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.9 (0.6\u0026ndash;1.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.9 (0.6\u0026ndash;1.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.01\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGlucose, mg/dL\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e141 (109\u0026ndash;191)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e143 (110\u0026ndash;189)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e147 (116\u0026ndash;200)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e132 (103\u0026ndash;178)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSodium, mEq/L\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e137 (133\u0026ndash;140)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e137 (134\u0026ndash;141)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e137 (133\u0026ndash;140)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e136 (133\u0026ndash;140)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePotassium, mEq/L\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4.0 (3.6\u0026ndash;4.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3.9 (3.5\u0026ndash;4.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e4.1 (3.7\u0026ndash;4.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e4.0 (3.6\u0026ndash;4.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.004\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eC-reactive protein, mg/dL\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e14.8 (6.2\u0026ndash;24.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e12.8 (4.5\u0026ndash;23.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e14.1 (5.5\u0026ndash;24.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e16.8 (9.0-24.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePT-INR\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.2 (1.1\u0026ndash;1.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.20 (1.07\u0026ndash;1.40)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1.20 (1.10\u0026ndash;1.40)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e1.30 (1.11\u0026ndash;1.50)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAPTT, sec\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e35.7 (30.3\u0026ndash;44.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e34.2 (29.1\u0026ndash;42.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e34.6 (29.6\u0026ndash;41.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e38.6 (32.8\u0026ndash;48.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eD-dimer, \u0026micro;g/mL\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e6.7 (3.0-15.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7.0 (2.9\u0026ndash;16.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e5.5 (2.6\u0026ndash;13.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e8.4 (3.9\u0026ndash;18.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFibrinogen, mg/mL\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e439 (313\u0026ndash;578)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e443 (313\u0026ndash;577)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e456 (317\u0026ndash;597)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e424 (300\u0026ndash;557)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.02\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003epH\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e7.39 (7.31\u0026ndash;7.45)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7.40 (7.32\u0026ndash;7.46)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e7.40 (7.31\u0026ndash;7.45)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e7.39 (7.30\u0026ndash;7.45)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.09\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePaCO2, mmHg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e34.0 (28.5\u0026ndash;41.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e33.0 (28.0-39.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e33.9 (28.8\u0026ndash;41.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e35.0 (29.0\u0026ndash;42.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.03\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePaO2, mmHg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e87.5 (69.8\u0026ndash;123)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e87.5 (68.8-128.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e84.1 (68.0-119.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e90.9 (73.0-126.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.03\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBase excess, mEq/L\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-3.2 (-7.0-0.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e-3.0 (-6.6-0.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-3.0 (-7.1- -0.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-3.5 (-7.5-0.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.55\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLactate, mmol/L\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2.9 (1.9-5.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3.1 (2.0-5.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e3.1 (2.0-5.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e2.6 (1.6\u0026ndash;4.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSOFA score, median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e8 (5\u0026ndash;11)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8 (5\u0026ndash;11)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e8 (5\u0026ndash;11)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e9 (6\u0026ndash;12)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.02\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRespiration\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2 (1\u0026ndash;3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2 (1\u0026ndash;2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e2 (1\u0026ndash;3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e2 (1\u0026ndash;3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.02\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCoagulation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1 (0\u0026ndash;2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0 (0\u0026ndash;1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0 (0\u0026ndash;1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e1 (0\u0026ndash;2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLiver\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0 (0\u0026ndash;1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0 (0\u0026ndash;1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0 (0\u0026ndash;1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0 (0\u0026ndash;1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.15\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCardiovascular\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3 (0\u0026ndash;4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1 (0\u0026ndash;4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e3 (0\u0026ndash;4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e3 (0\u0026ndash;4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCentral Nervous System\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1 (0\u0026ndash;3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1 (1\u0026ndash;3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1 (1\u0026ndash;3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e1 (0\u0026ndash;2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRenal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1 (0\u0026ndash;3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1 (0\u0026ndash;3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1 (0\u0026ndash;3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e1 (0\u0026ndash;3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.03\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAPACHE II score, median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e22 (16\u0026ndash;28)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e21 (16\u0026ndash;27)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e21 (16\u0026ndash;29)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e22 (17\u0026ndash;29)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.05\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSIRS, median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3 (2\u0026ndash;4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3 (2\u0026ndash;4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e3 (2\u0026ndash;4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e3 (2\u0026ndash;3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.05\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eJAAM DIC score, median (IQR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3 (2\u0026ndash;5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3 (2\u0026ndash;5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e3 (2\u0026ndash;5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e4 (2\u0026ndash;6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eManagement, \u003cem\u003en\u003c/em\u003e (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNoradrenaline\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1,000 (57.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e220 (52.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e387 (52.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e393 (65.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eVasopressin\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e209 (11.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e46 (11.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e83 (11.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e80 (13.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.42\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCorticosteroids\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e464 (27.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e86 (21.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e182 (25.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e196 (33.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eThrombomodulin\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e360 (21.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e73 (17.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e127 (17.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e160 (27.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAntithrombin III\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e333 (19.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e62 (15.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e107 (15.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e164 (27.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eImmunoglobulin\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e332 (19.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e60 (14.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e103 (14.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e169 (28.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePMX-DHP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e141 (8.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e25 (6.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e59 (8.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e57 (9.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.12\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eOutcomes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIn-hospital death, \u003cem\u003en\u003c/em\u003e (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e379 (22.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e158 (18.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e158 (22.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e146 (24.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.04\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"6\" nameend=\"c6\" namest=\"c1\"\u003e\u003cp\u003e\u003cem\u003eAbbreviation\u003c/em\u003e: \u003cem\u003enn\u003c/em\u003e number of neighbors, \u003cem\u003eSD\u003c/em\u003e standard deviation, \u003cem\u003eIQR\u003c/em\u003e interquartile range, \u003cem\u003eER\u003c/em\u003e emergency room, \u003cem\u003eAPTT\u003c/em\u003e Activated partial thromboplastin time, \u003cem\u003eSOFA score\u003c/em\u003e Sequential Organ Failure Assessment score, \u003cem\u003eAPACHE II score\u003c/em\u003e Acute Physiology and Chronic Health Evaluation II score, \u003cem\u003eSIRS\u003c/em\u003e Systemic Inflammatory Response Syndrome, \u003cem\u003eJAAM DIC score\u003c/em\u003e Japanese Association for Acute Medicine Disseminated Intravascular Coagulation score, \u003cem\u003ePMX-DHP\u003c/em\u003e polymyxin B-immobilized fiber column hemoperfusion\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eCluster analysis\u003c/h3\u003e\n\u003cp\u003eFigure \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e illustrates the ARI. In any number of subphenotypes, the ARIs of k-UMAP and k-prototype with high lambda tended to be high, and those of the KAMILA and k-prototype with low lambda tended to be high. A detailed description of each number of subphenotypes is provided in Additional file 1: Figure S4-9.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e shows the cluster consensus. We observed that the ranking of candidate clusters by cluster consensus was reversed when k was greater than 3. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e shows that fewer clusters tended to be more robust for all clustering methods, particularly for KAMILA and the k-prototype with lambda\u0026thinsp;=\u0026thinsp;1 and 0.01; however, we intended to explore as many characteristics as possible. Therefore, we selected the following three results of clustering conditions considering consensus and redundancy: two results of k-UMAP with more robustness (k-UMAP [number of neighbors (nn)\u0026thinsp;=\u0026thinsp;100, k\u0026thinsp;=\u0026thinsp;3], and k-UMAP [nn\u0026thinsp;=\u0026thinsp;200, k\u0026thinsp;=\u0026thinsp;5]) and one result of KAMILA (KAMILA [k\u0026thinsp;=\u0026thinsp;3]) because the KAMILA and k-prototype with low lambda had almost the same results of clustering as those of ARIs.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eAdditional file 1: Figure S13 illustrates the transitions between k-UMAP [nn\u0026thinsp;=\u0026thinsp;100, k\u0026thinsp;=\u0026thinsp;3] and k-UMAP [nn\u0026thinsp;=\u0026thinsp;200, k\u0026thinsp;=\u0026thinsp;5]. This figure indicates that k-UMAP [nn\u0026thinsp;=\u0026thinsp;200, k\u0026thinsp;=\u0026thinsp;5] is a divided population of k-UMAP [nn\u0026thinsp;=\u0026thinsp;100, k\u0026thinsp;=\u0026thinsp;3].\u003c/p\u003e\n\u003ch3\u003eCharacteristics of sepsis subphenotypes in each cluster analysis\u003c/h3\u003e\n\u003cp\u003eIn the k-UMAP [nn\u0026thinsp;=\u0026thinsp;100 k\u0026thinsp;=\u0026thinsp;3], patients with subphenotype 3 were likely to have high in-hospital mortality (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), but the severity of organ dysfunction was not significantly different among each subphenotype. The percentages of sex, admission route, steroid medication, and immunosuppressant use were particularly distinctive for each subphenotype. Subphenotype 1 included a lower percentage of men and a higher percentage of admission routes from the emergency room. Subphenotype 3 had a lower percentage of admission routes from the emergency room, a higher percentage of a medical history of collagen disease, and medication histories of steroids and immunosuppressants. In the k-UMAP [nn\u0026thinsp;=\u0026thinsp;200, k\u0026thinsp;=\u0026thinsp;5], patients with subphenotypes 3\u0026ndash;5 likely had high in-hospital mortality (Additional file 1: Table S5). However, the severity of organ dysfunction differed less among each subphenotype, similar to the k-UMAP [nn\u0026thinsp;=\u0026thinsp;100 k\u0026thinsp;=\u0026thinsp;3]. In KAMILA [k\u0026thinsp;=\u0026thinsp;3], patients with subphenotypes 2 and 3 likely had a high SOFA score, lactate levels, and in-hospital mortality (Additional file 1: Table S6).\u003c/p\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003eCluster analysis in the validation dataset\u003c/h2\u003e\u003cp\u003eAdditional file 1: Figure S8 and Table S7 show the clustering results and characteristics of the sepsis subphenotypes for k-UMAP [nn\u0026thinsp;=\u0026thinsp;100, k\u0026thinsp;=\u0026thinsp;3] in the validation dataset. The in-hospital mortality, SOFA, and Acute Physiology and Chronic Health Evaluation II scores, characteristics, admission route, medical histories, and blood test data for each subphenotype were similar to those in the derivation dataset. The clustering results of the derivation and validation data in k-UMAP are plotted in UMAP (Additional file 1: Figure S9), and a similar distribution was obtained.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003eTreatment effects in sepsis subphenotypes\u003c/h2\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e shows the odds ratios (ORs) for the interaction between treatment and subphenotypes for in-hospital mortality in each clustering method. In the k-UMAP [nn\u0026thinsp;=\u0026thinsp;100 k\u0026thinsp;=\u0026thinsp;3], rTM was significantly effective for sub phenotype 1 (OR 0.37, 95% Confidence interval [CI] 0.13\u0026ndash;0.80). In the k-UMAP [nn\u0026thinsp;=\u0026thinsp;200 k\u0026thinsp;=\u0026thinsp;5], rTM was significantly effective for subphenotype 1 (OR 0.30, 95% CI 0.09\u0026ndash;0.70) and AT III was significantly effective for subphenotype 2 (OR 0.49, 95% CI 0.20\u0026ndash;0.92). In contrast, no effective treatment was available for the subphenotypes in KAMILA [k\u0026thinsp;=\u0026thinsp;3].\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eAdditional file 1: Figure S10 shows the same odds for patients with septic shock and the overall trend was similar to all patients. AT Ⅲ was significantly effective for subphenotype 2 in the k-UMAP [nn\u0026thinsp;=\u0026thinsp;100, k\u0026thinsp;=\u0026thinsp;3] (OR 0.43, 95% CI 0.19\u0026ndash;0.80), subphenotype 2 in the k-UMAP [nn\u0026thinsp;=\u0026thinsp;200, k\u0026thinsp;=\u0026thinsp;5] (OR 0.41, 95% CI 0.16\u0026ndash;0.84) and subphenotype 1 in the KAMILA [k\u0026thinsp;=\u0026thinsp;3] (OR 0.31, 95% CI 0.09\u0026ndash;0.73). rTM was significantly effective for subphenotype1 in the k-UMAP [nn\u0026thinsp;=\u0026thinsp;200, k\u0026thinsp;=\u0026thinsp;5] (OR 0.42, 95% CI 0.13\u0026ndash;0.99).\u003c/p\u003e\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study, we explored the subphenotypes of Japanese patients with sepsis enrolled in cohort studies. For this purpose, we applied the following state-of-the-art clustering methods to the mixed data: k-prototype, KAMILA, and k-UMAP. We identified three or five subphenotypes with various features, and only k-UMAP could detect these subphenotypes in Japanese patients with sepsis who benefited from some treatments.\u003c/p\u003e\u003cp\u003eAs conventional clustering methods, such as the original k-means method, have not been developed for mixed data, previous studies have only used continuous variables in their analyses [\u003cspan additionalcitationids=\"CR22\" citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. In this study, we adopted state-of-the-art clustering methods for mixed data to consider categorical data as features. In the case of mixed data, one of the problems is the lack of consensus on determining the weight between the variables measured using different metrics, such as Euclidean or Hamming. Therefore, we indirectly investigated the clustering method for the weight balance between different metrics by comparing the k-prototype with several values of the lambda parameter using other clustering methods.\u003c/p\u003e\u003cp\u003eAccording to the ARI results, patients were classified into close subphenotypes when using k-UMAP and k-prototype with high lambda. As mentioned in Additional file 1, we used the same metrics for the k-prototype and k-UMAP. This metric indicates that k-UMAP assigned a larger weight to the categorical data than the k-prototype with a low lambda. In contrast, KAMILA and the k-prototype with low lambda values were classified as close subphenotypes. Thus, KAMILA tends to assign larger weights to numerical data than k-UMAP.\u003c/p\u003e\u003cp\u003eUMAP has been popular as an effective visualization method in machine learning and medicine and natural language processing, where improved clustering methods have been reported [\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e]. However, to our knowledge, the performance of k-UMAP has not been investigated for mixed datasets. Therefore, we examined this using the same simulation as that used in a previous study [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. The simulation indicated that the characteristics of k-UMAP were stable for a dataset heavily containing categorical data. Thus, k-UMAP is efficient when categorical data is clinically important. Otherwise, the k-UMAP algorithm differs from other partitioning clustering methods owing to the UMAP; therefore, k-UMAP can be used with other clustering methods to explore a wide range of patient characteristics.\u003c/p\u003e\u003cp\u003eBy comparing the subphenotypes in previous studies with those of k-UMAP, subphenotypes characterized by high mortality, cardiovascular failure, and coagulopathy were identified (Additional file 1: Table S8) [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e]. Moreover, k-UMAP could be characterized better than the subphenotypes in previous studies because it could reflect differences in elements of categorical variables, such as sex, background disease, medication history, and admission route. The three subphenotypes of k-UMAP were characterized as follows: subphenotype 1 was the older, female, emergency room admission type with the lowest in-hospital mortality; subphenotype 2 was the male, lifestyle-related disease type, such as cardiovascular disease, chronic lung disease, and peptic ulcer; and subphenotype 3 was the younger, immunocompromised type with the highest in-hospital mortality.\u003c/p\u003e\u003cp\u003eAnother contribution of this study is the evaluation of treatment effects in patients who received several treatments for sepsis. For subphenotype classification to be used in clinical practice, it should be impactful, such as recognizing differential treatment responses [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Generally, patients with sepsis are treated with multiple adjunct therapies depending on their medical condition. However, no studies have investigated the effects of these multiple therapies across subphenotypes. This study showed that the effects of multiple treatments varied based on the subphenotype. By examining treatment effects using the same subphenotype, the effects of each treatment can be compared among subphenotypes. A predictive model for subphenotypes could contribute to identify specific sepsis patients, and a novel treatment flow may be provided. In addition, this approach will be an important progress to individualized treatment strategies for sepsis.\u003c/p\u003e\u003cp\u003eThis study has several strengths. We used a reliable sepsis dataset with many variables and a few missing values. We investigated various clustering methods and compared their results. We adjusted covariate effects to exclude reverse causality.\u003c/p\u003e\u003cp\u003eThis study has some limitations. First, the statistical power was insufficient because of the small sample size for each subphenotype. Second, cluster analysis results may depend on a combination of variables. Third, the dataset excluded the timing of treatments; therefore, we could not precisely model the interactions between treatments. Therefore, future studies should use covariates as time-series data to build models that reflect multiple treatment effects and progress. Finally, because the datasets and patients were all Japanese, this study may have been influenced by the ethnic characteristics of the Japanese population and the trend toward more treatment in Japanese hospitals. Future studies in other populations and cultures are required to identify the global subphenotypes associated with treatment effects.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn this retrospective analysis of datasets from patients with sepsis, three state-of-the-art methods were applied to cluster analysis for mixed data, and three to five subphenotypes were identified to be associated with clinical information and outcomes. The effects of treatments differed for each subphenotype, and k-UMAP detected the subphenotypes of Japanese patients with sepsis who benefited from some treatments. Identifying patients for whom treatment is more effective can lead to precision medicine in critical care, and k-UMAP may be a good option for clustering methods using mixed clinical data.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eintensive care units (ICUs) , The Focused Outcomes Research in Emergency Care for Acute Respiratory Distress Syndrome, Sepsis, and Trauma (FORECAST), Japanese Association for Acute Medicine Sepsis Prognostication in the ICU and Emergency Room (JAAM SPICE-ICU), Japanese Association for Acute Medicine Multicenter Assessment for Sepsis Treatment and Outcome (JAAM MAESTRO), sepsis-related organ failure assessment (SOFA), recombinant thrombomodulin (rTM), antithrombin III (AT III), KAymeans for MIxed LArge data (KAMILA), Uniform Manifold Approximation and Projection for Dimension Reduction (UMAP), principal component analysis (PCA), adjusted rand indexes (ARIs), propensity score (PS), number of neighbors (nn), odds ratios (ORs)\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTwo original studies were approved, and the need for informed consent was waived by the Institutional Review Boards of the participating hospitals.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets generated and/or analyzed during the original studies are available at Mendeley Data, https://data.mendeley.com/datasets/vvv89kw3k5/1 (Tohoku Sepsis Registry). The datasets of the FORECAST sepsis study, JAAM SPICE-ICU, and JAAM MAESTRO are restrictions apply to the availability of the data, which were used under license for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission of the Japanese Association for Acute Medicine.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no conflicts of interest with respect to this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work (writing of the manuscript) was supported by departmental funds.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eYY, RS, GT, and SK designed the study. YY, RS, and GT wrote the analysis plan, and YY and RS analyzed the data. YY, RS, DK, and SK drafted the initial manuscript. All authors critically revised the manuscript. The corresponding author attests that all listed authors meet the authorship criteria and that no others meeting the criteria have been omitted.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe thank all investigators involved in the Tohoku Sepsis Registry, FORECAST Sepsis Study, JAAM SPICE-ICU, and JAAM MAESTRO for contributing to data collection and assessment. We thank Hikaru Matsuoka for technical advice. This work was supported by the MEXT/JSPS WISE Program and the Advanced Graduate Program for Future Medicine and Health Care at Tohoku University.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eFleischmann-Struzek C, Mellhammar L, Rose N, Cassini A, Rudd KE, Schlattmann P, Allegranzi B, Reinhart K: \u003cstrong\u003eIncidence and mortality of hospital- and ICU-treated sepsis: results from an updated and expanded systematic review and meta-analysis\u003c/strong\u003e. \u003cem\u003eIntensive Care Med \u003c/em\u003e2020, \u003cstrong\u003e46\u003c/strong\u003e:1552-1562.\u003c/li\u003e\n\u003cli\u003eEgi M, Ogura H, Yatabe T, Atagi K, Inoue S, Iba T, Kakihana Y, Kawasaki T, Kushimoto S, Kuroda Y\u003cem\u003e et al\u003c/em\u003e: \u003cstrong\u003eThe Japanese Clinical Practice Guidelines for Management of Sepsis and Septic Shock 2020 (J-SSCG 2020)\u003c/strong\u003e. \u003cem\u003eJ Intensive Care \u003c/em\u003e2021, \u003cstrong\u003e9\u003c/strong\u003e:53.\u003c/li\u003e\n\u003cli\u003eEvans L, Rhodes A, Alhazzani W, Antonelli M, Coopersmith CM, French C, Machado FR, McIntyre L, Ostermann M, Prescott HC\u003cem\u003e et al\u003c/em\u003e: \u003cstrong\u003eSurviving sepsis campaign: international guidelines for management of sepsis and septic shock 2021\u003c/strong\u003e. \u003cem\u003eIntensive Care Med \u003c/em\u003e2021, \u003cstrong\u003e47\u003c/strong\u003e:1181-1247.\u003c/li\u003e\n\u003cli\u003eInvestigators P, Rowan KM, Angus DC, Bailey M, Barnato AE, Bellomo R, Canter RR, Coats TJ, Delaney A, Gimbel E\u003cem\u003e et al\u003c/em\u003e: \u003cstrong\u003eEarly, Goal-Directed Therapy for Septic Shock - A Patient-Level Meta-Analysis\u003c/strong\u003e. \u003cem\u003eN Engl J Med \u003c/em\u003e2017, \u003cstrong\u003e376\u003c/strong\u003e:2223-2234.\u003c/li\u003e\n\u003cli\u003eSeymour CW, Gesten F, Prescott HC, Friedrich ME, Iwashyna TJ, Phillips GS, Lemeshow S, Osborn T, Terry KM, Levy MM: \u003cstrong\u003eTime to Treatment and Mortality during Mandated Emergency Care for Sepsis\u003c/strong\u003e. \u003cem\u003eN Engl J Med \u003c/em\u003e2017, \u003cstrong\u003e376\u003c/strong\u003e:2235-2244.\u003c/li\u003e\n\u003cli\u003ePeltan ID, Brown SM, Bledsoe JR, Sorensen J, Samore MH, Allen TL, Hough CL: \u003cstrong\u003eED Door-to-Antibiotic Time and Long-term Mortality in Sepsis\u003c/strong\u003e. \u003cem\u003eChest \u003c/em\u003e2019, \u003cstrong\u003e155\u003c/strong\u003e:938-946.\u003c/li\u003e\n\u003cli\u003eAvni T, Lador A, Lev S, Leibovici L, Paul M, Grossman A: \u003cstrong\u003eVasopressors for the Treatment of Septic Shock: Systematic Review and Meta-Analysis\u003c/strong\u003e. \u003cem\u003ePLoS One \u003c/em\u003e2015, \u003cstrong\u003e10\u003c/strong\u003e:e0129305.\u003c/li\u003e\n\u003cli\u003eIizuka Y, Sanui M, Sasabuchi Y, Lefor AK, Hayakawa M, Saito S, Uchino S, Yamakawa K, Kudo D, Takimoto K\u003cem\u003e et al\u003c/em\u003e: \u003cstrong\u003eLow-dose immunoglobulin G is not associated with mortality in patients with sepsis and septic shock\u003c/strong\u003e. \u003cem\u003eCrit Care \u003c/em\u003e2017, \u003cstrong\u003e21\u003c/strong\u003e:181.\u003c/li\u003e\n\u003cli\u003eRygard SL, Butler E, Granholm A, Moller MH, Cohen J, Finfer S, Perner A, Myburgh J, Venkatesh B, Delaney A: \u003cstrong\u003eLow-dose corticosteroids for adult patients with septic shock: a systematic review with meta-analysis and trial sequential analysis\u003c/strong\u003e. \u003cem\u003eIntensive Care Med \u003c/em\u003e2018, \u003cstrong\u003e44\u003c/strong\u003e:1003-1016.\u003c/li\u003e\n\u003cli\u003eVincent JL, Francois B, Zabolotskikh I, Daga MK, Lascarrou JB, Kirov MY, Pettila V, Wittebole X, Meziani F, Mercier E\u003cem\u003e et al\u003c/em\u003e: \u003cstrong\u003eEffect of a Recombinant Human Soluble Thrombomodulin on Mortality in Patients With Sepsis-Associated Coagulopathy: The SCARLET Randomized Clinical Trial\u003c/strong\u003e. \u003cem\u003eJAMA \u003c/em\u003e2019, \u003cstrong\u003e321\u003c/strong\u003e:1993-2002.\u003c/li\u003e\n\u003cli\u003eSantacruz CA, Pereira AJ, Celis E, Vincent JL: \u003cstrong\u003eWhich Multicenter Randomized Controlled Trials in Critical Care Medicine Have Shown Reduced Mortality? A Systematic Review\u003c/strong\u003e. \u003cem\u003eCrit Care Med \u003c/em\u003e2019, \u003cstrong\u003e47\u003c/strong\u003e:1680-1691.\u003c/li\u003e\n\u003cli\u003eFujii T, Luethi N, Young PJ, Frei DR, Eastwood GM, French CJ, Deane AM, Shehabi Y, Hajjar LA, Oliveira G\u003cem\u003e et al\u003c/em\u003e: \u003cstrong\u003eEffect of Vitamin C, Hydrocortisone, and Thiamine vs Hydrocortisone Alone on Time Alive and Free of Vasopressor Support Among Patients With Septic Shock: The VITAMINS Randomized Clinical Trial\u003c/strong\u003e. \u003cem\u003eJAMA \u003c/em\u003e2020, \u003cstrong\u003e323\u003c/strong\u003e:423-431.\u003c/li\u003e\n\u003cli\u003eDellinger RP, Bagshaw SM, Antonelli M, Foster DM, Klein DJ, Marshall JC, Palevsky PM, Weisberg LS, Schorr CA, Trzeciak S\u003cem\u003e et al\u003c/em\u003e: \u003cstrong\u003eEffect of Targeted Polymyxin B Hemoperfusion on 28-Day Mortality in Patients With Septic Shock and Elevated Endotoxin Level: The EUPHRATES Randomized Clinical Trial\u003c/strong\u003e. \u003cem\u003eJAMA \u003c/em\u003e2018, \u003cstrong\u003e320\u003c/strong\u003e:1455-1463.\u003c/li\u003e\n\u003cli\u003eJames A. Russell MD, Keith R. Walley, M.D., Joel Singer, Ph.D., Anthony C. Gordon, M.B., B.S., M.D.,, Paul C. H\u0026eacute;bert MD, D. James Cooper, B.M., B.S., M.D., Cheryl L. Holmes, M.D., Sangeeta Mehta, M.D.,, John T. Granton MD, Michelle M. Storms, B.Sc.N., Deborah J. Cook, M.D., Jeffrey J. Presneill, M.B., B.S., Ph.D.,, and Dieter Ayers MS, for the VASST Investigator: \u003cstrong\u003eVasopressin versus Norepinephrine Infusion in Patients with Septic Shock.pdf\u003c/strong\u003e. \u003cem\u003eN Engl J Med \u003c/em\u003e2008, \u003cstrong\u003e358\u003c/strong\u003e:877-887.\u003c/li\u003e\n\u003cli\u003eReddy K, Sinha P, O\u0026apos;Kane CM, Gordon AC, Calfee CS, McAuley DF: \u003cstrong\u003eSubphenotypes in critical care: translation into clinical practice\u003c/strong\u003e. \u003cem\u003eLancet Respir Med \u003c/em\u003e2020, \u003cstrong\u003e8\u003c/strong\u003e:631-643.\u003c/li\u003e\n\u003cli\u003eMaslove DM, Tang B, Shankar-Hari M, Lawler PR, Angus DC, Baillie JK, Baron RM, Bauer M, Buchman TG, Calfee CS\u003cem\u003e et al\u003c/em\u003e: \u003cstrong\u003eRedefining critical illness\u003c/strong\u003e. \u003cem\u003eNat Med \u003c/em\u003e2022, \u003cstrong\u003e28\u003c/strong\u003e:1141-1148.\u003c/li\u003e\n\u003cli\u003eWang W, Liu CF: \u003cstrong\u003eSepsis heterogeneity\u003c/strong\u003e. \u003cem\u003eWorld J Pediatr \u003c/em\u003e2023, \u003cstrong\u003e19\u003c/strong\u003e:919-927.\u003c/li\u003e\n\u003cli\u003eHarhay MO, Casey JD, Clement M, Collins SP, Gayat E, Gong MN, Jaber S, Laterre PF, Marshall JC, Matthay MA\u003cem\u003e et al\u003c/em\u003e: \u003cstrong\u003eContemporary strategies to improve clinical trial design for critical care research: insights from the First Critical Care Clinical Trialists Workshop\u003c/strong\u003e. \u003cem\u003eIntensive Care Med \u003c/em\u003e2020, \u003cstrong\u003e46\u003c/strong\u003e:930-942.\u003c/li\u003e\n\u003cli\u003eAnthony C. Gordon NA-L, Lieuwe D. Bos, Jose Dianti, Janet Victoria Diaz, Tomoko Fujii, Evangelos J. Giamarellos-Bourboulis, Jean-Louis Vincent, Carolyn S. Calfee: \u003cstrong\u003eFrom ICU Syndromes to ICU Subphenotypes: Consensus Report and Recommendations For Developing Precision Medicine in ICU\u003c/strong\u003e. \u003cem\u003eAm J Respir Crit Care Med \u003c/em\u003e2024, \u003cstrong\u003eOnline ahead of print.\u003c/strong\u003e\u003c/li\u003e\n\u003cli\u003eKnox DB, Lanspa MJ, Kuttler KG, Brewer SC, Brown SM: \u003cstrong\u003ePhenotypic clusters within sepsis-associated multiple organ dysfunction syndrome\u003c/strong\u003e. \u003cem\u003eIntensive Care Med \u003c/em\u003e2015, \u003cstrong\u003e41\u003c/strong\u003e:814-822.\u003c/li\u003e\n\u003cli\u003eSeymour CW, Kennedy JN, Wang S, Chang CH, Elliott CF, Xu Z, Berry S, Clermont G, Cooper G, Gomez H\u003cem\u003e et al\u003c/em\u003e: \u003cstrong\u003eDerivation, Validation, and Potential Treatment Implications of Novel Clinical Phenotypes for Sepsis\u003c/strong\u003e. \u003cem\u003eJAMA \u003c/em\u003e2019, \u003cstrong\u003e321\u003c/strong\u003e:2003-2017.\u003c/li\u003e\n\u003cli\u003eKudo D, Goto T, Uchimido R, Hayakawa M, Yamakawa K, Abe T, Shiraishi A, Kushimoto S: \u003cstrong\u003eCoagulation phenotypes in sepsis and effects of recombinant human thrombomodulin: an analysis of three multicentre observational studies\u003c/strong\u003e. \u003cem\u003eCrit Care \u003c/em\u003e2021, \u003cstrong\u003e25\u003c/strong\u003e:114.\u003c/li\u003e\n\u003cli\u003eBhavani SV, Semler M, Qian ET, Verhoef PA, Robichaux C, Churpek MM, Coopersmith CM: \u003cstrong\u003eDevelopment and validation of novel sepsis subphenotypes using trajectories of vital signs\u003c/strong\u003e. \u003cem\u003eIntensive Care Med \u003c/em\u003e2022, \u003cstrong\u003e48\u003c/strong\u003e:1582-1592.\u003c/li\u003e\n\u003cli\u003eZhang Z, Chen L, Liu X, Yang J, Huang J, Yang Q, Hu Q, Jin K, Celi LA, Hong Y: \u003cstrong\u003eExploring disease axes as an alternative to distinct clusters for characterizing sepsis heterogeneity\u003c/strong\u003e. \u003cem\u003eIntensive Care Med \u003c/em\u003e2023, \u003cstrong\u003e49\u003c/strong\u003e:1349-1359.\u003c/li\u003e\n\u003cli\u003eAhmad A, Khan SS: \u003cstrong\u003eSurvey of State-of-the-Art Mixed Data Clustering Algorithms\u003c/strong\u003e. \u003cem\u003eIEEE Access \u003c/em\u003e2019, \u003cstrong\u003e7\u003c/strong\u003e:31883-31902.\u003c/li\u003e\n\u003cli\u003eDellinger RP, Levy MM, Rhodes A, Annane D, Gerlach H, Opal SM, Sevransky JE, Sprung CL, Douglas IS, Jaeschke R\u003cem\u003e et al\u003c/em\u003e: \u003cstrong\u003eSurviving sepsis campaign: international guidelines for management of severe sepsis and septic shock: 2012\u003c/strong\u003e. \u003cem\u003eCrit Care Med \u003c/em\u003e2013, \u003cstrong\u003e41\u003c/strong\u003e:580-637.\u003c/li\u003e\n\u003cli\u003eManu Shankar-Hari M, MSc; Gary S. Phillips, MAS; Mitchell L. Levy, MD; ChristopherW. Seymour, MD, MSc; Vincent X. Liu, MD, MSc; Clifford S. Deutschman, MD; Derek C. Angus, MD, MPh; Gordon D. Rubenfeld, MD, MSc; Mervyn Singer, MD, FRCP; for the Sepsis Definitions Task Force: \u003cstrong\u003eDeveloping a NewDefinition and Assessing New Clinical Criteria for Septic Shock For the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3)\u003c/strong\u003e. \u003cem\u003eJAMA \u003c/em\u003e2016, \u003cstrong\u003e315\u003c/strong\u003e:775-787.\u003c/li\u003e\n\u003cli\u003eKudo D, Kushimoto S, Miyagawa N, Sato T, Hasegawa M, Ito F, Yamanouchi S, Honda H, Andoh K, Furukawa H\u003cem\u003e et al\u003c/em\u003e: \u003cstrong\u003eThe impact of organ dysfunctions on mortality in patients with severe sepsis: A multicenter prospective observational study\u003c/strong\u003e. \u003cem\u003eJ Crit Care \u003c/em\u003e2018, \u003cstrong\u003e45\u003c/strong\u003e:178-183.\u003c/li\u003e\n\u003cli\u003eAbe T, Ogura H, Shiraishi A, Kushimoto S, Saitoh D, Fujishima S, Mayumi T, Shiino Y, Nakada T-a, Tarui T\u003cem\u003e et al\u003c/em\u003e: \u003cstrong\u003eCharacteristics, management, and in-hospital mortality among patients with severe sepsis in intensive care units in Japan: the FORECAST study\u003c/strong\u003e. \u003cem\u003eCritical Care \u003c/em\u003e2018, 22:322.\u003c/li\u003e\n\u003cli\u003eAbe T, Yamakawa K, Ogura H, Kushimoto S, Saitoh D, Fujishima S, Otomo Y, Kotani J, Umemura Y, Sakamoto Y\u003cem\u003e et al\u003c/em\u003e: \u003cstrong\u003eEpidemiology of sepsis and septic shock in intensive care units between sepsis-2 and sepsis-3 populations: sepsis prognostication in intensive care unit and emergency room (SPICE-ICU)\u003c/strong\u003e. \u003cem\u003eJ Intensive Care \u003c/em\u003e2020, 8:44.\u003c/li\u003e\n\u003cli\u003eAbe T, Umemura Y, Ogura H, Kushimoto S, Fujishima S, Saitoh D, Gando S: \u003cstrong\u003eEarly versus delayed vasopressor administration in patients with septic shock\u003c/strong\u003e. \u003cem\u003eAcute Med Surg \u003c/em\u003e2023, \u003cstrong\u003e10\u003c/strong\u003e:e852.\u003c/li\u003e\n\u003cli\u003eJ L Vincent RM, J Takala, S Willatts, A De Mendon\u0026ccedil;a, H Bruining, C K Reinhart, P M Suter, L G Thijs: \u003cstrong\u003eThe SOFA (Sepsis.related Organ Failure Assessment) score to describe organ dysfunction/failure\u003c/strong\u003e. \u003cem\u003eIntensive Care Med \u003c/em\u003e1996, \u003cstrong\u003e22\u003c/strong\u003e:707-710.\u003c/li\u003e\n\u003cli\u003eHUANG Z: \u003cstrong\u003eExtensions to the k-Means Algorithm for Clustering Large DataSets with Categorical Values.pdf\u0026gt;\u003c/strong\u003e. \u003cem\u003eData Mining and Knowledge Discovery \u003c/em\u003e1998, \u003cstrong\u003e2\u003c/strong\u003e:283\u0026ndash;304.\u003c/li\u003e\n\u003cli\u003eFoss A, Markatou M, Ray B, Heching A: \u003cstrong\u003eA semiparametric method for clustering mixed data\u003c/strong\u003e. \u003cem\u003eMachine Learning \u003c/em\u003e2016, \u003cstrong\u003e105\u003c/strong\u003e:419-458.\u003c/li\u003e\n\u003cli\u003ePreud\u0026apos;homme G, Duarte K, Dalleau K, Lacomblez C, Bresso E, Smail-Tabbone M, Couceiro M, Devignes MD, Kobayashi M, Huttin O\u003cem\u003e et al\u003c/em\u003e: \u003cstrong\u003eHead-to-head comparison of clustering methods for heterogeneous data: a simulation-driven benchmark\u003c/strong\u003e. \u003cem\u003eSci Rep \u003c/em\u003e2021, \u003cstrong\u003e11\u003c/strong\u003e:4202.\u003c/li\u003e\n\u003cli\u003eMACQUEEN J: \u003cstrong\u003eSOME METHODS FOR CLASSIFICATION AND ANALYSIS OF MULTIVARIATE OBSERVATIONS\u003c/strong\u003e. \u003cem\u003eBerkeley Symp on Math Statist and Prob \u003c/em\u003e1967:281-297.\u003c/li\u003e\n\u003cli\u003eLeland McInnes JH, James Melville: \u003cstrong\u003eUMAP Uniform Manifold Approximation and Projection for Dimension Reduction\u003c/strong\u003e. \u003cem\u003earXiv \u003c/em\u003e2020(1802).\u003c/li\u003e\n\u003cli\u003eChris Ding XH: \u003cstrong\u003eAdvances in Knowledge Discovery and Data Mining\u003c/strong\u003e, vol. 3056. Sydney, Australia: 8th Pacific-Asia Conference; 2004.\u003c/li\u003e\n\u003cli\u003eHubert L: \u003cstrong\u003eComparing Partitions\u003c/strong\u003e. \u003cem\u003eJournal of Classification \u003c/em\u003e1985, \u003cstrong\u003e2\u003c/strong\u003e:193-218.\u003c/li\u003e\n\u003cli\u003eSTEFANO MONTI PT, JILL MESIROV, TODD GOLUB: \u003cstrong\u003eConsensus Clustering A Resampling-Based Method for Class Discovery and Visualization of Gene Expression Microarray Data.pdf\u003c/strong\u003e. \u003cem\u003eMachine Learning \u003c/em\u003e2003, \u003cstrong\u003e52\u003c/strong\u003e:91\u0026ndash;118.\u003c/li\u003e\n\u003cli\u003ePAUL R. ROSENBAUM DBR: \u003cstrong\u003eThe central role of the propensity score in observational studies for causal effects\u003c/strong\u003e. \u003cem\u003eBiometrika \u003c/em\u003e1983, \u003cstrong\u003e70\u003c/strong\u003e: 41\u0026ndash;55.\u003c/li\u003e\n\u003cli\u003eAustin PC: \u003cstrong\u003eAn Introduction to Propensity Score Methods for Reducing the Effects of Confounding in Observational Studies\u003c/strong\u003e. \u003cem\u003eMultivariate Behav Res \u003c/em\u003e2011, \u003cstrong\u003e46\u003c/strong\u003e:399-424.\u003c/li\u003e\n\u003cli\u003eAllaoui M, Kherfi ML, Cheriet A: \u003cstrong\u003eConsiderably Improving Clustering Algorithms Using UMAP Dimensionality Reduction Technique: A Comparative Study\u003c/strong\u003e. In: \u003cem\u003eImage and Signal Processing.\u003c/em\u003e edn.; 2020: 317-325.\u003c/li\u003e\n\u003cli\u003eSoussi S, Sharma D, Juni P, Lebovic G, Brochard L, Marshall JC, Lawler PR, Herridge M, Ferguson N, Del Sorbo L\u003cem\u003e et al\u003c/em\u003e: \u003cstrong\u003eIdentifying clinical subtypes in sepsis-survivors with different one-year outcomes: a secondary latent class analysis of the FROG-ICU cohort\u003c/strong\u003e. \u003cem\u003eCrit Care \u003c/em\u003e2022, \u003cstrong\u003e26\u003c/strong\u003e:114.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"critical-care","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"cric","sideBox":"Learn more about [Critical Care](http://ccforum.biomedcentral.com/)","snPcode":"13054","submissionUrl":"https://submission.nature.com/new-submission/13054/3","title":"Critical Care","twitterHandle":"@Crit_Care","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Sepsis, subphenotype, mixed data, cluster analysis, treatment effects","lastPublishedDoi":"10.21203/rs.3.rs-7716199/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7716199/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e\u003cp\u003eSepsis is a heterogeneous syndrome, and the treatment effects may vary depending on its subphenotype. Previous studies have not fully used mixed clinical data, nor have investigated the effects of these multiple treatments across subphenotypes. In this study, we aimed to classify patients with sepsis into subphenotypes based on mixed clinical data and examined the differences in treatment effectiveness by subphenotype.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eThis study was a secondary analysis of multicenter registries that enrolled patients with sepsis admitted to intensive care units in Japan. The patients aged 16 years or older admitted to the ICU due to a diagnosis of sepsis were included and fifty-two variables at admission were used in the cluster analysis. We applied the state-of-the-art clustering method named k-UMAP, which uses uniform manifold approximation and projection for dimensionality reduction, followed by clustering using k-means and the previous clustering methods k-prototype and KAMILA. To examine differences in the effectiveness of the six treatments by subphenotype, a logistic regression model was used for propensity score-based weighted data to calculate the odds ratios for the interaction of the treatments and subphenotype in each clustering method. The primary outcome was the in-hospital mortality rate.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eThe analysis included 1,756 patients. The results of three clustering methods using mixed clinical data led to the three conditions with high robustness being selected: k-UMAP [number of clusters (k)\u0026thinsp;=\u0026thinsp;3], k-UMAP [k\u0026thinsp;=\u0026thinsp;5], and KAMILA [k\u0026thinsp;=\u0026thinsp;3]. Hospital mortality, patient characteristics, and treatment effectiveness varied by subphenotype. Recombinant thrombomodulin was significantly effective for subphenotype 1 in the k-UMAP [k\u0026thinsp;=\u0026thinsp;3] and in the k-UMAP [k\u0026thinsp;=\u0026thinsp;5]. Antithrombin III was significantly effective for subphenotype 2 in the k-UMAP [k\u0026thinsp;=\u0026thinsp;5]. On the other hand, in KAMILA [k\u0026thinsp;=\u0026thinsp;3], no significant treatment differences were observed between subphenotypes.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e\u003cp\u003eUsing state-of-the-art clustering methods for mixed data were identified three to five subphenotypes which were associated with clinical information and outcomes. The effects of treatments differed for each subphenotype, and k-UMAP may reveal appropriate treatment targets that have not been proven effective.\u003c/p\u003e","manuscriptTitle":"Clinical subphenotypes of sepsis based on mixed data and differences in treatment effects: a cluster analysis of multicentre observational studies","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-15 02:09:26","doi":"10.21203/rs.3.rs-7716199/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-10-18T06:48:57+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-13T15:47:28+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-11T17:09:44+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"56748615292514452336609275913573901869","date":"2025-10-01T23:38:52+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"202537424522372406270266888871596189694","date":"2025-09-29T20:46:57+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-09-29T20:18:52+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-09-29T01:55:21+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-09-29T01:53:46+00:00","index":"","fulltext":""},{"type":"submitted","content":"Critical Care","date":"2025-09-25T21:37:44+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"critical-care","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"cric","sideBox":"Learn more about [Critical Care](http://ccforum.biomedcentral.com/)","snPcode":"13054","submissionUrl":"https://submission.nature.com/new-submission/13054/3","title":"Critical Care","twitterHandle":"@Crit_Care","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"af7d0eb3-8c1e-4420-9593-2a4833439261","owner":[],"postedDate":"October 15th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-01-19T16:47:34+00:00","versionOfRecord":{"articleIdentity":"rs-7716199","link":"https://doi.org/10.1186/s13054-025-05819-4","journal":{"identity":"critical-care","isVorOnly":false,"title":"Critical Care"},"publishedOn":"2026-01-15 16:30:14","publishedOnDateReadable":"January 15th, 2026"},"versionCreatedAt":"2025-10-15 02:09:26","video":"","vorDoi":"10.1186/s13054-025-05819-4","vorDoiUrl":"https://doi.org/10.1186/s13054-025-05819-4","workflowStages":[]},"version":"v1","identity":"rs-7716199","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7716199","identity":"rs-7716199","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}