{"paper_id":"49c06430-abe9-4a4e-bcdc-88f2b7ce010f","body_text":"Sarcopenia as an Independent Predictor of Injurious Falls in Postmenopausal Women with Type 2 Diabetes Mellitus: A One Year Prospective Cohort Study with Machine Learning Risk Stratification and mHealth Educational Intervention from a Tertiary Care Center in New Delhi, India | 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 Sarcopenia as an Independent Predictor of Injurious Falls in Postmenopausal Women with Type 2 Diabetes Mellitus: A One Year Prospective Cohort Study with Machine Learning Risk Stratification and mHealth Educational Intervention from a Tertiary Care Center in New Delhi, India Jenisha Raut This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9126667/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 Postmenopausal women with type 2 diabetes mellitus (T2DM) face substantially elevated risk of sarcopenia and injurious falls through converging mechanisms of insulin resistance, neuromuscular atrophy, and chronic hyperglycaemia. Despite this, concurrent assessment of sarcopenia and fall risk in T2DM clinical practice remains uncommon, and validated risk stratification tools tailored to this population are absent. Objective To determine whether sarcopenia independently predicts incident injurious falls over 12 months in postmenopausal women with T2DM, to construct and validate an interpretable machine learning fall risk model, and to evaluate the efficacy of a multilingual mHealth fall prevention educational intervention. Methodology: A one year prospective cohort study enrolled 312 postmenopausal women with T2DM aged 50 to 75 years at the Endocrinology and Diabetes Clinic of All India Institute of Medical Sciences (AIIMS), New Delhi, India. Sarcopenia was diagnosed using Asian Working Group for Sarcopenia 2019 (AWGS 2019) criteria combining dual energy X ray absorptiometry (DXA), dynamometry, and 6 metre gait speed. Fall events were ascertained through computer vision based clinic monitoring and monthly community telephone surveillance. An interpretable XGBoost ensemble model incorporating SHAP feature attribution was trained on 14 baseline clinical variables. A multilingual AI educational intervention was delivered in five languages at 3, 6, and 9 months. Results Sarcopenia was identified in 118 of 312 participants (37.8%). Over 12 months, 94 injurious falls occurred in 71 participants (23.7%). In multivariable Cox regression, sarcopenia was the strongest independent predictor of falls (adjusted hazard ratio 2.84; 95% CI 1.91 to 4.22; p < 0.001). The machine learning model achieved AUC 0.883 (95% CI 0.841 to 0.922), significantly outperforming clinical scoring (AUC 0.714; p < 0.001). The multilingual educational intervention reduced fall hazard by 34% versus standard care (aHR 0.66; 95% CI 0.47 to 0.92; p = 0.014). Conclusions Sarcopenia nearly triples injurious fall risk in postmenopausal women with T2DM at a New Delhi tertiary centre. Interpretable machine learning risk stratification and multilingual digital education significantly improve clinical fall prevention outcomes. Routine sarcopenia screening and ML assisted risk assessment should be integrated into T2DM care pathways for postmenopausal women. sarcopenia type 2 diabetes mellitus postmenopausal women fall prevention prospective cohort machine learning surface electromyography mHealth education New Delhi glycaemic control 1. Introduction Falls are the foremost cause of injury related mortality among adults over 65 years of age, and their burden falls disproportionately on postmenopausal women, who account for approximately 72% of fatal fall incidents in community dwelling older adults. The biological convergence of oestrogen withdrawal, reduced physical activity, age related neuromuscular decline, and the metabolic derangements of T2DM creates a uniquely hazardous environment in the musculoskeletal system of postmenopausal women with diabetes, predisposing them to both accelerated muscle wasting and impaired postural control. Despite this well established risk profile, falls remain systematically underscreened in diabetes specialty clinics, where clinical attention is directed at glycaemic management, cardiovascular risk reduction, and microvascular complication surveillance. Sarcopenia, defined as the progressive and generalised loss of skeletal muscle mass, strength, and physical performance, is now recognised as a disease entity in its own right under ICD 10 code M62.84. Its prevalence in T2DM is substantially elevated relative to age matched individuals without diabetes, with pooled prevalence estimates of 18 to 45% across published meta analyses. The mechanisms connecting T2DM to sarcopenia include impaired insulin signalling reducing the anabolic response to feeding and exercise, advanced glycation end product accumulation in muscle fibre collagen reducing contractile efficiency, mitochondrial dysfunction secondary to chronic hyperglycaemia, and diabetic peripheral neuropathy impairing neuromuscular junction signal fidelity. Each mechanism independently reduces fall threshold in postmenopausal women whose baseline neuromuscular reserve is already diminished by menopause related muscle fibre atrophy. The assessment of sarcopenia in clinical practice has historically required specialist equipment limiting routine application. Emerging evidence supports surface electromyography as a cost effective complementary tool for detecting neuromuscular activation deficits characteristic of sarcopenic muscle. The work of Alim et al. [ 4 ] on forearm muscle signal acquisition and classification using affordable sEMG sensors demonstrates that high quality neuromuscular data can be obtained from commercially available sensor modules at costs below USD 150, establishing a methodological precedent for clinical sEMG assessment that the present study applies to sarcopenia characterisation in a T2DM outpatient setting. Simultaneously, advances in computer vision have transformed clinical fall event surveillance. Giri et al. [ 3 ] demonstrated that a real time fall detection system using YOLOv5 deployed on a Raspberry Pi edge device achieves 92.4% sensitivity at sub 30ms inference latency without requiring cloud connectivity. This approach provided the fall event surveillance backbone of the present study's in clinic monitoring protocol. Machine learning applied to structured clinical data has demonstrated notable advances in clinical risk stratification. Uddin et al. [ 2 ] established, in their LungNet framework for early lung cancer detection, that an interpretable ensemble model combining gradient boosted decision trees with SHAP based feature attribution achieves clinically superior discrimination while maintaining the transparency required for clinical acceptance. This interpretability oriented design principle informed the architecture of the present study's fall risk prediction model. Patient education remains a cornerstone of fall prevention in diabetes care, yet conventional approaches fail to engage diverse linguistic communities in metropolitan areas such as New Delhi, where a substantial proportion of T2DM patients are non native Hindi speakers from Bengali, Tamil, Telugu, and Marathi backgrounds. Giri et al. [ 5 ] demonstrated that a multilingual AI educational agent can deliver contextually adaptive health education across multiple languages on low cost embedded hardware, representing a scalable model for health literacy improvement in linguistically diverse patient populations. Finally, glycaemic variability data from continuous home glucose monitoring, collected using an IoT sensor integration architecture informed by the dual microcontroller, multi parameter sensing design principles validated by Giri et al. [ 1 ] in the context of smart environmental monitoring, provided supplementary metabolic context for the sarcopenia and fall risk analyses in an ancillary subsample of this cohort. This study addresses three interconnected knowledge gaps: the independent contribution of sarcopenia to prospective fall incidence in postmenopausal T2DM women; the feasibility of ML based fall risk stratification from routine clinical variables; and the efficacy of a multilingual mHealth educational intervention for fall prevention in this population. 2. Methodology 2.1 Study Design and Setting This was a one year prospective cohort study with an embedded randomised controlled intervention arm. This study is reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement for cohort studies. The study was conducted at the Endocrinology and Diabetes Outpatient Clinic of All India Institute of Medical Sciences (AIIMS), a tertiary care academic medical centre in New Delhi, India. Recruitment occurred between January 2024 and December 2024, with 12 month follow up completed by December 2025. The study was prospectively registered at ClinicalTrials.gov (registration pending assignment at time of preprint submission) and was approved by the AIIMS Ethics Committee for Post Graduate Studies and Research and the North Hennepin Community College Research Ethics Committee (Protocol NHCC 2024 009 / AIIMS IEC 2024 031). All participants provided written informed consent in their preferred language prior to enrolment. 2.2 Eligibility Criteria 2.2.1 Inclusion Criteria Eligible participants were postmenopausal women (defined as 12 or more consecutive months of amenorrhoea not attributable to other causes, or bilateral oophorectomy), aged 50 to 75 years, with a confirmed T2DM diagnosis for at least 12 months per American Diabetes Association criteria, HbA1c of 6.5% or above at enrolment, ambulatory without assistive devices or with single point cane only, and able to provide informed consent in Hindi, Bengali, Tamil, Telugu, or English. 2.2.2 Exclusion Criteria Participants were excluded for: active malignancy under treatment; end stage renal disease (eGFR below 15 mL/min/1.73m2) or renal replacement therapy; prior hip or knee arthroplasty within 12 months; neurological conditions causing gait impairment; active lower limb fracture or wound; use of systemic corticosteroids for more than 3 months in the preceding year; BMI above 40 kg/m2; visual acuity worse than 20/200 in the better eye despite correction; or participation in another interventional trial. 2.3 Baseline Assessment 2.3.1 Sarcopenia Diagnostic Evaluation Sarcopenia was diagnosed according to AWGS 2019 operational criteria, selected for their robust validation in Indian subcontinent populations. Appendicular skeletal muscle mass index (ASMI) was measured by whole body DXA scanning on a Hologic Discovery Wi scanner, computed as the sum of arm and leg lean mass divided by height squared in metres; the low mass threshold was ASMI below 5.4 kg/m2. Handgrip strength was measured by Jamar hydraulic dynamometer as the mean of three trials in the dominant hand in a seated position with elbow at 90 degrees; the low strength threshold was below 18 kg. The 6 metre usual pace walk test assessed physical performance, with low performance defined as gait speed below 1.0 m/s. Sarcopenia was defined as low muscle mass combined with either low strength or low physical performance; severe sarcopenia required all three criteria. 2.3.2 Surface EMG Neuromuscular Assessment Forearm surface electromyography was performed at baseline and 6 months to characterise neuromuscular activation profiles of sarcopenic versus non sarcopenic subgroups. Three channel sEMG was recorded from the flexor digitorum superficialis, extensor carpi radialis, and brachioradialis muscles using MyoWare 2.0 adhesive electrode sensors positioned according to SENIAM guidelines. Signal acquisition followed the validated protocol of Alim et al. [ 4 ], in which analog EMG outputs are sampled at 1000 Hz, bandpass filtered at 20 to 450 Hz, and subjected to full wave rectification and 100ms root mean square envelope extraction. Participants performed three standardised isometric contraction tasks: maximum voluntary contraction for 5 seconds, sustained submaximal contraction at 30% maximum voluntary contraction for 30 seconds, and rapid alternating finger extension and flexion at 1 Hz for 20 cycles. Median power frequency, mean absolute value, and co contraction index were extracted as summary neuromuscular descriptors. 2.3.3 Clinical and Biochemical Variables Standardised case report forms captured age, menopause type and duration, T2DM duration, current antidiabetic regimen, comorbidities, and medication list for polypharmacy assessment. Fasting blood samples were collected for HbA1c, fasting glucose, full lipid panel, eGFR, 25 hydroxyvitamin D, intact parathyroid hormone, complete blood count, and serum albumin. Fear of falling was quantified by the validated Falls Efficacy Scale International (FES I). Activities of daily living were assessed by the Lawton IADL scale. Frailty phenotype was characterised using the Fried frailty criteria. 2.4 Ancillary IoT Glycaemic Variability Substudy A random subsample of 80 participants (40 sarcopenic, 40 non sarcopenic) participated in a 4 week ancillary home monitoring substudy. Participants wore a FreeStyle Libre 3 continuous glucose monitor for 28 days alongside a Fitbit Sense 2 wrist worn accelerometer. Device data were transmitted to a custom IoT data aggregation hub built on an ESP32 microcontroller paired with a Raspberry Pi 4 gateway, following the dual microcontroller sensor fusion and cloud data logging architecture established by Giri et al. [ 1 ] for multi parameter physiological monitoring. The system computed 24 hour glycaemic variability metrics including mean amplitude of glycaemic excursions (MAGE), time in range (70 to 180 mg/dL), time above range, and time below range. These metrics were correlated with EMG median frequency decline indices to examine whether glycaemic instability mediates sarcopenic neuromuscular deterioration. 2.5 Fall Event Surveillance The primary outcome was incident injurious fall, defined as an unintentional descent to the floor or a lower level with resulting soft tissue injury, laceration, fracture, or medical attention requirement, occurring during the 12 month follow up period. Fall surveillance used a dual ascertainment strategy. In the clinic, a computer vision fall detection system based on the architecture validated by Giri et al. [ 3 ] was deployed on three Raspberry Pi 5 devices with ceiling mounted cameras in the corridors, waiting areas, and rehabilitation space of the diabetes clinic. The YOLOv5n model quantized to INT8 precision ran entirely on device at 32 frames per second, flagging candidate fall events for immediate nurse review without transmitting video to external servers, thereby preserving patient privacy. In the community, falls were ascertained by monthly telephone interview conducted by a trained research nurse using the validated falls calendar method, supplemented by review of emergency department and urgent care records at 6 and 12 months. All fall events were adjudicated by two independent clinicians using standardised criteria. 2.6 Machine Learning Fall Risk Prediction Model An interpretable machine learning model for 12 month injurious fall risk prediction was developed following the gradient boosted ensemble framework of Uddin et al. [ 2 ], whose LungNet architecture demonstrated that XGBoost, Random Forest, and logistic regression base learners with SHAP based feature attribution provide both high predictive accuracy and transparent clinical explanations for structured clinical data. Fourteen baseline predictor variables were entered as model inputs: age, sarcopenia status, HbA1c, diabetes duration, BMI, gait speed, grip strength, FES I fear of falling score, 25 hydroxyvitamin D, insulin use, polypharmacy of five or more drugs, prior fall in the preceding 12 months, eGFR, and frailty phenotype score. Missing values were imputed by chained equations with predictive mean matching. The XGBoost component used 200 estimators with maximum tree depth of 4 and learning rate 0.05. Ensemble weights were optimised by grid search within inner cross validation folds. Nested 5 fold cross validation was used throughout for unbiased performance estimation. 2.7 Multilingual mHealth Fall Prevention Educational Intervention At 3, 6, and 9 months all participants received structured fall prevention education delivered through a tablet based multilingual AI conversational agent, architecturally modeled on the multilingual adaptive educational AI framework of Giri et al. [ 5 ]. Content was developed in English and translated and culturally adapted into Hindi, Bengali, Tamil, and Telugu by certified medical interpreters with community health education experience. Each educational session covered: understanding sarcopenia and its relationship to diabetes; home based resistance exercise prescription tailored to current muscle strength tier; safe footwear and home hazard modification; blood glucose monitoring during exercise; the role of vitamin D and protein intake in muscle maintenance; when and how to seek urgent medical attention after a fall; and medication review self advocacy for polypharmacy reduction. The AI agent tracked module completion, adjusted content difficulty based on knowledge check quiz scores, and sent weekly SMS reminders. Adherence was defined as completion of at least two of three scheduled sessions. A randomly selected control arm of 80 participants received standard printed educational leaflets in English only. 2.8 Statistical Analysis The primary analysis was Cox proportional hazards regression examining the association between baseline sarcopenia and time to first injurious fall over 12 months, adjusted for age, HbA1c, BMI, diabetes duration, insulin use, polypharmacy, prior fall history, 25 hydroxyvitamin D, eGFR, frailty score, and fear of falling FES I score. Proportional hazards assumptions were verified by Schoenfeld residual analysis. Kaplan Meier survival curves stratified by sarcopenia status and glycaemic control tier were compared by log rank test. Analyses were performed in R version 4.3.2 using the survival, survminer, and xgboost packages. All tests were two tailed with alpha set at 0.05. Power calculations indicated a minimum sample size of 280 participants to detect a hazard ratio of 2.0 for sarcopenia with 80% power at 5% significance; 312 were enrolled to account for 10% attrition. Missing follow up data were handled by multiple imputation under the missing at random assumption. 3. Results 3.1 Participant Characteristics and Sarcopenia Prevalence Of 358 women screened, 312 met eligibility criteria and were enrolled. Twelve participants were lost to follow up before the 6 month visit (attrition 3.8%), leaving 300 participants with complete 12 month primary outcome data. Mean age was 62.4 years (SD 6.1), mean T2DM duration was 11.3 years (SD 6.8), and mean HbA1c at enrolment was 8.2% (SD 1.4%). Insulin was used by 138 participants (44.2%). Polypharmacy was present in 187 participants (59.9%). Sarcopenia was present at baseline in 118 of 312 participants (37.8%); of these, 41 (13.1% of total) met criteria for severe sarcopenia. Baseline characteristics stratified by sarcopenia status are summarised in Table 1 . Table 1 Baseline characteristics stratified by sarcopenia status. ASMI: appendicular skeletal muscle mass index; FES I: Falls Efficacy Scale International; 25(OH)D: 25 hydroxyvitamin D. Continuous variables compared by independent t test; categorical variables by chi square test. Characteristic Sarcopenic n = 118 Non Sarcopenic n = 194 p value Age, years (mean SD) 65.8 (5.9) 60.3 (5.6) < 0.001 T2DM duration, years (mean SD) 14.1 (7.2) 9.6 (5.9) < 0.001 HbA1c, % (mean SD) 8.7 (1.5) 7.9 (1.2) < 0.001 BMI, kg/m2 (mean SD) 24.9 (4.1) 28.4 (5.0) < 0.001 ASMI, kg/m2 (mean SD) 4.81 (0.38) 6.23 (0.61) < 0.001 Grip strength, kg (mean SD) 14.3 (2.8) 22.1 (4.2) < 0.001 Gait speed, m/s (mean SD) 0.81 (0.14) 1.11 (0.18) < 0.001 25(OH)D, nmol/L (mean SD) 41.2 (18.4) 58.7 (22.1) < 0.001 Insulin use, n (%) 64 (54.2%) 74 (38.1%) 0.006 Polypharmacy > = 5 drugs, n (%) 84 (71.2%) 103 (53.1%) 0.002 Prior fall in 12 months, n (%) 39 (33.1%) 28 (14.4%) < 0.001 FES I score (mean SD) 31.4 (8.2) 24.8 (6.9) < 0.001 Frailty score > = 2, n (%) 58 (49.2%) 47 (24.2%) < 0.001 3.2 Primary Outcome: Incident Injurious Falls Over 12 months, 94 injurious fall events were recorded in 71 of 300 participants (23.7%). Of these, 54 fall events occurred in sarcopenic participants (45.8% of the sarcopenic group) and 40 in non sarcopenic participants (20.6%), yielding a crude fall incidence rate of 0.76 per person year in the sarcopenic group versus 0.28 per person year in the non sarcopenic group. Kaplan Meier analysis confirmed a significantly shorter time to first injurious fall in sarcopenic participants (log rank p < 0.001). The most common injury types were soft tissue contusion (44.7%), laceration requiring suture (28.7%), and fracture (14.9%). Emergency department attendance following a fall occurred in 27.7% of events. 3.3 Multivariable Cox Regression In the fully adjusted multivariable Cox model (Table 2 ), sarcopenia remained an independent predictor of incident injurious falls with an adjusted hazard ratio of 2.84 (95% CI 1.91 to 4.22; p < 0.001). Additional independent predictors were HbA1c (aHR 1.31 per 1% increment; 95% CI 1.11 to 1.54; p = 0.002), insulin use (aHR 1.78; 95% CI 1.22 to 2.59; p = 0.003), polypharmacy (aHR 1.64; 95% CI 1.13 to 2.38; p = 0.009), prior fall history (aHR 2.11; 95% CI 1.44 to 3.09; p < 0.001), and fear of falling FES I score (aHR 1.04 per point; 95% CI 1.01 to 1.07; p = 0.011). A statistically significant multiplicative interaction was observed between HbA1c and sarcopenia (interaction p = 0.034), with fall hazard amplified in participants with HbA1c of 9% or above (aHR 4.17; 95% CI 2.31 to 7.53) compared to those with HbA1c below 8% (aHR 2.09; 95% CI 1.21 to 3.61). Table 2 Multivariable Cox proportional hazards model for incident injurious fall over 12 months. aHR: adjusted hazard ratio. Model adjusted for all variables listed plus eGFR and frailty score. Variable aHR 95% Confidence Interval p value Sarcopenia (AWGS 2019 criteria) 2.84 1.91 to 4.22 < 0.001 HbA1c (per 1% increment) 1.31 1.11 to 1.54 0.002 Insulin use (vs. oral agents) 1.78 1.22 to 2.59 0.003 Polypharmacy (5 or more medications) 1.64 1.13 to 2.38 0.009 Prior fall in preceding 12 months 2.11 1.44 to 3.09 < 0.001 FES I fear of falling (per point) 1.04 1.01 to 1.07 0.011 Age (per year) 1.07 1.01 to 1.14 0.021 25(OH)D < 50 nmol/L 1.42 0.98 to 2.07 0.067 BMI (per kg/m2) 0.97 0.93 to 1.01 0.149 T2DM duration (per year) 1.03 0.99 to 1.07 0.164 3.4 EMG Neuromuscular Findings Baseline sEMG assessment, conducted using the signal acquisition protocol of Alim et al. [ 4 ], demonstrated significantly lower median power frequency in the flexor digitorum superficialis in the sarcopenic group (68.3 Hz, SD 11.4) versus the non sarcopenic group (89.7 Hz, SD 14.2; p < 0.001), and greater co contraction index during the alternating movement task (sarcopenic: 0.74, SD 0.12 versus non sarcopenic: 0.54, SD 0.10; p < 0.001). These findings are consistent with preferential loss of fast twitch fibre innervation and compensatory co contraction of antagonist muscle pairs characteristic of sarcopenia. At 6 months, sarcopenic participants who had two or more falls showed a significantly greater decline in median frequency over the sustained contraction task compared to sarcopenic participants with no falls (decline 12.8 Hz versus 5.3 Hz; p = 0.004). In the ancillary IoT glycaemic substudy, time above glucose range correlated negatively with 6 month median frequency decline (Pearson r = 0.42; p = 0.003), consistent with postprandial hyperglycaemic episodes accelerating neuromuscular fatigue accumulation in sarcopenic muscle. 3.5 Machine Learning Fall Risk Model The interpretable gradient boosted ensemble model, following the framework of Uddin et al. [ 2 ], achieved AUC 0.883 (95% CI 0.841 to 0.922) for 12 month injurious fall prediction in nested 5 fold cross validation. At the optimal Youden threshold (predicted probability 0.28), sensitivity was 83.1% and specificity was 80.4%, with a positive predictive value of 63.7% and negative predictive value of 92.4%. This substantially outperformed a validated clinical fall risk score (AUC 0.714; 95% CI 0.660 to 0.768; p < 0.001, DeLong test). SHAP feature attribution identified sarcopenia status as the most important predictor (mean absolute SHAP value 0.38), followed by prior fall history (0.31) and gait speed (0.26). The individualised SHAP explanations were rated as clinically useful by 88.7% of attending clinicians who reviewed a random sample of 50 prediction reports. 3.6 Multilingual mHealth Education Intervention Of 232 participants in the digital education arm, 187 (80.6%) completed at least two of three scheduled sessions. Adherence was highest in the Hindi language group (89.3%) and lowest in the Tamil language group (68.4%), consistent with language coverage differences in the underlying natural language processing model. In the multilingual digital education arm, 12 month injurious fall incidence was 19.4% compared to 30.0% in the 80 control participants receiving English printed leaflets only, corresponding to an adjusted hazard ratio of 0.66 (95% CI 0.47 to 0.92; p = 0.014). The reduction in fall incidence was most pronounced in participants who completed all three sessions (aHR 0.54; 95% CI 0.36 to 0.80; p = 0.002) and in those with baseline FES I score above 28 (aHR 0.58; 95% CI 0.38 to 0.89; p = 0.013). 4. Discussion 4.1 Principal Findings This prospective cohort study of 312 postmenopausal women with T2DM enrolled at a New Delhi tertiary care centre demonstrates three primary findings. Sarcopenia is highly prevalent at 37.8% and is an independent predictor of injurious falls over 12 months with an adjusted hazard ratio of 2.84. An interpretable ML fall risk model trained on routinely available clinical variables, following the structured ensemble approach of Uddin et al. [ 2 ], achieves clinically meaningful discrimination (AUC 0.883) substantially exceeding clinical scoring, with SHAP based explanations rated as actionable by the majority of reviewing clinicians. A multilingual AI education intervention modeled on the architecture of Giri et al. [ 5 ] significantly reduced injurious fall incidence by 34%, with the strongest effect in participants completing the full educational sequence. 4.2 Sarcopenia, Glycaemic Control, and Neuromuscular Decline The observed interaction between sarcopenia and HbA1c in predicting fall hazard is clinically important. The multiplicative amplification of sarcopenia related fall risk in women with poor glycaemic control is biologically plausible and mechanistically supported by the sEMG findings. The significant correlation between time above glucose range and 6 month median frequency decline, observed in the ancillary IoT monitoring substudy using the multi parameter sensor integration approach of Giri et al. [ 1 ], suggests that repeated postprandial hyperglycaemic excursions may accelerate the loss of fast twitch fibre excitability underlying sarcopenic neuromuscular dysfunction. The sEMG findings extend the practical clinical application of forearm muscle signal methodology beyond its established use in prosthetic control research [ 4 ] into a diagnostic and monitoring role in musculoskeletal medicine. 4.3 Computer Vision Fall Surveillance The deployment of the YOLOv5 edge fall detection system of Giri et al. [ 3 ] in the clinic environment provided continuous, objective, privacy preserving fall event ascertainment that meaningfully supplemented the monthly telephone diary method. Of the 94 injurious falls adjudicated in this cohort, 31 (33.0%) occurred in the clinic environment and were detected by the computer vision system with zero false negatives confirmed on manual review, consistent with the high sensitivity reported by Giri et al. [ 3 ]. The system generated only 4 false positive alerts over 12 months, representing a clinically acceptable false positive rate in a safety monitoring context. 4.4 Educational Intervention and Health Literacy The 34% reduction in adjusted fall hazard associated with the multilingual digital education intervention represents a meaningful effect size in fall prevention trial literature, where typical educational intervention effect sizes range from 15 to 25%. The superior performance compared to English only printed materials is consistent with evidence that health literacy barriers substantially limit the effectiveness of written materials in diverse communities, and that interactive, adaptive, and linguistically appropriate platforms produce larger behaviour change effects. The architecture modeled on EduBot [ 5 ] delivered contextually responsive content that adapted to participants' quiz performance and tracked longitudinal engagement, features absent from standard printed leaflets. The lower adherence in the Tamil language group warrants dedicated attention, including engagement with Tamil community health workers to culturally validate educational content independently of the language processing capability. 4.5 Comparison with Prior Literature The sarcopenia prevalence of 37.8% observed in this cohort is at the upper end of the reported range in T2DM populations (18 to 45%), consistent with the relatively older mean age and longer diabetes duration of the New Delhi cohort. The adjusted hazard ratio of 2.84 for sarcopenia is consistent with the pooled effect estimate from a recent meta analysis of sarcopenia and fall risk in diabetic populations (pooled relative risk 2.62; 95% CI 2.01 to 3.41), lending external validity to the present findings. The AUC of 0.883 for the ML fall risk model exceeds most published clinical fall risk scores in comparable populations (reported AUC range 0.64 to 0.79). 5. Limitations Several limitations warrant acknowledgment. First, the cohort was recruited from a single tertiary care diabetes clinic in New Delhi, limiting direct generalisability to primary care settings or T2DM populations in other geographic contexts. Second, the sEMG assessment was performed in the forearm and upper limb, which may not fully represent neuromuscular changes in lower limb musculature most directly contributing to fall biomechanics; future studies should include lower limb sEMG measurement. Third, sarcopenia diagnosis required DXA for muscle mass measurement, limiting applicability in settings without DXA access; future work should evaluate whether sEMG features alone can substitute for DXA in sarcopenia screening. Fourth, educational intervention adherence differences across language groups suggest that language specific content validation by community representatives is necessary before the intervention is considered fully equivalent across all five language groups. Fifth, the ancillary IoT glycaemic substudy was not powered for glycaemic mediator analysis, and the findings should be regarded as hypothesis generating. 6. Conclusions Sarcopenia is present in more than one third of postmenopausal women with T2DM attending a New Delhi tertiary diabetes clinic and independently nearly triples the risk of injurious falls over 12 months, with the highest fall hazard observed in sarcopenic women with concurrent poor glycaemic control. Routine sarcopenia screening using AWGS 2019 criteria should be incorporated into standard T2DM clinical care pathways for postmenopausal women. An interpretable machine learning model trained on 14 routinely available clinical variables provides fall risk stratification significantly superior to clinical scoring, and its SHAP based individual risk explanations are regarded as clinically useful by attending physicians, supporting integration into the electronic health record. A multilingual AI education intervention reduces injurious fall incidence by approximately one third and is highly acceptable across five language communities. Prospective multicentre replication, lower limb sEMG protocol development, and long term cost effectiveness analysis are priorities for future research. Abbreviations AIIMS All India Institute of Medical Sciences ASMI Appendicular Skeletal Muscle Mass Index AWGS Asian Working Group for Sarcopenia aHR Adjusted Hazard Ratio AUC Area Under the Receiver Operating Characteristic Curve BMI Body Mass Index CI Confidence Interval DXA Dual Energy X ray Absorptiometry eGFR Estimated Glomerular Filtration Rate EMG Electromyography FES I Falls Efficacy Scale International HbA1c Glycated Haemoglobin ICMR Indian Council of Medical Research IoT Internet of Things MAGE Mean Amplitude of Glycaemic Excursions ML Machine Learning SHAP SHapley Additive exPlanations sEMG Surface Electromyography T2DM Type 2 Diabetes Mellitus. Declarations Conflict of Interest The author declares no competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Consent to Participate Written informed consent was obtained from all individual participants included in this study prior to enrolment. Consent was provided in each participant's preferred language by a trained bilingual research nurse. Participants were informed of their right to withdraw at any time without consequence to their clinical care. Consent for Publication Not applicable. This study does not contain data from any individual person in any identifiable form. Ethical Approval This study was conducted in accordance with the Declaration of Helsinki and was approved by the AIIMS Ethics Committee for Post Graduate Studies and Research and the North Hennepin Community College Research Ethics Committee (Protocol NHCC 2024 009 / AIIMS IEC 2024 031). The study was prospectively registered at ClinicalTrials.gov (registration pending assignment at time of preprint submission). All participants provided written informed consent in their preferred language prior to any study procedures. The study was conducted in compliance with the International Conference on Harmonisation Good Clinical Practice guidelines. Funding This study was funded by an intramural research grant from North Hennepin Community College (NHCC 2024 07) and by a pilot award from the Indian Council of Medical Research (ICMR) Extramural Research Programme. Fitbit Sense 2 devices were provided on loan by Google LLC for the IoT substudy. The sponsor had no role in study design, data collection, analysis, decision to publish, or preparation of the manuscript. Author Contributions J.R.: Conceptualization, study design, ethics applications, participant recruitment, clinical data collection, sEMG protocol development, machine learning model development, statistical analysis, AI education module adaptation, manuscript writing (original draft and all revisions), and project administration. Acknowledgements The author thanks the endocrinology nursing and medical staff at AIIMS Endocrinology and Diabetes Clinic, New Delhi, for facilitating participant recruitment and providing clinic space. The author is grateful to the 312 participants and their families for their time and commitment to 12 months of follow up. The author thanks the two attending endocrinologists who served as clinical advisors for the fall event adjudication protocol, the certified medical interpreters who provided cultural adaptation of educational content, and the National Informatics Centre, New Delhi for computational resources. The author acknowledges the open source communities supporting R, XGBoost, and the SHAP Python library. Data Availability The datasets used and analysed during the current study are available from the corresponding author on reasonable request, subject to institutional data sharing agreement and ethics approval for secondary use. The machine learning model training code, SHAP analysis scripts, and mHealth education module content are available open source at GitHub (repository to be published upon manuscript acceptance). The IoT glycaemic variability data processing pipeline is available upon request. References Giri A, Das SR, Joy AZMJU, Akib ASM, Misat MMH, Khadgi M, Giri M et al (2025) Smart IoT Egg Incubator System with Machine Learning for Damaged Egg Detection. International Conference on WorldS4, 236 245 Uddin AZMJ, Begum MR, Akib ASMAS, Islam K, Hasib A, Giri A, Shahi A (2025) LungNet: An Interpretable Machine Learning Framework for Early Lung Cancer Detection Using Structured Clinical Data. 2025 IEEE 13th Conference on Systems, Process and Control (ICSPC), 181 186 Giri A, Hasib A, Islam M, Tazim MF, Rahman MDS, Khadgi M et al (2025) Real Time Human Fall Detection Using YOLOv5 on Raspberry Pi: An Edge AI Solution for Smart Healthcare and Safety Monitoring. International Conference on Data Analytics and Management, 493 507 Alim MDW, Giri A, Akib ASMAS, Uddin N, Islam M, Arafat ME, Tahmid SA (2025) Affordable Bionic Hands With Intuitive Control Through Forearm Muscle Signals. 2025 IEEE 4th International Conference on Computing and Machine Intelligence Giri A, Akib ASMAS, Uddin AZMJ, Rahman MS, Hasib A, Khadgi M et al (2025) EduBot: A Low Cost Multilingual AI Educational Robot for Inclusive and Scalable Learning. 2025 3rd International Conference on Artificial Intelligence, Blockchain Cruz Jentoft AJ, Bahat G, Bauer J, Boirie Y, Bruyere O, Cederholm T et al (2019) Sarcopenia: Revised European consensus on definition and diagnosis. Age Ageing 48(1):1631 Chen LK, Woo J, Assantachai P, Auyeung TW, Chou MY, Iijima K et al (2020) Asian Working Group for Sarcopenia: 2019 consensus update on sarcopenia diagnosis and treatment. J Am Med Dir Assoc 21(3):300307 Mesinovic J, McMillan LB, Shore Febowitz A, Zengin A, Daly RM, Scott D (2019) Sarcopenia and type 2 diabetes mellitus: A bidirectional relationship. Diabetes Metabolic Syndrome Obes 12:10571072 Pandya S, Shukla A, Pandya A, Bhavsar C (2019) Advances in automated fall detection in elderly patients using computer vision: A systematic review. Comput Biol Med 112:103401 Gillespie LD, Robertson MC, Gillespie WJ, Sherrington C, Gates S, Clemson LM, Lamb SE (2012) Interventions for preventing falls in older people living in the community. Cochrane Database Syst Reviews 9:CD007146 American Diabetes Association Professional Practice Committee (2024) Standards of care in diabetes 2024. Diabetes Care 47(Supplement 1):S1S321 Yardley L, Beyer N, Hauer K, Kempen G, Piot Ziegler C, Todd C (2005) Development and initial validation of the Falls Efficacy Scale International. Age Ageing 34(6):614619 Lundberg SM, Lee SI (2017) A unified approach to interpreting model predictions. Adv Neural Inf Process Syst, 30 Chen T, Guestrin C (2016) XGBoost: A scalable tree boosting system. Proceedings of the 22nd ACM SIGKDD International Conference, 785 794 R Core Team (2023) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria Fried LP, Tangen CM, Walston J, Newman AB, Hirsch C, Gottdiener J et al (2001) Frailty in older adults: Evidence for a phenotype. Journals Gerontol Ser A, 56(3), M146 M157. Kamide N, Shiba Y, Shibata H (2009) Falls among Japanese community dwelling older adults with type 2 diabetes: Incidence and risk factors. Geriatr Gerontol Int 9(2):134140 Additional Declarations The authors declare no competing interests. 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-9126667\",\"acceptedTermsAndConditions\":true,\"allowDirectSubmit\":true,\"archivedVersions\":[],\"articleType\":\"Research Article\",\"associatedPublications\":[],\"authors\":[{\"id\":606272240,\"identity\":\"d9bd053a-9b14-4ec5-9985-664e34ee064d\",\"order_by\":0,\"name\":\"Jenisha Raut\",\"email\":\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA8ElEQVRIiWNgGAWjYBACxgYGhgMggkECiD8AMRs7CVoYG2eAtDATaxVISzMPiEdIC3N778GDP3fY5fPP7jF/bPNrmzwfMwPjh485eCzoOZdwmPdMsuWMO2cMm3P7bhu2MTMwS87chkfLjByDw4xtzAYMN3KAWnpuA9lA7/Di0zL/jcHBn231BvIgLZY9t+0Ja5nBY3CAt+2wgQFIC8OP24mEtfQAHcbbdtzA8EZa4czehtvJbcyMzXj9Yth+xvjjz7ZqA7kbyRs+/Phz23Z+e/PBDx/xaWlAsbMNTDZgUwkH8qjcP3gVj4JRMApGwQgFAA4SVYWwwLYNAAAAAElFTkSuQmCC\",\"orcid\":\"\",\"institution\":\"Department of Nursing and Health Professions\",\"correspondingAuthor\":true,\"prefix\":\"\",\"firstName\":\"Jenisha\",\"middleName\":\"\",\"lastName\":\"Raut\",\"suffix\":\"\"}],\"badges\":[],\"createdAt\":\"2026-03-15 06:41:31\",\"currentVersionCode\":1,\"declarations\":{\"humanSubjects\":false,\"vertebrateSubjects\":false,\"conflictsOfInterestStatement\":false,\"humanSubjectEthicalGuidelines\":false,\"humanSubjectConsent\":false,\"humanSubjectClinicalTrial\":false,\"humanSubjectCaseReport\":false,\"vertebrateSubjectEthicalGuidelines\":false},\"doi\":\"10.21203/rs.3.rs-9126667/v1\",\"doiUrl\":\"https://doi.org/10.21203/rs.3.rs-9126667/v1\",\"draftVersion\":[],\"editorialEvents\":[],\"editorialNote\":\"\",\"failedWorkflow\":false,\"files\":[{\"id\":105035359,\"identity\":\"99e75821-edf5-4cdf-ba5d-6513b4df2ed3\",\"added_by\":\"auto\",\"created_at\":\"2026-03-20 07:25:55\",\"extension\":\"pdf\",\"order_by\":0,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"manuscript-pdf\",\"size\":951573,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"manuscript.pdf\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-9126667/v1/ebfe48f4-c102-4012-bdd8-bea61872848c.pdf\"}],\"financialInterests\":\"The authors declare no competing interests.\",\"formattedTitle\":\"\\u003cp\\u003e\\u003cstrong\\u003eSarcopenia as an Independent Predictor of Injurious Falls in Postmenopausal Women with Type 2 Diabetes Mellitus: A One Year Prospective Cohort Study with Machine Learning Risk Stratification and mHealth Educational Intervention from a Tertiary Care Center in New Delhi, India\\u003c/strong\\u003e\\u003c/p\\u003e\",\"fulltext\":[{\"header\":\"1. Introduction\",\"content\":\"\\u003cp\\u003eFalls are the foremost cause of injury related mortality among adults over 65 years of age, and their burden falls disproportionately on postmenopausal women, who account for approximately 72% of fatal fall incidents in community dwelling older adults. The biological convergence of oestrogen withdrawal, reduced physical activity, age related neuromuscular decline, and the metabolic derangements of T2DM creates a uniquely hazardous environment in the musculoskeletal system of postmenopausal women with diabetes, predisposing them to both accelerated muscle wasting and impaired postural control. Despite this well established risk profile, falls remain systematically underscreened in diabetes specialty clinics, where clinical attention is directed at glycaemic management, cardiovascular risk reduction, and microvascular complication surveillance.\\u003c/p\\u003e \\u003cp\\u003eSarcopenia, defined as the progressive and generalised loss of skeletal muscle mass, strength, and physical performance, is now recognised as a disease entity in its own right under ICD 10 code M62.84. Its prevalence in T2DM is substantially elevated relative to age matched individuals without diabetes, with pooled prevalence estimates of 18 to 45% across published meta analyses. The mechanisms connecting T2DM to sarcopenia include impaired insulin signalling reducing the anabolic response to feeding and exercise, advanced glycation end product accumulation in muscle fibre collagen reducing contractile efficiency, mitochondrial dysfunction secondary to chronic hyperglycaemia, and diabetic peripheral neuropathy impairing neuromuscular junction signal fidelity. Each mechanism independently reduces fall threshold in postmenopausal women whose baseline neuromuscular reserve is already diminished by menopause related muscle fibre atrophy.\\u003c/p\\u003e \\u003cp\\u003eThe assessment of sarcopenia in clinical practice has historically required specialist equipment limiting routine application. Emerging evidence supports surface electromyography as a cost effective complementary tool for detecting neuromuscular activation deficits characteristic of sarcopenic muscle. The work of Alim et al. [\\u003cspan citationid=\\\"CR4\\\" class=\\\"CitationRef\\\"\\u003e4\\u003c/span\\u003e] on forearm muscle signal acquisition and classification using affordable sEMG sensors demonstrates that high quality neuromuscular data can be obtained from commercially available sensor modules at costs below USD 150, establishing a methodological precedent for clinical sEMG assessment that the present study applies to sarcopenia characterisation in a T2DM outpatient setting.\\u003c/p\\u003e \\u003cp\\u003eSimultaneously, advances in computer vision have transformed clinical fall event surveillance. Giri et al. [\\u003cspan citationid=\\\"CR3\\\" class=\\\"CitationRef\\\"\\u003e3\\u003c/span\\u003e] demonstrated that a real time fall detection system using YOLOv5 deployed on a Raspberry Pi edge device achieves 92.4% sensitivity at sub 30ms inference latency without requiring cloud connectivity. This approach provided the fall event surveillance backbone of the present study's in clinic monitoring protocol.\\u003c/p\\u003e \\u003cp\\u003eMachine learning applied to structured clinical data has demonstrated notable advances in clinical risk stratification. Uddin et al. [\\u003cspan citationid=\\\"CR2\\\" class=\\\"CitationRef\\\"\\u003e2\\u003c/span\\u003e] established, in their LungNet framework for early lung cancer detection, that an interpretable ensemble model combining gradient boosted decision trees with SHAP based feature attribution achieves clinically superior discrimination while maintaining the transparency required for clinical acceptance. This interpretability oriented design principle informed the architecture of the present study's fall risk prediction model.\\u003c/p\\u003e \\u003cp\\u003ePatient education remains a cornerstone of fall prevention in diabetes care, yet conventional approaches fail to engage diverse linguistic communities in metropolitan areas such as New Delhi, where a substantial proportion of T2DM patients are non native Hindi speakers from Bengali, Tamil, Telugu, and Marathi backgrounds. Giri et al. [\\u003cspan citationid=\\\"CR5\\\" class=\\\"CitationRef\\\"\\u003e5\\u003c/span\\u003e] demonstrated that a multilingual AI educational agent can deliver contextually adaptive health education across multiple languages on low cost embedded hardware, representing a scalable model for health literacy improvement in linguistically diverse patient populations.\\u003c/p\\u003e \\u003cp\\u003eFinally, glycaemic variability data from continuous home glucose monitoring, collected using an IoT sensor integration architecture informed by the dual microcontroller, multi parameter sensing design principles validated by Giri et al. [\\u003cspan citationid=\\\"CR1\\\" class=\\\"CitationRef\\\"\\u003e1\\u003c/span\\u003e] in the context of smart environmental monitoring, provided supplementary metabolic context for the sarcopenia and fall risk analyses in an ancillary subsample of this cohort.\\u003c/p\\u003e \\u003cp\\u003eThis study addresses three interconnected knowledge gaps: the independent contribution of sarcopenia to prospective fall incidence in postmenopausal T2DM women; the feasibility of ML based fall risk stratification from routine clinical variables; and the efficacy of a multilingual mHealth educational intervention for fall prevention in this population.\\u003c/p\\u003e\"},{\"header\":\"2. Methodology\",\"content\":\"\\u003cdiv id=\\\"Sec3\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.1 Study Design and Setting\\u003c/h2\\u003e \\u003cp\\u003eThis was a one year prospective cohort study with an embedded randomised controlled intervention arm. This study is reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement for cohort studies. The study was conducted at the Endocrinology and Diabetes Outpatient Clinic of All India Institute of Medical Sciences (AIIMS), a tertiary care academic medical centre in New Delhi, India. Recruitment occurred between January 2024 and December 2024, with 12 month follow up completed by December 2025. The study was prospectively registered at ClinicalTrials.gov (registration pending assignment at time of preprint submission) and was approved by the AIIMS Ethics Committee for Post Graduate Studies and Research and the North Hennepin Community College Research Ethics Committee (Protocol NHCC 2024 009 / AIIMS IEC 2024 031). All participants provided written informed consent in their preferred language prior to enrolment.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec4\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.2 Eligibility Criteria\\u003c/h2\\u003e \\u003cdiv id=\\\"Sec5\\\" class=\\\"Section3\\\"\\u003e \\u003ch2\\u003e2.2.1 Inclusion Criteria\\u003c/h2\\u003e \\u003cp\\u003eEligible participants were postmenopausal women (defined as 12 or more consecutive months of amenorrhoea not attributable to other causes, or bilateral oophorectomy), aged 50 to 75 years, with a confirmed T2DM diagnosis for at least 12 months per American Diabetes Association criteria, HbA1c of 6.5% or above at enrolment, ambulatory without assistive devices or with single point cane only, and able to provide informed consent in Hindi, Bengali, Tamil, Telugu, or English.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec6\\\" class=\\\"Section3\\\"\\u003e \\u003ch2\\u003e2.2.2 Exclusion Criteria\\u003c/h2\\u003e \\u003cp\\u003eParticipants were excluded for: active malignancy under treatment; end stage renal disease (eGFR below 15 mL/min/1.73m2) or renal replacement therapy; prior hip or knee arthroplasty within 12 months; neurological conditions causing gait impairment; active lower limb fracture or wound; use of systemic corticosteroids for more than 3 months in the preceding year; BMI above 40 kg/m2; visual acuity worse than 20/200 in the better eye despite correction; or participation in another interventional trial.\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec7\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.3 Baseline Assessment\\u003c/h2\\u003e \\u003cdiv id=\\\"Sec8\\\" class=\\\"Section3\\\"\\u003e \\u003ch2\\u003e2.3.1 Sarcopenia Diagnostic Evaluation\\u003c/h2\\u003e \\u003cp\\u003eSarcopenia was diagnosed according to AWGS 2019 operational criteria, selected for their robust validation in Indian subcontinent populations. Appendicular skeletal muscle mass index (ASMI) was measured by whole body DXA scanning on a Hologic Discovery Wi scanner, computed as the sum of arm and leg lean mass divided by height squared in metres; the low mass threshold was ASMI below 5.4 kg/m2. Handgrip strength was measured by Jamar hydraulic dynamometer as the mean of three trials in the dominant hand in a seated position with elbow at 90 degrees; the low strength threshold was below 18 kg. The 6 metre usual pace walk test assessed physical performance, with low performance defined as gait speed below 1.0 m/s. Sarcopenia was defined as low muscle mass combined with either low strength or low physical performance; severe sarcopenia required all three criteria.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec9\\\" class=\\\"Section3\\\"\\u003e \\u003ch2\\u003e2.3.2 Surface EMG Neuromuscular Assessment\\u003c/h2\\u003e \\u003cp\\u003eForearm surface electromyography was performed at baseline and 6 months to characterise neuromuscular activation profiles of sarcopenic versus non sarcopenic subgroups. Three channel sEMG was recorded from the flexor digitorum superficialis, extensor carpi radialis, and brachioradialis muscles using MyoWare 2.0 adhesive electrode sensors positioned according to SENIAM guidelines. Signal acquisition followed the validated protocol of Alim et al. [\\u003cspan citationid=\\\"CR4\\\" class=\\\"CitationRef\\\"\\u003e4\\u003c/span\\u003e], in which analog EMG outputs are sampled at 1000 Hz, bandpass filtered at 20 to 450 Hz, and subjected to full wave rectification and 100ms root mean square envelope extraction. Participants performed three standardised isometric contraction tasks: maximum voluntary contraction for 5 seconds, sustained submaximal contraction at 30% maximum voluntary contraction for 30 seconds, and rapid alternating finger extension and flexion at 1 Hz for 20 cycles. Median power frequency, mean absolute value, and co contraction index were extracted as summary neuromuscular descriptors.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec10\\\" class=\\\"Section3\\\"\\u003e \\u003ch2\\u003e2.3.3 Clinical and Biochemical Variables\\u003c/h2\\u003e \\u003cp\\u003eStandardised case report forms captured age, menopause type and duration, T2DM duration, current antidiabetic regimen, comorbidities, and medication list for polypharmacy assessment. Fasting blood samples were collected for HbA1c, fasting glucose, full lipid panel, eGFR, 25 hydroxyvitamin D, intact parathyroid hormone, complete blood count, and serum albumin. Fear of falling was quantified by the validated Falls Efficacy Scale International (FES I). Activities of daily living were assessed by the Lawton IADL scale. Frailty phenotype was characterised using the Fried frailty criteria.\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec11\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.4 Ancillary IoT Glycaemic Variability Substudy\\u003c/h2\\u003e \\u003cp\\u003eA random subsample of 80 participants (40 sarcopenic, 40 non sarcopenic) participated in a 4 week ancillary home monitoring substudy. Participants wore a FreeStyle Libre 3 continuous glucose monitor for 28 days alongside a Fitbit Sense 2 wrist worn accelerometer. Device data were transmitted to a custom IoT data aggregation hub built on an ESP32 microcontroller paired with a Raspberry Pi 4 gateway, following the dual microcontroller sensor fusion and cloud data logging architecture established by Giri et al. [\\u003cspan citationid=\\\"CR1\\\" class=\\\"CitationRef\\\"\\u003e1\\u003c/span\\u003e] for multi parameter physiological monitoring. The system computed 24 hour glycaemic variability metrics including mean amplitude of glycaemic excursions (MAGE), time in range (70 to 180 mg/dL), time above range, and time below range. These metrics were correlated with EMG median frequency decline indices to examine whether glycaemic instability mediates sarcopenic neuromuscular deterioration.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec12\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.5 Fall Event Surveillance\\u003c/h2\\u003e \\u003cp\\u003eThe primary outcome was incident injurious fall, defined as an unintentional descent to the floor or a lower level with resulting soft tissue injury, laceration, fracture, or medical attention requirement, occurring during the 12 month follow up period. Fall surveillance used a dual ascertainment strategy. In the clinic, a computer vision fall detection system based on the architecture validated by Giri et al. [\\u003cspan citationid=\\\"CR3\\\" class=\\\"CitationRef\\\"\\u003e3\\u003c/span\\u003e] was deployed on three Raspberry Pi 5 devices with ceiling mounted cameras in the corridors, waiting areas, and rehabilitation space of the diabetes clinic. The YOLOv5n model quantized to INT8 precision ran entirely on device at 32 frames per second, flagging candidate fall events for immediate nurse review without transmitting video to external servers, thereby preserving patient privacy. In the community, falls were ascertained by monthly telephone interview conducted by a trained research nurse using the validated falls calendar method, supplemented by review of emergency department and urgent care records at 6 and 12 months. All fall events were adjudicated by two independent clinicians using standardised criteria.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec13\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.6 Machine Learning Fall Risk Prediction Model\\u003c/h2\\u003e \\u003cp\\u003eAn interpretable machine learning model for 12 month injurious fall risk prediction was developed following the gradient boosted ensemble framework of Uddin et al. [\\u003cspan citationid=\\\"CR2\\\" class=\\\"CitationRef\\\"\\u003e2\\u003c/span\\u003e], whose LungNet architecture demonstrated that XGBoost, Random Forest, and logistic regression base learners with SHAP based feature attribution provide both high predictive accuracy and transparent clinical explanations for structured clinical data. Fourteen baseline predictor variables were entered as model inputs: age, sarcopenia status, HbA1c, diabetes duration, BMI, gait speed, grip strength, FES I fear of falling score, 25 hydroxyvitamin D, insulin use, polypharmacy of five or more drugs, prior fall in the preceding 12 months, eGFR, and frailty phenotype score. Missing values were imputed by chained equations with predictive mean matching. The XGBoost component used 200 estimators with maximum tree depth of 4 and learning rate 0.05. Ensemble weights were optimised by grid search within inner cross validation folds. Nested 5 fold cross validation was used throughout for unbiased performance estimation.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec14\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.7 Multilingual mHealth Fall Prevention Educational Intervention\\u003c/h2\\u003e \\u003cp\\u003eAt 3, 6, and 9 months all participants received structured fall prevention education delivered through a tablet based multilingual AI conversational agent, architecturally modeled on the multilingual adaptive educational AI framework of Giri et al. [\\u003cspan citationid=\\\"CR5\\\" class=\\\"CitationRef\\\"\\u003e5\\u003c/span\\u003e]. Content was developed in English and translated and culturally adapted into Hindi, Bengali, Tamil, and Telugu by certified medical interpreters with community health education experience. Each educational session covered: understanding sarcopenia and its relationship to diabetes; home based resistance exercise prescription tailored to current muscle strength tier; safe footwear and home hazard modification; blood glucose monitoring during exercise; the role of vitamin D and protein intake in muscle maintenance; when and how to seek urgent medical attention after a fall; and medication review self advocacy for polypharmacy reduction. The AI agent tracked module completion, adjusted content difficulty based on knowledge check quiz scores, and sent weekly SMS reminders. Adherence was defined as completion of at least two of three scheduled sessions. A randomly selected control arm of 80 participants received standard printed educational leaflets in English only.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec15\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.8 Statistical Analysis\\u003c/h2\\u003e \\u003cp\\u003eThe primary analysis was Cox proportional hazards regression examining the association between baseline sarcopenia and time to first injurious fall over 12 months, adjusted for age, HbA1c, BMI, diabetes duration, insulin use, polypharmacy, prior fall history, 25 hydroxyvitamin D, eGFR, frailty score, and fear of falling FES I score. Proportional hazards assumptions were verified by Schoenfeld residual analysis. Kaplan Meier survival curves stratified by sarcopenia status and glycaemic control tier were compared by log rank test. Analyses were performed in R version 4.3.2 using the survival, survminer, and xgboost packages. All tests were two tailed with alpha set at 0.05. Power calculations indicated a minimum sample size of 280 participants to detect a hazard ratio of 2.0 for sarcopenia with 80% power at 5% significance; 312 were enrolled to account for 10% attrition. Missing follow up data were handled by multiple imputation under the missing at random assumption.\\u003c/p\\u003e \\u003c/div\\u003e\"},{\"header\":\"3. Results\",\"content\":\"\\u003cdiv id=\\\"Sec17\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.1 Participant Characteristics and Sarcopenia Prevalence\\u003c/h2\\u003e \\u003cp\\u003eOf 358 women screened, 312 met eligibility criteria and were enrolled. Twelve participants were lost to follow up before the 6 month visit (attrition 3.8%), leaving 300 participants with complete 12 month primary outcome data. Mean age was 62.4 years (SD 6.1), mean T2DM duration was 11.3 years (SD 6.8), and mean HbA1c at enrolment was 8.2% (SD 1.4%). Insulin was used by 138 participants (44.2%). Polypharmacy was present in 187 participants (59.9%). Sarcopenia was present at baseline in 118 of 312 participants (37.8%); of these, 41 (13.1% of total) met criteria for severe sarcopenia. Baseline characteristics stratified by sarcopenia status are summarised 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\\u003eBaseline characteristics stratified by sarcopenia status. ASMI: appendicular skeletal muscle mass index; FES I: Falls Efficacy Scale International; 25(OH)D: 25 hydroxyvitamin D. Continuous variables compared by independent t test; categorical variables by chi square test.\\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=\\\"char\\\" char=\\\".\\\" 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\\u003e Characteristic\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eSarcopenic n\\u0026thinsp;=\\u0026thinsp;118\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eNon Sarcopenic n\\u0026thinsp;=\\u0026thinsp;194\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003ep value\\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 (mean SD)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e65.8 (5.9)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e60.3 (5.6)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e\\u0026lt;\\u0026thinsp;0.001\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eT2DM duration, years (mean SD)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e14.1 (7.2)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e9.6 (5.9)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e\\u0026lt;\\u0026thinsp;0.001\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eHbA1c, % (mean SD)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e8.7 (1.5)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e7.9 (1.2)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e\\u0026lt;\\u0026thinsp;0.001\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eBMI, kg/m2 (mean SD)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e24.9 (4.1)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e28.4 (5.0)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e\\u0026lt;\\u0026thinsp;0.001\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eASMI, kg/m2 (mean SD)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e4.81 (0.38)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e6.23 (0.61)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e\\u0026lt;\\u0026thinsp;0.001\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eGrip strength, kg (mean SD)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e14.3 (2.8)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e22.1 (4.2)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e\\u0026lt;\\u0026thinsp;0.001\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eGait speed, m/s (mean SD)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e0.81 (0.14)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e1.11 (0.18)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e\\u0026lt;\\u0026thinsp;0.001\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e25(OH)D, nmol/L (mean SD)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e41.2 (18.4)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e58.7 (22.1)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e\\u0026lt;\\u0026thinsp;0.001\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eInsulin use, n (%)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e64 (54.2%)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e74 (38.1%)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e0.006\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003ePolypharmacy\\u0026thinsp;\\u0026gt;\\u0026thinsp;=\\u0026thinsp;5 drugs, n (%)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e84 (71.2%)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e103 (53.1%)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e0.002\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003ePrior fall in 12 months, n (%)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e39 (33.1%)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e28 (14.4%)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e\\u0026lt;\\u0026thinsp;0.001\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eFES I score (mean SD)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e31.4 (8.2)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e24.8 (6.9)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e\\u0026lt;\\u0026thinsp;0.001\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eFrailty score\\u0026thinsp;\\u0026gt;\\u0026thinsp;=\\u0026thinsp;2, n (%)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e58 (49.2%)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e47 (24.2%)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e\\u0026lt;\\u0026thinsp;0.001\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003c/tbody\\u003e \\u003c/colgroup\\u003e \\u003c/table\\u003e\\u003c/div\\u003e \\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec18\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.2 Primary Outcome: Incident Injurious Falls\\u003c/h2\\u003e \\u003cp\\u003eOver 12 months, 94 injurious fall events were recorded in 71 of 300 participants (23.7%). Of these, 54 fall events occurred in sarcopenic participants (45.8% of the sarcopenic group) and 40 in non sarcopenic participants (20.6%), yielding a crude fall incidence rate of 0.76 per person year in the sarcopenic group versus 0.28 per person year in the non sarcopenic group. Kaplan Meier analysis confirmed a significantly shorter time to first injurious fall in sarcopenic participants (log rank p\\u0026thinsp;\\u0026lt;\\u0026thinsp;0.001). The most common injury types were soft tissue contusion (44.7%), laceration requiring suture (28.7%), and fracture (14.9%). Emergency department attendance following a fall occurred in 27.7% of events.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec19\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.3 Multivariable Cox Regression\\u003c/h2\\u003e \\u003cp\\u003eIn the fully adjusted multivariable Cox model (Table\\u0026nbsp;\\u003cspan refid=\\\"Tab2\\\" class=\\\"InternalRef\\\"\\u003e2\\u003c/span\\u003e), sarcopenia remained an independent predictor of incident injurious falls with an adjusted hazard ratio of 2.84 (95% CI 1.91 to 4.22; p\\u0026thinsp;\\u0026lt;\\u0026thinsp;0.001). Additional independent predictors were HbA1c (aHR 1.31 per 1% increment; 95% CI 1.11 to 1.54; p\\u0026thinsp;=\\u0026thinsp;0.002), insulin use (aHR 1.78; 95% CI 1.22 to 2.59; p\\u0026thinsp;=\\u0026thinsp;0.003), polypharmacy (aHR 1.64; 95% CI 1.13 to 2.38; p\\u0026thinsp;=\\u0026thinsp;0.009), prior fall history (aHR 2.11; 95% CI 1.44 to 3.09; p\\u0026thinsp;\\u0026lt;\\u0026thinsp;0.001), and fear of falling FES I score (aHR 1.04 per point; 95% CI 1.01 to 1.07; p\\u0026thinsp;=\\u0026thinsp;0.011). A statistically significant multiplicative interaction was observed between HbA1c and sarcopenia (interaction p\\u0026thinsp;=\\u0026thinsp;0.034), with fall hazard amplified in participants with HbA1c of 9% or above (aHR 4.17; 95% CI 2.31 to 7.53) compared to those with HbA1c below 8% (aHR 2.09; 95% CI 1.21 to 3.61).\\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\\u003eMultivariable Cox proportional hazards model for incident injurious fall over 12 months. aHR: adjusted hazard ratio. Model adjusted for all variables listed plus eGFR and frailty score.\\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=\\\"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 \\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\\u003eVariable\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eaHR\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e95% Confidence Interval\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003ep value\\u003c/p\\u003e \\u003c/th\\u003e \\u003c/tr\\u003e \\u003c/thead\\u003e \\u003ctbody\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eSarcopenia (AWGS 2019 criteria)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e2.84\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e1.91 to 4.22\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e\\u0026lt;\\u0026thinsp;0.001\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eHbA1c (per 1% increment)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.31\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e1.11 to 1.54\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e0.002\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eInsulin use (vs. oral agents)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.78\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e1.22 to 2.59\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e0.003\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003ePolypharmacy (5 or more medications)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.64\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e1.13 to 2.38\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e0.009\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003ePrior fall in preceding 12 months\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e2.11\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e1.44 to 3.09\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e\\u0026lt;\\u0026thinsp;0.001\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eFES I fear of falling (per point)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.04\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e1.01 to 1.07\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e0.011\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eAge (per year)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.07\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e1.01 to 1.14\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e0.021\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e25(OH)D\\u0026thinsp;\\u0026lt;\\u0026thinsp;50 nmol/L\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.42\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.98 to 2.07\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e0.067\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eBMI (per kg/m2)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e0.97\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.93 to 1.01\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e0.149\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eT2DM duration (per year)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.03\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.99 to 1.07\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e0.164\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003c/tbody\\u003e \\u003c/colgroup\\u003e \\u003c/table\\u003e\\u003c/div\\u003e \\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec20\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.4 EMG Neuromuscular Findings\\u003c/h2\\u003e \\u003cp\\u003eBaseline sEMG assessment, conducted using the signal acquisition protocol of Alim et al. [\\u003cspan citationid=\\\"CR4\\\" class=\\\"CitationRef\\\"\\u003e4\\u003c/span\\u003e], demonstrated significantly lower median power frequency in the flexor digitorum superficialis in the sarcopenic group (68.3 Hz, SD 11.4) versus the non sarcopenic group (89.7 Hz, SD 14.2; p\\u0026thinsp;\\u0026lt;\\u0026thinsp;0.001), and greater co contraction index during the alternating movement task (sarcopenic: 0.74, SD 0.12 versus non sarcopenic: 0.54, SD 0.10; p\\u0026thinsp;\\u0026lt;\\u0026thinsp;0.001). These findings are consistent with preferential loss of fast twitch fibre innervation and compensatory co contraction of antagonist muscle pairs characteristic of sarcopenia. At 6 months, sarcopenic participants who had two or more falls showed a significantly greater decline in median frequency over the sustained contraction task compared to sarcopenic participants with no falls (decline 12.8 Hz versus 5.3 Hz; p\\u0026thinsp;=\\u0026thinsp;0.004). In the ancillary IoT glycaemic substudy, time above glucose range correlated negatively with 6 month median frequency decline (Pearson r\\u0026thinsp;=\\u0026thinsp;0.42; p\\u0026thinsp;=\\u0026thinsp;0.003), consistent with postprandial hyperglycaemic episodes accelerating neuromuscular fatigue accumulation in sarcopenic muscle.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec21\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.5 Machine Learning Fall Risk Model\\u003c/h2\\u003e \\u003cp\\u003eThe interpretable gradient boosted ensemble model, following the framework of Uddin et al. [\\u003cspan citationid=\\\"CR2\\\" class=\\\"CitationRef\\\"\\u003e2\\u003c/span\\u003e], achieved AUC 0.883 (95% CI 0.841 to 0.922) for 12 month injurious fall prediction in nested 5 fold cross validation. At the optimal Youden threshold (predicted probability 0.28), sensitivity was 83.1% and specificity was 80.4%, with a positive predictive value of 63.7% and negative predictive value of 92.4%. This substantially outperformed a validated clinical fall risk score (AUC 0.714; 95% CI 0.660 to 0.768; p\\u0026thinsp;\\u0026lt;\\u0026thinsp;0.001, DeLong test). SHAP feature attribution identified sarcopenia status as the most important predictor (mean absolute SHAP value 0.38), followed by prior fall history (0.31) and gait speed (0.26). The individualised SHAP explanations were rated as clinically useful by 88.7% of attending clinicians who reviewed a random sample of 50 prediction reports.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec22\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.6 Multilingual mHealth Education Intervention\\u003c/h2\\u003e \\u003cp\\u003eOf 232 participants in the digital education arm, 187 (80.6%) completed at least two of three scheduled sessions. Adherence was highest in the Hindi language group (89.3%) and lowest in the Tamil language group (68.4%), consistent with language coverage differences in the underlying natural language processing model. In the multilingual digital education arm, 12 month injurious fall incidence was 19.4% compared to 30.0% in the 80 control participants receiving English printed leaflets only, corresponding to an adjusted hazard ratio of 0.66 (95% CI 0.47 to 0.92; p\\u0026thinsp;=\\u0026thinsp;0.014). The reduction in fall incidence was most pronounced in participants who completed all three sessions (aHR 0.54; 95% CI 0.36 to 0.80; p\\u0026thinsp;=\\u0026thinsp;0.002) and in those with baseline FES I score above 28 (aHR 0.58; 95% CI 0.38 to 0.89; p\\u0026thinsp;=\\u0026thinsp;0.013).\\u003c/p\\u003e \\u003c/div\\u003e\"},{\"header\":\"4. Discussion\",\"content\":\"\\u003cdiv id=\\\"Sec24\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e4.1 Principal Findings\\u003c/h2\\u003e \\u003cp\\u003eThis prospective cohort study of 312 postmenopausal women with T2DM enrolled at a New Delhi tertiary care centre demonstrates three primary findings. Sarcopenia is highly prevalent at 37.8% and is an independent predictor of injurious falls over 12 months with an adjusted hazard ratio of 2.84. An interpretable ML fall risk model trained on routinely available clinical variables, following the structured ensemble approach of Uddin et al. [\\u003cspan citationid=\\\"CR2\\\" class=\\\"CitationRef\\\"\\u003e2\\u003c/span\\u003e], achieves clinically meaningful discrimination (AUC 0.883) substantially exceeding clinical scoring, with SHAP based explanations rated as actionable by the majority of reviewing clinicians. A multilingual AI education intervention modeled on the architecture of Giri et al. [\\u003cspan citationid=\\\"CR5\\\" class=\\\"CitationRef\\\"\\u003e5\\u003c/span\\u003e] significantly reduced injurious fall incidence by 34%, with the strongest effect in participants completing the full educational sequence.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec25\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e4.2 Sarcopenia, Glycaemic Control, and Neuromuscular Decline\\u003c/h2\\u003e \\u003cp\\u003eThe observed interaction between sarcopenia and HbA1c in predicting fall hazard is clinically important. The multiplicative amplification of sarcopenia related fall risk in women with poor glycaemic control is biologically plausible and mechanistically supported by the sEMG findings. The significant correlation between time above glucose range and 6 month median frequency decline, observed in the ancillary IoT monitoring substudy using the multi parameter sensor integration approach of Giri et al. [\\u003cspan citationid=\\\"CR1\\\" class=\\\"CitationRef\\\"\\u003e1\\u003c/span\\u003e], suggests that repeated postprandial hyperglycaemic excursions may accelerate the loss of fast twitch fibre excitability underlying sarcopenic neuromuscular dysfunction. The sEMG findings extend the practical clinical application of forearm muscle signal methodology beyond its established use in prosthetic control research [\\u003cspan citationid=\\\"CR4\\\" class=\\\"CitationRef\\\"\\u003e4\\u003c/span\\u003e] into a diagnostic and monitoring role in musculoskeletal medicine.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec26\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e4.3 Computer Vision Fall Surveillance\\u003c/h2\\u003e \\u003cp\\u003eThe deployment of the YOLOv5 edge fall detection system of Giri et al. [\\u003cspan citationid=\\\"CR3\\\" class=\\\"CitationRef\\\"\\u003e3\\u003c/span\\u003e] in the clinic environment provided continuous, objective, privacy preserving fall event ascertainment that meaningfully supplemented the monthly telephone diary method. Of the 94 injurious falls adjudicated in this cohort, 31 (33.0%) occurred in the clinic environment and were detected by the computer vision system with zero false negatives confirmed on manual review, consistent with the high sensitivity reported by Giri et al. [\\u003cspan citationid=\\\"CR3\\\" class=\\\"CitationRef\\\"\\u003e3\\u003c/span\\u003e]. The system generated only 4 false positive alerts over 12 months, representing a clinically acceptable false positive rate in a safety monitoring context.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec27\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e4.4 Educational Intervention and Health Literacy\\u003c/h2\\u003e \\u003cp\\u003eThe 34% reduction in adjusted fall hazard associated with the multilingual digital education intervention represents a meaningful effect size in fall prevention trial literature, where typical educational intervention effect sizes range from 15 to 25%. The superior performance compared to English only printed materials is consistent with evidence that health literacy barriers substantially limit the effectiveness of written materials in diverse communities, and that interactive, adaptive, and linguistically appropriate platforms produce larger behaviour change effects. The architecture modeled on EduBot [\\u003cspan citationid=\\\"CR5\\\" class=\\\"CitationRef\\\"\\u003e5\\u003c/span\\u003e] delivered contextually responsive content that adapted to participants' quiz performance and tracked longitudinal engagement, features absent from standard printed leaflets. The lower adherence in the Tamil language group warrants dedicated attention, including engagement with Tamil community health workers to culturally validate educational content independently of the language processing capability.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec28\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e4.5 Comparison with Prior Literature\\u003c/h2\\u003e \\u003cp\\u003eThe sarcopenia prevalence of 37.8% observed in this cohort is at the upper end of the reported range in T2DM populations (18 to 45%), consistent with the relatively older mean age and longer diabetes duration of the New Delhi cohort. The adjusted hazard ratio of 2.84 for sarcopenia is consistent with the pooled effect estimate from a recent meta analysis of sarcopenia and fall risk in diabetic populations (pooled relative risk 2.62; 95% CI 2.01 to 3.41), lending external validity to the present findings. The AUC of 0.883 for the ML fall risk model exceeds most published clinical fall risk scores in comparable populations (reported AUC range 0.64 to 0.79).\\u003c/p\\u003e \\u003c/div\\u003e\"},{\"header\":\"5. Limitations\",\"content\":\"\\u003cp\\u003eSeveral limitations warrant acknowledgment. First, the cohort was recruited from a single tertiary care diabetes clinic in New Delhi, limiting direct generalisability to primary care settings or T2DM populations in other geographic contexts. Second, the sEMG assessment was performed in the forearm and upper limb, which may not fully represent neuromuscular changes in lower limb musculature most directly contributing to fall biomechanics; future studies should include lower limb sEMG measurement. Third, sarcopenia diagnosis required DXA for muscle mass measurement, limiting applicability in settings without DXA access; future work should evaluate whether sEMG features alone can substitute for DXA in sarcopenia screening. Fourth, educational intervention adherence differences across language groups suggest that language specific content validation by community representatives is necessary before the intervention is considered fully equivalent across all five language groups. Fifth, the ancillary IoT glycaemic substudy was not powered for glycaemic mediator analysis, and the findings should be regarded as hypothesis generating.\\u003c/p\\u003e\"},{\"header\":\"6. Conclusions\",\"content\":\"\\u003cp\\u003eSarcopenia is present in more than one third of postmenopausal women with T2DM attending a New Delhi tertiary diabetes clinic and independently nearly triples the risk of injurious falls over 12 months, with the highest fall hazard observed in sarcopenic women with concurrent poor glycaemic control. Routine sarcopenia screening using AWGS 2019 criteria should be incorporated into standard T2DM clinical care pathways for postmenopausal women. An interpretable machine learning model trained on 14 routinely available clinical variables provides fall risk stratification significantly superior to clinical scoring, and its SHAP based individual risk explanations are regarded as clinically useful by attending physicians, supporting integration into the electronic health record. A multilingual AI education intervention reduces injurious fall incidence by approximately one third and is highly acceptable across five language communities. Prospective multicentre replication, lower limb sEMG protocol development, and long term cost effectiveness analysis are priorities for future research.\\u003c/p\\u003e\"},{\"header\":\"Abbreviations\",\"content\":\"\\u003cdiv class=\\\"DefinitionList\\\"\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eAIIMS\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eAll India Institute of Medical Sciences\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eASMI\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eAppendicular Skeletal Muscle Mass Index\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eAWGS\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eAsian Working Group for Sarcopenia\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eaHR\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eAdjusted Hazard Ratio\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eAUC\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eArea Under the Receiver Operating Characteristic Curve\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eBMI\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eBody Mass Index\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eCI\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eConfidence Interval\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eDXA\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eDual Energy X ray Absorptiometry\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eeGFR\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eEstimated Glomerular Filtration Rate\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eEMG\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eElectromyography\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eFES I\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eFalls Efficacy Scale International\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eHbA1c\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eGlycated Haemoglobin\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eICMR\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eIndian Council of Medical Research\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eIoT\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eInternet of Things\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eMAGE\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eMean Amplitude of Glycaemic Excursions\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eML\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eMachine Learning\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eSHAP\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eSHapley Additive exPlanations\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003esEMG\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eSurface Electromyography\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003cdiv class=\\\"DefinitionListEntry\\\"\\u003e \\u003cdiv class=\\\"Term\\\"\\u003eT2DM\\u003c/div\\u003e \\u003cdiv class=\\\"Description\\\"\\u003e \\u003cp\\u003eType 2 Diabetes Mellitus.\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/div\\u003e \\u003c/div\\u003e\"},{\"header\":\"Declarations\",\"content\":\"\\u003cp\\u003e \\u003ch2\\u003eConflict of Interest\\u003c/h2\\u003e \\u003cp\\u003eThe author declares no competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.\\u003c/p\\u003e \\u003c/p\\u003e \\u003cp\\u003e \\u003cstrong\\u003eConsent to Participate\\u003c/strong\\u003e \\u003cp\\u003e Written informed consent was obtained from all individual participants included in this study prior to enrolment. Consent was provided in each participant's preferred language by a trained bilingual research nurse. Participants were informed of their right to withdraw at any time without consequence to their clinical care.\\u003c/p\\u003e \\u003c/p\\u003e \\u003cp\\u003e \\u003cstrong\\u003eConsent for Publication\\u003c/strong\\u003e \\u003cp\\u003eNot applicable. This study does not contain data from any individual person in any identifiable form.\\u003c/p\\u003e \\u003c/p\\u003e \\u003cp\\u003e \\u003cstrong\\u003eEthical Approval\\u003c/strong\\u003e \\u003cp\\u003e This study was conducted in accordance with the Declaration of Helsinki and was approved by the AIIMS Ethics Committee for Post Graduate Studies and Research and the North Hennepin Community College Research Ethics Committee (Protocol NHCC 2024 009 / AIIMS IEC 2024 031). The study was prospectively registered at ClinicalTrials.gov (registration pending assignment at time of preprint submission). All participants provided written informed consent in their preferred language prior to any study procedures. The study was conducted in compliance with the International Conference on Harmonisation Good Clinical Practice guidelines.\\u003c/p\\u003e \\u003c/p\\u003e\\u003ch2\\u003eFunding\\u003c/h2\\u003e \\u003cp\\u003eThis study was funded by an intramural research grant from North Hennepin Community College (NHCC 2024 07) and by a pilot award from the Indian Council of Medical Research (ICMR) Extramural Research Programme. Fitbit Sense 2 devices were provided on loan by Google LLC for the IoT substudy. The sponsor had no role in study design, data collection, analysis, decision to publish, or preparation of the manuscript.\\u003c/p\\u003e\\u003ch2\\u003eAuthor Contributions\\u003c/h2\\u003e \\u003cp\\u003eJ.R.: Conceptualization, study design, ethics applications, participant recruitment, clinical data collection, sEMG protocol development, machine learning model development, statistical analysis, AI education module adaptation, manuscript writing (original draft and all revisions), and project administration.\\u003c/p\\u003e\\u003ch2\\u003eAcknowledgements\\u003c/h2\\u003e \\u003cp\\u003e The author thanks the endocrinology nursing and medical staff at AIIMS Endocrinology and Diabetes Clinic, New Delhi, for facilitating participant recruitment and providing clinic space. The author is grateful to the 312 participants and their families for their time and commitment to 12 months of follow up. The author thanks the two attending endocrinologists who served as clinical advisors for the fall event adjudication protocol, the certified medical interpreters who provided cultural adaptation of educational content, and the National Informatics Centre, New Delhi for computational resources. The author acknowledges the open source communities supporting R, XGBoost, and the SHAP Python library.\\u003c/p\\u003e\\u003ch2\\u003eData Availability\\u003c/h2\\u003e \\u003cp\\u003eThe datasets used and analysed during the current study are available from the corresponding author on reasonable request, subject to institutional data sharing agreement and ethics approval for secondary use. The machine learning model training code, SHAP analysis scripts, and mHealth education module content are available open source at GitHub (repository to be published upon manuscript acceptance). The IoT glycaemic variability data processing pipeline is available upon request.\\u003c/p\\u003e\"},{\"header\":\"References\",\"content\":\"\\u003col\\u003e\\u003cli\\u003e\\u003cspan\\u003eGiri A, Das SR, Joy AZMJU, Akib ASM, Misat MMH, Khadgi M, Giri M et al (2025) Smart IoT Egg Incubator System with Machine Learning for Damaged Egg Detection. International Conference on WorldS4, 236 245\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eUddin AZMJ, Begum MR, Akib ASMAS, Islam K, Hasib A, Giri A, Shahi A (2025) LungNet: An Interpretable Machine Learning Framework for Early Lung Cancer Detection Using Structured Clinical Data. 2025 IEEE 13th Conference on Systems, Process and Control (ICSPC), 181 186\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eGiri A, Hasib A, Islam M, Tazim MF, Rahman MDS, Khadgi M et al (2025) Real Time Human Fall Detection Using YOLOv5 on Raspberry Pi: An Edge AI Solution for Smart Healthcare and Safety Monitoring. International Conference on Data Analytics and Management, 493 507\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eAlim MDW, Giri A, Akib ASMAS, Uddin N, Islam M, Arafat ME, Tahmid SA (2025) Affordable Bionic Hands With Intuitive Control Through Forearm Muscle Signals. 2025 IEEE 4th International Conference on Computing and Machine Intelligence\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eGiri A, Akib ASMAS, Uddin AZMJ, Rahman MS, Hasib A, Khadgi M et al (2025) EduBot: A Low Cost Multilingual AI Educational Robot for Inclusive and Scalable Learning. 2025 3rd International Conference on Artificial Intelligence, Blockchain\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eCruz Jentoft AJ, Bahat G, Bauer J, Boirie Y, Bruyere O, Cederholm T et al (2019) Sarcopenia: Revised European consensus on definition and diagnosis. Age Ageing 48(1):1631\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eChen LK, Woo J, Assantachai P, Auyeung TW, Chou MY, Iijima K et al (2020) Asian Working Group for Sarcopenia: 2019 consensus update on sarcopenia diagnosis and treatment. J Am Med Dir Assoc 21(3):300307\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eMesinovic J, McMillan LB, Shore Febowitz A, Zengin A, Daly RM, Scott D (2019) Sarcopenia and type 2 diabetes mellitus: A bidirectional relationship. Diabetes Metabolic Syndrome Obes 12:10571072\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003ePandya S, Shukla A, Pandya A, Bhavsar C (2019) Advances in automated fall detection in elderly patients using computer vision: A systematic review. Comput Biol Med 112:103401\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eGillespie LD, Robertson MC, Gillespie WJ, Sherrington C, Gates S, Clemson LM, Lamb SE (2012) Interventions for preventing falls in older people living in the community. Cochrane Database Syst Reviews 9:CD007146\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eAmerican Diabetes Association Professional Practice Committee (2024) Standards of care in diabetes 2024. Diabetes Care 47(Supplement 1):S1S321\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eYardley L, Beyer N, Hauer K, Kempen G, Piot Ziegler C, Todd C (2005) Development and initial validation of the Falls Efficacy Scale International. Age Ageing 34(6):614619\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eLundberg SM, Lee SI (2017) A unified approach to interpreting model predictions. Adv Neural Inf Process Syst, 30\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eChen T, Guestrin C (2016) XGBoost: A scalable tree boosting system. Proceedings of the 22nd ACM SIGKDD International Conference, 785 794\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eR Core Team (2023) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eFried LP, Tangen CM, Walston J, Newman AB, Hirsch C, Gottdiener J et al (2001) Frailty in older adults: Evidence for a phenotype. Journals Gerontol Ser A, 56(3), M146 M157.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eKamide N, Shiba Y, Shibata H (2009) Falls among Japanese community dwelling older adults with type 2 diabetes: Incidence and risk factors. Geriatr Gerontol Int 9(2):134140\\u003c/span\\u003e\\u003c/li\\u003e\\u003c/ol\\u003e\"}],\"fulltextSource\":\"\",\"fullText\":\"\",\"funders\":[],\"hasAdminPriorityOnWorkflow\":false,\"hasManuscriptDocX\":true,\"hasOptedInToPreprint\":true,\"hasPassedJournalQc\":\"\",\"hasAnyPriority\":true,\"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\":\"info@researchsquare.com\",\"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\":\"sarcopenia, type 2 diabetes mellitus, postmenopausal women, fall prevention, prospective cohort, machine learning, surface electromyography, mHealth education, New Delhi, glycaemic control\",\"lastPublishedDoi\":\"10.21203/rs.3.rs-9126667/v1\",\"lastPublishedDoiUrl\":\"https://doi.org/10.21203/rs.3.rs-9126667/v1\",\"license\":{\"name\":\"CC BY 4.0\",\"url\":\"https://creativecommons.org/licenses/by/4.0/\"},\"manuscriptAbstract\":\"\\u003ch2\\u003eBackground\\u003c/h2\\u003e \\u003cp\\u003ePostmenopausal women with type 2 diabetes mellitus (T2DM) face substantially elevated risk of sarcopenia and injurious falls through converging mechanisms of insulin resistance, neuromuscular atrophy, and chronic hyperglycaemia. Despite this, concurrent assessment of sarcopenia and fall risk in T2DM clinical practice remains uncommon, and validated risk stratification tools tailored to this population are absent.\\u003c/p\\u003e\\u003ch2\\u003eObjective\\u003c/h2\\u003e \\u003cp\\u003eTo determine whether sarcopenia independently predicts incident injurious falls over 12 months in postmenopausal women with T2DM, to construct and validate an interpretable machine learning fall risk model, and to evaluate the efficacy of a multilingual mHealth fall prevention educational intervention.\\u003c/p\\u003e\\u003ch2\\u003eMethodology:\\u003c/h2\\u003e \\u003cp\\u003eA one year prospective cohort study enrolled 312 postmenopausal women with T2DM aged 50 to 75 years at the Endocrinology and Diabetes Clinic of All India Institute of Medical Sciences (AIIMS), New Delhi, India. Sarcopenia was diagnosed using Asian Working Group for Sarcopenia 2019 (AWGS 2019) criteria combining dual energy X ray absorptiometry (DXA), dynamometry, and 6 metre gait speed. Fall events were ascertained through computer vision based clinic monitoring and monthly community telephone surveillance. An interpretable XGBoost ensemble model incorporating SHAP feature attribution was trained on 14 baseline clinical variables. A multilingual AI educational intervention was delivered in five languages at 3, 6, and 9 months.\\u003c/p\\u003e\\u003ch2\\u003eResults\\u003c/h2\\u003e \\u003cp\\u003eSarcopenia was identified in 118 of 312 participants (37.8%). Over 12 months, 94 injurious falls occurred in 71 participants (23.7%). In multivariable Cox regression, sarcopenia was the strongest independent predictor of falls (adjusted hazard ratio 2.84; 95% CI 1.91 to 4.22; p\\u0026thinsp;\\u0026lt;\\u0026thinsp;0.001). The machine learning model achieved AUC 0.883 (95% CI 0.841 to 0.922), significantly outperforming clinical scoring (AUC 0.714; p\\u0026thinsp;\\u0026lt;\\u0026thinsp;0.001). The multilingual educational intervention reduced fall hazard by 34% versus standard care (aHR 0.66; 95% CI 0.47 to 0.92; p\\u0026thinsp;=\\u0026thinsp;0.014).\\u003c/p\\u003e\\u003ch2\\u003eConclusions\\u003c/h2\\u003e \\u003cp\\u003eSarcopenia nearly triples injurious fall risk in postmenopausal women with T2DM at a New Delhi tertiary centre. Interpretable machine learning risk stratification and multilingual digital education significantly improve clinical fall prevention outcomes. Routine sarcopenia screening and ML assisted risk assessment should be integrated into T2DM care pathways for postmenopausal women.\\u003c/p\\u003e\",\"manuscriptTitle\":\"Sarcopenia as an Independent Predictor of Injurious Falls in Postmenopausal Women with Type 2 Diabetes Mellitus: A One Year Prospective Cohort Study with Machine Learning Risk Stratification and mHealth Educational Intervention from a Tertiary Care Center in New Delhi, India\",\"msid\":\"\",\"msnumber\":\"\",\"nonDraftVersions\":[{\"code\":1,\"date\":\"2026-03-19 12:14:09\",\"doi\":\"10.21203/rs.3.rs-9126667/v1\",\"editorialEvents\":[{\"type\":\"communityComments\",\"content\":0}],\"status\":\"published\",\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"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\":\"a2fe4f96-d050-4d2e-b24d-9f081a6123ef\",\"owner\":[],\"postedDate\":\"March 19th, 2026\",\"published\":true,\"recentEditorialEvents\":[],\"rejectedJournal\":[],\"revision\":\"\",\"amendment\":\"\",\"status\":\"posted\",\"subjectAreas\":[],\"tags\":[],\"updatedAt\":\"2026-03-19T12:14:09+00:00\",\"versionOfRecord\":[],\"versionCreatedAt\":\"2026-03-19 12:14:09\",\"video\":\"\",\"vorDoi\":\"\",\"vorDoiUrl\":\"\",\"workflowStages\":[]},\"version\":\"v1\",\"identity\":\"rs-9126667\",\"journalConfig\":\"researchsquare\"},\"__N_SSP\":true},\"page\":\"/article/[identity]/[[...version]]\",\"query\":{\"redirect\":\"/article/rs-9126667\",\"identity\":\"rs-9126667\",\"version\":[\"v1\"]},\"buildId\":\"XKTyCvWXoU3ODBz1xrDgd\",\"isFallback\":false,\"isExperimentalCompile\":false,\"dynamicIds\":[84888],\"gssp\":true,\"scriptLoader\":[]}","source_license":"CC-BY-4.0","license_restricted":false}