Predicting deep vein thrombosis risk in patients with knee fractures: a nomogram-based study

Abstract

Objective

This study aims to develop a predictive model based on clinical and laboratory indicators to assess the risk of deep vein thrombosis (DVT) in patients with knee fractures, thereby facilitating early identification of high-risk individuals and guiding personalised interventions.

Method

Patients with knee fractures were randomly assigned to either a training or validation cohort in a 7:3 ratio. Discrepancy analysis was conducted to compare the variables between the two cohorts. A multiple logistic regression model was employed to identify the independent predictors of lower limb DVT in patients with knee fractures. Receiver operating characteristic (ROC) curves were subsequently utilized to evaluate the diagnostic precision of risk markers, with predictive efficacy assessed through area under the curve calculations.

Result

Among 923 hospitalised patients meeting the study criteria, we identified six independent factors—age, PTA, FIB, TT, D-dimer, and BMI—for nomogram construction from 21 variables. The area under the ROC curve for predicting lower extremity DVT was 0.75 (95% confidence interval [CI]: 0.71–0.79) in the training set and 0.76 in the validation set. Decision curve analysis confirmed the nomogram’s clinical applicability.

Conclusion

The nomogram developed in this study shows intense discrimination, calibration, and clinical utility, suggesting promise for predicting lower extremity DVT and functioning as a quantitative instrument in clinical settings.

Introduction

The development of deep venous clots in leg vessels constitutes a significant clinical concern following extremity trauma, with immobility during convalescence representing a primary predisposing factor. Characterised by insidious onset, elevated mortality and morbidity rates, and a lack of specific clinical manifestations, prompt identification and targeted intervention pose significant challenges(Nicholson et al. 2020). Consequently, the development of a DVT prediction tool is warranted to facilitate the clinical identification of DVT in patients for prompt diagnosis and intervention. While existing studies have explored risk factors for DVT following lower limb fractures(Xiang et al. 2025; Niu et al. 2021; Dou et al. 2022; Cheng et al. 2022), and nomogram-based models predict DVT risk after hip fracture surgery(Bo et al. 2024), there is a paucity of data on DVT risk factors upon hospital admission for knee fracture patients, and a practical risk assessment model is currently lacking. Our previous study found that knee fractures, a subtype of lower extremity fractures, have the highest incidence of DVT(Chang et al. 2021). As the leading orthopaedic referral centre in Eastern China, our hospital possesses specialised expertise in knee fracture management. This has positioned our institution as a primary destination for the surgical treatment of knee fractures within the region. Based on this, the present study aims to thoroughly examine the risk factors associated with DVT in knee fracture patients at the time of hospital admission. Nomograms have emerged as prominent predictive instruments in contemporary clinical practice, with validated applications spanning orthopaedics, oncology, and surgical specialties. In recent years, nomograms have gained widespread acceptance as a predictive method Ain oncology(Wang et al. 2025; Li et al. 2025a, b; Balachandran et al. 2015). It meets the requirements of integrated models and plays a role in advancing personalised medicine. The objective is to establish a nomogram model based on these risk factors to offer healthcare professionals a clinical tool for evaluating DVT risk in knee fracture patients.

Materials and methods

Patients and study design

This retrospective study was conducted at the Affiliated Hospital of Shandong University of Traditional Chinese Medicine, involving a statistical analysis of electronic medical records from January 2019 to December 2023 for patients who underwent surgical intervention for knee joint fractures. The study included consecutive patients aged ≥ 18 years with closed knee joint fractures (including distal femoral, patellar, or tibial plateau fractures). Exclusion criteria comprised recent anticoagulant/antiplatelet use, concomitant non-knee fractures, prior DVT history, active malignancies, incomplete medical documentation, and pregnancy. A total of 1,056 patients with knee joint fractures were initially enrolled, and following screening, 923 patients were included in the final analysis. This retrospective study was performed using the principles outlined in the Declaration of Helsinki and received approval from the Ethics Committee of the Affiliated Hospital of Shandong University of Traditional Chinese Medicine. All patient data have been subject to thorough and irreversible anonymisation procedures. All clinical cases met the inclusion and exclusion criteria. Based on the inclusion/exclusion criteria, the study incorporated a comprehensive set of variables—including lifestyle factors, comorbidities, and laboratory biomarkers—as potential risk factors for subsequent analysis. These variables were selected to ensure a thorough evaluation of their associations with DVT in patients with knee fractures. Patient data were collected, encompassing age, gender, and body mass index (BMI) computed from height and weight measurements. Lifestyle factors considered were smoking and alcohol consumption. Underlying diseases included hypertension and diabetes. Laboratory biomarkers analysed comprised hematocrit (HCT), platelet count (PLT), albumin (ALB), fasting blood glucose (GLU), triglycerides (TG), total cholesterol (CHOL), high-density lipoprotein (HDL), low-density lipoprotein (LDL), D-dimer, fibrinogen (FIB), activated partial thromboplastin time (APTT), thrombin time (TT), Prothrombin Time Activity (PTA), and prothrombin time (PT).

Diagnosis of DVT

Per our institution’s standardised protocol for reducing complications, all patients undergoing surgery for lower limb fractures must complete a mandatory Doppler ultrasound examination of the lower extremities within 24 h of admission. This policy has significantly improved the survival rate of patients with DVT at our hospital. Each ultrasound was performed by one physician and reviewed by a second. The diagnosis of DVT strictly adhered to the criteria outlined in the third edition of the Chinese Medical Association guidelines. Diagnostic criteria for DVT included: venous incompressibility, luminal obstruction or filling defects, loss of respiratory variation in veins above the knee, and insufficient blood flow enhancement in the calf and foot during compression manoeuvres.

Statistical methods

Differences between the DVT and non-DVT groups in terms of lifestyle habits, medical history, and laboratory findings were compared. Continuous variables with a normal distribution were presented as mean ± standard deviation (SD) and assessed using independent samples t-tests. Skewed continuous variables were expressed as median and interquartile range, with between-group comparisons conducted using the Mann-Whitney U test. Categorical variables were described by frequency and percentage (n, %) and compared using the chi-square test. Group allocation was determined solely based on Doppler ultrasonography findings. These findings were deliberately omitted from statistical comparisons to prevent diagnostic circularity. Variables with a significance level below 0.05 in univariate logistic analysis underwent multivariate logistic regression to identify independent risk factors associated with lower extremity DVT development.

A personalised risk prediction nomogram was developed based on the outcomes of multivariate logistic regression. The model’s efficacy and reliability were assessed through rigorous evaluations. Initially, internal validation was conducted utilising bootstrap resampling (1000 iterations) to determine calibration values for performance metrics. Subsequently, predictive accuracy was measured by the AUC for discriminative ability, calibration curve, and the Hosmer-Lemeshow test for calibration assessment. Lastly, DCA was utilised to evaluate the practical utility of the model in clinical settings.

Statistical analyses were performed utilising SPSS 25.0 software. A significance level of P < 0.05 was considered statistically significant.

Results

Patient enrollment process

The study initially included 1,056 cases, from which individuals were excluded based on predefined criteria: recent use of anticoagulant or antiplatelet medications (5 cases), concomitant fractures at other sites (26 cases), history of DVT (12 cases), and incomplete clinical data (90 cases). All DVT classifications analysed were confirmed exclusively through color Doppler ultrasound findings. Ultimately, 923 cases met the inclusion criteria, with 223 of them diagnosed with DVT, as depicted in Fig. 1. We performed Little’s MCAR test to assess the nature of the missing data in the pathology study. The results showed insufficient evidence to reject the hypothesis that the data were missing completely at random (p = 0.24). Consequently, we proceeded with a complete case analysis for the primary modeling, considering the potential risk of bias to be relatively low. The study participants were randomly assigned to either the training group (n = 646) or the validation group (n = 277) at a 7:3 ratio. Baseline comparisons of clinical data between the two groups revealed no statistically significant differences. Table 1 displays the comprehensive demographic and clinical profiles of the study cohort.

Fig. 1.

Data collection and grouping flowchart

Table 1.

A comparison of general data between the DVT group and the control group

Variables	Total (n = 923)	test (n = 277)	train (n = 646)	Statistic	P
Age, n(%)				χ²=5.98	0.426
20	47 (5.09)	12 (4.33)	35 (5.42)
30	135 (14.63)	51 (18.41)	84 (13.00)
40	160 (17.33)	41 (14.80)	119 (18.42)
50	251 (27.19)	73 (26.35)	178 (27.55)
60	188 (20.37)	56 (20.22)	132 (20.43)
70	101 (10.94)	31 (11.19)	70 (10.84)
80	41 (4.44)	13 (4.69)	28 (4.33)
SEX, n(%)				χ²=0.37	0.542
0	781 (84.62)	234 (84.48)	547 (84.67)
1	142 (15.38)	43 (15.52)	99 (15.33)
SMOKING, n(%)				χ²=0.35	0.556
0	797 (86.35)	242 (87.36)	555 (85.91)
1	126 (13.65)	35 (12.64)	91 (14.09)
DRIKING, n(%)				χ²=0.68	0.409
0	807 (87.43)	246 (88.81)	561 (86.84)
1	116 (12.57)	31 (11.19)	85 (13.16)
PT, Mean ± SD	13.55 ± 6.82	13.12 ± 1.00	13.73 ± 8.12	t=-1.26	0.209
PTA, Mean ± SD	101.85 ± 14.12	102.72 ± 14.30	101.48 ± 14.03	t = 1.22	0.223
APTT, Mean ± SD	36.51 ± 4.87	36.33 ± 4.59	36.59 ± 4.99	t=-0.73	0.465
TT, Mean ± SD	16.46 ± 1.72	16.45 ± 2.00	16.47 ± 1.58	t=-0.19	0.847
HCT, Mean ± SD	37.14 ± 5.20	37.27 ± 5.11	37.09 ± 5.24	t = 0.47	0.636
ALB, Mean ± SD	38.99 ± 3.94	39.00 ± 3.74	38.98 ± 4.03	t = 0.05	0.957
CHOL, Mean ± SD	4.68 ± 1.08	4.76 ± 1.10	4.65 ± 1.07	t = 1.48	0.140
HDLC, Mean ± SD	1.20 ± 0.34	1.23 ± 0.33	1.19 ± 0.34	t = 1.54	0.125
LDLC, Mean ± SD	2.87 ± 0.87	2.91 ± 0.88	2.85 ± 0.86	t = 0.94	0.350
BMI, Mean ± SD	24.77 ± 3.47	24.44 ± 3.54	24.92 ± 3.43	t=-1.90	0.058
FIB, M (Q₁, Q₃)	3.51 (3.05, 4.24)	3.62 (3.08, 4.33)	3.49 (3.01, 4.23)	Z=-1.14	0.255
D-dimer, M (Q₁, Q₃)	1.14 (0.56, 2.65)	1.00 (0.59, 2.53)	1.20 (0.55, 2.68)	Z=-0.83	0.409
PLT, M (Q₁, Q₃)	225.00 (193.00, 271.00)	224.00 (191.00, 268.00)	225.00 (193.00, 273.00)	Z=-0.92	0.358
GLU, M (Q₁, Q₃)	5.84 (5.33, 6.91)	5.74 (5.36, 6.53)	5.91 (5.32, 7.10)	Z=-1.36	0.173
TG, M (Q₁, Q₃)	1.09 (0.81, 1.55)	1.14 (0.80, 1.52)	1.08 (0.83, 1.57)	Z=-0.20	0.840

t: t-test, Z: Mann-Whitney test, χ²: Chi-square test

SD: standard deviation, M: Median, Q₁: 1st Quartile, Q₃: 3st Quartile

Univariate and multivariate analysis

Univariate logistic regression analysis of 21 variables in the training set identified statistically significant differences (p < 0.05) between individuals with DVT and those without DVT across eight parameters: age, prothrombin activity (PTA), D-dimer, hematocrit (HCT), fibrinogen(FIB), albumin (ALB), high-density lipoprotein cholesterol (HDLC), and body mass index (BMI). Conversely, no significant disparities (p>0.05) were found in gender, smoking and alcohol history, diabetes, hypertension, thrombin time (TT), total cholesterol (CHOL), prothrombin time (PT), triglycerides (TG), platelet count (PLT), blood glucose (GLU), activated partial thromboplastin time (APTT), or low-density lipoprotein cholesterol (LDLC) between the two groups.

Considering this, the eight abovementioned variables showing significant differences were incorporated into the multivariate logistic regression analysis. Subsequently, six factors were determined to be independent risk factors for DVT following knee fracture, all with a significance level of p < 0.05: Age (1.02; 95%CI: 1.01–1.03), PTA (0.98; 95%CI: 0.97–0.99), FIB (1.24; 95%CI: 1.05–1.46), TT (0.81; 95%CI: 0.70–0.94), D-dimer (1.17; 95%CI: 1.11–1.23) and BMI (1.10; 95%CI: 1.04–1.17).(Table 2).

Table 2.

Univariate and multivariate logistic regression analysis of risk factors for DVT in knee fracture patients

	Univariate		Multivariate
	OR (95%CI)	P	OR (95%CI)	P
SEX
0	1.00 (Reference)
1	1.06 (0.74 ~ 1.52)	0.748
Diabetes
0	1.00 (Reference)
1	1.15 (0.66 ~ 2.00)	0.633
SMOKING
0	1.00 (Reference)
1	0.79 (0.46 ~ 1.36)	0.392
Hypertension
0	1.00 (Reference)
1	0.92 (0.56 ~ 1.53)	0.758
DRIKING
0	1.00 (Reference)
1	0.94 (0.55 ~ 1.61)	0.831
Age	1.02 (1.01 ~ 1.03)	< 0.001	1.02 (1.01 ~ 1.03)	0.008
PT	0.99 (0.96 ~ 1.02)	0.549
PTA	0.97 (0.96 ~ 0.99)	< 0.001	0.98 (0.97 ~ 0.99)	0.025
FIB	1.24 (1.07 ~ 1.45)	0.005	1.24 (1.05 ~ 1.46)	0.011
APTT	1.03 (0.99 ~ 1.06)	0.134
D-dimer	1.20 (1.13 ~ 1.26)	< 0.001	1.17 (1.11 ~ 1.23)	< 0.001
HCT	0.93 (0.90 ~ 0.96)	< 0.001
PLT	1.00 (1.00 ~ 1.00)	0.496
GLU	1.05 (0.98 ~ 1.12)	0.151
TG	0.86 (0.71 ~ 1.02)	0.089
ALB	0.90 (0.86 ~ 0.95)	< 0.001
CHOL	0.85 (0.71 ~ 1.01)	0.057
HDLC	0.68 (0.39 ~ 1.20)	0.181
LDLC	0.91 (0.74 ~ 1.12)	0.381
BMI	1.09 (1.04 ~ 1.15)	0.001	1.10 (1.04 ~ 1.17)	< 0.001
TT	0.82 (0.71 ~ 0.93)	0.003	0.81 (0.70 ~ 0.94)	0.005

OR: Odds Ratio, CI: Confidence Interval

*P < 0.05 was considered statistically significant

Nomogram development and validation

This study identified age, prothrombin time activity (PTA), fibrinogen (FIB), thrombin time (TT), D-dimer, and body mass index (BMI) as independent predictors of DVT. Linear tests using four-node restricted cubic splines (RCS) were conducted on continuous variables. The results indicate no statistical significance for the nonlinear components of all variables (P > 0.05): Age (χ² = 2.15, P = 0.34), PTA (χ² = 1.82, P = 0.42), FIB (χ² = 3.47, P = 0.87), D-dimer (χ² = 4.28, P = 0.23), BMI (χ² = 1.23, P = 0.55), and TT (χ² = 0.89, P = 0.18). Consequently, all continuous variables are incorporated as linear terms in the final model. Cook’s distance was used to detect influential observations, with no significant outliers impacting model coefficients. A nomogram model was developed based on these predictors to forecast DVT occurrence. The model was constructed using data from the training cohort, depicted in Fig. 2. In the nomogram, age was represented as a value of 20 for the 18–29 age group and 80 for individuals aged 80 years and above, encompassing adults aged 18 and older. The bar chart visually displayed actual values of fibrinogen, thrombin time, D-dimer, PTA, and BMI.

Fig. 2.

Nomogram prediction model for risk of DVT in the lower limbs after knee joint fracture

Following 1000 bootstrap internal validations, the C-statistic for the training set was 0.817, while for the validation set, it was 0.783. The training set model demonstrated an AUC of 0.75 (95% CI 0.71–0.79), indicating a strong alignment between predicted and actual probabilities for the training set. In the validation set, the AUC was 0.76 (95% CI 0.70–0.83) (Fig. 3).The calibration curve reveals that the training set has a calibration intercept of -0.03(95% CI -0.15–0.09),and a calibration slope of 0.98(95% CI 0.92–1.04), the validation set exhibits a calibration intercept of -0.12 (95% CI -0.20 to -0.04) and a slope of 0.95 (95% CI 0.91 to 1.01). These results suggest the model may be slightly overfitted and tends to slightly overestimate risk. However, such deviations are generally considered acceptable in clinical practice (Fig. 4). Furthermore, the Hosmer-Lemeshow test yields P-values of 0.7842 for the training set and 0.4449 for the validation set, indicating no significant fitting bias. DCA revealed that within the probability threshold range of 0.10 to 0.70 for the training cohort and 0.10 to 0.90 for the validation cohort (Fig. 5), using the nomogram to predict DVT incidence provided net benefits when implementing interventions (such as repeat color Doppler ultrasonography, upgrading venous thromboembolism prophylaxis, and potentially delaying surgery) for knee fracture patients across both cohorts. These results underscore the model’s strength and clinical relevance.

Fig. 3.

ROC curve evaluation of the DVT risk stratification algorithm performance in derivation (A) and internal validation (B) cohorts following knee fractures

Fig. 4.

Risk calibration curve for lower extremity DVT following knee joint fractures in the training set (A) and validation set (B). The X-axis indicates the model’s predicted probability, while the Y-axis indicates the actual likelihood. The diagonal dashed line represents a perfect prediction of the ideal model. The solid line illustrates the nomogram’s performance, with closer proximity to the diagonal dashed line indicating enhanced predictive consistency of the nomogram

Fig. 5.

Decision curve analysis of the risk prediction model for DVT after knee joint fractures in the training set (A) and validation set (B)

Discussion

Despite standard preventive measures for DVT in hospitalised patients, including physical and pharmacological interventions, asymptomatic individuals often do not receive follow-up lower limb ultrasound screening. The potential for undetected DVT to cause severe complications underscores the necessity for additional practical diagnostic tools beyond routine ultrasound. This study introduces a nomogram model for diagnosing lower limb DVT, based on blood and biochemical tests conducted at admission. This model addresses the demand for straightforward and efficient diagnostic solutions by utilizing readily available predictive indicators in clinical settings.

Nomograms are extensively utilized in clinical prediction models to visually represent the combined predictive value of multiple risk factors(Yang et al. 2022; Bian et al. 2021). The nomogram created in this research functions as a visual aid that consolidates various independent risk factors into a quantifiable probability assessment system for the early detection of lower extremity DVT. With a substantial number of sample events (223 DVT events in total, with 6 predictors ultimately incorporated, resulting in an events per predictor (EPP) value of 37.2), and by effectively addressing potential nonlinear relationships between continuous variables and outcomes through restricted cubic splits (RCS), the model maintains statistical robustness and reliability. This tool integrates routinely collected clinical and biochemical parameters upon admission - such as D-dimer, platelet count, type of injury, and comorbidities - to simplify complex multivariate relationships into an intuitive graphical interface. This facilitates swift assessment of individual patient risk by clinicians without the need for specialized statistical software, bridging theoretical risk prediction with clinical decision-making. Consequently, this transforms abstract predictive analysis into actionable diagnostic and treatment recommendations.

A total of 923 patients with knee joint fractures were studied, among whom 223 cases experienced DVT, resulting in an incidence rate of 24.2%. Previous research has shown varying incidence rates of DVT post lower limb fractures(Niu et al. 2021; Chang et al. 2021; Wang et al. 2018; Paydar et al. 2016), likely influenced by study settings, demographics, and DVT detection timing. The study participants were predominantly from North China, with a significant number referred from other regions to our provincial-level tertiary medical institution, potentially leading to delayed hospital admissions post-fracture. This delay could contribute to the relatively high DVT incidence observed in this study.Moreover, 133 patients were excluded from the study for various reasons, which could have impacted the elevated DVT incidence rate. Among the excluded cases, 90 were omitted due to incomplete data, with a DVT incidence of 23.3% in this subgroup. Comparison of baseline characteristics between included and excluded patients revealed no significant differences (p > 0.05), indicating random occurrence of missing data. Hence, we assert that the complete case analysis is warranted in this study, ensuring the reliability of the results.

This study confirmed that advanced age is an independent risk factor for DVT after knee fracture, which is consistent with the majority of previous literature(Zhang et al. 2020; Wu et al. 2024; Cong et al. 2024). Furthermore, our results showed that the mean age of patients in the DVT group (53 years) was significantly higher than that in the non-DVT group (48.21 years), suggesting a gradual increase in DVT risk with age. This association may be attributed to reduced mobility, diminished venous return, and a higher burden of comorbidities among older patients. However, the odds ratio (OR) for each additional year of age found in our study was lower than that reported by Cong et al. (OR 1.27)(Cong et al. 2024). This discrepancy may stem from heterogeneity in the age distribution of the study populations: our cohort included a larger proportion of middle-aged patients (mean age ~ 50 years), whereas Cong et al. focused on an older population (mean age > 70 years), in whom age-related physiological decline may be more pronounced.

This study further supports the positive correlation between obesity (as characterized by BMI) and the risk of DVT, which is consistent with the consensus in this field(Klovaite et al. 2015; Horvei et al. 2016; Aggarwal et al. 2022). However, it is noteworthy that the effect size of BMI in this study (OR 1.10) was lower than that reported by Aggarwal in a similar population (OR 1.27)(Aggarwal et al. 2022). A possible explanation lies in the limitations of BMI as an indicator of obesity: it does not distinguish between fat and muscle mass, nor does it reflect body fat distribution. Zhang et al.‘s study incorporated waist circumference measurements and found a stronger association between central obesity and DVT, which may explain the higher OR value they reported. Therefore, while our results confirm the importance of obesity, they also suggest that future studies employing more precise body composition indicators—such as waist-to-hip ratio or visceral adipose area—may more clearly elucidate the true strength of the association between obesity and DVT.

The literature on the independent predictive value of coagulation parameters like PTA, TT, and FIB for deep vein thrombosis (DVT) presents conflicting conclusions(Li et al. 2024; Kobayashi et al. 2023; Huang et al. 2025; Feng et al. 2024).This study revealed that even after adjusting for factors such as age and BMI, these parameters exhibited significant independent predictive capability. Particularly, PTA (prothrombin time activity, presented as a percentage) displayed an inverse relationship with DVT risk, indicating that higher PTA levels were linked to a reduced DVT risk. This contradicts findings by Kobayashi(Kobayashi et al. 2023). PTA typically mirrors the performance of the extrinsic coagulation pathway, with elevated levels usually signaling enhanced coagulation function. The negative correlation between PTA and DVT risk in this study suggests that in some clinical settings, PTA may indicate better anticoagulation management or improved liver function, and thus protect against DVT. The differences between studies may arise from several factors. First, differences between standardization of detection methods (say differences in reagents and instruments) may directly affect comparison. Second, samples were collected at admission, and other studies might collect samples after surgery or after anticoagulation therapy, depending on coagulation marker levels. Third, since we used multiple predictors, the association may be a false positive based on multiple comparisons.

Elevated D-dimer levels typically indicate ongoing thrombotic activity and secondary hyperfibrinolysis within the body, which are commonly associated with conditions such as PE, DVT, and DIC(Tayal et al. 2024). However, despite its high sensitivity, D-dimer has limited specificity and is easily influenced by non-thrombotic factors—such as inflammation, pregnancy, and recent surgery—which constrains its predictive value when used alone. Therefore, in clinical risk assessment, integrating it with other predictors to develop a comprehensive predictive model is of great significance for improving diagnostic accuracy and risk stratification capability in DVT.

Several factors demonstrated statistical significance in univariate analysis (P < 0.05), such as PT, PLT, GLU, CHOL, HCT, ALB, and HDLC, aligning with findings from prior studies(Smith et al. 2025; Liu et al. 2025; Li et al. 2025a, b; Kusnezov et al. 2025). However, these factors did not maintain significance in multivariate analysis, potentially attributed to multicollinearity or insufficient statistical power. Therefore, this study requires a larger sample size and support from data from other research centres.

The diagnostic model we have developed can complement the findings obtained through color Doppler ultrasound in evaluating the risk of DVT in the lower extremities of individuals with knee joint fractures. Patients identified with higher scores should receive proactive physical and pharmacological preventive interventions, with the option of a repeat color Doppler ultrasound examination if deemed necessary. This approach aims to facilitate early detection and management of lower extremity DVT. The diagnostic model can provide a reference for clinical practice, particularly emphasising older patients, individuals with obesity, and those presenting with abnormal coagulation parameters, thereby aiding in DVT prevention. Early diagnosis of DVT can effectively reduce the economic burden of patients(Heerink et al. 2023). Given the elevated risk and potential implications of DVT, including surgical delays and increased financial burdens, the risk factors integrated into the model are straightforward and readily accessible in clinical settings. This approach aims to enhance the accuracy of lower limb DVT diagnosis while mitigating the risks associated with missed or delayed diagnoses.

Limitation

This study is subject to several limitations. First and foremost, the model was developed and validated on a single-center dataset. Although internal validation demonstrated good discriminative ability, the model’s performance is inherently tied to our local prehospital care systems, transfer protocols, and in-hospital treatment standards. And we did not calculate the sample size at the beginning of the experiment. This implies that its calibration might degrade when directly applied to centers with different practice patterns. We strongly recommend thorough external validation and model calibration within the target population before any clinical implementation. Secondly, the presence of missing data in certain patients, particularly concerning the incidence of DVT following knee fractures, could potentially impact the robustness of our conclusions. Thirdly, there is an ongoing discourse in the literature regarding the applicability of nomograms within specific ethnic populations. Fourthly, restricting DVT assessment to ≤ 24 h post-trauma precludes analysis of delayed thrombotic events, potentially underestimating late-onset DVT incidence and its distinct risk profiles. Finally, exclusion of clinical indicators and lifestyle factors such as physical activity levels and immobilisation duration necessitates further data to strengthen the findings. We are currently engaged in the process of multi-centre external validation through collaborative efforts with other Chinese institutions and intend to broaden the inclusion criteria in upcoming trials. Despite recognising these limitations, they delineate pivotal avenues for future research endeavours.

Conclusion

This study identifies age, PTA, TT, FIB, D-dimer, and BMI as independent risk factors for DVT after knee fractures. It develops a nomogram model for individualised risk assessment, offering a practical tool for identifying high-risk patients. The study shows strong application potential by focusing on a specific patient group and using readily available clinical variables. However, as a single-centre retrospective study, future validation and optimisation with large-sample, multi-centre data are necessary.

Authors’ contributions

Zhang Yongkui provided the article’s ideas, Gong Rui and Wang Cong collected the data, Ximin conducted the data analysis and wrote the article, and all authors have reviewed this article and are responsible for it.

Funding

The authors received no financial support for this article’s research, authorship, and/or publication.

Data availability

The datasets generated and/or analysed during the current study are not publicly available as they contain information that could compromise the privacy of research participants, but they are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study adheres to the Declaration of Helsinki. The Ethics Committee of Affiliated Hospital of Shandong University of Traditional Chinese Medicine has approved this study. Ethics Review Approval Number: AF/SC-08/03 − 0.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.