Predictive value of serum biochemical indicators combined with imaging scores for prognosis of traumatic liver rupture

Abstract

BACKGROUND

Traumatic liver rupture has high mortality (10%-15%) but current prognostic tools show limited accuracy. This study develops an integrated model combining serum biomarkers and imaging scores for improved risk stratification.

AIM

To investigate the predictive value of serum biochemical indicators combined with imaging scores for the prognosis of patients with traumatic liver rupture, establish a comprehensive prediction model, and provide scientific evidence for early clinical identification of high-risk patients.

METHODS

A single-center retrospective cohort study was conducted, including 186 patients with traumatic liver rupture treated from June 2020 to June 2024. Patient baseline data, serum biochemical indicators within 2 hours of admission (including liver function indicators, coagulation function, inflammatory indicators), and imaging scores were collected. The primary outcome was adverse prognosis within 30 days, including death, need for emergency surgical intervention, severe complications, or need for Intensive care unit treatment. Univariate and multivariate logistic regression analyses were used to screen independent predictive factors, establish a prediction model, and evaluate its predictive efficacy.

RESULTS

Among 186 patients, 142 cases (76.3%) had good prognosis and 44 cases (23.7%) had poor prognosis. Age, liver function indicators, coagulation function indicators, inflammatory indicators, and imaging scores in the poor prognosis group were significantly higher than those in the good prognosis group (all P < 0.001). Multivariate logistic regression analysis showed that aspartate aminotransferase, international normalized ratio, C-reactive protein, Glasgow Coma Scale score, and long-term services and supports (LTSS) total score were independent predictive factors for prognosis (all P < 0.05). The combined prediction model based on these 5 indicators had a receiver operating characteristic area under the curve (AUC) of 0.923 (95%confidence interval: 0.881-0.965), with sensitivity of 86.4%, specificity of 91.5%, and Youden index of 0.779, significantly superior to using serum biochemical indicators alone (AUC = 0.887) or long-term services and supports score alone (AUC = 0.845) (all P < 0.05). Decision curve analysis showed that the model had good clinical net benefit.

CONCLUSION

The comprehensive prediction model established by combining serum biochemical indicators with imaging scores can effectively predict the prognosis of patients with traumatic liver rupture, providing important scientific evidence for early clinical identification of high-risk patients.

Key Words: Traumatic liver rupture; Serum biochemical indicators; Imaging score; Prognosis prediction; Traumatic liver rupture

Core Tip: This study established a comprehensive prognostic prediction model for traumatic liver rupture by integrating serum biochemical indicators with clinical assessment and imaging evaluation. In a cohort of 186 patients, the integrated model demonstrated superior predictive performance compared to single-modality approaches. The model showed excellent discrimination and clinical utility, offering a practical tool for early identification of high-risk patients and guiding individualized treatment strategies in traumatic liver rupture management. All predictive variables are routinely available within 2 hours of hospital admission, ensuring rapid risk assessment without additional costs or specialized testing.

INTRODUCTION

Traumatic liver rupture is one of the most common and dangerous types of injury in abdominal trauma, accounting for 35%-45% of all blunt abdominal trauma, with a mortality rate of 10%-15%[1]. With the increasing incidence of accidental injuries such as traffic accidents and falls from height, the incidence of traumatic liver rupture is showing an increasing trend year by year, and has become an important clinical challenge faced by trauma surgery[2]. Early accurate assessment of patient prognosis is of great significance for formulating reasonable treatment strategies and improving patient survival rates.

Currently, prognostic assessment of traumatic liver rupture mainly relies on clinical scoring systems, such as the Injury Severity Score (ISS) and Glasgow Coma Scale (GCS)[3]. However, these traditional scoring systems are mainly based on anatomical injury severity and consciousness status, making it difficult to comprehensively reflect the physiological functional status and injury severity of the body, with limited predictive accuracy. In recent years, the value of serum biochemical indicators in trauma prognosis assessment has gradually received attention, as liver function indicators, coagulation function, and inflammatory indicators can sensitively reflect the degree of liver injury and the body's inflammatory response status[4].

Imaging examination, particularly contrast-enhanced computed tomography (CT) scanning, is the gold standard for diagnosing and evaluating traumatic liver rupture[5]. CT imaging-based scoring systems, such as the Liver Trauma CT Severity Score (LTSS), can objectively quantify the anatomical characteristics of liver injury, including injury extent, hematoma size, intraperitoneal bleeding volume, etc., providing important information for prognostic assessment[6]. However, single imaging scores often ignore the patient’s physiological functional status, which may lead to insufficient predictive accuracy.

Serum biochemical indicators can reflect the body’s physiological functional status, while imaging scores objectively assess anatomical injury severity, and both have good complementarity[7]. Combining serum biochemical indicators with imaging scores is expected to establish a more comprehensive and accurate prognostic prediction model, providing more reliable evidence for clinical decision-making[8]. However, current research on the value of their combined application in prognostic prediction of traumatic liver rupture is still relatively limited.

Therefore, this study aims to explore the predictive value of serum biochemical indicators combined with imaging scores for the prognosis of patients with traumatic liver rupture, establish a comprehensive prediction model, and provide scientific evidence for early identification of high-risk patients and formulation of individualized treatment plans in clinical practice.

MATERIALS AND METHODS

This study was a single-center retrospective cohort study. The study protocol was reviewed and approved by the Medical Ethics Committee of our hospital and complied with the ethical principles of the Declaration of Helsinki. As this was a retrospective study, patient informed consent was waived.

Study subjects

Study population: Patients with traumatic liver rupture admitted to our hospital from June 2020 to June 2024 were selected as study subjects.

Inclusion criteria: (1) Age ≥ 18 years; (2) Traumatic liver rupture confirmed by CT or surgery; (3) Time from trauma to hospital admission ≤ 24 hours; (4) Complete clinical data, including serum biochemical indicators and imaging examination data at admission; and (5) Complete follow-up data.

Exclusion criteria: (1) Previous history of liver disease (cirrhosis, liver cancer, chronic hepatitis, etc.); (2) Combined severe trauma to other organs (ISS score > 25 points and Hepatic Injury Scale (HIS) > 3 points)[9]; (3) Use of medications affecting liver function within 48 hours before admission; (4) Pregnant patients; and (5) Cardiac and respiratory arrest upon admission.

Data collection

Baseline data collection: The following patient data were collected through the HIS: (1) Demographic characteristics: Age, gender, body mass index; (2) Trauma mechanism: Traffic accident, fall from height, heavy object impact, others; (3) Past medical history: Hypertension, diabetes, heart disease, etc.; (4) Vital signs at admission: Systolic blood pressure, diastolic blood pressure, heart rate, respiratory rate, body temperature; (5) Consciousness status at admission (GCS); and (6) ISS.

Serum biochemical indicator testing: All patients had their first venous blood samples collected within 2 hours of admission, and the following indicators were tested using an automated biochemical analyzer (Beckman Coulter AU5800): (1) Liver function indicators: Alanine aminotransferase, aspartate aminotransferase (AST), total bilirubin, direct bilirubin, albumin, alkaline phosphatase, γ-glutamyl transferase; (2) Coagulation function: Prothrombin time, activated partial thromboplastin time, international normalized ratio (INR); (3) Inflammatory indicators: C-reactive protein (CRP), procalcitonin (PCT), white blood cell count (WBC), neutrophil percentage; and (4) Other indicators: Creatinine, blood urea nitrogen, lactate dehydrogenase, creatine kinase.

Imaging examination and scoring

All patients underwent abdominal contrast-enhanced CT examination after admission, using a multi-slice spiral CT scanner with scanning parameters: Slice thickness 2 mm, reconstruction slice thickness 1mm, contrast agent dose 1.5 mL/kg body weight, injection rate 3.5 mL/second.

Liver injury grading: According to the 2018 World Society of Emergency Surgery liver trauma grading standard[10], liver injuries were classified into grades I-VI: Grade I was subcapsular hematoma < 10 cm² or capsular tear depth < 1 cm; grade II was subcapsular hematoma 10-50 cm² or liver parenchymal tear depth 1-3 cm, length < 10 cm; grade III was subcapsular hematoma > 50 cm² or liver parenchymal tear depth > 3 cm; grade IV was liver parenchymal tear involving 25%-75% of liver lobe; grade V was liver parenchymal tear involving > 75% of liver lobe or > 3 Couinaud liver segments, or hepatic vein injury; grade VI was liver avulsion.

HIS: According to the American Association for the Surgery of Trauma standard, HIS was graded 1-6: Grade 1 was capsular tear or subcutaneous hematoma; grade 2 was capsular tear with active bleeding, liver parenchymal laceration < 3 cm deep; grade 3 was liver parenchymal laceration > 3 cm deep; grade 4 was liver parenchymal laceration involving 25%-75% of liver lobe; grade 5 was liver parenchymal laceration involving > 75% of liver lobe or hepatic vein tear; grade 6 was liver avulsion.

Imaging scoring system: The LTSS was used, which included liver injury extent (0-3 points: 0 points for no obvious injury, 1 point for injury involving < 25% liver parenchyma, 2 points for injury involving 25%-50% liver parenchyma, 3 points for injury involving > 50% liver parenchyma), hematoma size (0-2 points: 0 points for no hematoma, 1 point for hematoma diameter < 5 cm, 2 points for hematoma diameter ≥ 5 cm), intraperitoneal bleeding volume (0-3 points: 0 points for no intraperitoneal bleeding, 1 point for small amount of bleeding in pelvis only, 2 points for moderate amount of bleeding in pelvis + paracolic gutters, 3 points for large amount of bleeding in entire abdomen), active bleeding signs (0-2 points: 0 points for no active bleeding, 1 point for suspected active bleeding, 2 points for definite active bleeding). Total score 0-10 points, with higher scores indicating more severe injury. All imaging data were independently reviewed and scored by 2 radiologists with more than 10 years of experience in abdominal imaging diagnosis. When opinions differed, consensus was reached through discussion. Inter-rater reliability was assessed using intraclass correlation coefficient (ICC).

Treatment plans and follow-up

Conservative treatment or surgical treatment was adopted according to the severity of the patient's condition. (1) Conservative treatment included bed rest, fasting, fluid infusion, hemostasis, anti-infection, etc.; and (2) Surgical treatment included liver repair, partial hepatectomy, hepatic artery embolization, etc. Follow-up was conducted through outpatient visits, telephone follow-up, etc., with the follow-up endpoint being 6 months after discharge. All enrolled patients were followed up until January 2025 to ensure completeness of 6-month follow-up data.

Outcome indicator definitions

Primary outcome indicator was poor prognosis, defined as the occurrence of any of the following events within 30 days of admission: Death, need for emergency surgical intervention, occurrence of serious complications (massive intraperitoneal bleeding, septic shock, multiple organ failure), or need for intensive care unit (ICU) treatment. Secondary outcome indicator was 6-month survival rate.

Statistical analysis

SPSS 25.0 software was used for statistical analysis. Continuous variables were assessed for normality using the Shapiro-Wilk test. Normally distributed data were expressed as mean ± SD, with between-group comparisons using t-test; non-normally distributed data were expressed as median (interquartile range), with between-group comparisons using Mann-Whitney U test. Categorical variables were expressed as numbers and percentages, with between-group comparisons using χ² test or Fisher’s exact test.

Univariate analysis was used to screen factors related to prognosis, and variables with P < 0.1 were included in multivariate logistic regression analysis to identify independent predictors. A prediction model was established, and receiver operating characteristic curves were used to evaluate the predictive performance of the model, calculating area under the curve (AUC), sensitivity, specificity, positive predictive value, and negative predictive value. Hosmer-Lemeshow goodness-of-fit test was used to evaluate model calibration. Decision curve analysis (DCA) was used to evaluate the clinical net benefit of the model. Bootstrap resampling (1000 times) was used for internal validation.

Inter-rater reliability for imaging scoring was assessed using ICC, with ICC > 0.75 considered good reliability. All statistical tests were two-sided tests, with P < 0.05 considered statistically significant.

RESULTS

Baseline characteristics of study subjects

During the study period, a total of 186 patients with traumatic liver rupture were included, including 118 males (63.4%) and 68 females (36.6%), aged 18-78 years, with a mean age of (42.3 ± 15.7) years. According to whether poor prognosis occurred within 30 days, patients were divided into a good prognosis group (n = 142, 76.3%) and a poor prognosis group (n = 44, 23.7%). Comparison of baseline characteristics between the two groups (Table 1).

Table 1 Comparison of baseline characteristics between two groups, n (%).

Variable	Good prognosis group (n = 142)	Poor prognosis group (n = 44)	Statistical value	P value
Demographic characteristics
Age (years), mean ± SD	40.2 ± 14.8	48.5 ± 17.2	t = 3.15	0.002
Male	87 (61.3)	31 (70.5)	χ2 = 1.25	0.264
BMI (kg/m2), mean ± SD	23.4 ± 3.2	24.1 ± 3.8	t = 1.12	0.264
Trauma mechanism
Traffic accident	78 (54.9)	28 (63.6)	χ2 = 4.83	0.185
Fall from height	35 (24.6)	12 (27.3)	-	-
Heavy object impact	21 (14.8)	3 (6.8)	-	-
Others	8 (5.6)	1 (2.3)	-	-
Past medical history
Hypertension	23 (16.2)	12 (27.3)	χ2 = 2.58	0.108
Diabetes	11 (7.7)	6 (13.6)	χ2 = 1.42	0.233
Heart disease	8 (5.6)	4 (9.1)	χ2 = 0.72	0.397
Other diseases	15 (10.6)	8 (18.2)	χ2 = 1.86	0.172
Vital signs
Systolic BP (mmHg), mean ± SD	128.4 ± 18.3	108.2 ± 22.1	t = 5.98	< 0.001
Diastolic BP (mmHg), mean ± SD	78.5 ± 12.6	68.3 ± 15.2	t = 4.35	< 0.001
Heart rate (beats/minute), mean ± SD	88.7 ± 16.2	106.3 ± 21.4	t = 5.72	< 0.001
Respiratory rate (breaths/minute), mean ± SD	19.2 ± 3.4	22.8 ± 4.7	t = 5.21	< 0.001
Body temperature (°C), mean ± SD	36.8 ± 0.6	36.4 ± 0.8	t = 3.42	0.001
Consciousness and injury scores, mean ± SD
GCS score	14.2 ± 1.8	11.8 ± 2.7	t = 6.35	< 0.001
ISS score	12.4 ± 4.3	18.7 ± 6.2	t = 7.12	< 0.001

BMI: Body mass index; GCS: Glasgow Coma Scale; ISS: Injury Severity Score.

Comparison of serum biochemical indicators

Comparison of serum biochemical indicators between the two groups showed that liver function indicators, coagulation function indicators, and inflammatory indicators were all significantly worse in the poor prognosis group (Table 2).

Table 2 Comparison of serum biochemical indicators between two groups, mean ± SD.

Indicator	Good prognosis group (n = 142)	Poor prognosis group (n = 44)	Statistical value	P value
Liver function indicators
ALT (U/L)	156.3 ± 89.2	312.7 ± 145.6	t = 7.65	< 0.001
AST (U/L)	168.4 ± 102.3	387.9 ± 178.4	t = 8.92	< 0.001
TBIL (μmol/L)	18.7 ± 8.4	34.2 ± 15.6	t = 7.23	< 0.001
DBIL (μmol/L)	7.2 ± 3.8	15.4 ± 8.2	t = 7.81	< 0.001
ALB (g/L)	38.6 ± 5.2	31.4 ± 6.8	t = 6.94	< 0.001
ALP (U/L)	89.3 ± 32.1	118.7 ± 45.3	t = 4.65	< 0.001
GGT (U/L)	45.8 ± 28.3	78.2 ± 41.6	t = 5.87	< 0.001
Coagulation function
PT (second)	12.8 ± 1.6	16.4 ± 3.2	t = 8.76	< 0.001
APTT (second)	28.4 ± 4.2	38.7 ± 8.1	t = 9.34	< 0.001
INR	1.08 ± 0.14	1.42 ± 0.28	t = 9.12	< 0.001
Inflammatory indicators
CRP (mg/L)	23.6 ± 18.4	67.8 ± 32.5	t = 9.87	< 0.001
PCT (ng/mL)	0.34 ± 0.28	1.28 ± 0.84	t = 8.45	< 0.001
WBC (× 109/L)	11.2 ± 4.3	16.8 ± 6.7	t = 6.12	< 0.001
NEU% (%)	76.8 ± 9.2	85.4 ± 8.6	t = 5.38	< 0.001
Other indicators
Cr (μmol/L)	78.3 ± 18.2	112.6 ± 35.4	t = 7.23	< 0.001
BUN (mmol/L)	5.8 ± 2.1	9.4 ± 3.8	t = 6.89	< 0.001
LDH (U/L)	287.4 ± 84.6	468.3 ± 156.2	t = 8.76	< 0.001
CK (U/L)	345.2 ± 168.4	612.8 ± 284.7	t = 6.98	< 0.001

ALT: Alanine aminotransferase; AST: Aspartate aminotransferase; TBIL: Total bilirubin; DBIL: Direct bilirubin; ALB: Albumin; ALP: Alkaline phosphatase; GGT: Gamma-glutamyl transferase; PT: Prothrombin time; APTT: Activated partial thromboplastin time; INR: International normalized ratio; CRP: C-reactive protein; PCT: Procalcitonin; WBC: White blood cell count; NEU: Neutrophil count; Cr: Creatinine; BUN: Blood urea nitrogen; LDH: Lactate dehydrogenase; CK: Creatine kinase.

Imaging examination and scoring

All imaging data were independently reviewed and scored by 2 radiologists with more than 10 years of experience in abdominal imaging diagnosis. The inter-rater reliability ICC for imaging scoring was 0.87 (95%confidence interval: 0.82-0.91), indicating good inter-rater reliability. Imaging scoring results showed that all scores in the poor prognosis group were significantly higher than those in the good prognosis group (Table 3).

Table 3 Comparison of imaging scores between two groups, n (%).

Scoring system	Good prognosis group (n = 142)	Poor prognosis group (n = 44)	Statistical value	P value
WSES liver injury grading
Grade I	32 (22.5)	2 (4.5)	χ2 = 28.47	< 0.001
Grade II	45 (31.7)	8 (18.2)
Grade III	38 (26.8)	12 (27.3)
Grade IV	21 (14.8)	15 (34.1)
Grade V	6 (4.2)	6 (13.6)
Grade VI	0 (0)	1 (2.3)
HIS score, mean ± SD	2.4 ± 1.1	3.8 ± 1.3	t = 6.98	< 0.001
LTSS score, mean ± SD
Liver injury extent score	1.2 ± 0.8	2.1 ± 0.9	t = 6.12	< 0.001
Hematoma size score	0.8 ± 0.7	1.4 ± 0.6	t = 5.23	< 0.001
Intraperitoneal bleeding score	1.1 ± 0.9	2.3 ± 0.8	t = 7.86	< 0.001
Active bleeding score	0.6 ± 0.7	1.5 ± 0.7	t = 7.41	< 0.001
LTSS total score	3.7 ± 2.1	7.3 ± 2.0	t = 9.87	< 0.001

WSES: World Society of Emergency Surgery; HIS: Hepatic Injury Scale; LTSS: Long-term services and supports.

Treatment plans and prognostic analysis

The distribution of treatment plans between the two groups is shown in Table 4. The good prognosis group was mainly treated conservatively, while the poor prognosis group was mainly treated surgically, with a statistically significant difference (χ² = 25.34, P < 0.001) (Table 4).

Table 4 Comparison of treatment plans between two groups, n (%).

Treatment plan	Good prognosis group (n = 142)	Poor prognosis group (n = 44)	χ2	P value
Conservative treatment	98 (69.0)	8 (18.2)	25.34	< 0.001
Surgical treatment	44 (31.0)	36 (81.8)	-	-
Liver repair	28 (19.7)	18 (40.9)	7.82	0.020
Partial hepatectomy	12 (8.5)	12 (27.3)	9.15	0.002
Hepatic artery embolization	4 (2.8)	6 (13.6)	6.87	0.009

Among the 44 patients in the poor prognosis group, adverse events included: Death in 12 cases (27.3%), need for emergency surgical intervention in 23 cases (52.3%), serious complications in 16 cases (36.4%), and need for ICU treatment in 19 cases (43.2%). Among these, the 23 cases requiring emergency surgical intervention were all included in the aforementioned 36 surgical treatment patients, with the remaining 13 cases being elective surgery. Some patients experienced multiple adverse events simultaneously.

The 6-month follow-up results showed that the 6-month survival rate in the good prognosis group was 100% (142/142). In the poor prognosis group, 12 cases died within 30 days, and among the remaining 32 patients, 2 cases died during the 6-month follow-up period. Therefore, the 6-month survival rate in the poor prognosis group was 68.2% (30/44), with a statistically significant difference between the two groups (χ² = 42.86, P < 0.001).

Univariate analysis results

Univariate analysis was used to screen factors related to prognosis, with variables having P < 0.1 as candidate variables, among which indicators with P < 0.05 showing statistical significance (Table 5). A total of 28 variables met the inclusion criteria for multivariate analysis with P < 0.1.

Table 5 Univariate analysis results for traumatic liver rupture prognosis (P < 0.05).

Variable	β	SE	Wald χ2	P value	OR value	95%CI
Baseline characteristics
Age	0.031	0.010	9.63	0.002	1.032	1.012-1.053
Vital signs
Systolic BP	-0.034	0.007	23.47	< 0.001	0.967	0.953-0.981
Diastolic BP	-0.047	0.014	11.29	0.001	0.954	0.928-0.981
Heart rate	0.056	0.009	38.45	< 0.001	1.058	1.038-1.078
Respiratory rate	0.211	0.044	23.12	< 0.001	1.236	1.134-1.347
Body temperature	-0.793	0.243	10.64	0.001	0.453	0.281-0.730
Scoring systems
GCS score	-0.361	0.082	19.41	< 0.001	0.698	0.593-0.821
ISS score	0.171	0.031	30.45	< 0.001	1.187	1.118-1.260
Serum biochemical indicators
ALT	0.008	0.002	16.84	< 0.001	1.008	1.005-1.011
AST	0.009	0.002	20.25	< 0.001	1.009	1.006-1.012
TBIL	0.093	0.017	29.78	< 0.001	1.098	1.061-1.136
DBIL	0.136	0.029	22.04	< 0.001	1.146	1.084-1.212
ALB	-0.181	0.037	24.01	< 0.001	0.834	0.777-0.896
ALP	0.018	0.005	12.96	< 0.001	1.018	1.008-1.028
GGT	0.025	0.005	25.00	< 0.001	1.025	1.014-1.036
PT	0.563	0.095	35.12	< 0.001	1.756	1.458-2.115
APTT	0.181	0.029	39.07	< 0.001	1.198	1.132-1.269
INR	2.847	0.624	20.84	< 0.001	17.238	5.081-58.484
CRP	0.051	0.008	40.64	< 0.001	1.052	1.036-1.068
PCT	2.109	0.493	18.31	< 0.001	8.245	3.142-21.654
WBC	0.145	0.031	21.84	< 0.001	1.156	1.087-1.229
NEU%	0.093	0.020	21.56	< 0.001	1.098	1.055-1.143
Cr	0.030	0.005	36.00	< 0.001	1.031	1.020-1.042
BUN	0.261	0.053	24.24	< 0.001	1.298	1.168-1.441
LDH	0.008	0.001	64.00	< 0.001	1.008	1.006-1.010
CK	0.003	0.0005	36.00	< 0.001	1.003	1.002-1.004
Imaging scores
HIS score	0.758	0.158	23.04	< 0.001	2.134	1.562-2.915
LTSS total score	0.629	0.084	56.07	< 0.001	1.876	1.587-2.215

OR: Odds ratio; CI: Confidence interval; BP: Blood pressure; GCS: Glasgow Coma Scale; ISS: Injury Severity Score; ALT: Alanine aminotransferase; AST: Aspartate aminotransferase; TBIL: Total bilirubin; DBIL: Direct bilirubin; ALB: Albumin; ALP: Alkaline phosphatase; GGT: Gamma-glutamyl transferase; PT: Prothrombin time; APTT: Activated partial thromboplastin time; INR: International normalized ratio; CRP: C-reactive protein; PCT: Procalcitonin; WBC: White blood cell count; NEU: Neutrophil count; Cr: Creatinine; BUN: Blood urea nitrogen; LDH: Lactate dehydrogenase; CK: Creatine kinase; HIS: Hepatic Injury Scale; LTSS: Long-term services and supports.

Multivariate logistic regression analysis

Variables with P < 0.1 in univariate analysis were included in multivariate logistic regression analysis using forward stepwise method. Results showed that AST, INR, CRP, GCS score, and ICC total score were independent predictors of traumatic liver rupture prognosis (Table 6). The Hosmer-Lemeshow goodness-of-fit test for the model was χ² = 6.42 (P = 0.601), suggesting good model fit (Figure 1).

Figure 1 Hosmer-Lemeshow goodness-of-fit test for model calibration: Comparison of observed vs expected events across risk deciles. The bar chart displays the number of new psoriasis events (Y-axis) stratified by predicted risk deciles from lowest (decile 1) to highest (decile 10) risk. Green bars represent observed events while orange bars represent expected events based on the prediction model. The plot demonstrates good calibration, with observed and expected events closely aligned across all risk strata, and a clear risk gradient showing increasing event rates from low-risk to high-risk deciles.

Table 6 Multivariate logistic regression analysis for traumatic liver rupture prognosis.

Variable	β	SE	Wald χ2	P value	OR value	95%CI
AST (U/L)	0.008	0.002	12.34	< 0.001	1.008	1.003-1.013
INR	2.534	0.687	13.62	< 0.001	12.596	3.281-48.347
CRP (mg/L)	0.023	0.006	14.78	< 0.001	1.023	1.011-1.036
GCS score	-0.394	0.128	9.48	0.002	0.675	0.525-0.867
LTSS total score	0.542	0.134	16.37	< 0.001	1.720	1.322-2.238
Constant	-2.463	1.234	3.98	0.046	0.085	-

OR: Odds ratio; CI: Confidence interval; AST: Aspartate aminotransferase; INR: International normalized ratio; CRP: C-reactive protein; GCS: Glasgow Coma Scale; LTSS: Long-term services and supports.

Prediction model construction and validation

Based on multivariate analysis results, a prediction model was established: Logit(P) = -2.463 + 0.008 × AST + 2.534 × INR + 0.023 × CRP - 0.394 × GCS + 0.542 × LTSS receiver operating characteristic curve analysis showed that the prediction model had an AUC of 0.923 (95% confidence interval: 0.881-0.965), sensitivity of 86.4%, specificity of 91.5%, positive predictive value of 76.0%, negative predictive value of 95.6%, and Youden index of 0.779 (Figure 2).

Figure 2 Receiver operating characteristic curve of the combined prediction model: Diagnostic performance analysis. The purple curve demonstrates model discrimination performance with an area under the curve of 0.923, indicating excellent predictive accuracy. The diagonal dashed line represents random chance (area under the curve = 0.5). The orange point marks the optimal cutoff threshold that maximizes the balance between sensitivity (true positive rate, Y-axis) and specificity (1 minus false positive rate, X-axis) for clinical risk classification. ROC: Receiver operating characteristic; AUC: Area under the curve.

DCA showed that when the threshold probability was between 0.1-0.8, this prediction model had higher clinical net benefit compared to traditional scoring methods (Figure 3). Bootstrap resampling 1000 times for internal validation showed that the corrected AUC was 0.912, indicating that the model had good internal consistency (Figure 4).

Figure 3 Decision curve analysis of clinical net benefit: Comparison of treatment strategies across threshold probabilities. The teal curve represents the net benefit (Y-axis) of using the prediction model to guide treatment decisions across different threshold probabilities (X-axis). The orange dashed line shows the strategy of treating all patients, while the gray dashed line represents treating no patients. The model demonstrates positive net benefit compared to both default strategies across threshold probabilities ranging from approximately 0.05 to 0.65, with peak net benefit observed at a threshold probability of about 0.25, indicating substantial clinical value for risk-based decision making.

Figure 4 Bootstrap internal validation results (1000 iterations): Distribution of area under the curve values from bootstrap resampling. The histogram displays the distribution of area under the curve values obtained from 1000 bootstrap resamples. The distribution is approximately normal and centered around 0.90-0.91, with values ranging from 0.85 to 0.97. This internal validation demonstrates the stability and robustness of the model’s discriminative ability, with consistently high area under the curve values across bootstrap iterations confirming reliable predictive performance. AUC: Area under the curve.

Comparison of predictive performance of various indicators

Comparison results of predictive performance using serum biochemical indicators alone, imaging scores alone, and combined application (Table 7).

Table 7 Comparison of predictive performance of various indicators and combined models.

Predictive indicator/model	AUC	95%CI	Sensitivity (%)	Specificity (%)	Youden index
Single indicators
AST	0.798	0.728-0.868	75.0	78.2	0.532
INR	0.823	0.758-0.888	79.5	80.3	0.598
CRP	0.836	0.774-0.898	81.8	78.9	0.607
GCS score	0.742	0.665-0.819	72.7	71.8	0.445
LTSS total score	0.845	0.786-0.904	84.1	76.8	0.609
Combined models
Serum biochemical combined	0.887	0.836-0.938	84.1	85.2	0.693
Serum biochemical + imaging combined	0.923	0.881-0.965	86.4	91.5	0.779

AUC: Area under the curve; CI: Confidence interval; AST: Aspartate aminotransferase; INR: International normalized ratio; CRP: C-reactive protein; GCS: Glasgow Coma Scale; LTSS: Long-term services and supports.

DeLong test was used to compare the predictive performance of each model. The combined model of serum biochemical indicators and imaging scores was significantly superior to the serum biochemical indicators combined model (Z = 2.78, P = 0.005), significantly superior to LTSS total score alone (Z = 3.42, P < 0.001), and significantly superior to any single serum biochemical indicator alone (all P < 0.05).

DISCUSSION

This study analyzed clinical data from 186 patients with traumatic liver rupture and established a comprehensive prediction model combining serum biochemical indicators with imaging scores. The study found that this model has good predictive performance for predicting 30-day poor prognosis in patients, providing important scientific evidence for early identification of high-risk patients in clinical practice.

This study found that AST, INR, CRP, GCS score, and LTSS total score are independent predictors of prognosis in patients with traumatic liver rupture. The combined prediction model based on these five indicators achieved an AUC of 0.923, significantly superior to the predictive performance of any single indicator alone. This finding has important clinical significance. First, this model covers multiple dimensions reflecting liver injury severity, coagulation function, inflammatory response, consciousness status, and anatomical injury, enabling comprehensive assessment of the patient's pathophysiological status[11]. Second, all indicators in the model are routine clinical examination items that are easily obtainable and cost-effective, with good clinical operability[12].

AST, as a sensitive indicator of hepatocyte injury, rapidly increases after liver trauma, and its level is closely related to the degree of hepatocyte necrosis[13]. Previous studies have shown that the degree of AST elevation is significantly related to the severity and prognosis of traumatic liver rupture, which is consistent with the results of this study[14]. INR reflects coagulation status. Patients with traumatic liver rupture often experience coagulation dysfunction due to factors such as blood loss and liver function impairment. Elevated INR indicates deteriorating coagulation function and is an important warning signal for poor prognosis[15].

This study found that CRP is an independent predictor, suggesting that inflammatory response plays an important role in the prognosis of traumatic liver rupture. After trauma, the body initiates systemic inflammatory response syndrome, and CRP, as an acute-phase reactant protein, reflects the severity of the body’s inflammatory response[16]. Studies have shown that excessive inflammatory response can lead to microcirculatory dysfunction, tissue hypoxia, and multiple organ dysfunction, which are important causes of death in trauma patients[17]. Therefore, early monitoring of CRP levels helps identify patients with high inflammatory response and timely implementation of anti-inflammatory treatment measures.

Some studies have reported the value of PCT in trauma prognosis prediction, but in the multivariate analysis of this study, PCT did not enter the final model. This may be related to PCT mainly reflecting bacterial infection, while early trauma is primarily sterile inflammatory response[18]. In comparison, CRP can better reflect the inflammatory status in early trauma and therefore has stronger predictive value.

GCS score, as a classic tool for assessing consciousness status, showed good predictive value in this study. This finding is consistent with multiple study results, suggesting that the degree of nervous system injury is closely related to the overall prognosis of trauma patients[19]. Decreased GCS score not only reflects the severity of craniocerebral injury but may also indicate deterioration of overall condition, such as shock and hypoxia[20]. Additionally, changes in consciousness status affect patient cooperation and rehabilitation progress, indirectly affecting prognosis.

LTSS, as an imaging scoring system specifically for evaluating liver trauma, showed good predictive performance in this study. This scoring system comprehensively considers multiple dimensions such as liver injury extent, hematoma size, intraperitoneal bleeding volume, and active bleeding, enabling objective quantification of anatomical liver injury severity[21]. Compared with traditional World Society of Emergency Surgery grading and HIS scoring, LTSS scoring is more detailed and quantified, thus having stronger predictive value[22].

It is worth noting that the inter-rater reliability ICC for imaging scoring in this study was 0.87, indicating good reproducibility of scoring. This has important significance for clinical promotion and application of the scoring system, suggesting that imaging physicians with appropriate training can obtain consistent scoring results[23].

The most important finding of this study is that the comprehensive model combining serum biochemical indicators with imaging scores is significantly superior to using any single indicator alone. This result aligns with the concept of modern precision medicine, which improves disease prediction accuracy by integrating multi-dimensional information[24]. Serum biochemical indicators can reflect the body’s physiological functional status and metabolic changes, while imaging scores objectively assess anatomical injury severity, and both have good complementarity[25].

DCA showed that when the threshold probability was between 0.1-0.8, the combined model had higher clinical net benefit, indicating that this model has practical value in clinical decision-making. When the prediction probability exceeds a certain threshold, clinicians can consider adopting more aggressive treatment measures, such as close monitoring and preventive intervention[26].

Previous research on prognostic prediction of traumatic liver rupture has mainly focused on single indicators or traditional scoring systems. Some studies have reported the relationship between liver enzyme levels and prognosis, but with limited predictive performance[27]. Other studies have explored the predictive value of coagulation function indicators but lacked combined analysis with imaging indicators[28]. This study is the first to systematically integrate serum biochemical indicators and imaging scores, establishing a more comprehensive prediction model.

Compared with similar international studies, this study has a relatively large sample size, more comprehensive included indicators, and improved predictive performance. A similar study reported an AUC of 0.856, which is lower than the 0.923 in this study[29]. This may be related to differences in study population characteristics, included indicators, and modeling methods.

The prediction model established in this study has good clinical application prospects. First, all indicators in the model are routine clinical examination items that do not require additional special testing, with good accessibility[30]. Second, the model has high sensitivity (86.4%) and specificity (91.5%), effectively identifying high-risk patients and avoiding missed diagnosis and misdiagnosis. Third, DCA shows that the model has good clinical net benefit and can provide quantitative support for clinical decision-making.

In practical application, clinicians can calculate prediction probability based on the patient's AST, INR, CRP, GCS score, and LTSS total score. When the probability exceeds a certain threshold, more aggressive treatment measures should be considered, such as ICU monitoring and preventive surgical intervention. This helps improve the prognosis of high-risk patients and reduce mortality[31].

This study has some limitations that need to be recognized. First, as a single-center retrospective study, there is a possibility of selection bias, and the extrapolability of study results may be limited. Second, although Bootstrap internal validation was performed, external cohort validation is lacking, and the universality of the model needs further verification[32]. Third, the study only included short-term prognosis within 30 days and did not evaluate long-term prognostic indicators. While the 30-day endpoint captures the critical early period when most major adverse events occur and clinical decision-making has the most profound impact, future long-term follow-up studies are needed to assess functional recovery, quality of life, and late-onset complications. Fourth, some potential predictive factors such as genetic polymorphisms and nutritional status were not included in the analysis. Additionally, the exclusion of patients with pre-existing liver disease, while methodologically necessary to isolate the independent effect of traumatic injury and avoid confounding from chronic hepatic dysfunction, limits the generalizability of our model to patients with underlying liver conditions. Future studies should develop separate or modified prediction models for this important patient subgroup, as their baseline hepatic reserve and recovery trajectories may differ substantially.

Future research should address several important directions. First, multi-center prospective validation studies are essential to confirm the external validity of our model. Second, dynamic biomarker monitoring through serial measurements at 6 hours, 12 hours, 24 hours, and 48 hours post-admission could potentially enhance predictive accuracy and enable dynamic risk stratification. We are planning prospective studies to investigate this approach. Third, advanced modeling techniques such as machine learning algorithms (e.g., random forests and neural networks) may capture non-linear relationships and complex interactions among variables that traditional logistic regression may not fully elucidate. These comprehensive approaches will help further refine prognostic precision and improve the clinical utility of prediction models in traumatic liver rupture management.

CONCLUSION

In conclusion, this study successfully established a prognostic prediction model for traumatic liver rupture combining serum biochemical indicators with imaging scores. This model has good predictive performance and clinical practicality. AST, INR, CRP, GCS score, and LTSS total score are independent predictors of prognosis, and their combined application is significantly superior to single indicators. This model provides scientific evidence for early identification of high-risk patients and formulation of individualized treatment plans in clinical practice, and is expected to improve the prognosis of patients with traumatic liver rupture. However, multi-center prospective studies still need to be conducted for external validation to ensure the universality and reliability of the model.