A Machine Learning–Based Risk Stratification Model for Predicting Perioperative Blood Transfusion in Fracture Neck of Femur Surgery

A Machine Learning–Based Risk Stratification Model for Predicting Perioperative Blood Transfusion in Fracture Neck of Femur Surgery | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article A Machine Learning–Based Risk Stratification Model for Predicting Perioperative Blood Transfusion in Fracture Neck of Femur Surgery Olusegun Samson Oyagbesan, Abayomi Ojo¹, Bode¹ Afeniforo, Kehinde Oluwadiya¹, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8791028/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 6 You are reading this latest preprint version Abstract Background Hip fracture surgery in older adults is frequently associated with substantial blood loss, often necessitating perioperative blood transfusion. While transfusion can be life-saving, it is associated with increased morbidity, prolonged hospitalization, and higher healthcare costs. Accurate preoperative prediction of transfusion risk may improve patient blood management and perioperative planning. This study aimed to develop and internally validate a machine learning–based model to predict transfusion risk in patients undergoing hip fracture surgery. Methods We retrospectively studied 139 patients aged ≥65 years who underwent surgery for fracture neck of femur. Preoperative variables including age, packed cell volume (PCV), American Society of Anesthesiologists (ASA) grade, comorbidities, and surgical procedure type were used to develop a Random Forest model. Model performance was assessed using five-fold cross-validation and an independent test set. A multivariable logistic regression model was developed as a comparator. Patients were stratified into low-, intermediate-, and high-risk groups based on predicted probabilities. Results The Random Forest model demonstrated good discrimination during cross-validation (mean AUC–ROC 0.821) and excellent performance on the test set (AUC–ROC 0.966). At an optimal threshold of 0.60, sensitivity was 100% and specificity was 88.9%. Risk stratification revealed transfusion rates of 0%, 25%, and 85.7% in the low-, intermediate-, and high-risk groups, respectively. Age and preoperative PCV were the most influential predictors. The logistic regression model showed lower discriminative performance. Conclusion A Random Forest machine learning model using routinely available preoperative variables accurately predicts perioperative transfusion risk following fracture neck of femur surgery. This internally validated risk stratification approach supports individualized patient blood management and optimized perioperative resource allocation. Fracture neck of femur Blood transfusion Machine learning Risk stratification Random Forest Patient blood management Figures Figure 1 Figure 2 Figure 3 Figure 4 Background Fracture of the neck of the femur is one of the most serious injuries affecting older adults. It is linked to high rates of illness, death, and healthcare costs around the world. The global incidence of hip fractures is expected to rise significantly due to an aging population, especially in low- and middle-income countries where healthcare resources are often limited [ 1 ]. Surgery is the primary treatment; however, operating on a fractured neck of the femur frequently leads to substantial blood loss. As a result, many patients receive blood transfusions. Older patients are especially at risk for anemia during and after surgery due to existing health issues, nutritional deficiencies, and changes that come with age. This risk is worsened by blood loss from both the fracture and the surgery itself [ 3 , 4 ]. While transfusions can save lives, they have been linked to higher risks of infection, heart problems, longer hospital stays, and increased mortality [ 5 – 7 ]. Additionally, blood products are often limited and expensive in many healthcare systems. Patient blood management (PBM) focuses on identifying patients who are at high risk for needing transfusions early on. This allows for targeted interventions [ 8 ]. Traditional methods for predicting transfusion risk use linear statistical models and single predictors, which may not capture the complex relationships between patient conditions and surgical factors. Machine learning (ML) methods, especially ensemble algorithms like Random Forests, can model non-linear relationships and complex interactions without needing pre-determined assumptions [ 9 – 11 ]. Despite the growing use of ML in orthopedic surgery, there are still few ML-based models for predicting transfusion needs in surgeries for fracture neck of femur, particularly in low-resource settings. The aim of this study was to develop and internally validate a machine learning model for predicting perioperative blood transfusion in elderly patients undergoing surgery for fracture neck of femur. We also wanted to compare its performance to a traditional multivariable logistic regression model. Methods Study Design and Participants This retrospective observational study was conducted at a tertiary hospital following Institutional Review Board approval. Patients aged ≥ 65 years who underwent surgical treatment for fracture neck of femur between June, 2024 and June 2025 were screened. Pathological fractures, periprosthetic fractures, and records with missing key variables were excluded. After exclusions, 139 patients were included in the final analysis. This study represents a prediction model development and internal validation analysis using all available retrospective cases; no formal a priori sample size calculation was performed. Ethics Approval This study was conducted in accordance with the ethical standards of the institutional research committee and with the 1964 Declaration of Helsinki and its later amendments. Ethical approval was obtained from the institutional ethics committee with registration number IRB/IEC/Z0004553. Due to the retrospective nature of the study, the requirement for informed consent was waived. Data Collection and Outcome Definition Preoperative variables included age, sex, PCV, ASA grade, comorbidities (hypertension, diabetes mellitus, atrial flutter, stroke, Parkinson’s disease, dementia, obesity), and surgical procedure type. The primary outcome was receipt of at least one unit of allogeneic blood transfusion during the perioperative hospital admission (intraoperative and/or postoperative). Model Development and Statistical Analysis The dataset was randomly divided into a training cohort (n = 108) and an independent test cohort (n = 31). A Random Forest classifier was developed using the scikit-learn library (Python), following established ensemble learning methodology [ 11 , 12 ]. Hyperparameter tuning was performed using five-fold cross-validation. A multivariable binary logistic regression model using the same predictors was fitted as a comparator model . Model performance metrics included accuracy, sensitivity, specificity, precision, F1-score, and area under the receiver operating characteristic curve (AUC-ROC). The optimal probability threshold was determined using Youden’s J statistic [ 13 ]. The Random Forest model demonstrated good discriminative performance for predicting perioperative blood transfusion, with the receiver operating characteristic (ROC) curve shown in Fig. 1 . The Logistic Regression model also showed acceptable predictive performance, and its ROC curve is presented in Fig. 2 . Risk Stratification Patients were categorized into low- (< 0.30), intermediate- (0.30–0.59), and high-risk (≥ 0.60) groups. These thresholds were data-driven , informed by ROC analysis and Youden’s J statistic rather than arbitrary categorization Reporting Guideline This study was conducted and reported in accordance with the TRIPOD-AI reporting guideline for prediction model development and internal validation. Results Baseline Characteristics The mean age was 78.4 ± 8.6 years, and 66.2% of patients were female. Overall, 51 patients (36.7%) received perioperative blood transfusion. Table 1. Baseline Characteristics of the Study Population (n=139) Univariate Comparison Patients who received transfusion were significantly older, had lower preoperative PCV, higher ASA grade, and were more likely to undergo hemiarthroplasty (Table 2). Model Performance The Random Forest model achieved a cross-validated AUC-ROC of 0.821 and a test-set AUC-ROC of 0.966 (Figure 1). At the optimal threshold of 0.60, sensitivity was 100% and specificity was 88.9% (Table 3). The logistic regression comparator demonstrated lower discrimination on the test set (AUC approximately 0.65–0.70), indicating inferior performance relative to the Random Forest model. The Logistic Regression model ROC curve is presented in Fig. 2 . Risk Stratification Risk stratification revealed transfusion rates of 0%, 25%, and 85.7% in low-, intermediate-, and high-risk groups, respectively (Table 4). Observed transfusion rates across the low-, intermediate-, and high-risk groups are shown in Figure 3. Feature Importance Age and preoperative PCV were the most influential predictors, followed by hemiarthroplasty, based on mean decrease in Gini impurity—a standard feature importance metric in Random Forest models [11]. Feature importance derived from the Random Forest model is illustrated in Figure 4. Discussion This study shows that perioperative blood transfusion risk in elderly patients undergoing fracture neck of femur surgery can be accurately predicted by a Random Forest machine learning model using routinely available preoperative variables. During internal validation, the model outperformed a traditional logistic regression comparator and demonstrated strong discriminative performance. The model's key predictors, such as age, preoperative PCV, and type of surgical procedure, are in line with previous research, which supports the findings' clinical plausibility [6,15–18]. Crucially, complex non-linear interactions between predictors that are challenging to model with conventional statistical techniques are captured by the ML model. The translation of predicted probabilities into clinically intuitive risk categories represents a major strength. The absence of transfusion in the low-risk group suggests potential avoidance of routine cross-matching, while the high transfusion rate in the high-risk group supports early blood preparation and proactive anemia management. Implications for Low- and Middle-Income Settings In resource-constrained environments, efficient allocation of scarce blood products is critical. The proposed model relies solely on routinely collected preoperative data, making it feasible for implementation without additional infrastructure. Limitations Several limitations warrant consideration. First, this was a single-center retrospective study with a modest sample size; findings should therefore be interpreted as model development with internal validation , rather than definitive clinical implementation. Second, perioperative transfusion decisions are influenced by clinician judgment, institutional protocols, and blood availability, introducing subjectivity into the outcome. Third, some potentially relevant variables were not included. Fourth, confidence intervals for test-set performance metrics could not be estimated due to limited sample size and lack of resampling in the held-out dataset. External validation is required before clinical adoption. Conclusion A Random Forest machine learning model accurately predicts perioperative blood transfusion risk following fracture neck of femur surgery using simple preoperative variables. This internally validated risk stratification approach supports individualized patient blood management and optimized perioperative planning, particularly in resource-limited settings. Declarations Author Contributions OOS conceived the study, performed data analysis, and drafted the manuscript. BA and AO contributed to data interpretation and manuscript revision. OK and OEA provided senior supervision, methodological guidance, and critical revision of the manuscript. All authors approved the final version of the manuscript. Data Availability The datasets analyzed during this study are available from the corresponding author upon reasonable request. Funding This research received no external funding. Conflicts of Interest The authors declare no conflict of interest. Ethics Approval Ethical approval was obtained from the Health Research Ethics Committee of Obafemi Awolowo University Teaching Hospital Complex, Ile Ife, Nigeria with registration number IRB/IEC/Z0004553. Due to the retrospective nature of the study, the requirement for informed consent was waived by the same committee. Consent to Participate Not applicable. Consent for Publication Not applicable. References Veronese N, Maggi S. Epidemiology and social costs of hip fracture. Injury . 2018;49(8):1458–1460. Carson JL, Guyatt G, Heddle NM, et al. Clinical practice guidelines from the AABB: red blood cell transfusion thresholds and storage. JAMA . 2016;316(19):2025–2035. Wasserstein D, Farooq H, Govindarajan A, et al. Perioperative blood transfusion and postoperative complications following orthopaedic surgery. Bone Joint J . 2019;101-B(8):941–947. Frisch NB, Wessell NM, Charters MA, et al. Predictors and complications of blood transfusion in total hip and knee arthroplasty. J Arthroplasty . 2014;29(9 Suppl):189–192. Muñoz M, Acheson AG, Auerbach M, et al. International consensus statement on the peri-operative management of anaemia and iron deficiency. Anaesthesia . 2017;72(2):233–247. Foss NB, Kehlet H. Hidden blood loss after surgery for hip fracture. J Bone Joint Surg Br . 2006;88(8):1053–1059. Smith GH, Tsang J, Molyneux SG, White TO. The hidden blood loss after hip fracture. Injury . 2011;42(2):133–135. Sidey-Gibbons JAM, Sidey-Gibbons CJ. Machine learning in medicine: a practical introduction. BMC Med Res Methodol . 2019;19(1):64. Jans Ø, Jørgensen C, Kehlet H, Johansson PI. Role of preoperative anemia for risk of transfusion and postoperative morbidity in fast-track hip and knee arthroplasty. Transfusion . 2014;54(3):717–726. Spahn DR. Anemia and patient blood management in hip and knee surgery: a systematic review of the literature. Anesthesiology . 2010;113(2):482–495. Breiman L. Random forests. Machine Learning . 2001;45(1):5–32. Pedregosa F, Varoquaux G, Gramfort A, et al. Scikit-learn: machine learning in Python. J Mach Learn Res . 2011;12:2825–2830. Youden WJ. Index for rating diagnostic tests. Cancer . 1950;3(1):32–35. Tables Table 1. Baseline Characteristics of the Study Population (n = 139) Variable Overall Age, years (mean ± SD) 78.4 ± 8.6 Female sex, n (%) 92 (66.2) Preoperative PCV (%), mean ± SD 30.8 ± 4.9 ASA III, n (%) 82 (59.0) Hypertension, n (%) 88 (63.3) Diabetes mellitus, n (%) 34 (24.5) Dementia, n (%) 22 (15.8) Hemiarthroplasty, n (%) 54 (38.8) Perioperative transfusion, n (%) 51 (36.7) Table 2. Comparison of Patients With and Without Transfusion Variable Transfusion (n = 51) No Transfusion (n = 88) p-value Age (years), mean ± SD 82.1 ± 7.4 76.3 ± 8.5 <0.001 Preoperative PCV (%) 27.6 ± 4.2 32.7 ± 4.3 <0.001 ASA III, % 74.5 50.0 0.006 Hemiarthroplasty, % 58.8 27.3 <0.001 Hypertension, % 72.5 58.0 0.08 Table 3. Model Performance on Independent Test Set (n = 31) Metric Value Accuracy 93.5% Sensitivity 100% Specificity 88.9% Precision 86.7% F1-score 92.9% AUC–ROC 0.966 Table 4. Risk Stratification Based on Predicted Probability Risk Group Probability Range Transfusion Rate Low < 0.30 0% Intermediate 0.30–0.59 25% High ≥ 0.60 85.7% Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 25 Mar, 2026 Reviewers invited by journal 17 Mar, 2026 Editor assigned by journal 16 Mar, 2026 Editor invited by journal 19 Feb, 2026 Submission checks completed at journal 19 Feb, 2026 First submitted to journal 19 Feb, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8791028","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":607381169,"identity":"c1e10c2f-5b2f-4ebc-bd0c-b0e84714ddbc","order_by":0,"name":"Olusegun Samson Oyagbesan","email":"data:image/png;base64,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","orcid":"","institution":"Obafemi Awolowo University Teaching Hospitals Complex","correspondingAuthor":true,"prefix":"","firstName":"Olusegun","middleName":"Samson","lastName":"Oyagbesan","suffix":""},{"id":607381170,"identity":"6fdec05c-387d-4310-a5e0-6ffb95a0c96f","order_by":1,"name":"Abayomi Ojo¹","email":"","orcid":"","institution":"Obafemi Awolowo University Teaching Hospitals Complex","correspondingAuthor":false,"prefix":"","firstName":"Abayomi","middleName":"","lastName":"Ojo¹","suffix":""},{"id":607381171,"identity":"4ccae56c-e4c9-4b28-a62c-1c1f48fac336","order_by":2,"name":"Bode¹ Afeniforo","email":"","orcid":"","institution":"Obafemi Awolowo University Teaching Hospitals Complex","correspondingAuthor":false,"prefix":"","firstName":"Bode¹","middleName":"","lastName":"Afeniforo","suffix":""},{"id":607381174,"identity":"661fea20-0d57-4071-8942-0797832bd5f9","order_by":3,"name":"Kehinde Oluwadiya¹","email":"","orcid":"","institution":"Ekiti State University","correspondingAuthor":false,"prefix":"","firstName":"Kehinde","middleName":"","lastName":"Oluwadiya¹","suffix":""},{"id":607381175,"identity":"f97aed3b-eaee-4474-b1e0-54c6606a76e4","order_by":4,"name":"Ayodele Emmanuel Orimolade","email":"","orcid":"","institution":"Obafemi Awolowo University Teaching Hospitals Complex","correspondingAuthor":false,"prefix":"","firstName":"Ayodele","middleName":"Emmanuel","lastName":"Orimolade","suffix":""}],"badges":[],"createdAt":"2026-02-05 01:23:15","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8791028/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8791028/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104998232,"identity":"ee3f5a6d-4a60-47da-85b4-03a2fa6c0de3","added_by":"auto","created_at":"2026-03-19 16:25:50","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":237792,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eROC curve for the Random Forest model (test set)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eDescription:\u003c/em\u003eThe ROC curve demonstrates excellent discrimination on the independent test set (AUC = 0.966). The diagonal line indicates chance performance.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-8791028/v1/ca9439941801317dd7e5c017.png"},{"id":104998236,"identity":"7240ed73-b248-466d-bdda-03d555b41eb9","added_by":"auto","created_at":"2026-03-19 16:25:51","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":202821,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eROC curve for the logistic regression comparator\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eDescription:\u003c/em\u003eThe logistic regression model shows lower discrimination compared with the Random Forest model (AUC ≈ 0.65–0.70).\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-8791028/v1/686dbf534899ed6ef25577fc.png"},{"id":104998164,"identity":"1ae887b3-6fc4-49fa-b68c-fa167726f57e","added_by":"auto","created_at":"2026-03-19 16:25:37","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":124373,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eRisk stratification by predicted transfusion probability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eDescription:\u003c/em\u003eObserved transfusion rates across low-, intermediate-, and high-risk groups derived from data-driven probability thresholds.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-8791028/v1/02c36887dd5fc2c0363192bb.png"},{"id":104998240,"identity":"8a9b5883-77f2-46e2-8909-03a38fae50a7","added_by":"auto","created_at":"2026-03-19 16:25:53","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":127922,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFeature importance for the Random Forest model\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eDescription:\u003c/em\u003eFeature importance based on mean decrease in Gini impurity, highlighting age and preoperative PCV as dominant predictors.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-8791028/v1/12d5c1ba4e80d6a2aa313f69.png"},{"id":104998301,"identity":"218e313d-5778-4ced-b40f-993ea42bb963","added_by":"auto","created_at":"2026-03-19 16:26:17","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1545903,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8791028/v1/d559374f-5c06-4299-86c6-b80f43b4880b.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"A Machine Learning–Based Risk Stratification Model for Predicting Perioperative Blood Transfusion in Fracture Neck of Femur Surgery","fulltext":[{"header":"Background","content":"\u003cp\u003eFracture of the neck of the femur is one of the most serious injuries affecting older adults. It is linked to high rates of illness, death, and healthcare costs around the world. The global incidence of hip fractures is expected to rise significantly due to an aging population, especially in low- and middle-income countries where healthcare resources are often limited [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Surgery is the primary treatment; however, operating on a fractured neck of the femur frequently leads to substantial blood loss. As a result, many patients receive blood transfusions.\u003c/p\u003e \u003cp\u003eOlder patients are especially at risk for anemia during and after surgery due to existing health issues, nutritional deficiencies, and changes that come with age. This risk is worsened by blood loss from both the fracture and the surgery itself [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. While transfusions can save lives, they have been linked to higher risks of infection, heart problems, longer hospital stays, and increased mortality [\u003cspan additionalcitationids=\"CR6\" citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Additionally, blood products are often limited and expensive in many healthcare systems.\u003c/p\u003e \u003cp\u003ePatient blood management (PBM) focuses on identifying patients who are at high risk for needing transfusions early on. This allows for targeted interventions [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Traditional methods for predicting transfusion risk use linear statistical models and single predictors, which may not capture the complex relationships between patient conditions and surgical factors.\u003c/p\u003e \u003cp\u003eMachine learning (ML) methods, especially ensemble algorithms like Random Forests, can model non-linear relationships and complex interactions without needing pre-determined assumptions [\u003cspan additionalcitationids=\"CR10\" citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Despite the growing use of ML in orthopedic surgery, there are still few ML-based models for predicting transfusion needs in surgeries for fracture neck of femur, particularly in low-resource settings.\u003c/p\u003e \u003cp\u003eThe aim of this study was to develop and internally validate a machine learning model for predicting perioperative blood transfusion in elderly patients undergoing surgery for fracture neck of femur. We also wanted to compare its performance to a traditional multivariable logistic regression model.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStudy Design and Participants\u003c/h2\u003e \u003cp\u003e This retrospective observational study was conducted at a tertiary hospital following Institutional Review Board approval. Patients aged\u0026thinsp;\u0026ge;\u0026thinsp;65 years who underwent surgical treatment for fracture neck of femur between June, 2024 and June 2025 were screened. Pathological fractures, periprosthetic fractures, and records with missing key variables were excluded. After exclusions, 139 patients were included in the final analysis.\u003c/p\u003e \u003cp\u003eThis study represents a \u003cb\u003eprediction model development and internal validation analysis\u003c/b\u003e using all available retrospective cases; no formal a priori sample size calculation was performed.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eEthics Approval\u003c/h3\u003e\n\u003cp\u003e This study was conducted in accordance with the ethical standards of the institutional research committee and with the 1964 Declaration of Helsinki and its later amendments. Ethical approval was obtained from the institutional ethics committee with registration number IRB/IEC/Z0004553. Due to the retrospective nature of the study, the requirement for informed consent was waived.\u003c/p\u003e\n\u003ch3\u003eData Collection and Outcome Definition\u003c/h3\u003e\n\u003cp\u003ePreoperative variables included age, sex, PCV, ASA grade, comorbidities (hypertension, diabetes mellitus, atrial flutter, stroke, Parkinson\u0026rsquo;s disease, dementia, obesity), and surgical procedure type.\u003c/p\u003e \u003cp\u003eThe primary outcome was receipt of at least one unit of allogeneic blood transfusion during the perioperative hospital admission (intraoperative and/or postoperative).\u003c/p\u003e\n\u003ch3\u003eModel Development and Statistical Analysis\u003c/h3\u003e\n\u003cp\u003eThe dataset was randomly divided into a training cohort (n\u0026thinsp;=\u0026thinsp;108) and an independent test cohort (n\u0026thinsp;=\u0026thinsp;31). A Random Forest classifier was developed using the scikit-learn library (Python), following established ensemble learning methodology [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Hyperparameter tuning was performed using five-fold cross-validation.\u003c/p\u003e \u003cp\u003eA multivariable binary logistic regression model using the same predictors was fitted as a \u003cb\u003ecomparator model\u003c/b\u003e. Model performance metrics included accuracy, sensitivity, specificity, precision, F1-score, and area under the receiver operating characteristic curve (AUC-ROC). The optimal probability threshold was determined using Youden\u0026rsquo;s J statistic [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. The Random Forest model demonstrated good discriminative performance for predicting perioperative blood transfusion, with the receiver operating characteristic (ROC) curve shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. The Logistic Regression model also showed acceptable predictive performance, and its ROC curve is presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eRisk Stratification\u003c/h3\u003e\n\u003cp\u003ePatients were categorized into low- (\u0026lt;\u0026thinsp;0.30), intermediate- (0.30\u0026ndash;0.59), and high-risk (\u0026ge;\u0026thinsp;0.60) groups. These thresholds were \u003cb\u003edata-driven\u003c/b\u003e, informed by ROC analysis and Youden\u0026rsquo;s J statistic rather than arbitrary categorization\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eReporting Guideline\u003c/h2\u003e \u003cp\u003eThis study was conducted and reported in accordance with the \u003cb\u003eTRIPOD-AI reporting guideline\u003c/b\u003e for prediction model development and internal validation.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eBaseline Characteristics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe mean age was 78.4 ± 8.6 years, and 66.2% of patients were female. Overall, 51 patients (36.7%) received perioperative blood transfusion.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 1. Baseline Characteristics of the Study Population (n=139)\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eUnivariate Comparison\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePatients who received transfusion were significantly older, had lower preoperative PCV, higher ASA grade, and were more likely to undergo hemiarthroplasty (Table 2).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eModel Performance\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe Random Forest model achieved a cross-validated AUC-ROC of 0.821 and a test-set AUC-ROC of 0.966 \u003cstrong\u003e(Figure 1).\u003c/strong\u003e At the optimal threshold of 0.60, sensitivity was 100% and specificity was 88.9% (Table 3).\u003c/p\u003e\n\u003cp\u003eThe logistic regression comparator demonstrated lower discrimination on the test set (AUC approximately 0.65–0.70), indicating inferior performance relative to the Random Forest model. The Logistic Regression model ROC curve is presented in \u003cstrong\u003eFig. 2\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRisk Stratification\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eRisk stratification revealed transfusion rates of 0%, 25%, and 85.7% in low-, intermediate-, and high-risk groups, respectively (Table 4). Observed transfusion rates across the low-, intermediate-, and high-risk groups are shown in Figure 3.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFeature Importance\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAge and preoperative PCV were the most influential predictors, followed by hemiarthroplasty, based on mean decrease in Gini impurity—a standard feature importance metric in Random Forest models [11]. Feature importance derived from the Random Forest model is illustrated in Figure 4.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study shows that perioperative blood transfusion risk in elderly patients undergoing fracture neck of femur surgery can be accurately predicted by a Random Forest machine learning model using routinely available preoperative variables. During internal validation, the model outperformed a traditional logistic regression comparator and demonstrated strong discriminative performance.\u003c/p\u003e \u003cp\u003eThe model's key predictors, such as age, preoperative PCV, and type of surgical procedure, are in line with previous research, which supports the findings' clinical plausibility [6,15\u0026ndash;18]. Crucially, complex non-linear interactions between predictors that are challenging to model with conventional statistical techniques are captured by the ML model.\u003c/p\u003e \u003cp\u003eThe translation of predicted probabilities into clinically intuitive risk categories represents a major strength. The absence of transfusion in the low-risk group suggests potential avoidance of routine cross-matching, while the high transfusion rate in the high-risk group supports early blood preparation and proactive anemia management.\u003c/p\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eImplications for Low- and Middle-Income Settings\u003c/h2\u003e \u003cp\u003eIn resource-constrained environments, efficient allocation of scarce blood products is critical. The proposed model relies solely on routinely collected preoperative data, making it feasible for implementation without additional infrastructure.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eLimitations\u003c/h2\u003e \u003cp\u003eSeveral limitations warrant consideration. First, this was a single-center retrospective study with a modest sample size; findings should therefore be interpreted as \u003cb\u003emodel development with internal validation\u003c/b\u003e, rather than definitive clinical implementation. Second, perioperative transfusion decisions are influenced by clinician judgment, institutional protocols, and blood availability, introducing subjectivity into the outcome. Third, some potentially relevant variables were not included. Fourth, confidence intervals for test-set performance metrics could not be estimated due to limited sample size and lack of resampling in the held-out dataset. External validation is required before clinical adoption.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eA Random Forest machine learning model accurately predicts perioperative blood transfusion risk following fracture neck of femur surgery using simple preoperative variables. This internally validated risk stratification approach supports individualized patient blood management and optimized perioperative planning, particularly in resource-limited settings.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eOOS\u003c/strong\u003e conceived the study, performed data analysis, and drafted the manuscript.\u003cbr\u003e\u003cstrong\u003eBA and AO\u0026nbsp;\u003c/strong\u003econtributed to data interpretation and manuscript revision.\u003cbr\u003e\u003cstrong\u003eOK\u0026nbsp;\u003c/strong\u003eand\u0026nbsp;\u003cstrong\u003eOEA\u003c/strong\u003e provided senior supervision, methodological guidance, and critical revision of the manuscript.\u003cbr\u003e\u0026nbsp;All authors approved the final version of the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets analyzed during this study are available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003eFunding\u003c/p\u003e\n\u003cp\u003eThis research received no external funding.\u003c/p\u003e\n\u003cp\u003eConflicts of Interest\u003c/p\u003e\n\u003cp\u003eThe authors declare no conflict of interest.\u003c/p\u003e\n\u003cp\u003eEthics Approval\u003c/p\u003e\n\u003cp\u003eEthical approval was obtained from the Health Research Ethics Committee of Obafemi Awolowo University Teaching Hospital Complex, Ile Ife, Nigeria with registration number IRB/IEC/Z0004553. Due to the retrospective nature of the study, the requirement for informed consent was waived by the same committee.\u003c/p\u003e\n\u003cp\u003eConsent to Participate\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003eConsent for Publication\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eVeronese N, Maggi S. Epidemiology and social costs of hip fracture. \u003cem\u003eInjury\u003c/em\u003e. 2018;49(8):1458\u0026ndash;1460.\u003c/li\u003e\n \u003cli\u003eCarson JL, Guyatt G, Heddle NM, et al. Clinical practice guidelines from the AABB: red blood cell transfusion thresholds and storage. \u003cem\u003eJAMA\u003c/em\u003e. 2016;316(19):2025\u0026ndash;2035.\u003c/li\u003e\n \u003cli\u003eWasserstein D, Farooq H, Govindarajan A, et al. Perioperative blood transfusion and postoperative complications following orthopaedic surgery. \u003cem\u003eBone Joint J\u003c/em\u003e. 2019;101-B(8):941\u0026ndash;947.\u003c/li\u003e\n \u003cli\u003eFrisch NB, Wessell NM, Charters MA, et al. Predictors and complications of blood transfusion in total hip and knee arthroplasty. \u003cem\u003eJ Arthroplasty\u003c/em\u003e. 2014;29(9 Suppl):189\u0026ndash;192.\u003c/li\u003e\n \u003cli\u003eMu\u0026ntilde;oz M, Acheson AG, Auerbach M, et al. International consensus statement on the peri-operative management of anaemia and iron deficiency. \u003cem\u003eAnaesthesia\u003c/em\u003e. 2017;72(2):233\u0026ndash;247.\u003c/li\u003e\n \u003cli\u003eFoss NB, Kehlet H. Hidden blood loss after surgery for hip fracture. \u003cem\u003eJ Bone Joint Surg Br\u003c/em\u003e. 2006;88(8):1053\u0026ndash;1059.\u003c/li\u003e\n \u003cli\u003eSmith GH, Tsang J, Molyneux SG, White TO. The hidden blood loss after hip fracture. \u003cem\u003eInjury\u003c/em\u003e. 2011;42(2):133\u0026ndash;135.\u003c/li\u003e\n \u003cli\u003eSidey-Gibbons JAM, Sidey-Gibbons CJ. Machine learning in medicine: a practical introduction. \u003cem\u003eBMC Med Res Methodol\u003c/em\u003e. 2019;19(1):64.\u003c/li\u003e\n \u003cli\u003eJans \u0026Oslash;, J\u0026oslash;rgensen C, Kehlet H, Johansson PI. Role of preoperative anemia for risk of transfusion and postoperative morbidity in fast-track hip and knee arthroplasty. \u003cem\u003eTransfusion\u003c/em\u003e. 2014;54(3):717\u0026ndash;726.\u003c/li\u003e\n \u003cli\u003eSpahn DR. Anemia and patient blood management in hip and knee surgery: a systematic review of the literature. \u003cem\u003eAnesthesiology\u003c/em\u003e. 2010;113(2):482\u0026ndash;495.\u003c/li\u003e\n \u003cli\u003eBreiman L. Random forests. \u003cem\u003eMachine Learning\u003c/em\u003e. 2001;45(1):5\u0026ndash;32.\u003c/li\u003e\n \u003cli\u003ePedregosa F, Varoquaux G, Gramfort A, et al. Scikit-learn: machine learning in Python. \u003cem\u003eJ Mach Learn Res\u003c/em\u003e. 2011;12:2825\u0026ndash;2830.\u003c/li\u003e\n \u003cli\u003eYouden WJ. Index for rating diagnostic tests. \u003cem\u003eCancer\u003c/em\u003e. 1950;3(1):32\u0026ndash;35.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003e\u003cstrong\u003eTable 1. Baseline Characteristics of the Study Population (n = 139)\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"3\" cellpadding=\"0\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eVariable\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eOverall\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eAge, years (mean \u0026plusmn; SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e78.4 \u0026plusmn; 8.6\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eFemale sex, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e92 (66.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003ePreoperative PCV (%), mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e30.8 \u0026plusmn; 4.9\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eASA III, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e82 (59.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eHypertension, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e88 (63.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eDiabetes mellitus, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e34 (24.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eDementia, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e22 (15.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eHemiarthroplasty, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e54 (38.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003ePerioperative transfusion, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e51 (36.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003eTable 2. Comparison of Patients With and Without Transfusion\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"3\" cellpadding=\"0\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eVariable\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eTransfusion (n = 51)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eNo Transfusion (n = 88)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003ep-value\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eAge (years), mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e82.1 \u0026plusmn; 7.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e76.3 \u0026plusmn; 8.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003ePreoperative PCV (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e27.6 \u0026plusmn; 4.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e32.7 \u0026plusmn; 4.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eASA III, %\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e74.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e50.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.006\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eHemiarthroplasty, %\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e58.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e27.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eHypertension, %\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e72.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e58.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.08\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003eTable 3. Model Performance on Independent Test Set (n = 31)\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"3\" cellpadding=\"0\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eMetric\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eValue\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eAccuracy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e93.5%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eSensitivity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e100%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eSpecificity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e88.9%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003ePrecision\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e86.7%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eF1-score\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e92.9%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eAUC\u0026ndash;ROC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.966\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003eTable 4. Risk Stratification Based on Predicted Probability\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"3\" cellpadding=\"0\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eRisk Group\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eProbability Range\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eTransfusion Rate\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eLow\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026lt; 0.30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eIntermediate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.30\u0026ndash;0.59\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e25%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eHigh\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026ge; 0.60\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e85.7%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-informatics-and-decision-making","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"midm","sideBox":"Learn more about [BMC Medical Informatics and Decision Making](http://bmcmedinformdecismak.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/midm/default.aspx","title":"BMC Medical Informatics and Decision Making","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Fracture neck of femur, Blood transfusion, Machine learning, Risk stratification, Random Forest, Patient blood management","lastPublishedDoi":"10.21203/rs.3.rs-8791028/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8791028/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHip fracture surgery in older adults is frequently associated with substantial blood loss, often necessitating perioperative blood transfusion. While transfusion can be life-saving, it is associated with increased morbidity, prolonged hospitalization, and higher healthcare costs. Accurate preoperative prediction of transfusion risk may improve patient blood management and perioperative planning. This study aimed to develop and internally validate a machine learning–based model to predict transfusion risk in patients undergoing hip fracture surgery.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe retrospectively studied 139 patients aged ≥65 years who underwent surgery for fracture neck of femur. Preoperative variables including age, packed cell volume (PCV), American Society of Anesthesiologists (ASA) grade, comorbidities, and surgical procedure type were used to develop a Random Forest model. Model performance was assessed using five-fold cross-validation and an independent test set. A multivariable logistic regression model was developed as a comparator. Patients were stratified into low-, intermediate-, and high-risk groups based on predicted probabilities.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe Random Forest model demonstrated good discrimination during cross-validation (mean AUC–ROC 0.821) and excellent performance on the test set (AUC–ROC 0.966). At an optimal threshold of 0.60, sensitivity was 100% and specificity was 88.9%. Risk stratification revealed transfusion rates of 0%, 25%, and 85.7% in the low-, intermediate-, and high-risk groups, respectively. Age and preoperative PCV were the most influential predictors. The logistic regression model showed lower discriminative performance.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA Random Forest machine learning model using routinely available preoperative variables accurately predicts perioperative transfusion risk following fracture neck of femur surgery. This internally validated risk stratification approach supports individualized patient blood management and optimized perioperative resource allocation.\u003c/p\u003e","manuscriptTitle":"A Machine Learning–Based Risk Stratification Model for Predicting Perioperative Blood Transfusion in Fracture Neck of Femur Surgery","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-19 16:23:43","doi":"10.21203/rs.3.rs-8791028/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"80611517552680750342184525638148599305","date":"2026-03-26T02:24:12+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-03-17T07:00:45+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-03-16T10:13:24+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-02-20T04:26:43+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-02-19T16:27:55+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Informatics and Decision Making","date":"2026-02-19T14:39:57+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-informatics-and-decision-making","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"midm","sideBox":"Learn more about [BMC Medical Informatics and Decision Making](http://bmcmedinformdecismak.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/midm/default.aspx","title":"BMC Medical Informatics and Decision Making","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"cbc0ade3-df14-4e2e-9ef9-76cc67c7d6df","owner":[],"postedDate":"March 19th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-03-19T16:23:43+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-19 16:23:43","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8791028","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8791028","identity":"rs-8791028","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}