Development and validation of the CHI-CAT score: a novel tool for assessing cancer-associated venous thromboembolism risk in China Authors: Xiaoli Qin, Xiurong Gao, Qin He, Min Xu, Yujie Yang, Qian Jiang Published: 2025 ABSTRACT BackgroundIdentifying patients at risk for VTE and stratifying that risk is critical to determine who will benefit from thromboprophylaxis. China’s clinical practice still faces limitations, particularly regarding systematic and practical VTE risk assessment tools for cancer patients. This study aims to develop and validate the CHIna’s Cancer-Associated venous Thromboembolism risk assessment tool (CHI-CAT score) compared to the Khorana score.MethodsCancer patients aged 18 or older hospitalized in the Third People’s Hospital of Chengdu from March 2022 to September 2023 were retrospectively involved in this study. The cohort was divided into two groups based on whether they experienced VTE. Patients with VTE were compared to those without VTE according to their Khorana and CHI-CAT scores, respectively. Model discrimination performance was comprehensively evaluated using the standard sensitivity, specificity, maximum approximate entry index, and the area under the receiver operating characteristic curve(AUC).ResultsData from 868 patients were analyzed. The CHI-CAT score was developed from a derivation cohort (n = 428) and further assessed in the validation cohort (n = 440). The predictive threshold for the risk assessment tool was set at 8 points. The AUC of the CHI-CAT score was 0.808 (95% CI, 0.715–0.901), with a significance level of P < 0.001. The verification results indicate that the AUC of the CHI-CAT and Khorana scores were 0.782 and 0.662, respectively, with a significance level of P < 0.001. The sensitivity of these two models was 0.826 and 0.478, respectively, while the specificity was 0.640 and 0.751.ConclusionsThe CHI-CAT score has been developed and validated, demonstrating relatively comprehensive components and a predictive capability that exceeds the Khorana score. However, its reliability and validation need improvement through testing in larger internal and external retrospective cohorts. Introduction Venous thromboembolism (VTE), which encompasses both deep vein thrombosis (DVT) and pulmonary embolism (PE) [1], is a major cause of morbidity and mortality in patients with cancer and poses significant financial and resource usage challenges for the healthcare system [2, 3]. Research has shown that individuals with cancer are at a significantly higher risk of developing thrombosis, with the risk being 4 to 7 times greater than in those without cancer, and this group accounts for approximately 20% of all VTE cases [4, 5]. VTE in cancer patients is associated with several significant consequences, including increased inpatient mortality and a higher risk of VTE recurrence. Patients often require long-term anticoagulation, which carries a high bleeding risk. Additionally, these complications can lead to a diminished quality of life and greater utilization of healthcare resources [6]. Cancers and VTE have a reciprocal relationship: Cancers elevate the risk of VTE, which subsequently increases the mortality risk among cancer patients. However, the symptoms associated with cancer-related VTE are often non-specific and can be obscured by other cancer complications. In clinical practice, VTE typically undergoes examinations and treatments only when patients exhibit apparent clinical symptoms. Additionally, some medical staff may lack the experience necessary to recognize these symptoms promptly, leading to delays in diagnosis and treatment. This situation contributes to a high rate of missed diagnoses for VTE in cancer patients. So, Accurate risk assessment, along with prompt diagnosis and treatment of this process, is essential for preventing death in this high-risk population [7]. Nevertheless, routine primary prophylaxis is not feasible for all cancer patients due to the increased risk of bleeding and concerns regarding their quality of life. Therefore, it is critical to identify patients at risk for VTE and stratify that risk to determine who will benefit from thromboprophylaxis [7]. An ideal VTE risk assessment tool must identify VTE risk and predict the thrombus risk level accurately. The tool, which evaluates the number of risk factors alone, is inconvenient for users to interpret and needs more clinical operability. Hence, from 2008 to 2021, various cancer-related VTE assessment tools were developed, including five in the United States [8–12], two in Italy [13, 14], two in Spain [15, 16], and one in China [17]. In 2008, Khorana et al. developed the Khorana predictive score, the first VTE risk assessment tool for outpatient chemotherapy cancer patients [8], which was recommended by the International Society of Thrombosis and Haemostasis (ISTH) for those with a Khorana score ≥ 2, no drug interaction, and no risk of bleeding [18]. However, previous studies [19–21] have demonstrated that existing tools for assessing VTE risk are highly biased and yield inconsistent verification results. Addressing the lack of reliability and validity is crucial, as these factors can significantly influence patient outcomes, including morbidity and mortality. The existing tools are based on the Western population, and the gap between China and the West may lead to limited practice in China. In contrast, the tool based on the Chinese population is only used for lung cancer patients, which also needs further verification [17]. China’s clinical practice also faces challenges, particularly in terms of having systematic and practical VTE risk assessment tools specifically designed for cancer patients. So, In the early stages of our research, our group employed Delphi and the analytic hierarchy process(AHP) to construct VTE risk predictors related to cancer patients in China, including three primary indicators, ten secondary indicators, and 40 tertiary indicators [22]. However, it cannot distinguish high-risk patients with VTE. Therefore, this study aims to develop and validate the CHIna’s Cancer-Associated VTE risk assessment tool, namely the CHI-CAT score, and compare it to the Khorana score through a retrospective cohort study. Materials and methods Study design In the initial phase, we established a preliminary index pool through a systematic review and identified the risk predictors of cancer-related VTE using Delphi and AHP methodologies [22]. This study aims to validate clinically. The detailed design process is illustrated in Fig. 1. Fig. 1A study design A study design Study population and procedures The present study population was retrospectively analyzed, including cancer patients hospitalized in the Third People’s Hospital of Chengdu from March 2022 to September 2023. Inclusion and exclusion criteria Inclusion criteria included age ≥ 18; the first admission’s medical records were selected for patients with multiple admissions; hospital stay ≥ three days; and complete primary data, including white blood cell count, platelet count, hemoglobin count, and D-dimer level. Exclusion criteria included patients with VTE, blood or thyroid tumors, uterine fibroids, aneurysms, pituitary tumors, benign pharyngeal tumors, or cervical epitheliomatous lesions; incomplete case information; tumor diagnosis more than ten years ago or unknown; and participants in drug clinical trials. Diagnostic criteria Cancer: Cytological or pathological diagnosis. DVT can be confirmed through CT venography, magnetic resonance venography, lower limb color Doppler ultrasound, or venography. PE, CT pulmonary angiography, pulmonary angiography, nuclide pulmonary ventilation/perfusion scan, and magnetic resonance pulmonary angiography were positive in any of the tests. PE can also be diagnosed if the ultrasound shows a thrombus in the right atrium, right ventricle, and pulmonary artery, and the clinical manifestations are consistent with pulmonary embolism [23]. Data collection In statistical variable analysis, the test’s sample size was recommended to be 10 to 20 times that of the test items. The risk assessment tool developed in this study includes three primary, ten secondary, and 40 tertiary indicators, so the study population needs more than 400 cases. Patient basic information and relevant index data for the tool items were obtained through a retrospective review of medical records. Model discrimination performance was tested using the standard sensitivity and specificity [24]. SPSS calculated the sensitivity and specificity values for each tangent point. The tangent point with the closest entry index value was then chosen to determine the optimal critical value, balancing sensitivity and specificity [25]. Statistical analysis Descriptive statistics were employed to summarize patient characteristics. The cohort was divided into two groups based on the occurrence of VTE. Patients who developed VTE were compared to those who did not develop VTE in terms of individual Khorana and CHI-CAT scores, as well as baseline characteristics. The kappa coefficient was used to evaluate the results’ consistency test, usually from 0 to 1. A Kappa coefficient greater than 0.7 indicates highly consistent evaluation results, while a coefficient less than 0.4 suggests low consistency. The model’s discrimination performance was tested using sensitivity, specificity, and the area under the receiver operating characteristic curve(AUC) [26]. The AUC, indicating discrimination quality, ranged from 0.5 to 1.0. Prediction ability was classified as poor (0.5–0.7), moderate (0.7–0.9), and sound (0.9-1). All statistical analyses were performed using IBM SPSS Statistics 22 (New York, NY, USA). Results Construct the risk threshold of the CHI-CAT Essential characteristics of included cases Cancer patients hospitalized from March to December 2022 were selected to construct the tool risk threshold. 428 cases were reviewed, including 24 patients with VTE and 404 patients without VTE (P < 0.001), the difference being statistically significant. Table 1 for details. Table 1Essential characteristics of included casesBaseline featuresrisk scoreVTE events(n = 24)None VTE(n = 404)P-valueAsian population-124(100)404(100)BMI>24 kg/m213(12.5)114(28.2)BMI ≥ 28 kg/m220(0.00)36(8.9)Age ≥ 60 years old116(66.7)25(62.4)Female sex117(70.8)186(46)Permanent residence: Plateau (≥ 2500 m)11(4.2)8(2.0)Decreased activity ability26(25.0)8(2.0)Cardiovascular disease216(66.7)172(42.6)Lung or kidney disease13(12.5)72(17.8)Severe infection18(33.3)26(6.4)VTE Personal History10(0.00)0(0.0)VTE Family History10(0.00)0(0.0)Hospitalization history (≤ 1 month)17(29.2)62(15.3)History of lower limb varicose veins10(0.00)4(1.0)D-Dimer ≥ 500ug/L320(83.3)114(28.2)Plt ≥ 350 × 109/L22(8.3)16(4.0)TAT>4ng/mL220(83.3)0(0.0)Cancer diagnosis time ≤ 6months111(45.8)144(35.6)Stomach cancer52(8.3)38(9.4)Pancreatic cancer53(12.5)6(1.5)Brain tumor50(0.00)0(0.0)Lung cancer29(37.5)96(23.8)Lymphatic cancer30(0.00)30(7.4)Gynecological tumor21(4.2)30(7.4)Urogenital cancer20(0.00)40(9.9)Cancer compresses blood vessels/lymphatic tissue51(4.2)20(5.0)Transfer319(79.2)104(25.7)Invasion of lymph nodes26(25.0)98(24.3)Mediastinal involvement28(33.3)28(6.9)Platinum12(8.3)30(7.4)Hormonal replacement therapy30(0.00)0(0.0)High dose dexamethasone (≥ 160 mg/cycle)20(0.00)2(0.5)Antiangiogenic drugs22(8.3)12(3.0)Anticoagulants-64(16.7)13(3.2)Adriamycin11(4.2)0(0.0)Erythropoietic stimulating factor30(0.00)4(1.0)Surgical site (hip or knee surgery)50(0.00)0(0.0)Surgical time ≥ 2 h30(0.00)2(0.5)Surgical type (laparoscopic surgery, arthroscopic surgery)20(0.00)4(1.0)CVC10(0.00)8(2.0)VTE rating((mean ± SD)11.21 ± 4.686.23 ± 3.67<0.001TAT Thrombin antithrombin complex, CVC Central venous catheterization Table 1Essential characteristics of included casesBaseline featuresrisk scoreVTE events(n = 24)None VTE(n = 404)P-valueAsian population-124(100)404(100)BMI>24 kg/m213(12.5)114(28.2)BMI ≥ 28 kg/m220(0.00)36(8.9)Age ≥ 60 years old116(66.7)25(62.4)Female sex117(70.8)186(46)Permanent residence: Plateau (≥ 2500 m)11(4.2)8(2.0)Decreased activity ability26(25.0)8(2.0)Cardiovascular disease216(66.7)172(42.6)Lung or kidney disease13(12.5)72(17.8)Severe infection18(33.3)26(6.4)VTE Personal History10(0.00)0(0.0)VTE Family History10(0.00)0(0.0)Hospitalization history (≤ 1 month)17(29.2)62(15.3)History of lower limb varicose veins10(0.00)4(1.0)D-Dimer ≥ 500ug/L320(83.3)114(28.2)Plt ≥ 350 × 109/L22(8.3)16(4.0)TAT>4ng/mL220(83.3)0(0.0)Cancer diagnosis time ≤ 6months111(45.8)144(35.6)Stomach cancer52(8.3)38(9.4)Pancreatic cancer53(12.5)6(1.5)Brain tumor50(0.00)0(0.0)Lung cancer29(37.5)96(23.8)Lymphatic cancer30(0.00)30(7.4)Gynecological tumor21(4.2)30(7.4)Urogenital cancer20(0.00)40(9.9)Cancer compresses blood vessels/lymphatic tissue51(4.2)20(5.0)Transfer319(79.2)104(25.7)Invasion of lymph nodes26(25.0)98(24.3)Mediastinal involvement28(33.3)28(6.9)Platinum12(8.3)30(7.4)Hormonal replacement therapy30(0.00)0(0.0)High dose dexamethasone (≥ 160 mg/cycle)20(0.00)2(0.5)Antiangiogenic drugs22(8.3)12(3.0)Anticoagulants-64(16.7)13(3.2)Adriamycin11(4.2)0(0.0)Erythropoietic stimulating factor30(0.00)4(1.0)Surgical site (hip or knee surgery)50(0.00)0(0.0)Surgical time ≥ 2 h30(0.00)2(0.5)Surgical type (laparoscopic surgery, arthroscopic surgery)20(0.00)4(1.0)CVC10(0.00)8(2.0)VTE rating((mean ± SD)11.21 ± 4.686.23 ± 3.67<0.001TAT Thrombin antithrombin complex, CVC Central venous catheterization Essential characteristics of included cases TAT Thrombin antithrombin complex, CVC Central venous catheterization ROC curve value ROC curves were drawn according to the VTE risk scores of the two groups to determine the optimal threshold for predicting thrombus. Fig. 2 shows that the AUC of this study was 0.808, the standard error was 0.048, and the 95% confidence interval was 0.715–0.901, P < 0.001, indicating high discrimination capacity. Fig. 2Evaluation of the discriminating ability of the CHI-CAT score(receiver operating characteristic curve) Evaluation of the discriminating ability of the CHI-CAT score(receiver operating characteristic curve) Determining the tool risk threshold In this study, the maximum value of the Youden index is 0.500, corresponding to 70.8% sensitivity and 79.2% specificity (Table 2). The prediction threshold for the determination tool was initially set at 8.5 points. However, since the risk scores generated by this tool are all integers, the VTE risk diagnosis threshold was ultimately rounded to 8 points. Therefore, if a patient’s assessment score was eight or higher, they can be considered at risk for VTE. Table 2Maximum entry indexVTE ratingSensitivity1 - specificitySpecificityYouden index=sensitivity + specificity-1-21100-0.510.9780.0220.0220.510.9380.0620.0621.50.9580.9270.0730.0312.50.9580.8540.1460.1043.50.9580.7750.2250.1834.50.9170.6290.3710.2885.50.9170.5450.4550.3726.50.9170.4270.5730.4907.50.7920.3030.6970.4898.5 0.708 0.208 0.792 0.5009.50.6250.1740.8260.45110.50.50.1240.8760.37611.50.4580.090.910.36812.50.3750.0620.9380.31313.50.250.0390.9610.21114.50.250.0280.9720.22215.50.250.0170.9830.23316.50.1670.0060.9940.16117.50.1250.0060.9940.119190.042010.042210010 Table 2Maximum entry indexVTE ratingSensitivity1 - specificitySpecificityYouden index=sensitivity + specificity-1-21100-0.510.9780.0220.0220.510.9380.0620.0621.50.9580.9270.0730.0312.50.9580.8540.1460.1043.50.9580.7750.2250.1834.50.9170.6290.3710.2885.50.9170.5450.4550.3726.50.9170.4270.5730.4907.50.7920.3030.6970.4898.5 0.708 0.208 0.792 0.5009.50.6250.1740.8260.45110.50.50.1240.8760.37611.50.4580.090.910.36812.50.3750.0620.9380.31313.50.250.0390.9610.21114.50.250.0280.9720.22215.50.250.0170.9830.23316.50.1670.0060.9940.16117.50.1250.0060.9940.119190.042010.042210010 Maximum entry index Validation of the CHI-CAT Consistency comparison between two risk assessment tools 440 patients were included in the validation study. VTE risk assessment was conducted using the Khorana and CHI-CAT scores, respectively. According to the statistical results, the Kappa coefficient was 0.006, less than 0.4, indicating that the evaluation results of the two risk assessment tools need to be more consistent. P = 0.040, the difference between the two scores was statistically significant. Among the 440 patients, 180 (36.4%) were in the high-risk CHI-CAT category, 260 (63.6%) were in the low-risk category. The incidence of VTE was 10.5% (46/440) overall, occurring in 21.1% (38/180) of the high-and 3.1% (8/260) of the low-risk category. Meanwhile, among the 440 patients, 120 (27.3%) were in the intermediate and high-risk Khorana category, 320 (62.7%) were in the low-risk category. The incidence of VTE was 10.5% (46/440) overall, occurring in 18.3% (22/120) of the intermediate- and high-, 7.5% (24/320) of the low-risk category (Table 3). Table 3Risk stratification of two VTE risk assessment toolsCHI-CAT score categoryKhorana score categoryHigh risk, n(%)Low risk, n(%)Medium to high risk, n(%)Low risk, n(%)VTE(46)38(82.61)8(17.39)22(47.83)24(52.17)NO VTE(394 )142(36.04)252(63.96)98(24.87)296(75.13) Table 3Risk stratification of two VTE risk assessment toolsCHI-CAT score categoryKhorana score categoryHigh risk, n(%)Low risk, n(%)Medium to high risk, n(%)Low risk, n(%)VTE(46)38(82.61)8(17.39)22(47.83)24(52.17)NO VTE(394 )142(36.04)252(63.96)98(24.87)296(75.13) Risk stratification of two VTE risk assessment tools Predictive performance of the two risk assessment tools The CHI-CAT and Khorana model was utilized to assess the VTE of 440 patients. The sensitivity of the two tools was 0.826 and 0.478, respectively, while the specificity was 0.640 and 0.751. Table 4 for details. ROC curves were drawn for the scores of the risk assessment results of the two tools, as shown in Fig. 3. The AUC values were 0.782 and 0.662, respectively, with a P-value of less than 0.001, indicating that the difference was statistically significant. The VTE prediction accuracy of the CHI-CAT score was better than that of the Khorana risk assessment tool. Table 4Comparison of predictive performance indicators of the two toolsCategoryAUCSE P 95CI %SensitivitySpecificityYonden indexCHI-CAT0.7820.0320.0000.719 ~ 0.8460.8260.6400.461Khorana0.6620.0400.0000.583 ~ 0.7410.4780.7510.241 Table 4Comparison of predictive performance indicators of the two toolsCategoryAUCSE P 95CI %SensitivitySpecificityYonden indexCHI-CAT0.7820.0320.0000.719 ~ 0.8460.8260.6400.461Khorana0.6620.0400.0000.583 ~ 0.7410.4780.7510.241 Comparison of predictive performance indicators of the two tools Fig. 3ROC curve of the CHI-CAT and Khorana score ROC curve of the CHI-CAT and Khorana score Discussion Strengthens and limitations Assessing the VTE risk in cancer-associated patients is essential, as it has been linked to reduced morbidity and mortality. However, predicting the risk of VTE in hospitalized cancer patients remains a challenge. An excellent predictive tool holds great value in clinical settings. Guidelines for thromboprophylaxis and clinical practices vary significantly among countries and healthcare facilities [27]. Indeed, many recommendations for thromboprophylaxis are primarily based on small studies, including heterogeneous and selected populations. Therefore, developing a new prophylactic strategy to prevent VTE in this large patient group is essential. Using a risk-stratification model, low-risk patients can safely forgo thromboprophylaxis, avoiding potential downsides such as bleeding and costs. In contrast, high-risk patients can benefit from treatment to prevent VTE [27]. For this purpose, the CHI-CAT score serves as a valuable and relevant decision-making tool. International research on VTE risk assessment and prevention in cancer populations has produced mature and comprehensive tools for risk assessment, along with effective targeted prevention intervention systems. In comparison to China, which is still developing, statistics indicate that the VTE risk screening and assessment strategy has not yet been implemented across all medical institutions in China, including tertiary hospitals. Currently, the commonly used clinical assessment tools are primarily introduced from abroad or are minor adaptations of existing tools. Guidelines now recommend the Khorana model for cancer patients, noting that it has high sensitivity but certain limitations regarding specificity. Our research and previous studies have confirmed this point [28, 29]. The Khorana model’s effectiveness in screening for VTE varies across different studies and types of cancers. Furthermore, many external validation studies have indicated that the Khorana model’s ability to predict risk is not very promising. This may be related to the need to diversify current tumor treatment methods. The Khorana model has limited screening capabilities, particularly for high-risk cancers. Research indicates that it struggles to differentiate between patients with pancreatic cancer and those with lung cancer who are at high risk for VTE [30, 31]. A large sample study conducted in China is necessary to evaluate its effectiveness. Additionally, the Khorana model sets the body mass index (BMI) threshold at 35 kg/m². However, the overall size of the Asian population is relatively small, which may skew its applicability to the Chinese population. Studies have shown that when the BMI threshold is adjusted to 24.33 kg/m² or higher, the predictive ability of the Khorana model improves significantly [32]. Moreover, the risk stratification capability of the Khorana model is somewhat limited. Data indicates that the incidence of VTE is higher in middle and low-risk patients according to the scoring results of the Khorana model. In comparison, the high-risk group [33]. In addition, most tools used for assessing cancer patients’ risk of VTE have been enhanced by incorporating or adjusting specific risk factors based on the Khorana model. The Vienna model, for instance, adds two predictive factors: the D-dimer and the sp selector. The D-dimer has exceptionally high sensitivity and remains unaffected by chemotherapy drugs, increasing the Khorana model’s positive predictive value from 22.1% to 42.9% [34]. However, a drawback is that the sp selector is rarely utilized as a routine indicator in medical institutions, resulting in limited tool adoption. The Protecht model [13] introduces a new risk factor: the type of chemotherapy drugs administered, specifically platinum and gemcitabine. This model is primarily used for patients receiving non-surgical internal medicine. A more straightforward risk assessment tool is designed for outpatient cancer patients, focusing on just one risk factor—the tumor site or D-dimer—from the Khorana and Vienna models. However, this overly simplistic approach leads to low sensitivity, and few validation trials currently support its efficacy. Most existing enhanced tools have not undergone large-scale external validation, making them not recommended in clinical guidelines. The Saved score [11] and Impede VTE score [12] are also used to evaluate the risk of VTE in patients with multiple myeloma; however, they assign different weighting scores to the same risk factor—VTE history. For instance, the Saved score assigns a weight of 3, while the Impede VTE score assigns 5. This discrepancy can lead to varied VTE risk assessment results and ultimately affect clinical prevention decisions. This study established a risk threshold for the cancer-related VTE risk index system and improved the risk assessment tool. We conducted an exploratory empirical study on the new tool, comparing it with the classic Khorana tool recommended by the guidelines. This retrospective cohort study demonstrated that AUC was 0.808 (95%CI, 0.715 ~ 0.901), the standard error was 0.048, and P < 0.001 when the risk threshold of the tool was constructed. The ROC curve was plotted to evaluate the discrimination power of the model between the high-risk and the low-/intermediate-risk population for VTE. The AUC was 0.808, indicating excellent discrimination capacity. By identifying patients at the highest risk, we hope the CHI-CAT score will help guide clinical decision-making for early implementation of thromboprophylaxis to reduce morbidity and mortality associated with VTE in this patient population. Our study has several strengths. First and foremost, it addresses a significant gap in China, where foreign tools are currently the only available options for assessing VTE risk. This population may benefit from a course of thromboprophylaxis. Secondly, we have assigned a risk threshold for the tool, enhancing its usability. Thirdly, the CHI-CAT score incorporates up to 40 VTE risk factors, with thresholds for BMI tailored to the characteristics of the Chinese population, making it more comprehensive. In contrast, while the Khorana risk assessment tool considers six common VTE risk factors readily accessible in clinical practice, it has notable limitations. It fails to account for certain key VTE risk factors, such as high-risk tumor sites (like the CNS), prior history of VTE, thrombophilia, and other relevant biochemical parameters [35]. Additionally, the simplicity of the Khorana score might lead to an underestimation of VTE risk in some instances. In the future, it may be beneficial to incorporate the risk factors identified in the CHI-CAT score into an electronic alert system. This would enhance functionality and accessibility while raising awareness of VTE prevention among medical personnel. As a result, they would be better equipped to make informed decisions regarding VTE prevention strategies. Several limitations of the study should still be noted. First, the hospital is a comprehensive facility with relatively few cancer patients. Additionally, the tool developed in this research is validated solely in a single center, requiring further validation in other healthcare settings. Third, due to differences in patient characteristics—such as age, gender, and comorbidities—there may be significant variations in the incidence of VTE between the construction and validation groups. Our next steps include improving the indicator system and refining risk thresholds. We aim to expand the evaluation to include a broader range of diseases and drugs to assess the tool’s effectiveness and its correlation with VTE occurrence. We will also integrate time factors and practical testing. Following a detailed analysis to identify patterns, we will further design prospective studies to validate and refine the tool. Conclusions This study confirmed that the CHI-CAT score effectively predicts symptomatic VTE among cancer patients hospitalized for medical reasons. However, further validation of our model is warranted by testing in larger internal and external retrospective cohorts, including more types of tumors, while continuously optimizing the model.