Clinical Characteristics and Prognosis of Sepsis Subphenotypes Identified by Coagulation Indicator Trajectories: A Single-Center Retrospective Study

Abstract

Background

This study aimed to identify new sepsis subphenotypes on the basis of coagulation indicator trajectories and comprise clinical characteristics and prognosis.

Patients and Methods

This retrospective study included patients diagnosed with sepsis admitted to the intensive care unit of Peking Union Medical College Hospital from May 2016 to March 2023. Using group-based trajectory models, we classified patients into different subphenotypes on the basis of the dynamic daily changes in coagulation parameters within the first 7 days after sepsis diagnosis. Clinical characteristics and outcomes of patients were compared between subphenotypes.

Results

A total of 3,990 patients diagnosed with sepsis were included in this research. Patients were divided into four trajectory groups on the basis of indicator trajectory: Group 1 (n = 500), with high prothrombin times (PTs) and rapidly increasing D-dimer levels; Group 2 (n = 1,334) had normal PT, mildly increasing D-dimer levels and platelet counts. Group 3 (n = 1,013), with mildly elevated PT and D-dimer levels, along with lower platelet counts and fibrinogen levels. Group 4 (n = 1,143) had mildly elevated PT and D-dimer levels along with elevated platelet and fibrinogen levels. Four trajectory subphenotypes exhibit different 28-day mortality, overall in-hospital mortality, bleeding and thrombosis incidence, and the rate of patients with mechanical ventilation.

Conclusion

Coagulation trajectory subphenotypes offer a novel approach for stratifying sepsis heterogeneity, identifying high-risk patients, and refining prognostic assessment. The subphenotype with rapidly rising D-dimer levels warrants heightened clinical vigilance due to its association with the poorest outcomes.

 

MissingLinkText

Keywords

sepsis - subphenotype - disseminated intravascular coagulation - D-dimer - group-based trajectory models

Introduction

Sepsis is defined as life-threatening organ dysfunction caused by dysregulation of host response to infection.[1] It has resulted in approximately 11.0 million deaths worldwide, constituting an estimated 20% of all global deaths.[2] Its elevated mortality rate and considerable treatment costs constitute a significant economic and public health burden on society.[3] To date, a unified and effective therapeutic strategy for the management of sepsis is lacking. Sepsis exhibits significant heterogeneity among patients due to primary infection sites, pathogens, patient characteristics, and comorbidities, which leads to considerable challenges for clinical trials and treatment.[4] Therefore, investigating the subphenotypes of sepsis may have significant implications for the development of personalized treatment strategies for sepsis.[5]

Currently, several studies have defined various sepsis subphenotypes on the basis of trajectory models of changes in various indicators among hospitalized patients. Bhavani et al established subphenotypes on the basis of hourly changes in vital signs and compared outcomes among groups;[6] these subphenotypes have a positive impact on predicting patient outcomes and guiding personalized fluid resuscitation. Xu et al established subphenotypes of sepsis on the basis of trajectories of Sequential Organ Failure Assessment (SOFA) scores and reported that these subphenotypes were significant for identifying mortality rates in sepsis patients.[7] Furthermore, other researchers have also established sepsis subphenotypes on the basis of the trajectories of body temperature,[8] cortisol,[9] platelet count,[10] and other indicators in sepsis patients, which have played a positive role in estimating patient prognosis and guiding treatment. However, at present, no studies have been conducted to establish sepsis subphenotypes on the basis of the trajectories of coagulation indicators.

As one of the primary systems affected in sepsis, coagulation function may have a significant effect on patient prognosis.[11] In previous studies, the International Society on Thrombosis and Hemostasis disseminated intravascular coagulation (ISTH–DIC) and sepsis-induced coagulopathy (SIC) scores were applied to identify coagulopathy in sepsis patients and have been found to be associated with patient mortality rates.[12] [13] [14] However, these scores are not able to distinguish sepsis patients with different characteristics of coagulation abnormalities and cannot reflect the significance of the dynamic changes in coagulation indicators for the prognosis of sepsis patients. Therefore, identifying sepsis subphenotypes based on the trajectories of coagulation indicators may aid in predicting patient outcomes and positively influence treatment strategies.

The objectives of this study were to identify novel trajectory subphenotypes based on dynamic changes in coagulation indicators and to compare the differences in clinical characters and outcomes between different subphenotypes.

 

MissingLinkText

Methods

Study Design and Study Population

This is a retrospective study that included patients with sepsis who were treated at the intensive care unit (ICU) of Peking Union Medical College Hospital from May 2016 to March 2023. The diagnosis of sepsis was based on the Sepsis-3 criteria established in 2016.[1] This study was approved by the Ethics Committee of Peking Union Medical College Hospital, Chinese Academy of Medical Sciences. Patients were included if they met these criteria during their ICU admission, regardless of whether sepsis was present at admission or developed later. The exclusion criteria for the study include: (1) patients under the age of 18 years and (2) patients with an ICU hospital stay < 48 hours or fewer than two coagulation parameter tests. In this study, patients with anticoagulation requirements were routinely administered low-molecular-weight heparin sodium, which does not affect coagulation parameters.

 

MissingLinkText

Data Collection

The data collected in the study included the basic information and clinical characteristics of the patients, which primarily included sex, age, vital signs (the first values calculated from measurements taken within the first 24 hours following sepsis diagnosis), and comorbidities (hypertension, diabetes, cardiac disease, pulmonary disease, liver disease, and kidney disease). The first laboratory examination following sepsis diagnosis also included white blood cell count (WBC), hematocrit value (HCT), lymphocyte count, procalcitonin (PCT), alanine aminotransferase, albumin, total bilirubin (Tbil), and serum creatinine (Cr). Disease severity scores included the SOFA score, the Acute Physiology and Chronic Health Evaluation (APACHE) II score, and ISTH–DIC score. Missing data were imputed via the “mice” R package with predictive mean matching ([Supplementary Table S1], available in the online version).

For coagulation parameters, collected platelet counts, fibrinogen levels, prothrombin times (PTs), and D-dimer levels. One value per parameter per day was collected from patients within the first 7 days following the diagnosis of sepsis, regardless of whether patients remained in the ICU or were transferred to wards during this period. Each patient underwent at least three tests. Missing data were not imputed, as the Group-Based Trajectory Model (GBTM) can handle missing values.

 

MissingLinkText

Outcomes

The primary outcome of this study was the 28-day in-hospital mortality rate of the patients. The secondary outcomes include the overall in-hospital mortality rate, the incidence of major bleeding and thrombotic events during hospitalization, and the rate of patients with mechanical ventilation. Major bleeding was defined by ISTH criteria; fatal or symptomatic bleeding in critical organs (e.g., intracranial, intraspinal, intraocular, retroperitoneal, intra-articular, pericardial, or compartment syndrome in muscles); or a decrease in hemoglobin by ≥1.24 mmol/L (≥20 g/L). Thrombotic events are primarily encompassed by venous thromboembolism (ultrasound/computed tomography [CT]-confirmed deep venous thrombosis or pulmonary embolism) and arterial thrombosis (angiography/CT/magnetic resonance imaging-verified acute limb ischemia, stroke, or myocardial infarction).

 

MissingLinkText

Group-Based Trajectory Models

GBTM is a trajectory clustering approach that identifies subgroups of individuals with similar longitudinal patterns of one or more repeatedly measured variables over time.[15] To distinguish between different subtypes of coagulopathy in sepsis patients, we employed the GBTM algorithm and identified distinct groupings on the basis of dynamic monitoring of coagulation parameters, including PT, D-Dimer, platelet counts, and fibrinogen levels within the first 7 days following the diagnosis of sepsis ([Supplementary Table S2], available in the online version). Excluding patients with fewer than two coagulation parameter tests to ensure robust trajectory estimation. Each group of patients exhibited similar trajectory characteristics of coagulation parameters over time. The optimal trajectory model was determined through multiple criteria. First, each trajectory group included more than 5% of the study patients to ensure the clinical significance and statistical significance of the grouping. Second, the average posterior probability of the model should be greater than 0.7 to ensure its robustness. Third, the model should have lower values of the Akaike Information Criterion and the Bayesian Information Criterion. Additionally, the identified trajectory subgroups should be visually distinguishable and have good clinical interpretability. The GBTM calculations were completed via the “GBMT” R package.

 

MissingLinkText

Statistical Analysis

Continuous variables in the study that are normally distributed are expressed as the means ± standard deviations and were compared via Student's t-test. In addition, for continuous variables that were not normally distributed, medians and interquartile ranges were employed. Based on their distributional characteristics, parametric (Student's t-test) or nonparametric (Mann–Whitney U test) tests were used for between-group comparisons. Categorical variables are represented as frequencies (n) with percentages (%), and chi-square tests or Fisher's exact tests were applied for intergroup comparisons. Benjamini–Hochberg correction was applied across all variables to control the false discovery rate (FDR) at α = 0.05, with adjusted p-values reported. K-M survival analysis was employed to compare the differences in 28-day in-hospital mortality among sepsis patients with distinct coagulation profile trajectory groups. Logistic regression analysis was used to investigate the associations between distinct subphenotypes and outcomes such as in-hospital mortality within 28 days, overall in-hospital mortality, incidence of major bleeding, incidence of thromboembolic events, and the rate of mechanical ventilation, with adjustments for sex, SOFA score, APACHE II score, and comorbidities (including hypertension, diabetes mellitus, chronic heart disease, chronic pulmonary disease, chronic kidney disease, and liver disease). Statistical analyses were performed via R version 4.4.2 (http://www.R-project.org) and SPSS version 22.0.

 

MissingLinkText

 

MissingLinkText

Results

Study Population

In this study, a total of 6,647 patients admitted to the ICU at Peking Union Medical College Hospital from May 2016 to March 2023 were included. Among the study participants, 334 patients were under the age of 18, and 2,323 patients had a hospital stay of less than 48 hours or underwent fewer than two coagulation parameter tests. Therefore, 3,990 patients were ultimately included in the study ([Fig. 1]).

Fig. 1 Flowchart of the patient selection process.

 

MissingLinkText

Identification of Trajectory Subphenotypes

Four distinct trajectories were identified on the basis of the dynamic changes in coagulation indices over a 7-day period via the GBTM approach. Group 1 included 500 patients (12.5%), characterized by an initially increased and then gradually improved PT, a rapidly increasing D-dimer level, lower platelet counts, and a declining fibrinogen level. Group 2 included 1334 patients (33.4%), characterized by normal PT and platelet counts and mildly elevated D-dimer and fibrinogen levels. Group 3 included 1,013 patients (25.4%), characterized by mildly elevated PT and D-dimer levels, along with lower platelet counts and fibrinogen levels. Group 4 included 1,143 patients (28.7%), characterized by mildly elevated PT and D-dimer levels, along with elevated platelet and fibrinogen levels ([Fig. 2]). The four group trajectory models demonstrated a high degree of accuracy in the classification of sepsis subphenotypes, with an average posterior probability exceeding 90%. Cross-validation was used to internal validation. The dataset was randomly partitioned into 5-folds. GBTM was independently applied to each fold for subgroup identification. The results demonstrated that four distinct subgroups were consistently identified across all 5-folds, exhibiting similar trajectory characteristic patterns ([Supplementary Fig. S1], available in the online version). Sensitivity analysis was conducted by partitioning the cohort into two subgroups based on sepsis diagnosis time (2016–2019 and 2019–2023). GBTM was independently applied to each subgroup to assess the consistency of trajectory patterns across time periods ([Supplementary Fig. S2], available in the online version).

Fig. 2 Group-based trajectory modeling of coagulation indicators.

 

MissingLinkText

Clinical Characteristics of Subphenotypes

The proportion of males was greater in Group 2 (59.4%). There were no significant differences in age, length of the hospital stays, and the duration of ICU stays among patients across the trajectory groups. After Benjamini–Hochberg correction for multiple comparisons (FDR = 0.05), except for heart disease, there were significant differences in other comorbidities among the four groups of patients. In general, Group 3 had a relatively high proportion of patients with liver (7.1%) and kidney diseases (19.6%), whereas Group 2 and Group 4 had relatively high proportions of patients with hypertension (47.1%, 42.6%) and diabetes (25.5%, 27.1%). In terms of disease severity, patients in Group 1 and Group 3 had higher SOFA scores (p _adj  = 0.002) and ISTH–DIC scores (p _adj  = 0.002). Additionally, there were significant differences in the laboratory test results among the four groups. Patients in Group 1 and Group 3 presented lower HCT (p _adj  = 0.002), higher PCT (p _adj  = 0.002), and Tbil (p _adj  = 0.002) and Cr (p _adj  = 0.002) levels ([Table 1]). The characteristics of patients with sepsis excluded due to ICU length of stay < 48 hours but with comparatively complete coagulation data have been presented in the supplementary table ([Supplementary Table S5], available in the online version).

Table 1 Patient characteristics according to trajectory subphenotype

	Group 1	Group 2	Group 3	Group 4	p	p adj
N (%)	500 (12.5)	1,334 (33.4)	1,013 (25.4)	1,143 (28.7)
Characteristics
Sex (male) (%)	278 (55.6)	793 (59.4)	587 (57.9)	653 (57.1)	0.445	0.539
Age (y)	60 [46, 70]	61 [48, 71]	60 [46, 71]	60 [45, 70]	0.129	0.167
Hospital day (d)	15 [7, 29.25]	16[8] [28]	15[7] [29]	15[7] [28]	0.919	0.919
ICU day (d)	5[2] [12]	5[2] [11]	5[2] [12]	5[2] [12]	0.889	0.919
Vital signs
Body temperature	36.7 [36.2, 37.5]	36.8 [36.2, 37.5]	36.8 [36.1, 37.6]	36.7 [36.1, 37.5]	0.572	0.658
Heart rate	99 [84, 118.25]	98 [83, 115]	99 [83, 115]	100 [85, 117]	0.131	0.167
Comorbidities (%)
Hypertension	170 (34.0)	628 (47.1)	361 (35.6)	487 (42.6)	<0.001	0.002
Cardiovascular disease	89 (17.8)	232 (17.4)	191 (18.9)	209 (18.3)	0.826	0.905
Pulmonary disease	43 (8.6)	96 (7.2)	80 (7.9)	65 (5.7)	0.105	0.151
Liver disease	30 (6.0)	52 (3.9)	72 (7.1)	34 (3.0)	<0.001	0.002
Kidney disease	56 (11.2)	148 (11.1)	196 (19.6)	123 (10.8)	<0.001	0.002
Diabetes	95 (19.0)	340 (25.5)	213 (21.0)	310 (27.1)	<0.001	0.002
Laboratory results
WBC (×109/L)	10.49 [6.40, 16.14]	8.48 [5.23, 13.40]	8.65 [6.16, 12.50]	10.60 [7.02, 15.35]	<0.001	0.002
Lymphocyte (×109/L)	0.82 [0.51, 1.30]	0.72 [0.63, 1.53]	0.99 [0.60,1.51]	1.02 [0.62, 1.51]	<0.001	0.002
PCT (μg/L)	1.90 [0.40, 8.95]	0.25 [0.08, 1.28]	1.45 [0.29,6.90]	0.72 [0.18, 4.70]	<0.001	0.002
HCT (L/L)	30.2 [25.5, 36.3]	34.9 [29.7, 39.4]	28.1 [22.9,34.3]	32.8 [27.7, 37.9]	<0.001	0.002
ALT (U/L)	25.0 [13.0, 76.0]	20.0 [13.0, 36.0]	22.0 [12.0,51.0]	21.0 [13.0, 42.0]	<0.001	0.002
Alb (g/L)	30.0 [26.0, 35.0]	34.0 [29.0, 38.0]	29.0 [26.0,34.0]	31.0 [27.0, 36.0]	<0.001	0.002
Tbil (μmol/L)	16.15 [9.78, 24.58]	10.50 [7.53, 15.40]	16.10 [10.10, 31.30]	12.60 [8.70, 19.65]	<0.001	0.002
Cr (μmol/L)	97.0 [67.0, 188.0]	69.0 [54.0, 92.75]	99.0 [65.0, 224.0]	78.0 [58.0, 120.5]	<0.001	0.002
Disease severity scores
SOFA	8[5] [12]	6[4] [10]	8[5] [12]	7[4] [10]	<0.001	0.002
APACHE2	23[16] [26]	21[16] [26]	16[12] [21]	18[14] [23]	<0.001	0.002
ISTH–DIC	3[3] [5]	2[1] [2]	4[3] [4]	3[2] [3]	<0.001	0.002

Abbreviations: Alb, albumin; ALT, alanine aminotransferase; APACHE II, Acute Physiology and Chronic Health Evaluation II: Cr, creatinine; DIC, disseminated intravascular coagulation; HCT, hematocrit; ICU day, length of ICU stays; ISTH–DIC, International Society on Thrombosis and Hemostasis Disseminated Intravascular Coagulation. Hospital day: total hospitalization duration; p, original p-value; p _adj, Benjamini–Hochberg adjusted p-value for multiple comparisons (false discovery rate = 0.05); PCT, procalcitonin; SOFA, Sequential Organ Failure Assessment; Tbil, total bilirubin; WBC, white blood cell count.

 

MissingLinkText

Associations between Trajectory Subphenotypes and 28-day Mortality

There were significant differences in the 28-day mortality rates among the four trajectory subtypes (p < 0.001; [Table 2]). The results of the Kaplan-Meier (K-M) survival analysis revealed significant differences in the 28-day mortality rates among patients with the four trajectory subphenotypes (p < 0.001), with the highest to lowest rates observed in Group 1 (37.0%), Group 2 (11.2%), Group 3 (35.3%), and Group 4 (18.0%; [Fig. 3A]). After adjusting for sex, SOFA score, APACHE II score, and comorbidities (including hypertension, diabetes mellitus, chronic heart disease, chronic pulmonary disease, chronic kidney disease, and liver disease), with Group 2 as the reference group, the results of the logistic regression analysis showed that the 28-day mortality rate was highest in Group 1 (odds ratio [OR] = 4.42; 95% confidence interval [CI]: 3.41–5.75; p < 0.001), followed by Group 3 (OR = 3.97; 95% CI: 3.15–4.99; p < 0.001) and Group 4 (OR = 1.77; 95% CI: 1.39–2.25; p < 0.001; [Fig. 3B]). Interaction analyses between trajectory subphenotypes and comorbidities (hypertension, cardiovascular disease, pulmonary disease, liver disease, kidney disease, diabetes) showed no significant effect modification on 28-day mortality ([Supplementary Table S7], available in the online version). In cross-validation patients within the same subgroups exhibited similar 28-day mortality rates across all 5-folds ([Supplementary Table S4], available in the online version). In sensitivity analysis, subphenotypes in the two cohorts with different sepsis diagnosis time (2016–2019 and 2019–2023) showed similar 28-day mortality rate. Associations between trajectory subphenotypes and other outcomes.

Fig. 3 The 28-day survival probability curve (A) and odds ratios for 28-day hospital mortality after adjusting for sex, SOFA score, APACHE II score, and comorbidities (including hypertension, diabetes mellitus, chronic heart disease, chronic pulmonary disease, chronic kidney disease, and liver disease) (B). APACHE II, Acute Physiology and Chronic Health Evaluation II; SOFA, Sequential Organ Failure Assessment.

Table 2 Patient outcomes according to trajectory subphenotype

	Group 1	Group 2	Group 3	Group 4	p
N	500 (12.5)	1,334 (33.4)	1,013 (25.4)	1,143 (28.7)
28-d mortality	185 (37.0)	150 (11.2)	358 (35.3)	206 (18.0)	<0.001
Total hospital mortality	237 (47.4)	183 (13.7)	442 (43.6)	235 (20.6)	<0.001
Thromboembolism	43 (8.6)	84 (6.3)	68 (6.7)	93 (8.1)	0.176
Major bleeding	107 (21.4)	172 (12.9)	283 (27.9)	209 (18.3)	<0.001
Ventilation	434 (86.8)	847 (63.5)	753 (74.3)	885 (77.4)	<0.001

The results of the chi-square test revealed significant differences among the four trajectory subphenotypes in terms of overall hospital mortality (p < 0.001), incidence of major bleeding (p < 0.001), and the rate of patients with mechanical ventilation (p < 0.001; [Table 2]).

After further adjustment for sex, SOFA score, APACHE II score, and comorbidities, with Group 2 as the reference group, the results of the logistic regression analysis demonstrated that Group 1 had the highest overall in-hospital mortality (OR = 5.25; 95% CI: 4.09–6.73; p < 0.001), and the rate of mechanical ventilation (OR = 3.56; 95% CI: 2.65–4.78; p < 0.001) among the four groups. While no significant difference in thromboembolic incidence was observed (p = 0.176), regression analysis identified Group 1 as having a higher thrombotic risk (OR = 1.60; 95% CI: 1.07–2.40; p = 0.021). Patients in Group 3 presented the highest incidence of major bleeding (OR = 2.35; 95% CI: 1.87–2.95; p < 0.001) and the second highest overall hospital mortality (OR = 4.23; 95% CI: 3.41–5.25; p < 0.001) among the four trajectory subphenotypes. The incidence of thromboembolic events did not show any significant differences (OR = 1.18; 95% CI: 0.82–1.68; p = 0.372) between Groups 2 and 3 ([Fig. 4]). Patients in Group 4 also had relatively high rates of mortality (OR = 1.61; 95% CI: 1.29–2.02; p < 0.001) and bleeding incidence (OR = 1.49; 95% CI: 1.18–1.88; p < 0.001; [Fig. 4]).

Fig. 4 Odds ratios for overall hospital mortality (A), incidence of thromboembolism (B), incidence of major bleeding (C), and incidence of mechanical ventilation (D). After adjusting for sex, the SOFA score, APACHE II score, and comorbidities (including hypertension, diabetes mellitus, chronic heart disease, chronic pulmonary disease, chronic kidney disease, and liver disease) were calculated. APACHE II, Acute Physiology and Chronic Health Evaluation II; SOFA, Sequential Organ Failure Assessment.

 

MissingLinkText

 

MissingLinkText

Discussion

In this study, we employed the GBTM approach to identify novel trajectory subtypes on the basis of dynamic changes in the coagulation parameters of ICU patients. Patients classified into different trajectory subphenotypes presented distinct clinical characteristics and prognoses. Among the identified trajectory subphenotypes, the one characterized by a rapid increase in D-dimer levels, along with a higher PT and lower platelet and fibrinogen counts, was associated with the poorest outcomes. Patients with the trajectory subphenotype characterized by extremely low platelet counts and fibrinogen levels presented higher mortality rates and a greater incidence of major bleeding. These findings provide a novel perspective on the role of dynamic monitoring of coagulation parameters in sepsis patients admitted to the ICU.

Abnormal activation and dysfunction of the coagulation system play significant roles during sepsis.[16] Sepsis-induced DIC is strongly associated with adverse outcomes.[17] In addition to DIC, microcirculatory thrombosis is one of the primary mechanisms leading to multiple organ dysfunction in sepsis.[18] [19] However, microcirculation lacks effective monitoring and therapeutic interventions. Therefore, the surveillance of coagulation parameters may have significant implications for the prognosis of patients with sepsis. The ISTH–DIC score assesses coagulopathy using platelet count, prothrombin time (PT), fibrinogen, and D-dimer levels.[20] To facilitate early identification of sepsis-associated coagulopathy and guide early intervention, a more sensitive SIC score was subsequently proposed.[21] [22] Some clinical studies have indicated that both the ISTH–DIC score and the SIC are associated with adverse outcomes in sepsis patients,[23] [24] and early monitoring of these scores can help improve patient outcomes.[25] Currently, the ISTH–DIC and SIC scoring systems have a certain degree of positive impact on guiding the anticoagulant treatment of sepsis patients.[26] This study is consistent with these scoring systems, demonstrating that higher levels of PT and D-dimer and lower levels of platelet counts, and fibrinogen are correlated with a poorer prognosis in sepsis patients. In our study, there was no significant difference in the ISTH–DIC scores between Group 1 and Group 3, but the two subphenotypes represented different outcomes, particularly in major bleeding and thrombotic events. Furthermore, compared with ISTH–DIC scores, trajectory monitoring identifies progression. Four trajectory subphenotypes were specifically categorized on the basis of the heterogeneity in coagulation profiles observed among individuals with sepsis, with the goal of significantly increasing the predictive accuracy of sepsis outcomes and providing potential implications for guiding personalized therapeutic interventions. For example, patients in Group1, intensified anticoagulation may be warranted, balancing bleeding monitoring, and for patients in Group3, transfusion support (platelets/fibrinogen) and avoidance of aggressive anticoagulation could be prioritized.

Kudo et al employed the k-means clustering algorithm to establish sepsis subtypes on the basis of platelet counts, the prothrombin time-international normalized ratio, fibrinogen, fibrinogen/fibrin degradation products, D-dimer, and antithrombin activities in sepsis patients and compared the 28-day mortality rates between groups.[5] [27] The group characterized by high D-dimer levels, elevated PT, and low platelet counts and fibrinogen levels presented the highest 28-day mortality, followed by the group characterized by low D-dimer levels, elevated PT, and low platelet counts and fibrinogen.[27] Our study demonstrated a high degree of concordance with the sepsis subphenotypes identified in the above study. Furthermore, the subphenotypes established through the GBTM algorithm more accurately reflect the potential differences caused by the dynamic trends in coagulation parameters and may, to some extent, consider that multiple test results for each patient may provide greater accuracy. Some studies have established sepsis subphenotypes on the basis of the dynamic trajectory of platelet counts.[10] [28] [29] These studies indicate that a rapid decline and lower platelet counts are closely associated with 28-day mortality rates. Our study is largely consistent with these findings, suggesting that lower platelet counts are correlated with adverse outcomes in sepsis patients. However, our study examined the platelet count as one of several coagulation parameters rather than as an independent risk factor.

D-dimer, a soluble product of fibrin degradation, serves as an indicator of thrombotic events in blood vessels and microcirculation.[30] Research has indicated that elevated D-dimer levels can be an independent factor influencing the prognosis of sepsis patients[31] and may have a predictive role in determining outcomes.[32] However, some studies have also reported that lower D-dimer levels may be associated with poor outcomes.[33] Our study first identified a sepsis subphenotype with a rapid increase in D-dimer levels within the first 3 days, which was associated with the highest 28-day mortality rate. These findings suggest that monitoring D-dimer levels dynamically could be highly valuable for assessing patient prognosis and have potential implications for guiding anticoagulant treatment.

This study had several limitations that warrant elucidation. First, the investigation is a retrospective and may carry an inherent risk of bias in data collection, particularly concerning certain outcome measures. Thrombotic events may have been underestimated due to the potential presence of clinically silent thrombosis and the lack of comprehensive imaging in some patients. Furthermore, the higher incidence of bleeding events in our cohort is likely partially attributable to the inclusion of postsurgical sepsis patients and those with comorbidities affecting the hematological system. This retrospective study limits our ability to determine whether bleeding events resulted from confounding comorbidities. However, the findings within this broadly defined sepsis population may still reflect meaningful differences in clinical outcomes across distinct coagulation subphenotypes. Passage of time and the singular institution's perspective may skew the data, affecting the generalizability of the findings. Second, the study lacked comprehensive data on anticoagulant therapy specific to each subgroup, focusing solely on the comparative outcomes between different subgroups. While low-molecular-weight heparin sodium is routinely administered for anticoagulation at this center and is known to minimally impact coagulation parameters,[34] The specific data on anticoagulant therapy and blood products administered could not be obtained due to data accessibility limitations, which may act as additional confounding factors that could potentially influence the study findings, such as its effects on outcomes like thrombosis and bleeding. We will address this issue in subsequent research and compare the therapeutic efficacy of anticoagulant therapy across different subgroups. Third, owing to the methodological requirements of the GBTM algorithm, the study excluded patients with ICU stays shorter than 48 hours and those with fewer than two coagulation parameter tests, which may undermine the ability to assess early-stage sepsis. Consequently, further research is needed in the future to provide more clinical evidence for the role of coagulation parameter trajectory subphenotypes in assessing the prognosis of sepsis, particularly organs' dysfunction induced by sepsis, and guiding anticoagulant therapy.

 

MissingLinkText

Conclusion

In conclusion, this study employed the GBTM algorithm to establish four sepsis subtypes on the basis of the trajectories of dynamically monitored coagulation indicators. These distinct subtypes were significantly correlated with 28-day mortality, overall in-hospital mortality, the incidence of bleeding and thrombosis, and the rate of ventilation. Notably, the subphenotype with a rapid increase in D-dimer levels and high PT levels presented the highest 28-day mortality among the four subphenotypes.

 

MissingLinkText

 

MissingLinkText

Conflict of Interest

The authors declare that they have no conflict of interest.

Data Availability Statement

The data utilized in this study can be accessed by the corresponding author on reasonable request.

Contributors' Statement

All the authors participated in the conceptualization and design of the study. Data collection and analysis, as well as the writing of the manuscript, were completed by B.Y.W. All the authors were involved in the research process and the revision of the manuscript. All the authors have read the final draft and agreed to its publication.

Ethics Approval

This study received approval from the Institutional Review Board at Peking Union Medical College Hospital. The approval included a waiver for the informed consent of the patients.

• References

• 1 Singer M, Deutschman CS, Seymour CW. et al. The Third International Consensus definitions for sepsis and septic shock (sepsis-3). JAMA 2016; 315 (08) 801-810
• 2 Rudd KE, Johnson SC, Agesa KM. et al. Global, regional, and national sepsis incidence and mortality, 1990-2017: analysis for the Global Burden of Disease Study. Lancet 2020; 395 (10219): 200-211
• 3 Rhee C, Dantes R, Epstein L. et al; CDC Prevention Epicenter Program. Incidence and trends of sepsis in US hospitals using clinical vs claims data, 2009-2014. JAMA 2017; 318 (13) 1241-1249
• 4 Kheterpal S, Singh K, Topol EJ. Digitising the prediction and management of sepsis. Lancet 2022; 399 (10334): 1459
• 5 Santacroce E, D'Angerio M, Ciobanu AL. et al. Advances and challenges in sepsis management: modern tools and future directions. Cells 2024; 13 (05) 439
• 6 Bhavani SV, Semler M, Qian ET. et al. Development and validation of novel sepsis subphenotypes using trajectories of vital signs. Intensive Care Med 2022; 48 (11) 1582-1592
• 7 Xu Z, Mao C, Su C. et al. Sepsis subphenotyping based on organ dysfunction trajectory. Crit Care 2022; 26 (01) 197
• 8 Bhavani SV, Carey KA, Gilbert ER, Afshar M, Verhoef PA, Churpek MM. Identifying novel sepsis subphenotypes using temperature trajectories. Am J Respir Crit Care Med 2019; 200 (03) 327-335
• 9 Leng F, Gu Z, Pan S. et al. Novel cortisol trajectory sub-phenotypes in sepsis. Crit Care 2024; 28 (01) 290
• 10 Ye Q, Wang X, Xu X. et al. Serial platelet count as a dynamic prediction marker of hospital mortality among septic patients. Burns Trauma 2024; 12: tkae016
• 11 Amaral A, Opal SM, Vincent JL. Coagulation in sepsis. Intensive Care Med 2004; 30 (06) 1032-1040
• 12 Iba T, Levy JH. Sepsis-induced coagulopathy and disseminated intravascular coagulation. Anesthesiology 2020; 132 (05) 1238-1245
• 13 Ding R, Wang Z, Lin Y, Liu B, Zhang Z, Ma X. Comparison of a new criteria for sepsis-induced coagulopathy and International Society on Thrombosis and Haemostasis disseminated intravascular coagulation score in critically ill patients with sepsis 3.0: a retrospective study. Blood Coagul Fibrinolysis 2018; 29 (06) 551-558
• 14 Iba T, Helms J, Connors JM, Levy JH. The pathophysiology, diagnosis, and management of sepsis-associated disseminated intravascular coagulation. J Intensive Care 2023; 11 (01) 24
• 15 Nagin DS, Jones BL, Elmer J. Recent advances in group-based trajectory modeling for clinical research. Annu Rev Clin Psychol 2024; 20 (01) 285-305
• 16 Giustozzi M, Ehrlinder H, Bongiovanni D. et al. Coagulopathy and sepsis: pathophysiology, clinical manifestations and treatment. Blood Rev 2021; 50: 100864
• 17 Iba T, Levi M, Levy JH. Sepsis-induced coagulopathy and disseminated intravascular coagulation. Semin Thromb Hemost 2020; 46 (01) 89-95
• 18 Lelubre C, Vincent JL. Mechanisms and treatment of organ failure in sepsis. Nat Rev Nephrol 2018; 14 (07) 417-427
• 19 Iba T, Umemura Y, Wada H, Levy JH. Roles of coagulation abnormalities and microthrombosis in sepsis: pathophysiology, diagnosis, and treatment. Arch Med Res 2021; 52 (08) 788-797
• 20 Taylor Jr FB, Toh CH, Hoots WK, Wada H, Levi M. Scientific Subcommittee on Disseminated Intravascular Coagulation (DIC) of the International Society on Thrombosis and Haemostasis (ISTH). Towards definition, clinical and laboratory criteria, and a scoring system for disseminated intravascular coagulation. Thromb Haemost 2001; 86 (05) 1327-1330
• 21 Gando S, Meziani F, Levi M. What's new in the diagnostic criteria of disseminated intravascular coagulation?. Intensive Care Med 2016; 42 (06) 1062-1064
• 22 Dhainaut JF, Yan SB, Joyce DE. et al. Treatment effects of Drotrecogin alfa (activated) in patients with severe sepsis with or without overt disseminated intravascular coagulation. J Thromb Haemost 2004; 2 (11) 1924-1933
• 23 Gando S, Iba T, Eguchi Y. et al; Japanese Association for Acute Medicine Disseminated Intravascular Coagulation (JAAM DIC) Study Group. A multicenter, prospective validation of disseminated intravascular coagulation diagnostic criteria for critically ill patients: comparing current criteria. Crit Care Med 2006; 34 (03) 625-631
• 24 Iba T, Nisio MD, Levy JH, Kitamura N, Thachil J. New criteria for sepsis-induced coagulopathy (SIC) following the revised sepsis definition: a retrospective analysis of a nationwide survey. BMJ Open 2017; 7 (09) e017046
• 25 Umemura Y, Yamakawa K, Hayakawa M, Hamasaki T, Fujimi S. Japan Septic Disseminated Intravascular Coagulation (J-Septic DIC) study group. Screening itself for disseminated intravascular coagulation may reduce mortality in sepsis: a nationwide multicenter registry in Japan. Thromb Res 2018; 161: 60-66
• 26 Wiedermann CJ. Antithrombin concentrate use in disseminated intravascular coagulation of sepsis: meta-analyses revisited. J Thromb Haemost 2018; 16 (03) 455-457
• 27 Kudo D, Goto T, Uchimido R. et al. Coagulation phenotypes in sepsis and effects of recombinant human thrombomodulin: an analysis of three multicentre observational studies. Crit Care 2021; 25 (01) 114
• 28 Wang K, Lu D, Wang F. Subphenotypes of platelet count trajectories in sepsis from multi-center ICU data. Sci Rep 2024; 14 (01) 20187
• 29 Wang Y, Wu J, Shao T. et al. Prognostic implications of changes in platelet trajectories in patients with sepsis: a retrospective analysis using the medical information mart for intensive care-IV database. Shock 2024;
• 30 Weitz JI, Fredenburgh JC, Eikelboom JW. A test in context: D-dimer. J Am Coll Cardiol 2017; 70 (19) 2411-2420
• 31 Rodelo JR, De la Rosa G, Valencia ML. et al. D-dimer is a significant prognostic factor in patients with suspected infection and sepsis. Am J Emerg Med 2012; 30 (09) 1991-1999
• 32 Han YQ, Yan L, Zhang L. et al. Performance of D-dimer for predicting sepsis mortality in the intensive care unit. Biochem Med (Zagreb) 2021; 31 (02) 020709
• 33 Semeraro F, Ammollo CT, Caironi P. et al. Low D-dimer levels in sepsis: good or bad?. Thromb Res 2019; 174: 13-15
• 34 Chistolini A, Ruberto F, Alessandri F. et al; Policlinico Umberto I COVID-19 Group. Effect of low or high doses of low-molecular-weight heparin on thrombin generation and other haemostasis parameters in critically ill patients with COVID-19. Br J Haematol 2020; 190 (04) e214-e218

Publikationsverlauf

Eingereicht: 22. März 2025

Angenommen: 26. August 2025

Artikel online veröffentlicht:
09. September 2025