Development of a Fall-Related Injury Risk-Stratified Prediction Model for Older Adults

Development of a Fall-Related Injury Risk-Stratified Prediction Model for Older Adults | 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 Development of a Fall-Related Injury Risk-Stratified Prediction Model for Older Adults Xue Zhang, Yan Cai, Wei Zhu, Cong Wang, Shanshan Liu, Yan Jiang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8683642/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Falls and related injuries among older adults represent a significant public health issue, adversely affecting their quality of life and increasing the socioeconomic burden. This study aims to develop a risk stratification prediction model based on machine learning algorithms. Methods Using a convenience sampling method, a total of 402 older adults scheduled for discharge at the Center of Gerontology and Geriatrics of West China Hospital, Sichuan University from April to September 2024 were enrolled. The participants were followed for 6 months. The participants were categorized into three groups: non-fall, no/minor injury, and moderate to severe injury group. Least absolute shrinkage and selection operator (LASSO) regression was applied to screen candidate predictors. Multinomial logistic regression (MLR), random forest (RF), support vector machine (SVM), and naive Bayes (NB) were employed for model development, with internal validation performed via 5-fold cross-validation. Model performance was assessed via the macro-average AUC, F1 score and Brier score. The optimal model was interpreted using SHAP analysis. This study was approved by the Committee of Ethics of West China Hospital of Sichuan University (Approval No. 2024–923). Results The SVM demonstrated the best performance, with a macro-average AUC of 0.856 (95% CI: 0.837–0.878), and F1-score of 0.527 (95% CI: 0.410–0.619), and a Brier score of 0.086 (95% CI: 0.078–0.091). Conclusions This study developed a risk-stratified prediction model for fall-related injuries in older adults. The SVM model is capable of accurately predicting the risk levels of fall-related injuries in older adults. Registration www.chictr.org.cn ChiCTR2400085499. Registered 11/6/2024, first recruitment 13/6/2024. Older Adults Fall-related injuries Risk prediction model Machine learning Figures Figure 1 Figure 2 Figure 3 Figure 4 Background According to a report from the World Health Organization (WHO), approximately 28% to 42% of older adults experience a fall each year globally [ 1 ] . Among those who fall, 40% to 80% sustain physical injuries [ 1 , 2 ] , 32.40% require hospitalization [ 3 ] , and an estimated 684,000 older adults die annually from fall-related injuries and related complications [ 1 , 4 , 5 ] . Fall-related injuries have become the leading cause of incidental mortality among adults aged 60 years and above in China [ 6 ] and the second leading cause of incidental mortality worldwide. Fall-related injuries also impose a substantial economic burden on families and society [ 7 ] . The Decade of Healthy Ageing 2020–2030 report published by WHO in 2020 [ 8 ] emphasized that the prevention of fall-related injuries should be treated as a global priority in addressing the health challenges faced by older adults. Studies have shown that 32% to 40% of moderate to severe fall-related injuries can be prevented through measures such as exercise training [ 9 ] , thereby reducing the associated disease burden as well as the financial burden on families and society [ 10 , 11 ] . Accurate assessment of fall-related injury risk is crucial for early and proactive prevention. The World Guidelines for Falls Prevention and Management for Older Adults [ 12 ] emphasize that precise risk assessment is a prerequisite and foundation for risk stratification management. An intervention study conducted to prevent post-fall fractures demonstrated that by assessing the risk of fall-related injuries in older adults and implementing targeted preventive measures for high-risk groups, 49 fall-related injuries and seven deaths could be prevented per 1,000 individuals [ 13 ] . Additionally, research by Lachance [ 14 ] indicated that interventions targeting risk factors for fall-related injuries resulted in a reduction in economic burden of USD 1,237.89 per person. Therefore, assessing the risk of fall-related injuries in older adults and identifying high-risk individuals in a timely manner are essential for implementing effective early prevention measures. Risk prediction models, as essential quantitative tools for risk assessment, are increasingly being applied in the field of disease risk prediction. In the context of fall-related injuries among older adults, several studies have developed such prediction models [ 15 , 16 ] . However, in terms of outcome measures, most existing studies focus solely on predicting the occurrence of a fall-related injury and lack risk stratification by severity. This limitation hinders the efficient allocation of medical resources and the advancement of personalized prevention strategies. Methodologically, many studies exclude cases with missing data, which may introduce bias. Moreover, the use of random split-sample validation can result in unstable model performance estimates and an overestimation of its generalizability in real-world clinical settings [ 17 , 18 ] . Additionally, many models demonstrate area under the receiver operating characteristic curve (AUC-ROC) values below 0.80 [ 19 , 20 ] , indicating poor discriminative ability and restricting their utility in clinical practice. Therefore, this study aims to develop an optimal risk-stratified prediction model for fall-related injuries in older adults via multiple machine learning algorithms, with the goal of assisting healthcare professionals in effectively managing high-risk individuals and enhancing their safety. Materials and methods Participants From April to September 2024, a total of 402 older adults older adults scheduled for discharge were recruited via convenience sampling from the Geriatrics Center of West China Hospital, Sichuan University, one of the largest healthcare institutions in Southwest China. The inclusion criteria were as follows: (1) aged ≥ 60 years; (2) able to communicate effectively and cooperate with various assessments; and (3) provided informed consent for the research content, voluntarily participated and could actively cooperate. The exclusion criteria were as follows: (1) had a diagnosis of cognitive impairment such as dementia or a score below 7 on the Abbreviated Mental Test (AMT) or the presence of psychiatric disorders such as schizophrenia or neurodevelopmental delays; (2) were unable to walk due to severe conditions such as major organ failure (heart, liver, kidney) or malignant tumors [ 21 ] ; and (3) dependence on assistive devices such as walkers or wheelchairs for mobility. The sample size was calculated using the Events Per Variable (EPV) formula for clinical prediction model development [ 22 ] . The incidence of fall-related injuries among adults aged 60 and above is approximately 17% [ 23 ] , with 13 candidate predictors included and a requirement of 5 events per variable, the minimum sample size was calculated as N = 382. Accounting for an estimated 10% loss to follow-up, the baseline sample size required was 425. A total of 402 participants were retained in the final analysis, which satisfied the minimum sample size for developing the prediction model. Predictors The WHO's risk factor model for falls [ 24 , 25 ] acknowledges that fall-related injuries result from interacting factors across physiological, social, behavioral, and environmental dimensions. However, practical constraints limit the consistent measurement of environmental factors across diverse community and home settings. To bridge the gap between scientific rigor and practical applicability, this study focused on three measurable dimensions: physiological, social, and behavioral to develop a clinically feasible risk prediction model for fall-related injuries in older adults. Informed by systematic literature review and expert panel interviews, relevant predictors were identified into three domains: physiological, behavioral, and social [ 26 ] . The physiological dimensions included age, sex, body mass index (BMI), fall history in the past year, number of chronic diseases, history of stroke, Parkinson's disease, history of fragility fracture, activities of daily living (ADL) score, grip strength, and balance function. The number of chronic diseases included conditions diagnosed by a secondary or higher-level hospital: hypertension, diabetes, heart disease, osteoporosis, stroke, osteoarthritis, Parkinson's disease, malignant tumors, urinary system diseases, and lumbar disc herniation. A history of fragility fracture was defined as a fracture resulting from minor trauma, such as a fall from standing height or less [ 27 ] . The ADL score was assessed via the Barthel Index (BI) [ 28 ] with data extracted from the hospital electronic medical record system. The BI measures performance across 10 domains, including feeding, walking, and stair climbing. The total scores range from 0 to 100, with higher scores indicating greater independence. Grip strength was measured using a calibrated handheld electronic dynamometer. The participants sat upright in an armless chair with their feet flat on the floor, with the tested shoulder adducted neutrally, the elbow flexed at 90°, and the forearm and wrist in a neutral position. After demonstration and 1–2 practice trials, the maximum grip strength was measured 2–3 times per hand, each lasting 3–5 seconds. The highest value (kg) from each hand was recorded [ 29 ] . Balance function was assessed using the Timed Up and Go Test (TUGT) [ 30 ] . The behavioral dimension included the number of medications, referred to as the number of fall-risk-increasing drugs based on patient self-reports, and included antihypertensives (e.g., ACE inhibitors), hypoglycemics (e.g., metformin), anticoagulants/antiplatelets (e.g., heparin, warfarin), antidepressants (e.g., SSRIs), antipsychotics (e.g., phenothiazines), analgesics (e.g., opioids), sedative-hypnotics (e.g., benzodiazepines), and antiepileptics (e.g., phenytoin, carbamazepine). The social dimension included living alone, defined as residing without a spouse or partner. Outcome definition The outcome of this study was the occurrence of fall-related injuries within a 6-month follow-up period after the baseline assessment. Fall-related injuries were defined and graded on the basis of the Chinese National Database of Nursing Quality Indicators (CNDNQI) standard [ 31 ] , as shown in Table 1 . Table 1 Definition and grading standards for fall-related injuries Grade Definition & Clinical Management No injury No physical damage; requires no observation or treatment. Minor injury Requires only observation or basic care (e.g., abrasions, contusions). Moderate injury Requires nursing/medical intervention (e.g., bandaged sprains, sutured lacerations). Severe injury Requires hospitalization or specialist care (e.g., fractures, loss of consciousness). Death Fatal outcome due to the fall. Given their comparable clinical significance, moderate and severe injuries were combined. The participants were subsequently classified into three analysis groups: 1) non-fall; 2) no/minor injury group (after fall); and 3) moderate to severe injury (including death). Participant follow-up was conducted over a six-month period. At baseline, all participants received standardized instructions from trained researchers on maintaining a monthly fall calendar to document the date, severity, and management of any fall. To ensure timely and accurate outcome reporting and to mitigate recall bias, active verification was performed via telephone follow-ups at three-month intervals. The follow-up for any participant who experienced a moderate or severe fall-related injury or death was terminated, and the event was recorded as a study endpoint; all remaining participants completed the full six-month follow-up. Data analysis This study used SPSS 27.0 (version 27.0 Chicago, IL, USA) to conduct the statistical analysis. Descriptive statistics were used to summarize participant characteristics. Categorical variables are described as frequencies and percentages. Continuous variables with a normal distribution are presented as the mean ± standard deviation, while those without a normal distribution are expressed as the median and interquartile range. A two-sided significance level of α = 0.05 was set for all tests. The outliers were identified using 3σ rule and verified against original questionnaire entries. Records with implausible values or logical inconsistencies after confirmation via follow-up were excluded. For primary outcome measures, cases with missing data were excluded following verification, and sensitivity analysis was performed to evaluate potential bias. Predictor variables with missing values exceeding 20% were removed, whereas those with lower missing values (e.g., BMI: 1.24%; living alone status: 14.18%; grip strength: 5.47%) were imputed using MissForest. Potential multicollinearity was assessed using variance inflation factor (VIF). Variables with a VIF exceeding 5 were excluded from the model. Predictor selection was performed using LASSO regression with 5-fold cross-validation to determine the optimal λ, and predictors with nonzero coefficients were retained. Model training and optimization This study developed prediction models based on the multinomial logistic regression (MLR), random forest (RF), support vector machine (SVM), and naive Bayes (NB) algorithms. All data preprocessing and modeling procedures were implemented using Python 3.8 on the PyCharm (Community Edition) platform. The hyperparameters were tuned via grid search with 5-fold cross-validation. To address class imbalance, we applied algorithm-level adjustment by setting class_weight="balanced," which adjusts weights inversely proportional to class frequencies. Internal validation and model evaluation According to the PROBAST guidelines, studies developing prediction models should perform internal validation, preferably via methods such as cross-validation or resampling [ 32 ] . This study employed 5-fold cross-validation for internal validation. Model discrimination was assessed using the area under the receiver operating characteristic curve (AUC). As a multiclass problem, one-vs-rest ROC curves were plotted, and macro-averaged AUC was calculated. The formula is as follows: \(\:{\text{A}\text{U}\text{C}}_{\text{m}\text{a}\text{c}\text{r}\text{o}}=\frac{1}{3}{\sum\:}_{\text{i}}^{3}{\text{A}\text{U}\text{C}}_{\text{i}}\) . Calibration was evaluated using the Brier score, with lower scores indicating better agreement between the predicted probabilities and observed outcomes. Additional metrics included accuracy, macro-average precision, recall, and the F1-score. The F1- score, as the harmonic mean of precision and recall, was particularly emphasized for evaluating performance on imbalanced data. Model explanation Given that machine learning models often function as "black boxes," making it difficult to intuitively understand the basis of their predictions, this study employed the SHAP method to interpret the optimal predictive model [ 33 ] . Results Patient characteristics Among the 447 participants enrolled in this study, 45 were lost to follow-up after discharge and did not complete the full follow-up process; 402 participants completed the follow-up survey. The median age of the included older adults was 69 years (interquartile range: 63–75), with an overall age range of 60 to 102 years. The cohort included 192 males (47.76%) and 210 females (50.48%). The baseline characteristics of the study participants are detailed in Table 2 . Table 2 Baseline characteristics of the study participants (n = 402) Characteristic N \(\:\stackrel{-}{\mathbf{x}}\pm\:\mathbf{s}\) /M(Q1, Q3)/% Age (years) 402 69(63, 75) Sex Female 210 52.24 Male 192 47.76 BMI (kg/m²) 402 23.50 ± 3.32 Fall history in the past year No falls 319 79.35 1 fall 65 16.17 ≥ 2 falls 18 4.48 Number of chronic diseases 402 2(2, 4) History of stroke Yes 50 12.44 No 352 87.56 Parkinson's disease Yes 21 5.22 No 381 94.78 History of fragility fracture Yes 29 7.21 No 373 92.79 Activities of daily living (ADL) score (points) 402 100(85, 100) Grip strength (kg) 402 20.00(16.00, 27.00) Balance function Mobile without assistance 281 69.90 Mostly independent 81 20.15 Unsteady mobility 30 7.46 Mobility impaired 10 2.49 Number of medications 402 1(0, 2) Living alone Yes 11 3.19 No 334 96.81 Incidence of fall-related injuries During the 6-month follow-up, 402 older adults completed the study. 89 participants (22.1%) experienced falls during the follow-up period, 43 participants (10.7%) experienced no/minor injury group after fall, 46 participants (11.4%) resulted in moderate or severe injuries after fall, accounting for 11.4% of the fall-related injury rate. Feature selection and model comparison The collinearity test revealed variance inflation factor (VIF) values ranging from 1.02–1.89, indicating no severe multicollinearity. LASSO regression with 5-fold cross-validation identified an optimal lambda of 0.004 (Fig. 1 ). The 11 variables selected with nonzero coefficients included age, sex, BMI, fall history in the past year, number of chronic diseases, history of stroke, fragility fracture history, ADL score, grip strength, balance function, and number of medications. The coefficient path is shown in Fig. 2 . The optimal parameters for each model obtained through 5-fold cross-validation combined with a grid search are listed in Table 2 . Table 2 The best-tuned hyperparameters for each model Model Hyperparameters MLR 'C'=0.001, 'max_iter'=100, 'solver'='newton-cg' RF 'class_weight'='balanced', 'max_depth'=None, 'min_samples_split'=2, 'n_estimators'=100 SVM 'svc__C'=0.01, 'svc__gamma'='scale', 'svc__kernel'='linear' NB 'var_smoothing'=(0.001) Note: Multinomial logistic regression (MLR); random forest (RF); support vector machine (SVM); naive Bayes (NB) A comparison of the discriminative performance of the four models obtained through internal validation is presented in Table 3 , and the ROC curves of the models are shown in Fig. 3. Among the four machine learning algorithms, SVM demonstrated the best discriminative ability, achieving a macro-average AUC of 0.856 (95% CI: 0.837–0.878) and a Brier score of 0.086 (95% CI: 0.078–0.091). RF and NB showed moderate performance, while MLR had the poorest calibration. RF achieved the highest accuracy (0.811, 95% CI: 0.789–0.833), followed by SVM and NB, with MLR being the lowest. In terms of precision, RF also ranked first (0.716, 95% CI: 0.517–0.915), ahead of NB and SVM, while MLR performed worst. NB exhibited the highest recall (0.580, 95% CI: 0.432–0.729), followed by SVM and MLR, with RF having the lowest. In terms of the F1-score, the NB (0.560, 95% CI: 0.407–0.713) and SVM (0.527, 95% CI: 0.410–0.619) achieved a better balance, outperforming the MLR and RF. Table 3 The performance of each model Metric Model Value 95% CI Macro-verage AUC MLR 0.821 0.771–0.872 RF 0.832 0.783–0.881 SVM 0.856 0.837–0.878 NB 0.832 0.775–0.889 Brier Score MLR 0.133 0.066–0.201 RF 0.09 0.077–0.103 SVM 0.086 0.078–0.091 NB 0.122 0.087–0.158 Accuracy MLR 0.704 0.644–0.764 RF 0.811 0.789–0.833 SVM 0.739 0.694–0.774 NB 0.774 0.655–0.892 Precision MLR 0.502 0.398–0.606 RF 0.716 0.517–0.915 SVM 0.523 0.410–0.635 NB 0.554 0.392–0.715 Recall MLR 0.57 0.380–0.761 RF 0.466 0.381–0.550 SVM 0.559 0.416–0.667 NB 0.58 0.432–0.729 F1-Score MLR 0.514 0.385–0.643 RF 0.47 0.369–0.571 SVM 0.527 0.410–0.619 NB 0.56 0.407–0.713 Note: Multinomial logistic regression (MLR); random forest (RF); support vector machine (SVM); naive Bayes (NB) a.b. c.d. Figure 3. Receiver operating characteristic (ROC) curves of the models Note (a) Multinomial logistic regression (MLR); (b) random forest (RF); (c) support vector machine (SVM); (d) naive Bayes (NB); class 1: non-fall; class 2: no/minor injury; class 3: moderate-to-severe injury; dashed line: baseline (reference line) Interpretation of the optimal prediction model A comparison of macro-average AUC and F1-score revealed that the SVM model demonstrated the highest discriminative ability, whereas both the NB and SVM showed superior performance in terms of the F1-score. Therefore, SVM was selected as the final optimal prediction model. The SHAP summary plot (Fig. 4 ) illustrates the impact of each feature on the model output, with the bar length representing the magnitude of feature importance. SHAP analysis revealed that the three most influential features were the activities of daily living (ADL) score, age, and grip strength. The importance of predictors varied across different fall-related injury outcomes: the no/minor injury group was influenced primarily by the ADL score, age, grip strength, and number of chronic diseases, whereas the moderate-to-severe injury group was associated mainly with age, grip strength, and number of chronic diseases. Class 1: non-fall; class 2: no/minor injury; class 3: moderate-to-severe injury Discussion Through the development of our prediction model, this study identified several risk factors for fall-related injuries, among which SHAP analysis of the optimal SVM model revealed that the ADL score, age, and grip strength were the most influential predictors. ADL is a key indicator for assessing the self-care ability of older adults and objectively reflects their physiological and functional health status. Declines in ADL will lead to limitations in daily activities, impair physical health, affect postural control and balance maintenance, and increase the risk of fall-related injuries [ 34 , 35 ] . With advancing age, older adults often experience a decline in physical function, including reduced vision and hearing, diminished balance and gait control, and weakened muscle strength. These factors collectively contribute to a progressive increase in fall risk among older adults [ 36 ] . Existing studies confirm that age-related decline in muscle strength is a significant factor contributing to falls and functional impairment in older adults. As a key indicator of overall muscle strength, grip strength not only reflects upper limb force but also serves as a comprehensive measure of physical function [ 37 ] [ 38 ] . Moreover, an increase in the number of chronic diseases is significantly associated with a greater risk of fall-related injuries in older adults [ 35 ] . Notably, the importance of these predictors varied by injury severity, supporting a risk-stratified approach to fall-related injury prevention. For falls resulting in no/minor injury group, the model highlighted ADL score, age, grip strength, and chronic disease count as key predictors [ 39 ] . This suggests that in older adults with relatively preserved function, minor injury risk is closely linked to baseline functional status and cumulative health burden [ 40 ] . ADL score reflects overall daily functioning and recovery capacity, while age and chronic disease count indicate physiological decline; grip strength serves as a proxy for muscle function and resilience. For moderate-to-severe injuries, age, grip strength, and chronic disease count remained dominant, whereas ADL score played a diminished role. This implies that injury severity is driven more by intrinsic physiological vulnerabilities, such as sarcopenia, multimorbidity, and frailty [ 26 ] . The reduced predictive value of ADL in this group suggests that even functionally independent older adults remain susceptible to severe injury if underlying physiological reserves are compromised. These findings underscore the clinical importance of injury stratification: for lower-risk individuals, interventions may focus on maintaining function through ADL support and chronic disease management; for higher-risk individuals, priority should be given to muscle strengthening, multimorbidity coordination, and environmental modifications to reduce injury severity. This study developed a fall-related injury risk prediction model using multiple machine learning algorithms, incorporating 11 readily obtainable clinical predictors to identify older adults at risk of non-fall, mild injury and moderate injury, thereby addressing a limitation in existing prediction tools that typically lack severity stratification. The four models exhibited varying performance levels. Among them, the SVM model demonstrated superior discriminative performance, achieving the macro-average AUC of 0.856, F1-score of 0.527, and a Brier score of 0.086 (95% CI: 0.078–0.091). This can be attributed to the SVM’s ability to handle high-dimensional and nonlinear data, as well as its suitability for small sample sizes [ 41 ] . This model, which incorporates 11 readily obtainable clinical predictors, provides a practical tool for identifying older adults at risk of no fall, minor injury, or moderate-to-severe injury, thereby addressing a significant gap in existing prediction tools that typically lack severity stratification. The prediction model developed in this study is easy to evaluate and relies solely on readily obtainable indicators that require no invasive or complex procedures, such as blood tests. This simplicity enhances its suitability for various assessment settings and offers a practical tool for predicting fall-related injury risk in older adults, thereby improving the model's overall applicability. Strengths and limitations of the research This study develops a machine learning-based prediction model that enables risk stratification for fall-related injuries. This model provides a practical tool for healthcare providers and uses only 11 readily available clinical indicators to facilitate stratified management and precision prevention. There were several limitations in our study. First, this study developed prediction models using single machine learning algorithms (e.g., logistic regression, SVM), which may limit predictive power. Future research could employ ensemble methods such as stacking or boosting to combine multiple base models and improve overall performance. Second, our data were drawn from only a single center with 402 older adults. This may have resulted in overfitting of the models. Expanding the sample size and conducting external validation across diverse regions or timeframes would enhance its applicability and generalizability. Conclusions The results of this study revealed the following predictors of future fall-related injuries in older adults: ADL score, age, grip strength, number of chronic diseases, BMI, balance function, number of medications, fall history in the past year, history of stroke, sex, and history of fragility fracture. Using multiple machine learning algorithms, we developed a risk prediction model for fall-related injuries in older adults. Compared with existing models developed domestically and internationally, the SVM-based model demonstrated reliable accuracy and superior predictive performance. Moreover, the tool requires only 11 assessment items, making it practical for healthcare providers to screen and manage fall-related injury risk in older adults. Authors’ contributions Xue Zhang, Yan Cai, Wei Zhu, Cong Wang and Shanshan Liu conceived and designed the project. Xue Zhang, Yan Cai and Wei Zhu contributed to the data curation, analysis and interpretation. Xue Zhang drafted the original manuscript. Yan Cai, Wei Zhu, Cong Wang, Shanshan Liu and Yan Jiang performed the quality assessment and revised the manuscript critically. All authors have read and approved the submitted version. Declarations Ethics approval and consent to participate This study was conducted in accordance with the ethical principles of the Declaration of Helsinki. The study protocol was reviewed and approved by the Committee of Ethics of West China Hospital of Sichuan University (Approval No. 2024–923). Written informed consent was obtained from all individual participants included in the study prior to their enrollment. Competing interests The authors declare that they have no competing interests. Funding This study was supported by National Key Research and Development Program of China (2023YFC3605900). The funding sources were not involved in the study design, data collection, data analysis, data interpretation, report writing, or decision to submit the article for publication. Author Contribution Xue Zhang, Yan Cai, Wei Zhu, Cong Wang and Shanshan Liu conceived and designed the project. Xue Zhang, Yan Cai and Wei Zhu contributed to the data curation, analysis and interpretation. Xue Zhang drafted the original manuscript. Yan Cai, Wei Zhu, Cong Wang, Shanshan Liu and Yan Jiang performed the quality assessment and revised the manuscript critically. All authors have read and approved the submitted version. Data Availability The data that support the findings of this study are available on reasonable request from the corresponding author. The data are not publicly available due to privacy and ethical restrictions. References United UND, Nations, Department of Economic and Social Affairs, Population Division. World Population Prospects: The2017 Revision, Key Findings and Advance, Tables. [EB/OL]. https://esa.un.org/unpd/wpp/Publications/ . [J/OL]. WOO N, KIM SH. Sarcopenia influences fall-related injuries in community-dwelling older adults [J]. Geriatr Nurs. 2014;35(4):279–82. ER YULIANG D L, WANG YUAN, JI CUIRONG, et al. Analysis on data from Chinese National Injury Surveillance System, 2008–2013 on the characteristics of Falls [J]. Chin J Epidemiol. 2015;36(1):12–6. VIEIRA E R, PALMER R C, CHAVES PH. Prevention of falls in older people living in the community [J]. BMJ. 2016;353:i1419. KIM S, KIM S. WOO STemporal trends and patterns in mortality from falls across 59 high-income and upper-middle-income countries, 1990–2021, with projections up to 2040: a global time-series analysis and modelling study [J]. Volume 6. The lancet Healthy longevity; 2025. p. 100672. 1. LEILEI D, PENGPENG Y. The burden of injury in China, 1990–2017: findings from the Global Burden of Disease Study 2017 [J]. Lancet Public Health. 2019;4(9):e449–61. HAAGSMA J A,. YE P, ER Y, WANG H, et al. Burden of falls among people aged 60 years and older in mainland China, 1990–2019: findings from the Global Burden of Disease Study 2019 [J]. Lancet Public Health. 2021;6(12):e907–18. RUDNICKA E, NAPIERAŁA P. The World Health Organization (WHO) approach to healthy ageing [J]. Maturitas. 2020;139:6–11. CAMPBELL W W DIPIETROL, BUCHNER D M, et al. Physical Activity, Injurious Falls, and Physical Function in Aging: An Umbrella Review [J]. Med Sci Sports Exerc. 2019;51(6):1303–13. ONG M F, SOH K L, SAIMON R, et al. Fall prevention education to reduce fall risk among community-dwelling older persons: A systematic review [J]. J Nurs Adm Manag. 2021;29(8):2674–88. DYKES PC, BURNS Z. Evaluation of a Patient-Centered Fall-Prevention Tool Kit to Reduce Falls and Injuries: A Nonrandomized Controlled Trial [J]. JAMA Netw Open. 2020;3(11):e2025889. MONTERO-ODASSO M, VAN DER VELDE N, MARTIN, F C et al. World guidelines for falls prevention and management for older adults: a global initiative [J]. Age Ageing, 2022, 51(9). PAPAIOANNOU A, SANTESSO N, MORIN. et al. Recommendations for preventing fracture in long-term care [J]. Volume 187. CMAJ: Canadian Medical Association journal = journal de l'Association medicale canadienne; 2015. pp. 1135–44. 15S N,. LACHANCE CC, JURKOWSKI MP, DYMARZ AC, et al. Compliant flooring to prevent fall-related injuries in older adults: A scoping review of biomechanical efficacy, clinical effectiveness, cost-effectiveness, and workplace safety [J]. PLoS ONE. 2017;12(2):e0171652. SAI A, TANAKA K, OHASHI Y, et al. Quantitative sonographic assessment of quadriceps muscle thickness for fall injury prediction in patients undergoing maintenance hemodialysis: an observational cohort study [J]. BMC Nephrol. 2021;22(1):191. NILSSON M, ERIKSSON J, LARSSON B, et al. Fall Risk Assessment Predicts Fall-Related Injury, Hip Fracture, and Head Injury in Older Adults [J]. J Am Geriatr Soc. 2016;64(11):2242–50. SPEISER JL, CALLAHAN KE, HOUSTON DK, et al. Machine Learning in Aging: An Example of Developing Prediction Models for Serious Fall Injury in Older Adults [J]. J Gerontol Biol Sci Med Sci. 2021;76(4):647–54. EK S, CALDERóN-LARRAñAGA RIZZUTOD. Predicting First-Time Injurious Falls in Older Men and Women Living in the Community: Development of the First Injurious Fall Screening Tool [J]. J Am Med Dir Assoc. 2019;20(9):1163–e83. LI W, RAO Z, FU Y, et al. Value of the short physical performance battery (SPPB) in predicting fall and fall-induced injury among old Chinese adults [J]. BMC Geriatr. 2023;23(1):574. YAITA S, TAGO M. A Simple and Accurate Model for Predicting Fall Injuries in Hospitalized Patients: Insights from a Retrospective Observational Study in Japan [J]. Med Sci monitor: Int Med J experimental Clin Res. 2023;29:e941252. KATSUKI N E,. SM B, VL N. Australian Diabetes Foot Network: practical guideline on the provision of footwear for people with diabetes [J]. J Foot Ankle Res. 2013;6(1):1757–146. L W, W B, KG M, et al. A simulation study of sample size demonstrated the importance of the number of events per variable to develop prediction models in clustered data [J]. J Clin Epidemiol, 2015, 68(12): 1406–14. JIANG Y, XIA Q. Study on the epidemical characteristies and disease burden of fall-related imiury among community-dwelling elderly adults in Changning District, Shanghai [J]. Chin J Disease Control Prev. 2013;17(02):134–7. QIAN XX, CHAU PH, KWAN CW, et al. Investigating Risk Factors for Falls among Community-Dwelling Older Adults According to WHO's Risk Factor Model for Falls [J]. J Nutr Health Aging. 2021;25(4):425–32. CHOCK J. Identification of Fall Risk Factors and Development of a Falls Prediction Model for Community-Dwelling Older Adults Receiving Public Health Nursing Services in Hawai´i [D]. United States -- Hawaii; University of Hawai'i at Manoa, 2023. SRIVASTAVA S, MUHAMMAD T. Prevalence and risk factors of fall-related injury among older adults in India: evidence from a cross-sectional observational study [J]. BMC Public Health. 2022;22(1):550. ARLETTAZ Y. Augmented osteosynthesis in fragility fracture [J]. Orthopaedics & traumatology, surgery & research: OTSR. 2023, 109(1s): 103461. MAHONEY F I, BARTHEL D W. FUNCTIONAL. EVALUATION: THE BARTHEL INDEX [J]. Maryland State Med J. 1965;14:61–5. LIU H, HOU Y, LI H, et al. Influencing factors of weak grip strength and fall: a study based on the China Health and Retirement Longitudinal Study (CHARLS) [J]. BMC Public Health. 2022;22(1):2337. D P, S R. - The timed Up & Go: a test of basic functional mobility for frail elderly [J]. J Am Geriatr Soc, 1991, 39(2): 142–8. ZHIJUN WU W S, WEIYAN JAN. Incidence of Falls among Inpatient in China:A Survey of 490 Tertiary Hospitals [J]. Chin Health Qual Manage. 2019;26(3):14–7. MOONS K G M, WOLFF R F, RILEY, R D, et al. PROBAST: A Tool to Assess Risk of Bias and Applicability of Prediction Model Studies: Explanation and Elaboration [J]. Ann Intern Med. 2019;170(1):W1–33. LUNDBERG S, LEE S I.. A Unified Approach to Interpreting Model Predictions [M]. 2017. ZHAO M, LI S, XU Y, et al. Developing a Scoring Model to Predict the Risk of Injurious Falls in Elderly Patients: A Retrospective Case-Control Study in Multicenter Acute Hospitals [J]. Clin Interv Aging. 2020;15:1767–78. CHEN X, HE L, SHI K, et al. Interpretable Machine Learning for Fall Prediction Among Older Adults in China [J]. Am J Prev Med. 2023;65(4):579–86. SPEISER JL, CALLAHAN K E, HOUSTON D K, et al. Machine Learning in Aging: An Example of Developing Prediction Models for Serious Fall Injury in Older Adults [J]. Journals Gerontology: Ser A. 2020;76(4):647–54. AJ C-J GB. Sarcopenia: revised European consensus on definition and diagnosis [J]. Age Ageing. 2019;48(1):16–31. PORTO JM, NAKAISHI A P M, CANGUSSU-OLIVEIRA L M, et al. Relationship between grip strength and global muscle strength in community-dwelling older people [J]. Arch Gerontol Geriatr. 2019;82:273–8. LIANG X Z, CHAI J L. A fall risk prediction model based on the CHARLS database for older individuals in China [J]. BMC Geriatr. 2025;25(1):170. LI G Z,. SHAO L, SHI Y, XIE X Y, et al. Incidence and Risk Factors of Falls Among Older People in Nursing Homes: Systematic Review and Meta-Analysis [J]. J Am Med Dir Assoc. 2023;24(11):1708–17. CARRACEDO-REBOREDO P, LIñARES-BLANCO J, RODRíGUEZ-FERNáNDEZ N, et al. A review on machine learning approaches and trends in drug discovery [J]. Comput Struct Biotechnol J. 2021;19:4538–58. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted 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-8683642","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":597677146,"identity":"1b476dd8-4f13-47a9-b679-d3001be1aa33","order_by":0,"name":"Xue Zhang","email":"","orcid":"","institution":"Sichuan University","correspondingAuthor":false,"prefix":"","firstName":"Xue","middleName":"","lastName":"Zhang","suffix":""},{"id":597677148,"identity":"21f5febc-4626-4ed0-ada2-83a12dc43920","order_by":1,"name":"Yan Cai","email":"","orcid":"","institution":"Sichuan University","correspondingAuthor":false,"prefix":"","firstName":"Yan","middleName":"","lastName":"Cai","suffix":""},{"id":597677149,"identity":"1d89393d-16d0-4b82-83df-6fa6a7c627ca","order_by":2,"name":"Wei Zhu","email":"","orcid":"","institution":"Sichuan University","correspondingAuthor":false,"prefix":"","firstName":"Wei","middleName":"","lastName":"Zhu","suffix":""},{"id":597677150,"identity":"4e9c9658-52b7-4215-9c58-84819d1e9d8e","order_by":3,"name":"Cong Wang","email":"","orcid":"","institution":"Sichuan University","correspondingAuthor":false,"prefix":"","firstName":"Cong","middleName":"","lastName":"Wang","suffix":""},{"id":597677157,"identity":"c07a4efd-fe52-4d6e-8d63-a66df75121fd","order_by":4,"name":"Shanshan Liu","email":"","orcid":"","institution":"Sichuan University","correspondingAuthor":false,"prefix":"","firstName":"Shanshan","middleName":"","lastName":"Liu","suffix":""},{"id":597677159,"identity":"21b2f3db-cba7-4035-97e8-c75e129680d8","order_by":5,"name":"Yan Jiang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA2klEQVRIie3RPQrCMBTA8VeE1OGhmzwR2iukdFDBw7SLUwUnLyDEpe6Kl/AIkYAutbPgIvQCGR0ETXFwsxkF85/fj5cPAJfrB+OAniSYIPOXUuq7HQFDpkEHj+lhm9sSABUHlMWqzSzIkGZSjkUrFVhoZXzY7cnvZJyXiewLlgp/vVfzEUTbXdJwsFPODUGz5bxXG4SEX5uIwppQKii7KWQ25L2Fx4wysCRFwSWVScDwyM0jk8VdLllU0eKJ4WpZaX2fhN1BA6lr9T/fQc3jdZ5+2A26XC7Xn/YCMQJJyERpTbEAAAAASUVORK5CYII=","orcid":"","institution":"Sichuan University","correspondingAuthor":true,"prefix":"","firstName":"Yan","middleName":"","lastName":"Jiang","suffix":""}],"badges":[],"createdAt":"2026-01-24 04:23:15","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8683642/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8683642/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":103728549,"identity":"2a7b8328-41fa-410f-b435-96aa9c0d3905","added_by":"auto","created_at":"2026-03-02 08:43:20","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":98463,"visible":true,"origin":"","legend":"\u003cp\u003eLASSO cross-validation curve\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8683642/v1/903e9a758e8df11b3589ab2f.png"},{"id":103728552,"identity":"a2ef647b-9f5c-43d4-a7b0-b8ae93362a7f","added_by":"auto","created_at":"2026-03-02 08:43:22","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":73019,"visible":true,"origin":"","legend":"\u003cp\u003eLASSO Coefficient Path Plot\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8683642/v1/c0f8ada15fc58c0a302cfdfd.png"},{"id":103728548,"identity":"4cfa1267-9a92-4514-bd2b-4809e4c4012c","added_by":"auto","created_at":"2026-03-02 08:43:20","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":105858,"visible":true,"origin":"","legend":"\u003cp\u003eReceiver operating characteristic (ROC) curves of the models\u003c/p\u003e\n\u003cp\u003eNote: \u003cstrong\u003e(a)\u003c/strong\u003e Multinomial logistic regression (MLR); \u003cstrong\u003e(b)\u003c/strong\u003e random forest (RF); \u003cstrong\u003e(c)\u003c/strong\u003e support vector machine (SVM); \u003cstrong\u003e(d)\u003c/strong\u003e naive Bayes (NB); class 1: non-fall; class 2: no/minor injury; class 3: moderate-to-severe injury; dashed line: baseline (reference line)\u003c/p\u003e","description":"","filename":"1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8683642/v1/2c765fe87386dd3a923cf61a.jpg"},{"id":103728421,"identity":"e885a762-c572-467d-acb2-67c4da211b09","added_by":"auto","created_at":"2026-03-02 08:43:05","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":31663,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSHAP values of the features.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eClass 1: non-fall; class 2: no/minor injury; class 3: moderate-to-severe injury\u003c/p\u003e","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-8683642/v1/ec7667204cbc90ee93cb43fe.png"},{"id":105562911,"identity":"405a6fea-82ea-4758-9125-962a0738af7f","added_by":"auto","created_at":"2026-03-27 12:45:13","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1099440,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8683642/v1/3b4a60a7-b8a3-48d9-8b21-a64762a306c8.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Development of a Fall-Related Injury Risk-Stratified Prediction Model for Older Adults","fulltext":[{"header":"Background","content":"\u003cp\u003eAccording to a report from the World Health Organization (WHO), approximately 28% to 42% of older adults experience a fall each year globally\u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]\u003c/sup\u003e. Among those who fall, 40% to 80% sustain physical injuries\u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]\u003c/sup\u003e, 32.40% require hospitalization\u003csup\u003e[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/sup\u003e, and an estimated 684,000 older adults die annually from fall-related injuries and related complications\u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]\u003c/sup\u003e. Fall-related injuries have become the leading cause of incidental mortality among adults aged 60 years and above in China\u003csup\u003e[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/sup\u003e and the second leading cause of incidental mortality worldwide. Fall-related injuries also impose a substantial economic burden on families and society\u003csup\u003e[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/sup\u003e. The \u003cem\u003eDecade of Healthy Ageing 2020\u0026ndash;2030\u003c/em\u003e report published by WHO in 2020\u003csup\u003e[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]\u003c/sup\u003e emphasized that the prevention of fall-related injuries should be treated as a global priority in addressing the health challenges faced by older adults.\u003c/p\u003e \u003cp\u003eStudies have shown that 32% to 40% of moderate to severe fall-related injuries can be prevented through measures such as exercise training\u003csup\u003e[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/sup\u003e, thereby reducing the associated disease burden as well as the financial burden on families and society\u003csup\u003e[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]\u003c/sup\u003e. Accurate assessment of fall-related injury risk is crucial for early and proactive prevention. The World Guidelines for Falls Prevention and Management for Older Adults\u003csup\u003e[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]\u003c/sup\u003e emphasize that precise risk assessment is a prerequisite and foundation for risk stratification management. An intervention study conducted to prevent post-fall fractures demonstrated that by assessing the risk of fall-related injuries in older adults and implementing targeted preventive measures for high-risk groups, 49 fall-related injuries and seven deaths could be prevented per 1,000 individuals\u003csup\u003e[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/sup\u003e. Additionally, research by Lachance\u003csup\u003e[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]\u003c/sup\u003e indicated that interventions targeting risk factors for fall-related injuries resulted in a reduction in economic burden of USD 1,237.89 per person. Therefore, assessing the risk of fall-related injuries in older adults and identifying high-risk individuals in a timely manner are essential for implementing effective early prevention measures.\u003c/p\u003e \u003cp\u003eRisk prediction models, as essential quantitative tools for risk assessment, are increasingly being applied in the field of disease risk prediction. In the context of fall-related injuries among older adults, several studies have developed such prediction models\u003csup\u003e[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u003c/sup\u003e. However, in terms of outcome measures, most existing studies focus solely on predicting the occurrence of a fall-related injury and lack risk stratification by severity. This limitation hinders the efficient allocation of medical resources and the advancement of personalized prevention strategies. Methodologically, many studies exclude cases with missing data, which may introduce bias. Moreover, the use of random split-sample validation can result in unstable model performance estimates and an overestimation of its generalizability in real-world clinical settings\u003csup\u003e[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]\u003c/sup\u003e. Additionally, many models demonstrate area under the receiver operating characteristic curve (AUC-ROC) values below 0.80\u003csup\u003e[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]\u003c/sup\u003e, indicating poor discriminative ability and restricting their utility in clinical practice.\u003c/p\u003e \u003cp\u003eTherefore, this study aims to develop an optimal risk-stratified prediction model for fall-related injuries in older adults via multiple machine learning algorithms, with the goal of assisting healthcare professionals in effectively managing high-risk individuals and enhancing their safety.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eParticipants\u003c/h2\u003e \u003cp\u003eFrom April to September 2024, a total of 402 older adults older adults scheduled for discharge were recruited via convenience sampling from the Geriatrics Center of West China Hospital, Sichuan University, one of the largest healthcare institutions in Southwest China. The inclusion criteria were as follows: (1) aged\u0026thinsp;\u0026ge;\u0026thinsp;60 years; (2) able to communicate effectively and cooperate with various assessments; and (3) provided informed consent for the research content, voluntarily participated and could actively cooperate.\u003c/p\u003e \u003cp\u003eThe exclusion criteria were as follows: (1) had a diagnosis of cognitive impairment such as dementia or a score below 7 on the Abbreviated Mental Test (AMT) or the presence of psychiatric disorders such as schizophrenia or neurodevelopmental delays; (2) were unable to walk due to severe conditions such as major organ failure (heart, liver, kidney) or malignant tumors\u003csup\u003e[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/sup\u003e; and (3) dependence on assistive devices such as walkers or wheelchairs for mobility.\u003c/p\u003e \u003cp\u003eThe sample size was calculated using the Events Per Variable (EPV) formula for clinical prediction model development\u003csup\u003e[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/sup\u003e. The incidence of fall-related injuries among adults aged 60 and above is approximately 17%\u003csup\u003e[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]\u003c/sup\u003e, with 13 candidate predictors included and a requirement of 5 events per variable, the minimum sample size was calculated as N\u0026thinsp;=\u0026thinsp;382. Accounting for an estimated 10% loss to follow-up, the baseline sample size required was 425. A total of 402 participants were retained in the final analysis, which satisfied the minimum sample size for developing the prediction model.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003ePredictors\u003c/h3\u003e\n\u003cp\u003eThe WHO's risk factor model for falls\u003csup\u003e[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]\u003c/sup\u003e acknowledges that fall-related injuries result from interacting factors across physiological, social, behavioral, and environmental dimensions. However, practical constraints limit the consistent measurement of environmental factors across diverse community and home settings. To bridge the gap between scientific rigor and practical applicability, this study focused on three measurable dimensions: physiological, social, and behavioral to develop a clinically feasible risk prediction model for fall-related injuries in older adults.\u003c/p\u003e \u003cp\u003eInformed by systematic literature review and expert panel interviews, relevant predictors were identified into three domains: physiological, behavioral, and social\u003csup\u003e[\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]\u003c/sup\u003e. The physiological dimensions included age, sex, body mass index (BMI), fall history in the past year, number of chronic diseases, history of stroke, Parkinson's disease, history of fragility fracture, activities of daily living (ADL) score, grip strength, and balance function. The number of chronic diseases included conditions diagnosed by a secondary or higher-level hospital: hypertension, diabetes, heart disease, osteoporosis, stroke, osteoarthritis, Parkinson's disease, malignant tumors, urinary system diseases, and lumbar disc herniation. A history of fragility fracture was defined as a fracture resulting from minor trauma, such as a fall from standing height or less\u003csup\u003e[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]\u003c/sup\u003e. The ADL score was assessed via the Barthel Index (BI)\u003csup\u003e[\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]\u003c/sup\u003e with data extracted from the hospital electronic medical record system. The BI measures performance across 10 domains, including feeding, walking, and stair climbing. The total scores range from 0 to 100, with higher scores indicating greater independence. Grip strength was measured using a calibrated handheld electronic dynamometer. The participants sat upright in an armless chair with their feet flat on the floor, with the tested shoulder adducted neutrally, the elbow flexed at 90\u0026deg;, and the forearm and wrist in a neutral position. After demonstration and 1\u0026ndash;2 practice trials, the maximum grip strength was measured 2\u0026ndash;3 times per hand, each lasting 3\u0026ndash;5 seconds. The highest value (kg) from each hand was recorded\u003csup\u003e[\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]\u003c/sup\u003e. Balance function was assessed using the Timed Up and Go Test (TUGT) \u003csup\u003e[\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe behavioral dimension included the number of medications, referred to as the number of fall-risk-increasing drugs based on patient self-reports, and included antihypertensives (e.g., ACE inhibitors), hypoglycemics (e.g., metformin), anticoagulants/antiplatelets (e.g., heparin, warfarin), antidepressants (e.g., SSRIs), antipsychotics (e.g., phenothiazines), analgesics (e.g., opioids), sedative-hypnotics (e.g., benzodiazepines), and antiepileptics (e.g., phenytoin, carbamazepine).\u003c/p\u003e \u003cp\u003eThe social dimension included living alone, defined as residing without a spouse or partner.\u003c/p\u003e\n\u003ch3\u003eOutcome definition\u003c/h3\u003e\n\u003cp\u003eThe outcome of this study was the occurrence of fall-related injuries within a 6-month follow-up period after the baseline assessment. Fall-related injuries were defined and graded on the basis of the Chinese National Database of Nursing Quality Indicators (CNDNQI) standard\u003csup\u003e[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]\u003c/sup\u003e, as shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDefinition and grading standards for fall-related injuries\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGrade\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDefinition \u0026amp; Clinical Management\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo injury\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo physical damage; requires no observation or treatment.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMinor injury\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRequires only observation or basic care (e.g., abrasions, contusions).\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModerate injury\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRequires nursing/medical intervention (e.g., bandaged sprains, sutured lacerations).\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSevere injury\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRequires hospitalization or specialist care (e.g., fractures, loss of consciousness).\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDeath\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFatal outcome due to the fall.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eGiven their comparable clinical significance, moderate and severe injuries were combined. The participants were subsequently classified into three analysis groups: 1) non-fall; 2) no/minor injury group (after fall); and 3) moderate to severe injury (including death).\u003c/p\u003e \u003cp\u003eParticipant follow-up was conducted over a six-month period. At baseline, all participants received standardized instructions from trained researchers on maintaining a monthly fall calendar to document the date, severity, and management of any fall. To ensure timely and accurate outcome reporting and to mitigate recall bias, active verification was performed via telephone follow-ups at three-month intervals. The follow-up for any participant who experienced a moderate or severe fall-related injury or death was terminated, and the event was recorded as a study endpoint; all remaining participants completed the full six-month follow-up.\u003c/p\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eData analysis\u003c/h2\u003e \u003cp\u003eThis study used SPSS 27.0 (version 27.0 Chicago, IL, USA) to conduct the statistical analysis. Descriptive statistics were used to summarize participant characteristics. Categorical variables are described as frequencies and percentages. Continuous variables with a normal distribution are presented as the mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation, while those without a normal distribution are expressed as the median and interquartile range. A two-sided significance level of α\u0026thinsp;=\u0026thinsp;0.05 was set for all tests.\u003c/p\u003e \u003cp\u003eThe outliers were identified using 3σ rule and verified against original questionnaire entries. Records with implausible values or logical inconsistencies after confirmation via follow-up were excluded. For primary outcome measures, cases with missing data were excluded following verification, and sensitivity analysis was performed to evaluate potential bias. Predictor variables with missing values exceeding 20% were removed, whereas those with lower missing values (e.g., BMI: 1.24%; living alone status: 14.18%; grip strength: 5.47%) were imputed using MissForest.\u003c/p\u003e \u003cp\u003ePotential multicollinearity was assessed using variance inflation factor (VIF). Variables with a VIF exceeding 5 were excluded from the model. Predictor selection was performed using LASSO regression with 5-fold cross-validation to determine the optimal λ, and predictors with nonzero coefficients were retained.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eModel training and optimization\u003c/h3\u003e\n\u003cp\u003eThis study developed prediction models based on the multinomial logistic regression (MLR), random forest (RF), support vector machine (SVM), and naive Bayes (NB) algorithms. All data preprocessing and modeling procedures were implemented using Python 3.8 on the PyCharm (Community Edition) platform. The hyperparameters were tuned via grid search with 5-fold cross-validation. To address class imbalance, we applied algorithm-level adjustment by setting class_weight=\"balanced,\" which adjusts weights inversely proportional to class frequencies.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eInternal validation and model evaluation\u003c/h2\u003e \u003cp\u003eAccording to the PROBAST guidelines, studies developing prediction models should perform internal validation, preferably via methods such as cross-validation or resampling \u003csup\u003e[\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]\u003c/sup\u003e. This study employed 5-fold cross-validation for internal validation.\u003c/p\u003e \u003cp\u003eModel discrimination was assessed using the area under the receiver operating characteristic curve (AUC). As a multiclass problem, one-vs-rest ROC curves were plotted, and macro-averaged AUC was calculated. The formula is as follows: \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\text{A}\\text{U}\\text{C}}_{\\text{m}\\text{a}\\text{c}\\text{r}\\text{o}}=\\frac{1}{3}{\\sum\\:}_{\\text{i}}^{3}{\\text{A}\\text{U}\\text{C}}_{\\text{i}}\\)\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e Calibration was evaluated using the Brier score, with lower scores indicating better agreement between the predicted probabilities and observed outcomes. Additional metrics included accuracy, macro-average precision, recall, and the F1-score. The F1- score, as the harmonic mean of precision and recall, was particularly emphasized for evaluating performance on imbalanced data.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eModel explanation\u003c/h3\u003e\n\u003cp\u003eGiven that machine learning models often function as \"black boxes,\" making it difficult to intuitively understand the basis of their predictions, this study employed the SHAP method to interpret the optimal predictive model\u003csup\u003e[\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003ePatient characteristics\u003c/p\u003e \u003cp\u003eAmong the 447 participants enrolled in this study, 45 were lost to follow-up after discharge and did not complete the full follow-up process; 402 participants completed the follow-up survey. The median age of the included older adults was 69 years (interquartile range: 63\u0026ndash;75), with an overall age range of 60 to 102 years. The cohort included 192 males (47.76%) and 210 females (50.48%). The baseline characteristics of the study participants are detailed in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eBaseline characteristics of the study participants (n\u0026thinsp;=\u0026thinsp;402)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCharacteristic\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\stackrel{-}{\\mathbf{x}}\\pm\\:\\mathbf{s}\\)\u003c/span\u003e\u003c/span\u003e/M(Q1, Q3)/%\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge (years)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e402\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e69(63, 75)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSex\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e210\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e52.24\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e192\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e47.76\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBMI (kg/m\u0026sup2;)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e402\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e23.50\u0026thinsp;\u0026plusmn;\u0026thinsp;3.32\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFall history in the past year\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo falls\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e319\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e79.35\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1 fall\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e65\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e16.17\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026ge;\u0026thinsp;2 falls\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4.48\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNumber of chronic diseases\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e402\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2(2, 4)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHistory of stroke\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e50\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e12.44\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e352\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e87.56\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eParkinson's disease\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5.22\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e381\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e94.78\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHistory of fragility fracture\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e29\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e7.21\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e373\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e92.79\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eActivities of daily living (ADL) score (points)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e402\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e100(85, 100)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGrip strength (kg)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e402\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e20.00(16.00, 27.00)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBalance function\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMobile without assistance\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e281\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e69.90\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMostly independent\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e81\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e20.15\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUnsteady mobility\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e30\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e7.46\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMobility impaired\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.49\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNumber of medications\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e402\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1(0, 2)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLiving alone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3.19\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e334\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e96.81\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"3\"\u003eIncidence of fall-related injuries\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eDuring the 6-month follow-up, 402 older adults completed the study. 89 participants (22.1%) experienced falls during the follow-up period, 43 participants (10.7%) experienced no/minor injury group after fall, 46 participants (11.4%) resulted in moderate or severe injuries after fall, accounting for 11.4% of the fall-related injury rate.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eFeature selection and model comparison\u003c/h2\u003e \u003cp\u003eThe collinearity test revealed variance inflation factor (VIF) values ranging from 1.02\u0026ndash;1.89, indicating no severe multicollinearity. LASSO regression with 5-fold cross-validation identified an optimal lambda of 0.004 (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The 11 variables selected with nonzero coefficients included age, sex, BMI, fall history in the past year, number of chronic diseases, history of stroke, fragility fracture history, ADL score, grip strength, balance function, and number of medications. The coefficient path is shown 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 \u003cp\u003eThe optimal parameters for each model obtained through 5-fold cross-validation combined with a grid search are listed in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eThe best-tuned hyperparameters for each model\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHyperparameters\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e'C'=0.001, 'max_iter'=100, 'solver'='newton-cg'\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e'class_weight'='balanced', 'max_depth'=None, 'min_samples_split'=2, 'n_estimators'=100\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSVM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e'svc__C'=0.01, 'svc__gamma'='scale', 'svc__kernel'='linear'\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e'var_smoothing'=(0.001)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"2\"\u003eNote: Multinomial logistic regression (MLR); random forest (RF); support vector machine (SVM); naive Bayes (NB)\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eA comparison of the discriminative performance of the four models obtained through internal validation is presented in Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e3\u003c/span\u003e, and the ROC curves of the models are shown in Fig.\u0026nbsp;3. Among the four machine learning algorithms, SVM demonstrated the best discriminative ability, achieving a macro-average AUC of 0.856 (95% CI: 0.837\u0026ndash;0.878) and a Brier score of 0.086 (95% CI: 0.078\u0026ndash;0.091). RF and NB showed moderate performance, while MLR had the poorest calibration. RF achieved the highest accuracy (0.811, 95% CI: 0.789\u0026ndash;0.833), followed by SVM and NB, with MLR being the lowest. In terms of precision, RF also ranked first (0.716, 95% CI: 0.517\u0026ndash;0.915), ahead of NB and SVM, while MLR performed worst. NB exhibited the highest recall (0.580, 95% CI: 0.432\u0026ndash;0.729), followed by SVM and MLR, with RF having the lowest. In terms of the F1-score, the NB (0.560, 95% CI: 0.407\u0026ndash;0.713) and SVM (0.527, 95% CI: 0.410\u0026ndash;0.619) achieved a better balance, outperforming the MLR and RF.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eThe performance of each model\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMetric\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eValue\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e95% CI\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMacro-verage AUC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.821\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.771\u0026ndash;0.872\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.832\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.783\u0026ndash;0.881\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSVM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.856\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.837\u0026ndash;0.878\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.832\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.775\u0026ndash;0.889\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBrier Score\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.133\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.066\u0026ndash;0.201\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.09\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.077\u0026ndash;0.103\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSVM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.086\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.078\u0026ndash;0.091\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.122\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.087\u0026ndash;0.158\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAccuracy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.704\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.644\u0026ndash;0.764\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.811\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.789\u0026ndash;0.833\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSVM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.739\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.694\u0026ndash;0.774\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.774\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.655\u0026ndash;0.892\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePrecision\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.502\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.398\u0026ndash;0.606\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.716\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.517\u0026ndash;0.915\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSVM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.523\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.410\u0026ndash;0.635\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.554\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.392\u0026ndash;0.715\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRecall\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.57\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.380\u0026ndash;0.761\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.466\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.381\u0026ndash;0.550\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSVM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.559\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.416\u0026ndash;0.667\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.58\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.432\u0026ndash;0.729\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eF1-Score\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.514\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.385\u0026ndash;0.643\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.47\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.369\u0026ndash;0.571\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSVM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.527\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.410\u0026ndash;0.619\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.407\u0026ndash;0.713\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"4\"\u003eNote: Multinomial logistic regression (MLR); random forest (RF); support vector machine (SVM); naive Bayes (NB)\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003ea.b.\u003c/p\u003e \u003cp\u003ec.d.\u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;3. Receiver operating characteristic (ROC) curves of the models\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eNote\u003c/strong\u003e \u003cp\u003e \u003cb\u003e(a)\u003c/b\u003e Multinomial logistic regression (MLR); \u003cb\u003e(b)\u003c/b\u003e random forest (RF); \u003cb\u003e(c)\u003c/b\u003e support vector machine (SVM); \u003cb\u003e(d)\u003c/b\u003e naive Bayes (NB); class 1: non-fall; class 2: no/minor injury; class 3: moderate-to-severe injury; dashed line: baseline (reference line)\u003c/p\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eInterpretation of the optimal prediction model\u003c/h2\u003e \u003cp\u003eA comparison of macro-average AUC and F1-score revealed that the SVM model demonstrated the highest discriminative ability, whereas both the NB and SVM showed superior performance in terms of the F1-score. Therefore, SVM was selected as the final optimal prediction model. The SHAP summary plot (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e4\u003c/span\u003e) illustrates the impact of each feature on the model output, with the bar length representing the magnitude of feature importance. SHAP analysis revealed that the three most influential features were the activities of daily living (ADL) score, age, and grip strength. The importance of predictors varied across different fall-related injury outcomes: the no/minor injury group was influenced primarily by the ADL score, age, grip strength, and number of chronic diseases, whereas the moderate-to-severe injury group was associated mainly with age, grip strength, and number of chronic diseases.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eClass 1: non-fall; class 2: no/minor injury; class 3: moderate-to-severe injury\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThrough the development of our prediction model, this study identified several risk factors for fall-related injuries, among which SHAP analysis of the optimal SVM model revealed that the ADL score, age, and grip strength were the most influential predictors. ADL is a key indicator for assessing the self-care ability of older adults and objectively reflects their physiological and functional health status. Declines in ADL will lead to limitations in daily activities, impair physical health, affect postural control and balance maintenance, and increase the risk of fall-related injuries \u003csup\u003e[\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]\u003c/sup\u003e. With advancing age, older adults often experience a decline in physical function, including reduced vision and hearing, diminished balance and gait control, and weakened muscle strength. These factors collectively contribute to a progressive increase in fall risk among older adults\u003csup\u003e[\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]\u003c/sup\u003e. Existing studies confirm that age-related decline in muscle strength is a significant factor contributing to falls and functional impairment in older adults. As a key indicator of overall muscle strength, grip strength not only reflects upper limb force but also serves as a comprehensive measure of physical function\u003csup\u003e[\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e] [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]\u003c/sup\u003e. Moreover, an increase in the number of chronic diseases is significantly associated with a greater risk of fall-related injuries in older adults\u003csup\u003e[\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eNotably, the importance of these predictors varied by injury severity, supporting a risk-stratified approach to fall-related injury prevention. For falls resulting in no/minor injury group, the model highlighted ADL score, age, grip strength, and chronic disease count as key predictors\u003csup\u003e[\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]\u003c/sup\u003e. This suggests that in older adults with relatively preserved function, minor injury risk is closely linked to baseline functional status and cumulative health burden\u003csup\u003e[\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e]\u003c/sup\u003e. ADL score reflects overall daily functioning and recovery capacity, while age and chronic disease count indicate physiological decline; grip strength serves as a proxy for muscle function and resilience. For moderate-to-severe injuries, age, grip strength, and chronic disease count remained dominant, whereas ADL score played a diminished role. This implies that injury severity is driven more by intrinsic physiological vulnerabilities, such as sarcopenia, multimorbidity, and frailty\u003csup\u003e[\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]\u003c/sup\u003e. The reduced predictive value of ADL in this group suggests that even functionally independent older adults remain susceptible to severe injury if underlying physiological reserves are compromised. These findings underscore the clinical importance of injury stratification: for lower-risk individuals, interventions may focus on maintaining function through ADL support and chronic disease management; for higher-risk individuals, priority should be given to muscle strengthening, multimorbidity coordination, and environmental modifications to reduce injury severity.\u003c/p\u003e \u003cp\u003eThis study developed a fall-related injury risk prediction model using multiple machine learning algorithms, incorporating 11 readily obtainable clinical predictors to identify older adults at risk of non-fall, mild injury and moderate injury, thereby addressing a limitation in existing prediction tools that typically lack severity stratification. The four models exhibited varying performance levels. Among them, the SVM model demonstrated superior discriminative performance, achieving the macro-average AUC of 0.856, F1-score of 0.527, and a Brier score of 0.086 (95% CI: 0.078\u0026ndash;0.091). This can be attributed to the SVM\u0026rsquo;s ability to handle high-dimensional and nonlinear data, as well as its suitability for small sample sizes \u003csup\u003e[\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e]\u003c/sup\u003e. This model, which incorporates 11 readily obtainable clinical predictors, provides a practical tool for identifying older adults at risk of no fall, minor injury, or moderate-to-severe injury, thereby addressing a significant gap in existing prediction tools that typically lack severity stratification.\u003c/p\u003e \u003cp\u003eThe prediction model developed in this study is easy to evaluate and relies solely on readily obtainable indicators that require no invasive or complex procedures, such as blood tests. This simplicity enhances its suitability for various assessment settings and offers a practical tool for predicting fall-related injury risk in older adults, thereby improving the model's overall applicability.\u003c/p\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eStrengths and limitations of the research\u003c/h2\u003e \u003cp\u003eThis study develops a machine learning-based prediction model that enables risk stratification for fall-related injuries. This model provides a practical tool for healthcare providers and uses only 11 readily available clinical indicators to facilitate stratified management and precision prevention.\u003c/p\u003e \u003cp\u003eThere were several limitations in our study. First, this study developed prediction models using single machine learning algorithms (e.g., logistic regression, SVM), which may limit predictive power. Future research could employ ensemble methods such as stacking or boosting to combine multiple base models and improve overall performance. Second, our data were drawn from only a single center with 402 older adults. This may have resulted in overfitting of the models. Expanding the sample size and conducting external validation across diverse regions or timeframes would enhance its applicability and generalizability.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusions","content":"\u003cp\u003eThe results of this study revealed the following predictors of future fall-related injuries in older adults: ADL score, age, grip strength, number of chronic diseases, BMI, balance function, number of medications, fall history in the past year, history of stroke, sex, and history of fragility fracture. Using multiple machine learning algorithms, we developed a risk prediction model for fall-related injuries in older adults. Compared with existing models developed domestically and internationally, the SVM-based model demonstrated reliable accuracy and superior predictive performance. Moreover, the tool requires only 11 assessment items, making it practical for healthcare providers to screen and manage fall-related injury risk in older adults.\u003c/p\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eAuthors\u0026rsquo; contributions\u003c/h2\u003e \u003cp\u003eXue Zhang, Yan Cai, Wei Zhu, Cong Wang and Shanshan Liu conceived and designed the project. Xue Zhang, Yan Cai and Wei Zhu contributed to the data curation, analysis and interpretation. Xue Zhang drafted the original manuscript. Yan Cai, Wei Zhu, Cong Wang, Shanshan Liu and Yan Jiang performed the quality assessment and revised the manuscript critically. All authors have read and approved the submitted version.\u003c/p\u003e \u003c/div\u003e"},{"header":"Declarations","content":" \u003cp\u003e \u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e \u003cp\u003e This study was conducted in accordance with the ethical principles of the Declaration of Helsinki. The study protocol was reviewed and approved by the Committee of Ethics of West China Hospital of Sichuan University (Approval No. 2024\u0026ndash;923). Written informed consent was obtained from all individual participants included in the study prior to their enrollment.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eCompeting interests\u003c/strong\u003e \u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eThis study was supported by National Key Research and Development Program of China (2023YFC3605900). The funding sources were not involved in the study design, data collection, data analysis, data interpretation, report writing, or decision to submit the article for publication.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eXue Zhang, Yan Cai, Wei Zhu, Cong Wang and Shanshan Liu conceived and designed the project. Xue Zhang, Yan Cai and Wei Zhu contributed to the data curation, analysis and interpretation. Xue Zhang drafted the original manuscript. Yan Cai, Wei Zhu, Cong Wang, Shanshan Liu and Yan Jiang performed the quality assessment and revised the manuscript critically. All authors have read and approved the submitted version.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe data that support the findings of this study are available on reasonable request from the corresponding author. The data are not publicly available due to privacy and ethical restrictions.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eUnited UND, Nations, Department of Economic and Social Affairs, Population Division. World Population Prospects: The2017 Revision, Key Findings and Advance, Tables. [EB/OL]. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://esa.un.org/unpd/wpp/Publications/\u003c/span\u003e\u003cspan address=\"https://esa.un.org/unpd/wpp/Publications/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. [J/OL].\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWOO N, KIM SH. Sarcopenia influences fall-related injuries in community-dwelling older adults [J]. Geriatr Nurs. 2014;35(4):279\u0026ndash;82.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eER YULIANG D L, WANG YUAN, JI CUIRONG, et al. Analysis on data from Chinese National Injury Surveillance System, 2008\u0026ndash;2013 on the characteristics of Falls [J]. Chin J Epidemiol. 2015;36(1):12\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVIEIRA E R, PALMER R C, CHAVES PH. Prevention of falls in older people living in the community [J]. BMJ. 2016;353:i1419.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKIM S, KIM S. WOO STemporal trends and patterns in mortality from falls across 59 high-income and upper-middle-income countries, 1990\u0026ndash;2021, with projections up to 2040: a global time-series analysis and modelling study [J]. Volume 6. The lancet Healthy longevity; 2025. p. 100672. 1.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLEILEI D, PENGPENG Y. The burden of injury in China, 1990\u0026ndash;2017: findings from the Global Burden of Disease Study 2017 [J]. Lancet Public Health. 2019;4(9):e449\u0026ndash;61. HAAGSMA J A,.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYE P, ER Y, WANG H, et al. Burden of falls among people aged 60 years and older in mainland China, 1990\u0026ndash;2019: findings from the Global Burden of Disease Study 2019 [J]. Lancet Public Health. 2021;6(12):e907\u0026ndash;18.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRUDNICKA E, NAPIERAŁA P. The World Health Organization (WHO) approach to healthy ageing [J]. Maturitas. 2020;139:6\u0026ndash;11.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCAMPBELL W W DIPIETROL, BUCHNER D M, et al. Physical Activity, Injurious Falls, and Physical Function in Aging: An Umbrella Review [J]. Med Sci Sports Exerc. 2019;51(6):1303\u0026ndash;13.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eONG M F, SOH K L, SAIMON R, et al. Fall prevention education to reduce fall risk among community-dwelling older persons: A systematic review [J]. J Nurs Adm Manag. 2021;29(8):2674\u0026ndash;88.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDYKES PC, BURNS Z. Evaluation of a Patient-Centered Fall-Prevention Tool Kit to Reduce Falls and Injuries: A Nonrandomized Controlled Trial [J]. JAMA Netw Open. 2020;3(11):e2025889.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMONTERO-ODASSO M, VAN DER VELDE N, MARTIN, F C et al. World guidelines for falls prevention and management for older adults: a global initiative [J]. Age Ageing, 2022, 51(9).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePAPAIOANNOU A, SANTESSO N, MORIN. et al. Recommendations for preventing fracture in long-term care [J]. Volume 187. CMAJ: Canadian Medical Association journal\u0026thinsp;=\u0026thinsp;journal de l'Association medicale canadienne; 2015. pp. 1135\u0026ndash;44. 15S N,.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLACHANCE CC, JURKOWSKI MP, DYMARZ AC, et al. Compliant flooring to prevent fall-related injuries in older adults: A scoping review of biomechanical efficacy, clinical effectiveness, cost-effectiveness, and workplace safety [J]. PLoS ONE. 2017;12(2):e0171652.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSAI A, TANAKA K, OHASHI Y, et al. Quantitative sonographic assessment of quadriceps muscle thickness for fall injury prediction in patients undergoing maintenance hemodialysis: an observational cohort study [J]. BMC Nephrol. 2021;22(1):191.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNILSSON M, ERIKSSON J, LARSSON B, et al. Fall Risk Assessment Predicts Fall-Related Injury, Hip Fracture, and Head Injury in Older Adults [J]. J Am Geriatr Soc. 2016;64(11):2242\u0026ndash;50.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSPEISER JL, CALLAHAN KE, HOUSTON DK, et al. Machine Learning in Aging: An Example of Developing Prediction Models for Serious Fall Injury in Older Adults [J]. J Gerontol Biol Sci Med Sci. 2021;76(4):647\u0026ndash;54.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEK S, CALDER\u0026oacute;N-LARRA\u0026ntilde;AGA RIZZUTOD. Predicting First-Time Injurious Falls in Older Men and Women Living in the Community: Development of the First Injurious Fall Screening Tool [J]. J Am Med Dir Assoc. 2019;20(9):1163\u0026ndash;e83.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLI W, RAO Z, FU Y, et al. Value of the short physical performance battery (SPPB) in predicting fall and fall-induced injury among old Chinese adults [J]. BMC Geriatr. 2023;23(1):574.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYAITA S, TAGO M. A Simple and Accurate Model for Predicting Fall Injuries in Hospitalized Patients: Insights from a Retrospective Observational Study in Japan [J]. Med Sci monitor: Int Med J experimental Clin Res. 2023;29:e941252. KATSUKI N E,.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSM B, VL N. Australian Diabetes Foot Network: practical guideline on the provision of footwear for people with diabetes [J]. J Foot Ankle Res. 2013;6(1):1757\u0026ndash;146.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eL W, W B, KG M, \u003cem\u003eet al.\u003c/em\u003e A simulation study of sample size demonstrated the importance of the number of events per variable to develop prediction models in clustered data [J]. J Clin Epidemiol, 2015, 68(12): 1406\u0026ndash;14.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJIANG Y, XIA Q. Study on the epidemical characteristies and disease burden of fall-related imiury among community-dwelling elderly adults in Changning District, Shanghai [J]. Chin J Disease Control Prev. 2013;17(02):134\u0026ndash;7.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQIAN XX, CHAU PH, KWAN CW, et al. Investigating Risk Factors for Falls among Community-Dwelling Older Adults According to WHO's Risk Factor Model for Falls [J]. J Nutr Health Aging. 2021;25(4):425\u0026ndash;32.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCHOCK J. Identification of Fall Risk Factors and Development of a Falls Prediction Model for Community-Dwelling Older Adults Receiving Public Health Nursing Services in Hawai\u0026acute;i [D]. United States -- Hawaii; University of Hawai'i at Manoa, 2023.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSRIVASTAVA S, MUHAMMAD T. Prevalence and risk factors of fall-related injury among older adults in India: evidence from a cross-sectional observational study [J]. BMC Public Health. 2022;22(1):550.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eARLETTAZ Y. Augmented osteosynthesis in fragility fracture [J]. Orthopaedics \u0026amp; traumatology, surgery \u0026amp; research: OTSR. 2023, 109(1s): 103461.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMAHONEY F I, BARTHEL D W. FUNCTIONAL. EVALUATION: THE BARTHEL INDEX [J]. Maryland State Med J. 1965;14:61\u0026ndash;5.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLIU H, HOU Y, LI H, et al. Influencing factors of weak grip strength and fall: a study based on the China Health and Retirement Longitudinal Study (CHARLS) [J]. BMC Public Health. 2022;22(1):2337.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eD P, S R. - The timed Up \u0026amp; Go: a test of basic functional mobility for frail elderly [J]. J Am Geriatr Soc, 1991, 39(2): 142\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZHIJUN WU W S, WEIYAN JAN. Incidence of Falls among Inpatient in China:A Survey of 490 Tertiary Hospitals [J]. Chin Health Qual Manage. 2019;26(3):14\u0026ndash;7.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMOONS K G M, WOLFF R F, RILEY, R D, et al. PROBAST: A Tool to Assess Risk of Bias and Applicability of Prediction Model Studies: Explanation and Elaboration [J]. Ann Intern Med. 2019;170(1):W1\u0026ndash;33.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLUNDBERG S, LEE S I.. A Unified Approach to Interpreting Model Predictions [M]. 2017.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZHAO M, LI S, XU Y, et al. Developing a Scoring Model to Predict the Risk of Injurious Falls in Elderly Patients: A Retrospective Case-Control Study in Multicenter Acute Hospitals [J]. Clin Interv Aging. 2020;15:1767\u0026ndash;78.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCHEN X, HE L, SHI K, et al. Interpretable Machine Learning for Fall Prediction Among Older Adults in China [J]. Am J Prev Med. 2023;65(4):579\u0026ndash;86.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSPEISER JL, CALLAHAN K E, HOUSTON D K, et al. Machine Learning in Aging: An Example of Developing Prediction Models for Serious Fall Injury in Older Adults [J]. Journals Gerontology: Ser A. 2020;76(4):647\u0026ndash;54.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAJ C-J GB. Sarcopenia: revised European consensus on definition and diagnosis [J]. Age Ageing. 2019;48(1):16\u0026ndash;31.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePORTO JM, NAKAISHI A P M, CANGUSSU-OLIVEIRA L M, et al. Relationship between grip strength and global muscle strength in community-dwelling older people [J]. Arch Gerontol Geriatr. 2019;82:273\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLIANG X Z, CHAI J L. A fall risk prediction model based on the CHARLS database for older individuals in China [J]. BMC Geriatr. 2025;25(1):170. LI G Z,.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSHAO L, SHI Y, XIE X Y, et al. Incidence and Risk Factors of Falls Among Older People in Nursing Homes: Systematic Review and Meta-Analysis [J]. J Am Med Dir Assoc. 2023;24(11):1708\u0026ndash;17.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCARRACEDO-REBOREDO P, LI\u0026ntilde;ARES-BLANCO J, RODR\u0026iacute;GUEZ-FERN\u0026aacute;NDEZ N, et al. A review on machine learning approaches and trends in drug discovery [J]. Comput Struct Biotechnol J. 2021;19:4538\u0026ndash;58.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"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":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Older Adults, Fall-related injuries, Risk prediction model, Machine learning","lastPublishedDoi":"10.21203/rs.3.rs-8683642/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8683642/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\u003eFalls and related injuries among older adults represent a significant public health issue, adversely affecting their quality of life and increasing the socioeconomic burden. This study aims to develop a risk stratification prediction model based on machine learning algorithms.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eUsing a convenience sampling method, a total of 402 older adults scheduled for discharge at the Center of Gerontology and Geriatrics of West China Hospital, Sichuan University from April to September 2024 were enrolled. The participants were followed for 6 months. The participants were categorized into three groups: non-fall, no/minor injury, and moderate to severe injury group. Least absolute shrinkage and selection operator (LASSO) regression was applied to screen candidate predictors. Multinomial logistic regression (MLR), random forest (RF), support vector machine (SVM), and naive Bayes (NB) were employed for model development, with internal validation performed via 5-fold cross-validation. Model performance was assessed via the macro-average AUC, F1 score and Brier score. The optimal model was interpreted using SHAP analysis. This study was approved by the Committee of Ethics of West China Hospital of Sichuan University (Approval No. 2024–923).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe SVM demonstrated the best performance, with a macro-average AUC of 0.856 (95% CI: 0.837–0.878), and F1-score of 0.527 (95% CI: 0.410–0.619), and a Brier score of 0.086 (95% CI: 0.078–0.091).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study developed a risk-stratified prediction model for fall-related injuries in older adults. The SVM model is capable of accurately predicting the risk levels of fall-related injuries in older adults.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRegistration\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ewww.chictr.org.cn ChiCTR2400085499. Registered 11/6/2024, first recruitment 13/6/2024.\u003c/p\u003e","manuscriptTitle":"Development of a Fall-Related Injury Risk-Stratified Prediction Model for Older Adults","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-02 08:42:49","doi":"10.21203/rs.3.rs-8683642/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"59aeab47-daf5-43e9-b22b-6640766be3b0","owner":[],"postedDate":"March 2nd, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-03-20T17:39:31+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-02 08:42:49","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8683642","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8683642","identity":"rs-8683642","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}