Advertisement

Validation of the IMPROVE bleeding risk assessment model in surgical patients: Results from the DissolVE-2 Study

  • Author Footnotes
    1 Drs Zhu Zhang and Kaiyuan Zhen contributed equally to this study.
    Zhu Zhang
    Footnotes
    1 Drs Zhu Zhang and Kaiyuan Zhen contributed equally to this study.
    Affiliations
    Department of Pulmonary and Critical Care Medicine, Centre of Respiratory Medicine, China-Japan Friendship Hospital, Beijing, China.

    National Centre for Respiratory Medicine, Beijing, China.

    National Clinical Research Centre for Respiratory Diseases, Beijing, China.
    Search for articles by this author
  • Author Footnotes
    1 Drs Zhu Zhang and Kaiyuan Zhen contributed equally to this study.
    Kaiyuan Zhen
    Footnotes
    1 Drs Zhu Zhang and Kaiyuan Zhen contributed equally to this study.
    Affiliations
    Department of Pulmonary and Critical Care Medicine, Centre of Respiratory Medicine, China-Japan Friendship Hospital, Beijing, China.

    Peking University China-Japan Friendship School of Clinical Medicine, Beijing, China.

    National Centre for Respiratory Medicine, Beijing, China.

    National Clinical Research Centre for Respiratory Diseases, Beijing, China.
    Search for articles by this author
  • Weimin Li
    Affiliations
    Department of Respiratory and Critical Care Medicine, Department of Pulmonary and Critical Care Medicine, West China Hospital, Sichuan University, Sichuan, China.
    Search for articles by this author
  • Xinyu Qin
    Affiliations
    Department of General Surgery, Zhongshan Hospital, Fudan University, Shanghai, China.
    Search for articles by this author
  • Jieming Qu
    Affiliations
    Department of Respiratory and Critical Care Medicine, Ruijin Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai, China.

    Department of Respiratory Medicine; Huadong Hospital affiliated to Fudan University, Shanghai, China.
    Search for articles by this author
  • Yuankai Shi
    Affiliations
    Department of Medical Oncology, National Cancer Center/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing Key Laboratory of Clinical Study on Anticancer Molecular Targeted Drugs, Beijing, China.
    Search for articles by this author
  • Ruihua Xu
    Affiliations
    Department of Medical Oncology, Sun Yat-sen University Cancer Center, State Key Laboratory of Oncology in South China, Collaborative Innovation Center for Cancer Medicine, Guangdong, China.
    Search for articles by this author
  • Yuming Xu
    Affiliations
    Department of Pharmacy, The First Affiliated Hospital of Zhengzhou University, Henan, China.
    Search for articles by this author
  • Dan Shen
    Affiliations
    Sanofi China, 17-19 Floor, Jing'an Kerry Centre Tower 3, Jing'an District, Shanghai, China
    Search for articles by this author
  • Jingjing Du
    Affiliations
    Sanofi China, 17-19 Floor, Jing'an Kerry Centre Tower 3, Jing'an District, Shanghai, China
    Search for articles by this author
  • Changbin Cai
    Affiliations
    Sanofi China, 17-19 Floor, Jing'an Kerry Centre Tower 3, Jing'an District, Shanghai, China
    Search for articles by this author
  • Zhenguo Zhai
    Correspondence
    Corresponding authors at: No. 2, East Yinghua Road, Chaoyang District, Beijing 100029, China.
    Affiliations
    Department of Pulmonary and Critical Care Medicine, Centre of Respiratory Medicine, China-Japan Friendship Hospital, Beijing, China.

    Peking University China-Japan Friendship School of Clinical Medicine, Beijing, China.

    National Centre for Respiratory Medicine, Beijing, China.

    National Clinical Research Centre for Respiratory Diseases, Beijing, China.
    Search for articles by this author
  • Chen Wang
    Correspondence
    Corresponding authors at: No. 2, East Yinghua Road, Chaoyang District, Beijing 100029, China.
    Affiliations
    Department of Pulmonary and Critical Care Medicine, Centre of Respiratory Medicine, China-Japan Friendship Hospital, Beijing, China.

    Peking University China-Japan Friendship School of Clinical Medicine, Beijing, China.

    Chinese Academy of Medical Sciences, Peking Union Medical College, Beijing, China.

    WHO Collaboration Center for Tobacco Cessation and Respiratory Diseases Prevention, Beijing, China.

    National Centre for Respiratory Medicine, Beijing, China.

    National Clinical Research Centre for Respiratory Diseases, Beijing, China.
    Search for articles by this author
  • on behalf of theDissolVE-2 investigators
  • Author Footnotes
    1 Drs Zhu Zhang and Kaiyuan Zhen contributed equally to this study.
Open AccessPublished:January 16, 2023DOI:https://doi.org/10.1016/j.thromres.2023.01.013

      Abstract

      Introduction

      IMPROVE Bleeding Risk Score (BRS) is known to be validated and widely accepted in medical patients. However, its relevance in surgical patients has so far not been explored. External validation of the IMPROVE BRS on bleeding in surgical patients can hopefully improve clinical practice (for surgical patients).

      Methods

      Data from 6986 surgical patients were collected from the DissolVE-2 cohort. The Kaplan-Meier method was used to assess the incidences of major bleeding and any bleeding among surgical patients within 14 days of admission. A cut-off value of BRS ≥7 indicated a higher risk of bleeding. Risk factors associated with major and any bleeding were analysed by the Cox regression method. Model discrimination was evaluated by area under the receiver operator characteristic curves (AUC). Calibration curves and Hosmer-Lemeshow χ2 statistics were used to measure the difference between predicted and observed bleeding risks.

      Results

      A total of 6399 surgical patients were included in the final validation cohort. The cumulative incidence rate of any bleeding was 3.9 % (95 % confidence interval [CI], 3.4–4.5), of which the incidence rate of major bleeding was 1.2 % (95 % CI, 0.9–1.6). Among patients with a BRS of ≥7, 16.3 % reported any bleeding, and 26.3 % reported major bleeding. The IMPROVE BRS had a better discriminative power (AUC = 0.69) and excellent goodness of fit (Hosmer-Lemeshow test, P = 0.208) for the prediction of major bleeding events as compared with any bleeding (AUC = 0.55; Hosmer-Lemeshow test, P = 0.004). The calibration plot suggested a more accurate prediction for major bleeding events. Moreover, the IMPROVE BRS had a higher AUC value of 0.83 and better goodness of fit (P = 0.2616) for major bleeding in patients undergoing abdominal surgery than other surgery types.

      Conclusion

      The IMPROVE BRS is a simple and practical technique that can help in predicting the risk of major bleeding in surgical patients, improving functional and safety outcomes of hospitalized patients with surgery.

      Keywords

      1. Introduction

      In real-world clinical practice, bleeding is a major surgical complication among hospitalized patients due to multiple complexities encountered during an intervention, including hemodilution, activation of fibrinolytic and inflammatory pathways, consumption of coagulation factor in extracorporeal circuits, and acquired platelet and hemostatic dysfunctions [
      • Ghadimi K.
      • Levy J.H.
      • Welsby I.J.
      Perioperative management of the bleeding patient.
      ]. Acquired defects of hemostasis may occur in surgical patients due to administration of anticoagulants that are prescribed to prevent venous thromboembolism (VTE), another common clinical complication that is anticipated in hospitalized patients, especially in those undergoing clinical interventions such as surgery or other invasive procedures [
      • Hansrani V.
      • Khanbhai M.
      • McCollum C.
      The prevention of venous thromboembolism in surgical patients.
      ].
      The type and severity of bleeding in surgical patients are propelled by a combination of several risk factors, such as the type of surgery, patient characteristics (age, gender, and comorbidities), and the use of certain drugs that affects hemostasis, such as prophylactics, antibiotics, and lipid-lowering drugs [
      • Curnow J.
      • Pasalic L.
      • Favaloro E.
      Why do patients bleed?.
      ]. Severe and uncontrollable bleeding during surgery can increase the mortality from <1 % to 20 %, and if conventional methods fail to arrest bleeding, the consequences can lead to an 8-fold increase in the odds of death [
      • Marietta M.
      • Facchini L.
      • Pedrazzi P.
      • Busani S.
      • Torelli G.
      Pathophysiology of bleeding in surgery.
      ]. Furthermore, complications due to in-hospital bleeding are associated with increased critical care utilization and length of stay, resulting in heavy clinical and economic burdens [
      • Marietta M.
      • Facchini L.
      • Pedrazzi P.
      • Busani S.
      • Torelli G.
      Pathophysiology of bleeding in surgery.
      ]. This highlights the criticality of evaluating risk in those who are more prone to surgical hemorrhage to mitigate the serious consequences of bleeding in surgical patients.
      Various national and international guidelines endorse the use of clinically validated VTE risk assessment models for the appropriate selection of thromboprophylaxis in surgical patients. Despite evidence that pharmacological prophylaxis does not increase the risk of hemorrhage in surgical patients [
      • Vasilakis V.
      • Klein G.M.
      • Trostler M.
      • Mukit M.
      • Marquez J.E.
      • Dagum A.B.
      • Pannucci C.J.
      • Khan S.U.
      Postoperative venous thromboembolism prophylaxis utilizing enoxaparin does not increase bleeding complications after abdominal body contouring surgery.
      ,
      • Parker S.G.
      • McGlone E.R.
      • Knight W.R.
      • Sufi P.
      • Khan O.A.
      Enoxaparin venous thromboembolism prophylaxis in bariatric surgery: a best evidence topic.
      ], fear of bleeding events is one of the main reasons for its suboptimal use in China and globally [
      • Zhang Z.
      • Zhai Z.
      • Li W.
      • Qin X.
      • Qu J.
      • Shi Y.
      • Xu R.
      • Xu Y.
      • Wang C.
      Validation of the IMPROVE bleeding risk score in Chinese medical patients during hospitalization: findings from the dissolve-2 study.
      ]. Nonetheless, no standardized risk assessment model exists till date to predict the risk of bleeding in surgical patients. The IMPROVE study investigators recognized various risk factors and set a model to predict the bleeding risk for hospitalized patients who are at risk for VTE [
      • Decousus H.
      • Tapson V.F.
      • Bergmann J.-F.
      • Chong B.H.
      • Froehlich J.B.
      • Kakkar A.K.
      • Merli G.J.
      • Monreal M.
      • Nakamura M.
      • Pavanello R.
      • Pini M.
      • Piovella F.
      • Spencer F.A.
      • Spyropoulos A.C.
      • Turpie A.G.G.
      • Zotz R.B.
      • Fitzgerald G.
      • Anderson F.A.
      IMPROVE Investigators, Factors at admission associated with bleeding risk in medical patients: findings from the IMPROVE investigators.
      ]. In this study, a scoring system was obtained by examining 11 factors that are independently associated with major bleeding and any bleeding (major + clinically relevant non major [CRNM] bleeding) events [
      • Decousus H.
      • Tapson V.F.
      • Bergmann J.-F.
      • Chong B.H.
      • Froehlich J.B.
      • Kakkar A.K.
      • Merli G.J.
      • Monreal M.
      • Nakamura M.
      • Pavanello R.
      • Pini M.
      • Piovella F.
      • Spencer F.A.
      • Spyropoulos A.C.
      • Turpie A.G.G.
      • Zotz R.B.
      • Fitzgerald G.
      • Anderson F.A.
      IMPROVE Investigators, Factors at admission associated with bleeding risk in medical patients: findings from the IMPROVE investigators.
      ]. Based on the major and any bleeding incidence in their study, a score threshold of ≥7 predicted a higher bleeding risk.
      The DissolVE-2 study analysed medical and surgical patients in China to identify VTE and bleeding risk factors. The study reported that 2766 (39.6 %) of the 6986 surgical patients had a high bleeding risk compared with 10.7 % of medical patients. Although IMPROVE Bleeding risk assessment model (RAM) is validated in various populations of medical inpatients, its relevance in surgical patients has not been explored yet. Herein, we aim to bridge this gap by externally validating the IMPROVE score for bleeding risk in surgical patients from the DissolVE-2 study that involved a nationwide population to facilitate tailored decisions for the administration of prophylactic or other preoperative interventions in patients with low-bleeding risk.

      2. Methods

      2.1 Study design and participants

      The DissolVE-2 study previously reported was an observational, cross-sectional, multicenter study, which included patients from 60 hospitals across 44 major cities in China [
      • Zhai Z.
      • Kan Q.
      • Li W.
      • Qin X.
      • Qu J.
      • Shi Y.
      • Xu R.
      • Xu Y.
      • Zhang Z.
      • Wang C.
      • Zhu F.
      • Jiang M.
      • Deng Y.
      • Li X.
      • Geng D.
      • Zhai J.
      • Wang C.
      • Zhang J.
      • Cao J.
      • Chen S.
      • Li G.
      • Wang L.
      • Zhang Y.
      • Liu X.
      • Zhang X.
      • Luo X.
      • Li X.
      • Zhao L.
      • Zu N.
      • Lu X.
      • Yang L.
      • Zhang N.
      • Pang W.
      • Ji Y.
      • Yao Y.
      • Sun X.
      • Xu W.
      • Zhao J.
      • Leng H.
      • Yu Z.
      • Kuang G.
      • Huang Y.
      • Chen L.
      • Shi J.
      • Xin S.
      • Xiao P.
      • Du Z.
      • Wei Y.
      • Liu Y.
      • Zhang C.
      • Xu Y.
      • Yang Z.
      • Liu R.
      • Liu Y.
      • Chen Y.
      • Qin J.
      • Liu S.
      • Zhou H.
      • Zhang Z.
      • Wei L.
      • Ren T.
      • Wang B.
      • Chen H.
      • Mao G.
      • Ying Y.
      • Ding G.
      • Dui D.
      • Du T.
      • Li X.
      • Lu J.
      • Zheng D.
      • He T.
      VTE risk profiles and prophylaxis in medical and surgical inpatients.
      ]. Adult (>18 years) patients hospitalized for general medicine, cardiac, neurovascular, oncology, respiratory, rheumatic, and general surgery and orthopedic wards were enrolled in the study (Supplementary Table 1). The study was approved by the respective institutional ethical review boards. Written informed consent from included patients was waived off due to the study's retrospective nature.
      This analysis only included surgical patients who had undergone a surgical operation as per CHEST guidelines (9th edition) and/or were admitted with a major traumatic event not requiring an operation. Furthermore, patients with missing values for the 11 IMPROVE bleeding risk factor were excluded from the validation cohort, as the main agenda of this study was to validate IMPROVE bleeding risk score (BRS) in surgical patients.

      2.2 Study procedure

      Clinical data were collected and stored using the electronic data capture system, including demographic details, surgical and procedural details, risk factors for VTE and bleeding outcome of patients at discharge, and details of VTE prophylaxis. All data related to bleeding were pulled out for this study.
      The IMPROVE Bleeding RAM consisted of 11 risk factors, including age, male gender, moderate or severe renal failure (glomerular filtration rate [GFR] <60 mL/min/m2), current cancer, rheumatic disease, central venous catheter, intensive care unit/cardiac care unit (ICU/CCU), hepatic failure (international normalized ratio [INR] >1.5), platelet count <50 × 109, bleeding in the 3 months before admission, and active gastroduodenal ulcer and each risk factor was given 1 to 4.5 points. Each surgical patient in the validation cohort were calculated the IMPROVE BRS by summing up total points from the 11 risk factors and classified into 2 risk groups (IMPROVE BRS <7 and ≥ 7), in which an IMPROVE BRS ≥7 indicated a higher bleeding risk.

      2.3 Definition of bleeding

      The main outcomes of this study were the occurrence of major bleeding and any bleeding (major bleeding + CRNM bleeding) within 14 days of admission. Major bleeding was defined as a bleeding event that causes death, clinically severe hemorrhage requiring a blood transfusion of ≥2 units of red blood cells (RBCs), causing a decrease of hemoglobin levels by ≥2 g/dL, or hemorrhage within a major organ (which includes intraocular, intracranial, pericardial, and retroperitoneal hemorrhage). CRNM bleeding is defined as hemorrhoid bleeding, epistaxis, or gingival bleeding lasting for >5 min, gross hematuria for >24 h, subcutaneous hematoma for >25 cm2, wound hematoma >100 cm2, hematoma that requires drainage, and other bleeding (which includes clinically overt acute bleeding events important enough to be recorded, for e.g., epistaxis requiring intervention or spontaneous macroscopic hematuria) [
      • Büller H.R.
      Fondaparinux or enoxaparin for the initial treatment of symptomatic deep venous thrombosis: a randomized trial.
      ,
      • Bahit M.C.
      • Lopes R.D.
      • Wojdyla D.M.
      • Held C.
      • Hanna M.
      • Vinereanu D.
      • Hylek E.M.
      • Verheugt F.
      • Goto S.
      • Alexander J.H.
      • Wallentin L.
      • Granger C.B.
      Non-major bleeding with apixaban versus warfarin in patients with atrial fibrillation.
      ].

      2.4 Statistical analyses

      All statistical analyses were performed using the Statistical Analysis System (SAS) University Edition. Continuous variables were described as the number of observations, median and interquartile range (IQR), or mean and standard deviation (SD), whereas categorical variables were presented as numbers and percentages. The cumulative bleeding rate within 14 days from the day of admission till the occurrence of bleeding was described using the Kaplan-Meier (KM) method. Multivariable analyses were performed by the Cox regression model to analyse the risk factors associated with bleeding events. For every risk stratification factor, hazard ratios and 95 % confidence intervals (CIs) for major bleeding and any bleeding were computed. Calibration curves were plotted to display the relationship of predicted bleeding risk divided into 15 deciles against observed outcomes. The slope of calibration curves that equals 1 was associated with a perfect calibration of an ideal model. The Hosmer-Lemeshow χ2 statistics [
      • Hosmer D.W.
      • Lemeshow S.
      Applied Logistic Regression: Hosmer/Applied Logistic Regression.
      ] were performed, where P > 0.05 indicated a good fit of the predictive model. Discrimination of the model was assessed using receiver operating characteristic (ROC) curves and the area under the curve (AUC). Furthermore, the model's competency to discriminate between high- and low-risk patients was measured by comparing the proportions and bleeding rates for each risk group (IMPROVE BRS <7 and ≥ 7). Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated by a standard formula.

      3. Results

      3.1 Patient characteristics

      The DissolVE-2 study screened a total of 51,835 patients between March and September 2016, of which 14,000 were enrolled, and 13,609 patients were included in the full analysis set. Of the total, there were 6986 (51.3 %) surgical patients. After excluding 587 (8.4 %) patients with missing values for platelet count, international normalized ratio (INR), and glomerular filtration rate (GFR), a total of 6399 surgical patients were finally included in the validation cohort of this study (Supplementary Fig. 1).
      In the validation cohort, approximately two-thirds of the patients were female, approximately three-quarters were aged between 40 and 84 years, and < 10 % had severe or moderate renal failure. The most frequent surgical intervention was abdominal surgery (28.9 %). VTE prophylaxis was administered to only 803 (12.6 %) surgical patients, and the median duration of hospitalization was 11 (7–17) days (Table 1). As compared with patients with CRNM bleeding, major bleeding was more observed in patients of male (44.1 % vs. 56.1 %), ICU/CCU admitted (9.9 % vs 29.8 %), central venous catheter (9.2 % vs. 35.1 %), current cancer (15.15 % vs. 26.3 %), common comorbidities such as diabetes (5.9 % vs. 14 %) and hypertension (16.5 % vs. 22.8 %) and extended hospital stay (22 days vs. 8 days). Nevertheless, patients with CRNM bleeding were more prone to having a bleeding history 3 months before admission (19.7 % vs. 1.8 %). Over half of the patients having CRNM bleeding underwent abdominal surgery (57.9 %), followed by lumbar puncture, epidural, or spinal anesthesia (11.2 %). Patients with major bleeding commonly underwent abdominal surgery (38.6 %), craniotomy (21.1 %), and spinal surgery or spinal trauma (7 %; Table 1).
      Table 1Demographic patient characteristics stratified by bleeding status.
      Total patientsAny bleeding
      Any bleeding is defined as either major bleeding or CRNM bleeding.
      Major bleedingCRNM bleeding
      (N = 6399)(N = 209)(N = 57)(N = 152)
      Patient characteristics
      Age groups, n(%)
       <40 years1572(24.6)65(31.1)4(7.0)61(40.1)
       ≥40 and <85 years4759(74.4)140(67.0)51(89.5)89(58.6)
       ≥85 years68(1.1)4(1.9)2(3.5)2(1.3)
      Male, n(%)2471(38.6)99(47.4)32(56.1)67(44.1)
      Medical conditions
      Active gastroduodenal ulcer, n (%)53(0.8)2(1.0)2(3.5)0(0.0)
      Bleeding in 3 months before admission, n (%)200(3.1)31(14.8)1(1.8)30(19.7)
      Platelet count <50 × 109, n (%)30(0.5)0(0.0)0(0.0)0(0.0)
      Thrombocytopenia, n (%)21(0.3)2(1.0)1(1.8)1(0.7)
      Hepatic failure (INR > 1.5), n (%)41(0.6)1(0.5)1(1.8)0(0.0)
      Severe renal failure, n (%)
       GFR <30 mL/min/m252(0.8)2(1.0)1(1.8)1(0.7)
       GFR 30–59 mL/min/m2196(3.1)6(2.9)4(7.0)2(1.3)
       GFR ≥60 mL/min/m26151(96.1)201(96.2)52(91.2)149(98.0)
      ICU/CCU, n (%)556(8.7)32(15.3)17(29.8)15(9.9)
      Central venous catheter, n (%)802(12.5)34(16.3)20(35.1)14(9.2)
      Rheumatic disease, n (%)46(0.7)2(1.0)0(0.0)2(1.3)
      Current cancer, n (%)1458(22.8)38(18.2)15(26.3)23(15.1)
      Diabetes, n (%)419(6.6)17(8.1)8(14.0)9(5.9)
      Coagulation disorders, n (%)24(0.4)1(0.5)1(1.8)0(0.0)
      Acute stroke, n (%)212(3.3)6(2.9)5(8.8)1(0.7)
      Hypertension, n (%)1362(21.3)38(18.2)13(22.8)25(16.5)
      Concomitant use of anticoagulants, antiplatelet therapy, NSAIDs, or thrombolytic drugs, n (%)918(14.4)83(39.7)10(17.5)73(48.0)
      Surgery type
      Lumbar puncture, epidural, or spinal anesthesia, n (%)377(5.9)20(9.6)3(5.3)17(11.2)
      Abdominal surgery, n (%)1851(28.9)110(52.6)22(38.6)88(57.9)
      Pancreaticoduodenectomy, n (%)25(0.4)0(0.0)0(0.0)0(0.0)
      Hepatic resection, n (%)108(1.7)7(3.4)2(3.5)5(3.3)
      Cardiac surgery, n(%)47(0.7)3(1.4)2(3.5)1(0.7)
      Thoracic surgery390(6.1)7(3.4)1(1.8)6(4.0)
      Pneumonectomy or extended resection, n (%)227(3.6)2(1.0)0(0.0)2(1.3)
      Craniotomy, n (%)343(5.4)12(5.7)12(21.1)0(0.0)
      Spinal surgery or spinal trauma, n (%)463(7.2)11(5.3)8(14.0)3(2.0)
      VTE prophylaxis803(12.6)15(7.2)4(7.0)11(7.2)
      Length of hospital stay (days), median (IQR)11(7–17)12(6–20)22(15–29)8(5–14)
      CRNM, clinically relevant nonmajor; INR, International Normalized Ratio; GFR, glomerular filtration rate; ICU, intensive care unit; CCU, cardiac care unit; NSAIDs, nonsteroidal anti-inflammatory drugs; VTE, venous thromboembolism; IQR, interquartile range.
      a Any bleeding is defined as either major bleeding or CRNM bleeding.

      3.2 Cumulative bleeding incidence and reasons for bleeding

      A total of 57 (0.9 %) major bleeding and 152 (2.4 %) CRNM bleeding events were noted in the validation cohort. The cumulative incidence of bleeding within 14 days from admission to hospital indicated the any bleeding incidence rate of 3.9 % (95 % CI, 3.4–4.5), of which the incidence rate of major bleeding was 1.2 % (95 % CI, 0.9–1.6; Fig. 1). The major bleeding types included bleeding leading to a drop in hemoglobin level by ≥2 g/dL (2.9 %), bleeding leading to a transfusion of ≥2 RBC units (21.5 %), intracranial bleeding (1.4 %), epidural hematoma (1.0 %), and bleeding in the retroperitoneal, intraocular, or pericardial organs (0.5 %). CRNM bleeding types included hemorrhoidal bleeding (9.1 %), hematoma either >100 cm3 (2.4 %) or hematoma requiring drainage (1.4 %), gross hematuria that lasted for >24 h (0.5 %), and other bleeding significant enough to be reported in the hospital chart (59.3 %; Supplementary Table 2).
      Fig. 1
      Fig. 1Kaplan-Meier curves demonstrating the cumulative incidence rates of any bleeding and major bleeding within 14 days from admission.

      3.3 IMPROVE BRS in surgical patients

      As per the original IMPROVE BRS, the 11 factors that demonstrated a higher risk of bleeding were evaluated for their association with surgical patients. In the validation cohort, any bleeding and major bleeding events were significantly associated with patients aged 40 to 84 years (any bleeding HR: 0.62 [P = 0.002]; major bleeding HR: 3.04 [P = 0.034]) and ICU/CCU (any bleeding HR: 1.68 [P = 0.012]; major bleeding HR: 2.33 [P = 0.009]). Male gender (HR: 1.39 [P = 0.022]), bleeding in 3 months before admission (HR: 6.25 [P < 0.0001]), and current cancer (HR: 0.63 [P = 0.016]) were significantly associated with any bleeding only, whereas central venous catheter (HR: 2.26 [P = 0.01]) was significantly associated with only major bleeding (Table 2).
      Table 2Association between bleedings and IMPROVE bleeding risk factors (N = 6399).
      Any bleedingMajor bleeding
      HR[95 % CI]p-ValueHR[95 % CI]p-Value
      Patient characteristics
      Age groups
       ≥40 and <85 vs <40 years0.62[0.46–0.84]0.0022
      P < 0.05.
      3.04[1.09-8.51]0.0338
      P < 0.05.
       ≥85 vs <40 years1.04[0.37–2.93]0.94015.66[1.00–32.10]0.0502
      Male vs Female1.39[1.05–1.83]0.0215
      P < 0.05.
      1.43[0.83-2.45]0.1945
      Medical conditions
      Active gastroduodenal ulcer1.06[0.25–4.41]0.93872.79[0.66–11.79]0.1633
      Bleeding in 3 months before admission6.25[4.25–9.19]<0.0001
      P < 0.05.
      0.54[0.07-3.93]0.5421
      Platelet count <50 × 1090.00[−]0.96950.00[−]0.9874
      Hepatic failure (INR > 1.5)0.71[0.10–5.08]0.73101.70[0.23–12.62]0.6020
      Severe renal failure
       GFR 30–59 vs ≥60 mL/min/m20.72[0.31–1.65]0.43851.14[0.39–3.30]0.8093
       GFR <30 vs ≥60 mL/min/m20.73[0.18–2.99]0.65961.45[0.20–10.69]0.7137
      ICU/CCU1.68[1.12–2.52]0.0124
      P < 0.05.
      2.33[1.24-4.37]0.0089
      P < 0.05.
      Central venous catheter1.16[0.77–1.76]0.47582.26[1.22–4.19]0.0098
      P < 0.05.
      Rheumatic disease1.02[0.24–4.26]0.98120.00[−]0.9927
      Current cancer0.63[0.43–0.92]0.0163
      P < 0.05.
      0.71[0.38-1.33]0.2837
      INR, International Normalized Ratio; GFR, glomerular filtration rate; ICU, intensive care unit; CCU, cardiac care unit; HR, hazard ratio; CI, confidence interval.
      low asterisk P < 0.05.
      Out of the 209 patients who reported any bleeding event, 175 (83.7 %) and 34 (16.3 %) patients had an IMPROVE bleeding score of <7 and ≥ 7, respectively. Major bleeding event was observed in 57 patients, of which 42 (73.7 %) patients had an IMPROVE score of <7, and 15 (26.3 %) patients had a score of ≥7 (Fig. 2a ). The incidence of any bleeding was 3.0 % and 5.9 %, whereas the incidence of major bleeding was 0.7 % and 2.6 % for IMPROVE BRS <7 and ≥ 7, respectively (Fig. 2b).
      Fig. 2
      Fig. 2Implications of IMPROVE BRS for clinical decision.
      a. Proportions of surgical patients with IMPROVE BRS <7 and ≥ 7. The bar chart shows that patients with IMPROVE BRS ≥7 were more likely to experience major bleeding events than any bleeding events (26.3 % vs 16.3 %). b. Bleeding rates of surgical patients with IMPROVE BRS <7 and ≥ 7. The bar chart shows that patients with IMPROVE BRS ≥7 had approximately 2 to 3-fold increased risk of bleeding compared with BRS <7 (5.9 % vs 3.0 % for any bleeding rates; 2.6 % vs 0.7 % for major bleeding rates).

      3.4 Model calibration and discrimination

      The IMPROVE bleeding RAM underestimated any bleeding risk in patients with a predicted risk of >5 %, whereas it slightly overestimated any bleeding risk in patients with a predicted risk of <4 % and major bleeding risk. Compared with any bleeding, the calibration curve of major bleeding was closer to the diagonal dashed line, which revealed a better calibration, indicating the competency of the model to predict major bleeding events (Fig. 3). The Hosmer-Lemeshow χ2 statistic for major bleeding was 8.44, with a P-value of 0.208, indicating no significant deviation between the predicted and observed events, implying excellent goodness of fit. However, any bleeding with a P-value of 0.004 for the Hosmer-Lemeshow test indicated poor goodness of fit. The ROC curves reflected possibly helpful discrimination of the predictive model with an AUC value of 0.69 for major bleeding, and relatively poor discrimination for any bleeding, with an AUC value of 0.55 (Fig. 4).
      Fig. 3
      Fig. 3Calibration curves displaying the relationship of mean predicted bleeding risk divided into 15 deciles against observed bleeding risk. The diagonal dashed line showed an ideal calibration. Compared with any bleeding, the calibration curve of major bleeding was closer to the diagonal dashed line, which suggested a better calibration.
      Fig. 4
      Fig. 4Area under ROC curves measure the capacity of model discrimination. The diagonal dashed line shows the classification due to chance (AUC = 0.5). The IMPROVE bleeding RAM reflected possibly helpful discrimination to predict major bleeding events (AUC = 0.69) in comparison with any bleeding (AUC = 0.55).
      For major bleeding, the IMPROVE BRS revealed a high specificity of 91.4 % and sensitivity of 26.3 %, with positive predictive value (PPV) and negative predictive value (NPV) of 2.7 % and 99.3 %, respectively. Sensitivity and specificity for any bleeding event were 16.3 % and 91.5 %, whereas the PPV and NPV values for any bleeding were 6.1 % and 97 %, respectively (Supplementary Table 3).
      Furthermore, ROC curves were analysed Hosmer-Lemeshow test was performed for different surgery types by excluding those with <10 bleeding events. The IMPROVE BRS had a higher AUC value of 0.83 and better goodness of fit (χ2 = 8.88, P = 0.2616) for major bleeding in patients undergoing abdominal surgery than other surgery types. The craniotomy and spinal surgery/trauma had less impact on the model calibration and discrimination power (Supplementary Table 4).

      4. Discussion

      The current study represents for the first time the extensive external validation of the IMPROVE BRS in a nationwide cohort of Chinese surgical patients. The RAM reported herein was based on the previously classified variables evaluated upon hospital admission for medical patients, along with an IMPROVE risk calculator to derive an individualized prognostic score that evaluates the bleeding risk for every patient. The resulting nomogram in the present study was able to explicitly discriminate between surgical patients who did or did not develop major bleeding events and was appropriately calibrated. Moreover, our overall results show that the IMPROVE BRS has a high predictive accuracy for anticipating major bleeding events in surgical patients with significant specificity, sensitivity, and predictive values, which was generally on par with those previously validated among medical patients. Although few efforts have been made to develop BRS for different types of surgical patients [
      • Klein A.A.
      • Collier T.
      • Yeates J.
      • Miles L.F.
      • Fletcher S.N.
      • Evans C.
      • Richards T.
      The ACTA PORT-score for predicting perioperative risk of blood transfusion for adult cardiac surgery.
      ,
      • Algattas H.
      • Kimmell K.T.
      • Vates G.E.
      Risk of reoperation for hemorrhage in patients after craniotomy.
      ,
      • Biancari F.
      • Brascia D.
      • Onorati F.
      • Reichart D.
      • Perrotti A.
      • Ruggieri V.G.
      • Santarpino G.
      • Maselli D.
      • Mariscalco G.
      • Gherli R.
      • Rubino A.S.
      • De Feo M.
      • Gatti G.
      • Santini F.
      • Dalén M.
      • Saccocci M.
      • Kinnunen E.-M.
      • Airaksinen J.K.E.
      • D’Errigo P.
      • Rosato S.
      • Nicolini F.
      Prediction of severe bleeding after coronary surgery: the WILL-BLEED Risk Score.
      ,
      • Sławek-Szmyt S.
      • Araszkiewicz A.
      • Grygier M.
      • Szmyt K.
      • Seniuk W.
      • Waśniewski M.
      • Smukowski T.
      • Chmielewska-Michalak L.
      • Lesiak M.
      • Mitkowski P.
      PACE DRAP: a simple score for predicting significant bleeding complications after cardiac implantable electronic device surgery.
      ], it has not been externally validated in a large population. An added benefit of the current IMPROVE tool is that the 11 variables used to calculate the BRS are readily available pre-operatively, making it practical to incorporate in real-world clinical practice. This will guide surgeons and resident doctors to employ adequate actions to prevent bleeding in high-risk patients and facilitate personalized prophylactic interventions during surgery.
      An important finding of our study is that patients with an IMPROVE BRS ≥7 had approximately 2- to 3-fold increased bleeding incidence, validating that the cut-off score of 7 can distinguish well between a high- and low-bleeding risk in surgical patients. A similar trend was observed among medical patients, where a score of ≥7 resulted in over a 2-fold increased risk of major bleeding, from 1.6 % to 5.4 % [
      • Rosenberg D.J.
      • Press A.
      • Fishbein J.
      • Lesser M.
      • McCullagh L.
      • McGinn T.
      • Spyropoulos A.C.
      External validation of the IMPROVE bleeding risk assessment model in medical patients.
      ]. Furthermore, our study revealed that patients with BRS ≥7 were more likely to experience major bleeding events (26.3 % vs. 16.3 %), which confirmed the model's competency to finely discriminate among patients stratified according to bleeding risk.
      The IMPROVE BRS in our cohort of surgical patients exhibited a good predictive power for major bleeding events and was in concordance with other IMPROVE BRS in medical patients (AUC = 0.69 vs. 0.63–0.71) [
      • Zhang Z.
      • Zhai Z.
      • Li W.
      • Qin X.
      • Qu J.
      • Shi Y.
      • Xu R.
      • Xu Y.
      • Wang C.
      Validation of the IMPROVE bleeding risk score in Chinese medical patients during hospitalization: findings from the dissolve-2 study.
      ,
      • Decousus H.
      • Tapson V.F.
      • Bergmann J.-F.
      • Chong B.H.
      • Froehlich J.B.
      • Kakkar A.K.
      • Merli G.J.
      • Monreal M.
      • Nakamura M.
      • Pavanello R.
      • Pini M.
      • Piovella F.
      • Spencer F.A.
      • Spyropoulos A.C.
      • Turpie A.G.G.
      • Zotz R.B.
      • Fitzgerald G.
      • Anderson F.A.
      IMPROVE Investigators, Factors at admission associated with bleeding risk in medical patients: findings from the IMPROVE investigators.
      ,
      • Rosenberg D.J.
      • Press A.
      • Fishbein J.
      • Lesser M.
      • McCullagh L.
      • McGinn T.
      • Spyropoulos A.C.
      External validation of the IMPROVE bleeding risk assessment model in medical patients.
      ]. Moreover, the discrimination capacity was also comparable with those reported by other BRS in patients undergoing surgery such as the ACTA-PORT [
      • Klein A.A.
      • Collier T.
      • Yeates J.
      • Miles L.F.
      • Fletcher S.N.
      • Evans C.
      • Richards T.
      The ACTA PORT-score for predicting perioperative risk of blood transfusion for adult cardiac surgery.
      ] score in cardiac surgical patients (AUC = 0.76), and BRS in craniotomy patients (AUC = 0.68) [
      • Algattas H.
      • Kimmell K.T.
      • Vates G.E.
      Risk of reoperation for hemorrhage in patients after craniotomy.
      ]. Other models such as Kucher (AUC = 0.76), IMPROVE 4-factor (AUC = 0.69), Zakai model 2 (AUC = 0.73), Intermountain (AUC = 0.87), and NAVAL (AUC = 0.72) have also been validated in the medical patients [
      • Cobben M.R.R.
      • Nemeth B.
      • Lijfering W.M.
      • Cannegieter S.C.
      Validation of risk assessment models for venous thrombosis in hospitalized medical patients.
      ]. Of note, we observed more favorable AUC values for major bleeding events compared with any bleeding (AUC = 0.69 vs 0.55), potentially suggesting better applicability of the model to selectively predict the risk of major bleeding in surgical patients. Further analysis of the model performance confirmed better consistency of the nomogram calibration curve to predict major bleeding events as compared with any bleeding.
      The IMPROVE BRS shares a number of risk factors for major bleeding complications common to other models [

      Venous thromboembolism: reducing the risk for patients in hospital | Guidance | NICE, (n.d.). https://www.nice.org.uk/guidance/cg92 (accessed December 22, 2021).

      ,
      • Ruíz-Giménez N.
      • Suárez C.
      • González R.
      • Nieto J.
      • Todolí J.
      • Samperiz Á.
      • Monreal M.
      the RIETE Investigators
      Predictive variables for major bleeding events in patients presenting with documented acute venous thromboembolism.Findings from the RIETE Registry.
      ,
      • Gould M.K.
      • Garcia D.A.
      • Wren S.M.
      • Karanicolas P.J.
      • Arcelus J.I.
      • Heit J.A.
      • Samama C.M.
      Prevention of VTE in nonorthopedic surgical patients: Antithrombotic Therapy and Prevention of Thrombosis, 9th ed: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines.
      ]. Contextually, older age and malignancy are well-accepted risk factors for peri-operative major bleeding, and are common elements of the current BRS, including the RIETE (Registro Informatizado de Enfermedad Tromboembolica) score, and risk factors listed in the ACCP guidelines, 9th edition [
      • Ruíz-Giménez N.
      • Suárez C.
      • González R.
      • Nieto J.
      • Todolí J.
      • Samperiz Á.
      • Monreal M.
      the RIETE Investigators
      Predictive variables for major bleeding events in patients presenting with documented acute venous thromboembolism.Findings from the RIETE Registry.
      ,
      • Kearon C.
      • Akl E.A.
      • Comerota A.J.
      • Prandoni P.
      • Bounameaux H.
      • Goldhaber S.Z.
      • Nelson M.E.
      • Wells P.S.
      • Gould M.K.
      • Dentali F.
      • Crowther M.
      • Kahn S.R.
      Antithrombotic therapy for VTE disease.
      ]. Apart from age-specific fragility, bleeding may also be exacerbated in surgical patients with cancer due to the tumor itself, or by immunotherapies or chemotherapies that induce thrombocytopenia [
      • Johnstone C.
      • Rich S.E.
      Bleeding in cancer patients and its treatment: a review.
      ]. In contrast to the original and the external validation of the IMPROVE studies in medical patients [
      • Decousus H.
      • Tapson V.F.
      • Bergmann J.-F.
      • Chong B.H.
      • Froehlich J.B.
      • Kakkar A.K.
      • Merli G.J.
      • Monreal M.
      • Nakamura M.
      • Pavanello R.
      • Pini M.
      • Piovella F.
      • Spencer F.A.
      • Spyropoulos A.C.
      • Turpie A.G.G.
      • Zotz R.B.
      • Fitzgerald G.
      • Anderson F.A.
      IMPROVE Investigators, Factors at admission associated with bleeding risk in medical patients: findings from the IMPROVE investigators.
      ,
      • Rosenberg D.J.
      • Press A.
      • Fishbein J.
      • Lesser M.
      • McCullagh L.
      • McGinn T.
      • Spyropoulos A.C.
      External validation of the IMPROVE bleeding risk assessment model in medical patients.
      ], the present study revealed that bleeding events were significantly associated only with age, gender, current cancer, ICU stay, and central venous catheter. A plausible reason for this is that our study reported Cox regression to estimate HR, which may have prevented the overestimation of certain risk factors [
      • George A.
      • Stead T.S.
      • Ganti L.
      What's the risk: differentiating risk ratios, odds ratios, and hazard ratios?.
      ], as compared with the logistic regression analysis reported in their studies. In the current study, the age group between 40 and 85 years and current cancer had a statistically significant association with any bleeding with HR values of 0.62 and 0.63, respectively. The observed low HR values might be due to a small number of bleeding events (any bleeding = 209) among 6399 surgical patients and thus, impacting the estimation of HR values using Cox model. Another possible reason might be the definition of any bleeding, where 59.3 % of patients were classified as other types of bleeding, that might lead to information bias. Whereas, for major bleeding, patients aged between 40 and 85 years had a statistically significant increased risk of major bleeding (HR: 3.04, P = 0.0338) and current cancer had a lower HR value but was not statistically significant (HR: 0.71, P = 0.2837) with major bleeding.
      The Dissolve-2 cohort had similar specificities between surgical and medical patients for both, any bleeding (91.5 % vs. 90.8 %) and major bleeding (91.4 % vs. 90.3 %), at the cost of lower sensitivity among surgical patients for predicting major bleeding events (26.3 % vs. 51 %). This could be due to a higher number of patients administered with VTE prophylaxis in the surgical cohort of the DissolVE-2 population (12.6 % vs. 8.5 %). However, the proportion of patients on VTE prophylaxis is still low which could be due to physicians' fear of bleeding. Nevertheless, the sensitivity found in the present study was moderately on par with the validation study by Rosenberg et al. (33.3 %) on medical patients. The NPV in our cohort was comparable with that reported by other IMPROVE studies on medical patients [
      • Zhang Z.
      • Zhai Z.
      • Li W.
      • Qin X.
      • Qu J.
      • Shi Y.
      • Xu R.
      • Xu Y.
      • Wang C.
      Validation of the IMPROVE bleeding risk score in Chinese medical patients during hospitalization: findings from the dissolve-2 study.
      ,
      • Decousus H.
      • Tapson V.F.
      • Bergmann J.-F.
      • Chong B.H.
      • Froehlich J.B.
      • Kakkar A.K.
      • Merli G.J.
      • Monreal M.
      • Nakamura M.
      • Pavanello R.
      • Pini M.
      • Piovella F.
      • Spencer F.A.
      • Spyropoulos A.C.
      • Turpie A.G.G.
      • Zotz R.B.
      • Fitzgerald G.
      • Anderson F.A.
      IMPROVE Investigators, Factors at admission associated with bleeding risk in medical patients: findings from the IMPROVE investigators.
      ,
      • Rosenberg D.J.
      • Press A.
      • Fishbein J.
      • Lesser M.
      • McCullagh L.
      • McGinn T.
      • Spyropoulos A.C.
      External validation of the IMPROVE bleeding risk assessment model in medical patients.
      ]. The predictive ability of various other RAMs in medical patients revealed a notable PPV and NPV, such as Padua (PPV: 1.4 %; NPV: 99.1 %), Kucher (PPV: 2.3 %; NPV: 99 %), Lecumberri (PPV: 1.4 %; NPV: 99 %), NICE (PPV: 1.5 %; NPV: 99.3 %), PRETEMED (PPV: 1.3 %; NPV: 99 %), IMPROVE 4-factor(PPV: 2.3 %; NPV: 99 %), Geneva (PPV: 1.4 %; NPV: 99.1 %), Zakai model 2 (PPV: 1.1 %; NPV: 98.6 %),Intermountain(PPV: 3.2 %; NPV: 99 %), Caprini(PPV: 1.4 %; NPV: 99.4 %), and NAVAL(PPV: 3.1 %; NPV: 98.9 %) [
      • Cobben M.R.R.
      • Nemeth B.
      • Lijfering W.M.
      • Cannegieter S.C.
      Validation of risk assessment models for venous thrombosis in hospitalized medical patients.
      ].High NPV and specificity values of our study indicated that the RAM accurately predicts the majority of the surgical patients who render a low bleeding risk [
      • Rosenberg D.J.
      • Press A.
      • Fishbein J.
      • Lesser M.
      • McCullagh L.
      • McGinn T.
      • Spyropoulos A.C.
      External validation of the IMPROVE bleeding risk assessment model in medical patients.
      ], and contrary to a consistently high NPV, lower PPV is characterized as a common limitation of clinical bleeding prediction techniques [
      • Ueki Y.
      • Bär S.
      • Losdat S.
      • Otsuka T.
      • Zanchin C.
      • Zanchin T.
      • Gragnano F.
      • Gargiulo G.
      • Siontis G.C.M.
      • Praz F.
      • Lanz J.
      • Hunziker L.
      • Stortecky S.
      • Pilgrim T.
      • Heg D.
      • Valgimigli M.
      • Windecker S.
      • Räber L.
      Validation of the Academic Research Consortium for High Bleeding Risk (ARC-HBR) criteria in patients undergoing percutaneous coronary intervention and comparison with contemporary bleeding risk scores.
      ].
      Of note, our data showed that the rate of major bleeding incidence by 14 days after admission was in concordance with other studies, whereas the rate of any bleeding incidence was comparatively higher than those reported by other IMPROVE reports on medical patients [
      • Zhang Z.
      • Zhai Z.
      • Li W.
      • Qin X.
      • Qu J.
      • Shi Y.
      • Xu R.
      • Xu Y.
      • Wang C.
      Validation of the IMPROVE bleeding risk score in Chinese medical patients during hospitalization: findings from the dissolve-2 study.
      ,
      • Hostler D.C.
      • Marx E.S.
      • Moores L.K.
      • Petteys S.K.
      • Hostler J.M.
      • Mitchell J.D.
      • Holley P.R.
      • Collen J.F.
      • Foster B.E.
      • Holley A.B.
      Validation of the international medical prevention registry on venous thromboembolism bleeding risk score.
      ]. This could be due to the inherent tendency of increased bleeding risk in patients undergoing surgery. Moreover, higher reporting of clinically significant minor bleeding may have contributed to the observed higher rate of overall bleeding. It is also worth noting that the DissolVE-2 cohort demonstrated a similar number of major bleeding events in surgical (0.9 %) and medical (0.7 %) patients. In fact, in patients with an IMPROVE score ≥ 7, much lower major bleeding rates were observed among surgical patients compared to the medical cohort (2.6 % vs 4.5 %) of the DissolVE-2 study. Interestingly, this was in spite of more patients undergoing VTE prophylaxis in the surgical compared with the medical group (12.6 % vs 8.5 %) [
      • Zhang Z.
      • Zhai Z.
      • Li W.
      • Qin X.
      • Qu J.
      • Shi Y.
      • Xu R.
      • Xu Y.
      • Wang C.
      Validation of the IMPROVE bleeding risk score in Chinese medical patients during hospitalization: findings from the dissolve-2 study.
      ]. This seems to indicate that appropriate administration of VTE prophylaxis may not increase the risk of major bleeding events, as also reported by other studies [
      • Vasilakis V.
      • Klein G.M.
      • Trostler M.
      • Mukit M.
      • Marquez J.E.
      • Dagum A.B.
      • Pannucci C.J.
      • Khan S.U.
      Postoperative venous thromboembolism prophylaxis utilizing enoxaparin does not increase bleeding complications after abdominal body contouring surgery.
      ,
      • Parker S.G.
      • McGlone E.R.
      • Knight W.R.
      • Sufi P.
      • Khan O.A.
      Enoxaparin venous thromboembolism prophylaxis in bariatric surgery: a best evidence topic.
      ].
      The strengths of our study include a large, representative, nationwide sample size of surgical patients from the DissolVE-2 study, along with sufficient bleeding events to facilitate the robust assessment of both discrimination and calibration of the BRS. Although this study presented the analysis of surgical compared with medical patients from the previous IMPROVE studies, most predictive parameters for major bleeding events are on par with the other IMPROVE reports, expanding the validity of the model to surgical patients. Limitations of the study include its retrospective nature, where no follow-up data were available as it was assembled from the patients' medical records. Furthermore, 29 bleeding events were reported from the 587 patients who were excluded from this study, which may have introduced a selection bias. Nevertheless, the incidence rate was similar in our validation cohort and in the overall cohort of surgical patients (3.9 % vs 3.4 %). Hence, the effect of selection bias, if any, may be negligible. Undeniably, the bleeding risk assessment model of this study lacks specific surgical factors.

      5. Conclusion

      In conclusion, this was the first study to investigate the IMPROVE BRS in a large Chinese cohort of surgical patients, validating a simple and useful technique to predict the risk of major bleeding. This expands the applicability of the BRS for medical and surgical patients, which will enhance the safety and functional outcomes of all hospitalized patients. Future studies to externally validate IMPROVE BRS in surgical patients of other ethnicities will enable broader clinical application and will be based on IMPROVE BRS modified and validated to be more suitable for surgical patients.

      Author contribution

      Z.Z.G., C. W. is the guarantor of the article and takes responsibility for the content of the manuscript, including integrity of the data and accuracy of the analysis. Z. Z, K.Y.Z., Z. Z. G. and C. W. conceptualized and designed the study. Z. G. Z., X. Y. Q., Y. K. S., R. H. X., W. M. L., Q. C. K., Y. M. X., J. M. Q., Z. Z., and C. W. participated in data acquisition; Z. Z, K.Y.Z., Z. Z. G. analysed and interpreted the data; and Z. Z, K.Y.Z., Z. Z. G. and C. W. drafted the manuscript. All authors provided final approval of the version to be published.

      Role of the sponsors

      The authors are individually and collectively responsible for all content and editorial decisions and received no payment from Sanofi related to the development/presentation of this publication. The authors declare that there is no conflict of interest.

      Other contributions

      Medical writing support, under the direction of authors was provided by Dr. Amit Bhat (Indegene Pvt., Ltd., Bengaluru, India).and funded by Sanofi.

      Data availability

      Qualified researchers may request access to patient-level data and related study documents, including the clinical study report, study protocol with any amendments, blank case report form, statistical analysis plan, and dataset specifications. Patient-level data will be anonymized, and study documents will be redacted to protect the privacy of our trial participants. Further details on Sanofi's data sharing criteria, eligible studies, and process for requesting access can be found at: https://www.vivli.org/.

      Declaration of competing interest

      The authors are individually and collectively responsible for all content and editorial decisions and received no payment from Sanofi related to the development/presentation of this publication. The authors declare that there is no conflict of interest. Dan Shen, Jingjing Du, and Changbin Cai are all employees of Sanofi and may hold shares and/or stock options in the company.

      Acknowledgment

      The study was funded by CAMS Innovation Fund for Medical Sciences (CIFMS) (2021-I2M-1-049) and Sanofi.

      Appendix A. Supplementary data

      References

        • Ghadimi K.
        • Levy J.H.
        • Welsby I.J.
        Perioperative management of the bleeding patient.
        Br. J. Anaesth. 2016; 117: iii18-iii30https://doi.org/10.1093/bja/aew358
        • Hansrani V.
        • Khanbhai M.
        • McCollum C.
        The prevention of venous thromboembolism in surgical patients.
        in: Islam Md.S. Thrombosis and Embolism: From Research to Clinical Practice. Springer International Publishing, Cham2016: 1-8https://doi.org/10.1007/5584_2016_100
        • Curnow J.
        • Pasalic L.
        • Favaloro E.
        Why do patients bleed?.
        Surg J. 2016; 02: e29-e43https://doi.org/10.1055/s-0036-1579657
        • Marietta M.
        • Facchini L.
        • Pedrazzi P.
        • Busani S.
        • Torelli G.
        Pathophysiology of bleeding in surgery.
        Transplant. Proc. 2006; 38: 812-814https://doi.org/10.1016/j.transproceed.2006.01.047
        • Vasilakis V.
        • Klein G.M.
        • Trostler M.
        • Mukit M.
        • Marquez J.E.
        • Dagum A.B.
        • Pannucci C.J.
        • Khan S.U.
        Postoperative venous thromboembolism prophylaxis utilizing enoxaparin does not increase bleeding complications after abdominal body contouring surgery.
        Aesthet. Surg. J. 2020; 40: 989-995https://doi.org/10.1093/asj/sjz274
        • Parker S.G.
        • McGlone E.R.
        • Knight W.R.
        • Sufi P.
        • Khan O.A.
        Enoxaparin venous thromboembolism prophylaxis in bariatric surgery: a best evidence topic.
        Int. J. Surg. 2015; 23: 52-56https://doi.org/10.1016/j.ijsu.2015.09.005
        • Zhang Z.
        • Zhai Z.
        • Li W.
        • Qin X.
        • Qu J.
        • Shi Y.
        • Xu R.
        • Xu Y.
        • Wang C.
        Validation of the IMPROVE bleeding risk score in Chinese medical patients during hospitalization: findings from the dissolve-2 study.
        Lancet Reg.Health West.Pac. 2020; 4100054https://doi.org/10.1016/j.lanwpc.2020.100054
        • Decousus H.
        • Tapson V.F.
        • Bergmann J.-F.
        • Chong B.H.
        • Froehlich J.B.
        • Kakkar A.K.
        • Merli G.J.
        • Monreal M.
        • Nakamura M.
        • Pavanello R.
        • Pini M.
        • Piovella F.
        • Spencer F.A.
        • Spyropoulos A.C.
        • Turpie A.G.G.
        • Zotz R.B.
        • Fitzgerald G.
        • Anderson F.A.
        IMPROVE Investigators, Factors at admission associated with bleeding risk in medical patients: findings from the IMPROVE investigators.
        Chest. 2011; 139: 69-79https://doi.org/10.1378/chest.09-3081
        • Zhai Z.
        • Kan Q.
        • Li W.
        • Qin X.
        • Qu J.
        • Shi Y.
        • Xu R.
        • Xu Y.
        • Zhang Z.
        • Wang C.
        • Zhu F.
        • Jiang M.
        • Deng Y.
        • Li X.
        • Geng D.
        • Zhai J.
        • Wang C.
        • Zhang J.
        • Cao J.
        • Chen S.
        • Li G.
        • Wang L.
        • Zhang Y.
        • Liu X.
        • Zhang X.
        • Luo X.
        • Li X.
        • Zhao L.
        • Zu N.
        • Lu X.
        • Yang L.
        • Zhang N.
        • Pang W.
        • Ji Y.
        • Yao Y.
        • Sun X.
        • Xu W.
        • Zhao J.
        • Leng H.
        • Yu Z.
        • Kuang G.
        • Huang Y.
        • Chen L.
        • Shi J.
        • Xin S.
        • Xiao P.
        • Du Z.
        • Wei Y.
        • Liu Y.
        • Zhang C.
        • Xu Y.
        • Yang Z.
        • Liu R.
        • Liu Y.
        • Chen Y.
        • Qin J.
        • Liu S.
        • Zhou H.
        • Zhang Z.
        • Wei L.
        • Ren T.
        • Wang B.
        • Chen H.
        • Mao G.
        • Ying Y.
        • Ding G.
        • Dui D.
        • Du T.
        • Li X.
        • Lu J.
        • Zheng D.
        • He T.
        VTE risk profiles and prophylaxis in medical and surgical inpatients.
        Chest. 2019; 155: 114-122https://doi.org/10.1016/j.chest.2018.09.020
        • Büller H.R.
        Fondaparinux or enoxaparin for the initial treatment of symptomatic deep venous thrombosis: a randomized trial.
        Ann. Intern. Med. 2004; 140: 867https://doi.org/10.7326/0003-4819-140-11-200406010-00007
        • Bahit M.C.
        • Lopes R.D.
        • Wojdyla D.M.
        • Held C.
        • Hanna M.
        • Vinereanu D.
        • Hylek E.M.
        • Verheugt F.
        • Goto S.
        • Alexander J.H.
        • Wallentin L.
        • Granger C.B.
        Non-major bleeding with apixaban versus warfarin in patients with atrial fibrillation.
        Heart. 2017; 103: 623-628https://doi.org/10.1136/heartjnl-2016-309901
        • Hosmer D.W.
        • Lemeshow S.
        Applied Logistic Regression: Hosmer/Applied Logistic Regression.
        John Wiley & Sons, Inc., Hoboken, NJ, USA2000https://doi.org/10.1002/0471722146
        • Klein A.A.
        • Collier T.
        • Yeates J.
        • Miles L.F.
        • Fletcher S.N.
        • Evans C.
        • Richards T.
        The ACTA PORT-score for predicting perioperative risk of blood transfusion for adult cardiac surgery.
        Br. J. Anaesth. 2017; 119: 394-401https://doi.org/10.1093/bja/aex205
        • Algattas H.
        • Kimmell K.T.
        • Vates G.E.
        Risk of reoperation for hemorrhage in patients after craniotomy.
        World Neurosurg. 2016; 87: 531-539https://doi.org/10.1016/j.wneu.2015.09.020
        • Biancari F.
        • Brascia D.
        • Onorati F.
        • Reichart D.
        • Perrotti A.
        • Ruggieri V.G.
        • Santarpino G.
        • Maselli D.
        • Mariscalco G.
        • Gherli R.
        • Rubino A.S.
        • De Feo M.
        • Gatti G.
        • Santini F.
        • Dalén M.
        • Saccocci M.
        • Kinnunen E.-M.
        • Airaksinen J.K.E.
        • D’Errigo P.
        • Rosato S.
        • Nicolini F.
        Prediction of severe bleeding after coronary surgery: the WILL-BLEED Risk Score.
        Thromb. Haemost. 2017; 117: 445-456https://doi.org/10.1160/TH16-09-0721
        • Sławek-Szmyt S.
        • Araszkiewicz A.
        • Grygier M.
        • Szmyt K.
        • Seniuk W.
        • Waśniewski M.
        • Smukowski T.
        • Chmielewska-Michalak L.
        • Lesiak M.
        • Mitkowski P.
        PACE DRAP: a simple score for predicting significant bleeding complications after cardiac implantable electronic device surgery.
        Pol. Arch. Intern. Med. 2020; 130: 206-215https://doi.org/10.20452/pamw.15180
        • Rosenberg D.J.
        • Press A.
        • Fishbein J.
        • Lesser M.
        • McCullagh L.
        • McGinn T.
        • Spyropoulos A.C.
        External validation of the IMPROVE bleeding risk assessment model in medical patients.
        Thromb. Haemost. 2016; 116: 530-536https://doi.org/10.1160/TH16-01-0003
        • Cobben M.R.R.
        • Nemeth B.
        • Lijfering W.M.
        • Cannegieter S.C.
        Validation of risk assessment models for venous thrombosis in hospitalized medical patients.
        Res. Pract. Thromb. Haemost. 2019; 3: 217-225https://doi.org/10.1002/rth2.12181
      1. Venous thromboembolism: reducing the risk for patients in hospital | Guidance | NICE, (n.d.). https://www.nice.org.uk/guidance/cg92 (accessed December 22, 2021).

        • Ruíz-Giménez N.
        • Suárez C.
        • González R.
        • Nieto J.
        • Todolí J.
        • Samperiz Á.
        • Monreal M.
        • the RIETE Investigators
        Predictive variables for major bleeding events in patients presenting with documented acute venous thromboembolism.Findings from the RIETE Registry.
        Thromb. Haemost. 2008; 100: 26-31https://doi.org/10.1160/TH08-03-0193
        • Gould M.K.
        • Garcia D.A.
        • Wren S.M.
        • Karanicolas P.J.
        • Arcelus J.I.
        • Heit J.A.
        • Samama C.M.
        Prevention of VTE in nonorthopedic surgical patients: Antithrombotic Therapy and Prevention of Thrombosis, 9th ed: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines.
        Chest. 2012; 141: e227S-e277Shttps://doi.org/10.1378/chest.11-2297
        • Kearon C.
        • Akl E.A.
        • Comerota A.J.
        • Prandoni P.
        • Bounameaux H.
        • Goldhaber S.Z.
        • Nelson M.E.
        • Wells P.S.
        • Gould M.K.
        • Dentali F.
        • Crowther M.
        • Kahn S.R.
        Antithrombotic therapy for VTE disease.
        Chest. 2012; 141: e419S-e496Shttps://doi.org/10.1378/chest.11-2301
        • Johnstone C.
        • Rich S.E.
        Bleeding in cancer patients and its treatment: a review.
        Ann. Palliat. Med. 2018; 7: 265-273https://doi.org/10.21037/apm.2017.11.01
        • George A.
        • Stead T.S.
        • Ganti L.
        What's the risk: differentiating risk ratios, odds ratios, and hazard ratios?.
        Cureus. 2020; https://doi.org/10.7759/cureus.10047
        • Ueki Y.
        • Bär S.
        • Losdat S.
        • Otsuka T.
        • Zanchin C.
        • Zanchin T.
        • Gragnano F.
        • Gargiulo G.
        • Siontis G.C.M.
        • Praz F.
        • Lanz J.
        • Hunziker L.
        • Stortecky S.
        • Pilgrim T.
        • Heg D.
        • Valgimigli M.
        • Windecker S.
        • Räber L.
        Validation of the Academic Research Consortium for High Bleeding Risk (ARC-HBR) criteria in patients undergoing percutaneous coronary intervention and comparison with contemporary bleeding risk scores.
        EuroIntervention. 2020; 16: 371-379https://doi.org/10.4244/EIJ-D-20-00052
        • Hostler D.C.
        • Marx E.S.
        • Moores L.K.
        • Petteys S.K.
        • Hostler J.M.
        • Mitchell J.D.
        • Holley P.R.
        • Collen J.F.
        • Foster B.E.
        • Holley A.B.
        Validation of the international medical prevention registry on venous thromboembolism bleeding risk score.
        Chest. 2016; 149: 372-379https://doi.org/10.1378/chest.14-2842