Multidimensional Predictors of Health-Related Quality of Life in Parkinson’s Disease Using Ensemble Learning and Network Analysis

Multidimensional Predictors of Health-Related Quality of Life in Parkinson’s Disease Using Ensemble Learning and Network Analysis | 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 Article Multidimensional Predictors of Health-Related Quality of Life in Parkinson’s Disease Using Ensemble Learning and Network Analysis Juseon Hwang, Changhong Youm, Hwayoung Park, Bohyun Kim, Hyejin Choi, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6755283/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 21 Oct, 2025 Read the published version in Scientific Reports → Version 1 posted 10 You are reading this latest preprint version Abstract Parkinson’s disease (PD) causes motor, nonmotor, and mental health challenges that significantly impact health-related quality of life (HRQoL), and most previous studies relied on subjective assessments. Several factors are associated with poor HRQoL; however, the causal relationships remain unclear. Therefore, we aimed to identify key symptom predictors of HRQoL and its subdomains in PD and examine the structure of their interaction networks. We assessed 101 individuals with PD in the ON medication state. HRQoL was measured using a weighted ensemble of the Least Absolute Shrinkage and Selection Operator and Extreme Gradient Boosting for feature selection, followed by stepwise multivariate linear regression. Network analysis was conducted to explore variable interrelations. The HRQoL total score was predicted by the Beck Anxiety Inventory (BAI), Fall Efficacy Scale-Korean (FES-K), Hoehn and Yahr scale, treatment duration, maximum jerk, and sample entropy, with an explanatory power of 65.7%. Regarding subdimensions, anxiety, fear of falling, and sample entropy of acceleration were key determinants. This study identifies key motor and nonmotor predictors of HRQoL in PD and reveals domain-specific networks. These findings may inform targeted interventions and clinical decision-making to improve HRQoL outcomes in people with PD. Health sciences/Biomarkers Health sciences/Health care/Quality of life Parkinson’s disease HRQoL gait motor symptom nonmotor symptom machine learning Figures Figure 1 Figure 2 Figure 3 Figure 4 INTRODUCTION As healthcare systems increasingly prioritize patient-centered care, quality of life (QoL) has become a critical metric for evaluating health outcomes [ 1 ]. This shift has expanded the focus of medical research beyond traditional treatment outcomes to include the broader impacts of disease on individuals' daily lives [ 2 ]. Research on QoL has primarily focused on chronic illnesses—particularly cancer—where long-term survivors often continue to face persistent challenges even after completing treatment [ 2 ]. Similarly, neurodegenerative disorders such as Huntington’s and Alzheimer’s diseases impose compounded physical and psychological burdens that severely compromise patients’ QoL [ 3 ]. Managing these conditions remains challenging due to their multifaceted symptomology and the lack of effective disease-modifying therapies [ 3 , 4 ]. Parkinson’s disease (PD) is a progressive neurodegenerative disorder caused by the loss of dopaminergic neurons in the substantia nigra [ 5 , 6 ]. It is characterized by motor symptoms—including tremor, bradykinesia, rigidity, postural instability, and gait dysfunction—and nonmotor symptoms, such as neuropsychiatric, autonomic, sensory, and sleep disturbances. These symptoms profoundly impact health-related QoL (HRQoL) and progressively deteriorate over time [ 7 ]. Consequently, improving HRQoL is a key objective in PD management [ 8 ]. Numerous studies have explored the predictors of HRQoL in PD, identifying motor and nonmotor contributors [ 7 , 9 , 10 ]. Depression, anxiety, apathy, cognitive decline, falls, and motor impairment (Unified Parkinson’s Disease Rating Scale [UPDRS]-II scores) are all associated with poor HRQoL [ 11 ]. However, reliance on the Parkinson’s Disease Questionnaire (PDQ)-39 summary index rather than its subscales can obscure the complexity and multidimensionality of HRQoL [ 12 ]. Identifying domain-specific predictors may better guide targeted interventions. Previous studies have attempted to do so, revealing domain-specific associations (such as fatigue predicting emotional well-being, UPDRS-III predicting activities of daily living (ADL), and communication) [ 13 ]; however, most previous studies relied heavily on subjective assessments and are prone to recall bias and clinical misinterpretation, especially regarding motor symptoms [ 11 ]. Gait and balance impairments are major determinants of reduced QoL in patients with PD but are often inadequately captured in brief clinical exams [ 14 , 15 ]. Integrating sensor-based gait analysis offers a promising, objective complement to clinical evaluations. Recent studies show that as QoL declines, distinctive changes in gait—such as reduced pace, increased variability, and asymmetry—become more pronounced [ 16 ], indicating that sensor data may be valuable for continuous, real-world monitoring and intervention. Despite these insights, most research has focused on associated factors (correlates) rather than determinant factors. The associated factors do not show a clear causal link and are less consistent than the determinant factors [ 17 ]. Understanding how different symptoms contribute—directly or indirectly—to HRQoL is essential, and traditional statistical methods often fall short of revealing these complex interdependencies [ 18 ]. The emergence of machine learning (ML) techniques offers new opportunities to model and predict HRQoL using diverse data sources, including clinical assessments and sensor-based metrics [ 19 ]. In particular, network analysis has emerged as a powerful approach to visualize and quantify symptom interactions [ 20 ], allowing for a more comprehensive understanding of symptom interrelations and potential intervention targets [ 21 ]. Therefore, in this study, we aimed to identify key symptom predictors of overall HRQoL and its subdomains in individuals with PD using ML approaches and examine the structure of symptom-HRQoL networks to reveal interdependencies that could inform more precise, domain-specific interventions. RESULTS Participant characteristics The cohort selection process is illustrated in Fig. 1. Of the 102 individuals with PD, 101 completed the HRQoL assessments and were included in the analysis. Table 1 summarizes the cohort’s demographic and clinical characteristics (n = 101). Table 1 Demographic and clinical characteristics of people with Parkinson’s disease (n = 101) Characteristics Mean ± SD or n (%) Male 47 (46.5) Female 54 (53.5) Age (years) 68.07 ± 7.23 Height (cm) 160.48 ± 8.39 Body weight (kg) 63.00 ± 10.63 BMI (kg/m 2 ) 24.38 ± 3.08 K-MMSE (scores) 27.77 ± 2.09 MoCA (scores) 25.88 ± 2.97 Hoehn and Yahr scale Stage 1 28 (27.7) Stage 2 50 (49.5) Stage 3 23 (22.8) UPDRS total (scores) 54.20 ± 23.06 UPDRS part Ⅰ (scores) 10.19 ± 5.15 UPDRS part Ⅱ (scores) 13.02 ± 7.14 UPDRS part Ⅲ (scores) 29.26 ± 15.19 UPDRS part Ⅳ (scores) 1.73 ± 2.68 L-Dopa equivalent dose (mg/day) 546.90 ± 291.04 Treatment duration (years) 4.74 ± 4.26 Symptom duration (years) 5.69 ± 4.37 BDI (scores) 14.65 ± 8.95 BAI (scores) 30.67 ± 7.62 PDSS2 (scores) 15.56 ± 10.66 ESS (scores) 6.21 ± 4.17 FSS (scores) 32.58 ± 14.01 Fall 80 (79.1) FESK (scores) 25.91 ± 10.54 SEE (scores) 62.73 ± 21.26 NFOGQ (scores) 3.68 ± 7.12 NMSS (scores) 44.62 ± 25.26 PDQ39_total (scores) 34.94 ± 18.17 PDQ39_Mobility (scores) 9.23 ± 7.38 PDQ39_ Activities of daily living (scores) 6.70 ± 4.89 PDQ39_Emotional (scores) 5.50 ± 5.12 PDQ39_Stigma (scores) 4.59 ± 4.06 PDQ39_Socialsupport (scores) 0.71 ± 1.48 PDQ39_Cognition (scores) 3.62 ± 2.75 PDQ39_Communication (scores) 2.24 ± 2.16 PDQ39_Bodilydiscomfort (scores) 2.35 ± 2.68 Grip_L (scores) 25.86 ± 7.89 Grip_R (scores) 26.65 ± 7.87 FTSST (seconds) 9.62 ± 3.91 6MWT distance (m) 389.82 ± 101.20 SPPB (scores) 10.96 ± 1.69 Mini-BEST (scores) 23.03 ± 3.49 NQ 66.04 ± 11.09 IPAQ 2792.29 ± 4912.22 All the data are indicated as mean ± standard deviation for continuous variables and counts and percentages for categorical data; BMI: Body mass index; K-MMSE: Korean mini-mental state examination; MoCA: Montreal cognitive assessment; UPDRS: Unified Parkinson’s disease rating score; BDI: Beck depression inventory; BAI: Beck anxiety inventory; PDSS2: Parkinson’s disease sleep scale 2nd version; ESS: Epworth sleepiness scale; FSS: Fatigue severity scale; FESK: Fall efficacy scale-Korean; SEE: Self-efficacy for exercise; NFOG-Q: New freezing of gait questionnaire; NMSS: Nonmotor symptoms scale; PDQ39: the 39-item Parkinson’s disease questionnaire; FTSST: five times sit-to-stand test; 6MWT: 6-minute walk test; SPPB: Short physical performance battery; Mini-BEST: Mini balance evaluation systems test; NQ: Nutrition quotient; IPAQ: international physical activity questionnaire. Performance evaluation of machine learning models The model's performance evaluation included reporting the average performance across five folds for each model. The performance of each model and a weighted average for each model using the coefficient of determination (R 2 ) matrix are summarized in Table 2. Table 2 Model evaluation metrics and their corresponding weights. Domain ML models MSE R 2 Weights PDQ-39 total LASSO 0.54 0.42 0.60 XGBoost 0.67 0.29 0.40 Mobility LASSO 0.63 0.51 0.44 XGBoost 0.45 0.66 0.56 ADL LASSO 0.40 0.51 0.44 XGBoost 0.45 0.66 0.56 Emotional well-being LASSO 0.83 0.33 0.60 XGBoost 0.97 0.22 0.40 PDQ-39: Parkinson’s disease questionnaire-39; ADL: Activity of daily living; ML: Machine learning; LASSO: Least absolute shrinkage and selection operator; XGBoost: Extreme gradient boosting. Predictors of overall HRQoL Supplementary Table S1 presents each feature importance derived from Least Absolute Shrinkage and Selection Operator (LASSO) regression and Extreme Gradient Boosting (XGBoost) algorithms. Figure 2 illustrates the average permutation importance of 41 selected features across models. After feature reduction using linear regression, the most significant predictors of the Parkinson’s disease questionnaire-39 (PDQ-39) total score were Beck Anxiety Inventory (BAI) (β = 0.341, p < 0.001), Fall Efficacy Scale-Korean (FES-K) (β = 0.446, p < 0.001), Hoehn and Yahr (H&Y) (β = 0.264, p < 0.001), treatment duration (β = 0.145, p = 0.023), Left turning at faster speed (TurnFL)_posterior superior iliac spine (PSI) sample entropy for gyroscope (SampEnGyr) (β = 0.170, p = 0.007), forward walking (FW)_right ankle (RANK) sample entropy for acceleration (SampEnAcc) (β = 0.190, p = 0.008), FW_PSI SampEnAcc (β = 0.147, p = 0.028), and FW_RANK maximum jerk value (Max Jerk) (β = − 0.194, p = 0.004) (Table 3). Table 3 Association between features and PDQ-39 and its domain using linear regression analysis Variables Β (SE) t-value 95% CI for OR p-value R N 2 PDQ-39 total BAI 0.341 (0.062) 5.505 0.218–0.464 < 0.001 0.657 FESK 0.446 (0.068) 6.521 0.310–0.582 < 0.001 FW_RANK_SampEnAcc 0.190 (0.070) 2.724 0.052–0.329 0.008 H&Y 0.264 (0.065) 4.032 0.134–0.394 < 0.001 TurnFL_PSI_SampEnGyr 0.170 (0.062) 2.762 0.048–0.293 0.007 FW_RANK_MaxJerk −0.194 (0.065) −2.974 −0.324 – −0.065 0.004 Treatment_duration 0.145 (0.063) 2.304 0.020–0.269 0.023 FW_PSI_SampEnAcc 0.147 (0.066) 2.240 0.017–0.277 0.028 ADL UPDRS_total 0.409 (0.076) 5.377 1.067–2.732 < 0.001 0.610 TurnPR_LANK_MaxAcc 0.175 (0.071) 2.481 0.082–0.341 0.015 FESK 0.375 (0.078) 4.784 0.219–0.530 < 0.001 SEE 0.224 (0.068) 3.309 0.090–0.359 0.001 BW_RANK_SampEnAcc 0.186 (0.067) 2.789 0.053–0.318 0.006 SPPB −0.236 (0.073) −3.206 −0.381 – −0.090 0.002 TurnFL_T10_SampEnGyr 0.182 (0.068) 2.685 0.047–0.317 0.009 FW_DSL −0.177(0.073) −2.403 −0.322 – −0.031 0.018 Mobility FESK 0.443 (0.068) 6.542 0.309–0.578 < 0.001 0.687 UPDRS_total 0.259 (0.076) 3.397 0.108–0.411 0.001 NFOGQ 0.156 (0.065) 2.386 0.026–0.285 0.019 FW_RANK_MaxJerk −0.219 (0.060) −3.645 −0.338 – − 0.100 < 0.001 FW_LANK_SampEnAcc 0.237 (0.062) 3.820 0.114–0.360 < 0.001 6MWT −0.182 (0.064) −2.868 −0.309 – −0.056 0.005 BAI 0.146 (0.059) 2.460 0.028–0.264 0.016 Emotional well-being BAI 0.363 (0.090) 4.025 0.184–0.542 < 0.001 0.408 BDI 0.323 (0.090) 3.570 0.143–0.502 0.001 TurnPL_LANK_MaxGyr 0.254 (0.081) 3.125 0.093–0.415 0.002 6MWT −0.167 (0.082) −2.044 −0.329 – −0.005 0.044 PDQ-39: Parkinson’s disease questionnaire-39; ADL: Activity of daily living; OR: Odds ratio; B: Logistic regression coefficient; SE: Standard error; 95% CI: 95% confidence interval; R N 2 is the fit statistic for the Nagelkerke model; BAI: Beck anxiety inventory; FESK: Falls efficacy scale-Korea: FW: forward walking; SampEn: Sample entropy; Acc: Acceleration; RANK: right ankle; H&Y: Hoehn & yahr scale; TurnFL: Left turning at faster speed; Gyr: gyroscope; UPDRS: Unified Parkinson’s disease rating scale; TurnPR: Right turning at preferred speed; LANK: Left ankle; FESK: Fall efficacy scale-korean; SEE: Self-efficacy for exercise; BW: Backward walking; SPPB: Short Physical Performance Battery; DSL: Left double support phase; NFOGQ: New freezing of gait questionnaire; 6MWT: 6-minute walk test; BDI: Beck depression inventory, p < 0.05. The network structure between these predictors and the PDQ-39 total score is depicted in Fig. 3. All predictors were associated with the PDQ-39 total score, with the strongest connection observed between FES-K and PDQ-39 ( r = 0.446). Additional key associations included BAI ( r = 0.341), H&Y ( r = 0.264), FW_RANK Max Jerk ( r = 0.194), FW_RANK SampEnAcc ( r = 0.190), TurnFL_PSI SampEnGyr ( r = 0.170), FW_PSI SampEnAcc ( r = 0.147), and treatment duration ( r = 0.145). Within the network of predictor variables, FES-K was linked to H&Y ( r = 0.204) and FW_RANK SampEnAcc ( r = 0.116). FW_RANK SampEnAcc was associated with FW_PSI SampEnAcc ( r = 0.184) and FW_RANK Max Jerk ( r = 0.184) and was also connected to TurnFL_PSI SampEnGyr ( r = 0.103). Predictors of subdimensions The permutation importance of the variables identified through LASSO regression and XGBoost algorithms for PDQ-39 subdimensions is presented in Supplementary Table S1. For the mobility dimension, Fig. 2 illustrates the average permutation importance of 43 selected features across models. Linear regression analyses identified the most significant predictors of mobility as FES-K (β = 0.438, p < 0.001), UPDRS total score (β = 0.157, p = 0.001), New freezing of gait questionnaire (NFOGQ) (β = 0.138, p = 0.019), FW_RANK Max Jerk (β = − 0.229, p < 0.001), FW_left ankle (LANK) SampEnAcc (β = 0.243, p < 0.001), 6-Minute Walk Test (6MWT) (β = − 0.159, p = 0.005), and BAI (β = 0.145, p = 0.016) (Table 3). In the network analysis, all predictors were associated with mobility, with the strongest connection observed between FES-K and mobility ( r = 0.438). Additional key associations included FW_LANK SampEnAcc ( r = 0.243), FW_RANK Max Jerk ( r = − 0.229), 6MWT ( r = − 0.159), UPDRS total score ( r = 0.157), BAI ( r = 0.145), and NFOGQ ( r = 0.138). Within the network of predictor variables, the UPDRS total score was linked to H&Y ( r = 0.471), FES-K ( r = 0.215), NFOGQ ( r = 0.182), 6MWT ( r = 0.179), and BAI ( r = 0.114). FES-K was associated with NFOGQ ( r = 0.127), while 6MWT was connected to H&Y ( r = 0.153). In addition, FW_RANK Max Jerk was linked to FW_LANK SampEnAcc ( r = 0.163). For the ADL dimension, Fig. 2 illustrates the average permutation importance of 28 selected features across models. Linear regression analyses identified the most significant predictors of ADL as UPDRS total score (β = 0.409, p < 0.001), right turning at preferred speed (TurnPR)_LANK maximum acceleration value (MaxAcc) (β = 0.175, p = 0.015), FES-K (β = 0.375, p < 0.001), Self-Efficacy for Exercise (SEE) (β = 0.224, p = 0.001), backward walking (BW)_RANK SampEnAcc (β = 0.186, p = 0.006), Short Physical Performance Battery (SPPB) (β = − 0.236, p = 0.002), TurnFL_T10 SampEnGyr (β = 0.182, p = 0.009), and FW_left double support phase (DSL) (β = − 0.177, p = 0.018) (Table 3). In the network analysis, all predictors were associated with ADL, with the strongest connection observed between UPDRS total score and ADL ( r = 0.409). Additional key associations included FES-K ( r = 0.375), SEE ( r = 0.224), SPPB ( r = − 0.236), TurnPR_LANK MaxAcc ( r = 0.175), BW_RANK SampEnAcc ( r = 0.186), TurnFL_T10 SampEnGyr ( r = 0.182), and FW_Double Support (L) ( r = − 0.177). Within the network of predictor variables, the UPDRS total score was linked to FES-K ( r = 0.319) and SPPB ( r = 0.204). TurnPR_LANK MaxAcc was associated with TurnFL_T10 SampEnGyr ( r = 0.223) and FW_DSL ( r = 0.143). FES-K was further connected to SEE ( r = 0.235) and BW_RANK SampEnAcc ( r = 0.138). In addition, SPPB and FW_DSL exhibited a strong association ( r = 0.288). For the emotional dimension, Fig. 2 illustrates the average permutation importance of 34 selected features across models. Linear regression analyses identified the most significant predictors of emotional well-being dimension as BAI (β = 0.363, p < 0.001), Beck Depression Inventory (BDI) (β = 0.323, p = 0.001), left turning at preferred speed (TurnPL)_LANK maximum gyroscope value (MaxGyr) (β = 0.254, p = 0.002), and 6MWT (β = − 0.167, p = 0.044) (Table 3). In the network analysis, all predictors were associated with the emotional domain, with the strongest connection observed between BAI and emotional well-being ( r = 0.363). Additional key associations included BDI ( r = 0.323), TurnPL_LANK MaxGyr ( r = 0.254), and 6MWT ( r = − 0.167). Within the network of predictor variables, BAI exhibited a strong association with BDI ( r = 0.429). In addition, TurnPL_LANK MaxGyr was connected to 6MWT ( r = 0.228). DISCUSSION This study identified non-motor and motor symptoms as key predictors of HRQoL in individuals with PD, with an overall explanatory power of 65.7%. Of the non-motor variables, anxiety (BAI), fear of falling (FOF; FESK), H&Y stage, and treatment duration emerged as significant predictors. Motor predictors included sample entropy of acceleration, angular velocity, and maximum jerk. Specifically, TurnFL_PSI SampEnGyr (β = 0.170), FW_RANK SampEnAcc (β = 0.190), and FW_PSI SampEnAcc (β = 0.147) were positively associated with higher PDQ-39 scores, indicating poorer HRQoL. While higher maximal jerk in FW was associated with better HRQoL in this study, previous studies show that individuals with PD typically exhibit reduced jerk, which may contribute to lower perceived mobility [ 22 ]. This finding suggests that as ankle acceleration and angular velocity are more irregular, which reflected in increased sample entropy, HRQoL become declines [ 23 ]. A healthy gait is marked by flexibility and adaptability, reflected in complex, fractal-like stride patterns. Sample entropy, derived from approximate entropy, quantifies movement regularity and is an effective marker of gait dysfunction [ 24 , 25 ]. In PD, degeneration of motor control leads to increased entropy and loss of complexity in gait, signifying impaired adaptability [ 26 ]. This study highlights the relevance of PSI and ankle-based metrics in capturing real-world gait characteristics [ 27 ]. Depression has traditionally been cited as the strongest HRQoL predictor in PD [ 9 , 11 ]; however, this study emphasized the importance of anxiety, FOF, treatment duration, and H&Y stage [ 28 , 29 ]. These results align with the known “wearing-off” effect of long-term dopaminergic therapy, which contributes to worsening motor and non-motor symptoms and reduced QoL [ 30 ]. Treatment fluctuations often provoke anxiety, particularly fear of falling [ 31 ]. In this study, our results underscore FOF (FES-K) as a dominant predictor of overall HRQoL, consistent with earlier research [ 31 ]. FOF—a lack of confidence in safely performing daily tasks—leads to activity restriction and social isolation, worsening QoL [ 32 ]. Network analysis revealed notable associations between FOF and both the H&Y stage (r = 0.204) and SampEnAcc (r = 0.116). As PD progresses, declining postural stability and heightened FOF reduce mobility and independence. FOF, although often underrecognized, plays a critical role in limiting function by diminishing confidence in walking ability [ 32 ]. Thus, interventions targeting gait irregularities and FOF may positively impact HRQoL. The mobility dimension of PDQ-39 was predicted by BAI (β = 0.145), FESK (β = 0.438), UPDRS total score (β = 0.157), NFOGQ (β = 0.138), 6MWT distance (β = -0.159), FW_RANK Max Jerk (β = -0.229), and FW_LANK SampEnAcc (β = 0.234), with an explanatory power of 68.7%. These findings are consistent with previous research linking mobility-related QoL to anxiety [ 33 ], FOF [ 31 ], motor severity [ 34 ], FOG [ 35 ], and walking performance [ 36 ]. These results suggest that reduced walking ability and gait instability strongly influence perceptions of mobility and QoL [ 34 , 36 ]. Notably, NFOGQ, a measure of FOG, was closely linked to FOF in the network analysis, indicating their joint role in diminishing functional independence. FOG-related motor instability, paired with psychological fear, may cause individuals to avoid walking and limit social interaction, contributing to physical deconditioning and reduced HRQoL [ 34 , 35 ]. The ADL dimension was predicted by UPDRS total score (β = 0.409), TurnPR_LANK MaxAcc (β = 0.175), FESK (β = 0.375), SEE (β = -0.224), BW_RANK SampEnAcc (β = 0.186), SPPB (β = -0.236), TurnFL_T10 SampEnGyr (β = 0.182), and FW DSL (β = -0.177), explaining 60.1% of the variance. Reduced walking capacity and balance—reflected in prolonged double support time and irregular gait patterns—negatively influenced self-perceptions of functional ability [ 22 , 34 ]. Network analysis highlighted UPDRS as a central node, linking disease severity with reduced physical performance and increased FOF. As PD progresses, motor decline, functional impairment, and fear of movement interact to restrict independence in daily activities [ 37 , 38 ]. These results emphasize the need for interventions that address motor control and non-motor symptoms to improve ADL-related QoL. The emotional well-being domain was predicted by BAI (β = 0.363), BDI (β = 0.323), TurnPL_LANK MaxGyr (β = 0.254), and 6MWT (β = -0.167), with an explanatory power of 40.8%. Consistent with the findings of previous studies, anxiety had a stronger impact than motor severity or disease duration [ 39 ]. Network analysis revealed a strong correlation between anxiety and depression, which are often comorbid in PD and associated with embarrassment about visible symptoms and resulting social withdrawal [ 33 ]. In addition, reduced 6MWT distance reflects impaired aerobic capacity and mobility [ 36 ], which may contribute to emotional distress and diminished self-efficacy. Our findings also show that higher maximal angular velocity during turning—measured using gyroscopes—was linked to lower HRQoL. Individuals with PD often reduce turning speed to maintain balance, and such compensatory behavior may stem from FOF in public, further impacting emotional well-being [ 40 , 41 ]. These findings suggest that emotional well-being interventions should go beyond treating anxiety and depression and incorporate physical strategies, such as improving aerobic capacity and promoting safer turning techniques [ 42 ]. This study has some limitations. First, its cross-sectional design precludes causal inference or longitudinal analysis of HRQoL changes over time. Future longitudinal studies are needed to determine whether modifying key predictors improves HRQoL. Second, all assessments were conducted during the ON-medication state. Evaluating participants during OFF-medication states may yield deeper insights into the relationship between motor function and QoL [ 27 ]. Third, despite adjusting the hyperparameter in the LASSO model, the performance for certain subdomains of PDQ-39, including stigma, communication, bodily discomfort, cognition, and social support, remained negative. This result may be attributable to the relatively weak correlation between these subdomains and the PDQ-39 total score (see Supplementary Table S2 for correlations between the PDQ-39 total score and its subdimensions) and may explain the lack of association with variables. Finally, this study is based on a single-center Korean cohort, which may limit the generalizability of our results depending on the clinical and cultural backgrounds. Therefore, future studies should include a larger validation multiethnic cohort [ 13 ]. In conclusion, this study identified key predictors of HRQoL and its domains in individuals with PD, integrating clinical, psychological, and sensor-based motor data through machine learning and network analysis. Anxiety, fear of falling, and sample entropy of acceleration emerged as central predictors of overall HRQoL and domain-specific outcomes. These findings underscore the importance of combining objective motor assessments with psychological and clinical evaluations to inform targeted interventions. Addressing motor irregularities and psychological barriers may be essential for improving QoL in individuals living with PD. METHODS Participants This study included 101 individuals with PD. Participants were selected based on the following criteria: (a) a clinician-confirmed diagnosis of idiopathic PD according to the UK Parkinson’s Disease Society Brain Bank Clinical Diagnostic Criteria [ 43 ], (b) age between 40 and 85 years, and (c) H&Y stages 1–3, indicating independent walking and mobility without assistive devices. Individuals were excluded if they had a history of cardiovascular, musculoskeletal, vestibular, or other neurological disorders, required mobility-assistive devices, or had pharmacotherapy-refractory dyskinesia. All participants were in the ON medication state during experimental protocols, ensuring they experienced the full effects of their medication. All study procedures involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards. The study protocol and supplementary information files were approved by the Institutional Review Board of Dong-A University Hospital (IRB number: DAUHIRB-22-089) (see ethics approval letter in the supplementary file). All patients were informed of the experimental procedures and provided written informed consent before data collection. The study is registered with the Clinical Research Information Service in the Republic of Korea (KCT0009353). Experimental protocol Ethical consent was collected after participants’ recruitment, and all participants underwent physical assessments and clinician-administered evaluations of their clinical characteristics. Cognitive function was assessed using the Korean Mini-Mental State Examination (K-MMSE) and the Montreal Cognitive Assessment (MoCA). Gait and non-motor symptoms were evaluated using the NFOGQ and the Non-Motor Symptoms Scale in PD (NMSS), respectively. Fall-related concerns were measured with the FES-K, while exercise self-efficacy was assessed using the SEE. Psychological and sleep disturbances were evaluated using the BDI, BAI, Parkinson’s Disease Sleep Scale 2nd version (PDSS-2), ESS, and Fatigue Severity Scale (FSS). Disease severity and motor symptoms were assessed using the UPDRS, which includes Parts I (non-motor aspects of daily living), II (motor aspects of daily living), III (motor examination), and IV (motor complications), as well as the H&Y scale. They also responded to lifestyle questionnaires, including nutrition quotient (NQ) and the international physical activity questionnaire (IPAQ). Participants’ physical characteristics, including height, body weight, shoulder offset, elbow width, wrist width, hand thickness, leg length, knee width, and ankle width measurements, were measured in the second session. Participants completed a 5-min warm-up, including stretching exercises, before beginning the experiment. We used nine infrared cameras (Vicon MX-T10, Oxford Metrics, UK) and wearable sensors with a stretchable belt (Xsens DOT, Movella Technologies, Enschede, Netherlands) for this experiment. Participants were instructed to walk at a comfortable and fast pace. The walking tasks included (a) forward walking (FW), (b) backward walking (BW) at a self-selected speed, and (c) 360° turns in the right (Turn_PSR, Turn_FSR) and left (Turn_PSL, Turn_FSL) directions at self-selected and maximum speeds [ 44 ]. Each measurement was repeated three times, and the mean value was calculated. Finally, participants underwent postural and functional assessments, including the Short Physical Performance Battery (SPPB), Mini Balance Evaluation Systems Test (Mini-BEST), grip strength measurement, the Five Times Sit-to-Stand Test (FTSST), and the 6-Minute Walk Test (6MWT). Data analysis The physical model was constructed using 39 round reflective markers based on the Plug-in Gait Full Body Model (Vicon Motion Systems Ltd., Oxford, UK), a modified version of the Helen Hayes Marker Set. A global coordinate system was established, with the positive X-axis to the right, the positive Y-axis facing anteriorly, and the Z-axis defined as the cross-product between the X-axis and Y-axis, with the positive Z-axis facing superiorly. Data was recorded at a sampling rate of 100 Hz and filtered using a fourth-order Butterworth low-pass filter at 10 Hz [ 44 ]. Spatiotemporal gait variables were extracted from three-dimensional marker data and included walking speed (m/s), stride length (m), single-support phase (%), double-support phase (%), and contralateral temporal coordination. Left–right contralateral temporal coordination (CLL) was defined as the average difference between the left elbow and right knee, and right–left contralateral temporal coordination (CLR) used as that between the right elbow and left knee. In addition, wearable sensors were attached to six anatomical landmarks, including the left and right upper arms (LELB, RELB) (5 cm above the lateral humeral epicondyle), left and right lateral malleolus (LANK, RANK), 10th thoracic spine (T10), and lumbar spine (center of the left and right posterior superior iliac spine [PSI]) [ 45 ]. The raw data obtained from the wearable sensor using inertial measurement units (IMU) was recorded at a sampling rate of 60 Hz in each axis (x-axis: vertical, y-axis: mediolateral, and z-axis: anteroposterior). Extracted gait features from IMU sensors included maximum jerk (Maxjerk) and angular velocity jerk (Maxangveljerk); mean acceleration (Meanacc) and gyroscope values (Meangyr); maximum acceleration (Maxacc) and gyroscope values (Maxgyr); root mean square (RMS) acceleration (RMSacc) and gyroscope values (RMSgyr); and sample entropy for acceleration (SampEnacc) and gyroscope (SampEngyr) signals. Supplementary Table S3 provides details on the estimation of these metrics. Pre-processing The dataset was pre-processed to identify its structure and any inconsistencies. Pre-processing involved two main steps. First, we used k-nearest neighbor (k-NN) imputation to address the missing values, which were estimated and replaced using a k-NN imputer with five neighbors and leveraged the average of the nearest data points to ensure informed imputations. Lastly, this dataset was normalized with z-score normalization, which enabled the comparability of different features by consistent scaling. Statistical analysis All data were tested for normality using the Shapiro–Wilk test, and statistical analyses were conducted using parametric and nonparametric methods. Demographic and clinical characteristics were summarized using mean and standard deviation (SD) and counts and percentages for categorical data. Spearman’s rho correlation analysis was performed to identify subdomains of the PDQ-39 that were significantly associated with the total PDQ-39 score. Statistical analyses were performed using SPSS 21.0 (IBM Corp., Armonk, NY, USA), and data preprocessing and analysis were conducted using MATLAB R2017b (MathWorks, Natick, MA, USA) and Python (version 3.10) with Pandas, NumPy, Scikit-learn. Prediction models Least Absolute Shrinkage and Selection Operator The Least Absolute Shrinkage and Selection Operator (LASSO) algorithm was employed to analyze and identify variables influencing the total PDQ-39 score and its dimensions. LASSO is a linear regression technique that incorporates a penalty term based on the L1 norm to constrain model coefficients, effectively setting some coefficients to zero and enabling automatic feature selection [ 46 ]. The optimization objective of LASSO is expressed as follows: $$\:\text{arg}min\sum\:_{n=1}^{N}{\left({\mathcal{Y}}^{n}-\sum\:_{q=1}^{m}{\beta\:}_{q}{\mathcal{x}}_{q}^{n}\right)}^{2}+\:\lambda\:\sum\:_{q=1}^{m}\left|{\beta\:}_{q}\right|$$ \(\:{\mathcal{Y}}^{n}\) represents the class label for the n -th trial, \(\:N\) is the total number of trials, \(\:{\mathcal{x}}_{q}^{n}\) denoted the q -th clinical factor of the n -th sample, \(\:{\beta\:}_{q}\) is the regression coefficient of the q -th factor, and \(\:\lambda\:\) is the regularization parameter, where a larger \(\:\lambda\:\) results in a sparser model, setting more \(\:{\beta\:}_{q}\) values to zero. We applied a 5-fold cross-validation approach to optimize the LASSO regularization parameter (alpha), which was selected based on the highest mean R² across the five folds to assess model selection stability. The data were split into training (80%) and test (20%) sets, ensuring model evaluation was performed on unseen data. After identifying the optimal alpha, the model was retrained using the entire training set, and the final evaluation was performed on the independent test set (20% of the data). The final model was evaluated using mean squared error (MSE) and coefficient of determination (R 2 ), and the final selected variables were defined as those with non-zero regression coefficients. Extreme Gradient Boosting The Extreme Gradient Boosting (XGBoost) classification algorithm is a high-performance boosting technique that minimizes the loss function through iterative model addition. XGBoost sequentially adds models to the ensemble in each iteration. The algorithm generates feature importance scores based on the gain score during the training process [ 47 ]. The model's output is described as follows: $$\:{\widehat{\mathcal{Y}}}_{i}=\sum\:_{k=1}^{K}{f}_{k}\left({\mathcal{x}}_{\mathcal{i}}\right),{f}_{k}\in\:F,\:\in\:F,\:k=1$$ Where \(\:k\) represents the total number of predictors, and \(\:{f}_{k}\) is the function corresponding to the \(\:k\) -th predictor in the functional space \(\:F\) . We used the training feature \(\:{\mathcal{x}}_{\mathcal{i}}\) to predict the target variable \(\:{\mathcal{Y}}_{i}\) . We ranked the variables in order of importance and selected the variables that covered at least 80% of the cumulative score. The dataset was split into training (80%) and testing (20%) sets. Hyperparameter tuning was performed using the grid search method with 3-fold cross-validation to identify the optimal parameters. In particular, the number of estimators (50, 100, 200), learning rate (0.01, 0.1, 0.2), and maximum tree depth (3, 5, 7) were tuned. In addition, the generalization performance of the optimized model was evaluated using 5-fold cross-validation on the entire dataset. The final model was evaluated using MSE and R 2 . Weighted Average Ensemble Method The weighted average ensemble method enhances prediction accuracy by assigning greater importance to models with better performance. To generate the final prediction, a weighted average of each model's output was calculated using the following Equation, where the predictions of all models were combined with their corresponding weights [ 48 ]. The final feature importance was obtained by summing the feature importance values from each model. $$\:{P}_{final}=\frac{{\mathcal{W}}_{1}\bullet\:\:{P}_{1}+{\mathcal{W}}_{2}\bullet\:\:{P}_{2}+{\mathcal{W}}_{3}\bullet\:\:{P}_{3}+\dots\:+{\mathcal{W}}_{N}\bullet\:\:{P}_{N}}{{\mathcal{W}}_{1}+{\mathcal{W}}_{2}+{\mathcal{W}}_{3}+\dots\:+{\mathcal{W}}_{N}}$$ In the ensemble classifier, N models generate predictions for a given input, where \(\:{P}_{1}\) , \(\:{P}_{2}\) , \(\:{P}_{3}\) , …, \(\:{P}_{N}\) represent the predictions from Models 1, 2, 3, and so on. The predictions from all models were combined with their corresponding weights, denoted as \(\:{\mathcal{W}}_{1}\) , \(\:{\mathcal{W}}_{2}\) , \(\:{\mathcal{W}}_{3}\) ,…, \(\:{\mathcal{W}}_{N}\) , to create the weighted average ensemble classifier. Finally, the optimal variables for evaluating the determinants of HRQoL in individuals with PD were identified using a forward stepwise linear regression. Network analysis Network analysis was employed to explore the associations between selected variables and HRQoL based on partial correlations. We applied the graphical lasso model [ 49 ] to estimate the partial correlations between variables and shrink the less relevant edges to zero. Therefore, this network structure represents the regularized partial correlations, highlighting the most relevant relationships while minimizing irrelevant connections. The nodes with the darkest color correspond to the raw scores of the PDQ-39 total score and its eight subscales. In addition, all nodes were positioned using the spring_layout of the Fruchterman-Reingold force-directed algorithm based on the strength of the connections between them [ 50 ]. Edge thickness reflects the strength of the relationships, using regularized partial correlations between nodes (Fig. 4 ). Declarations ACKNOWLEDGMENTS The authors thank all the participants of this study. This work was supported by a Dong-A University Research Fund and grants from the National Research Foundation of Korea (NRF) funded by the Korean government (MSIT) (No. 2022R1A2C100933711; Changhong Youm), the Basic Science Research Program through the NRF funded by the Ministry of Education (No. 2022R1A6A3A0108756411; Hwayoung Park), and the Ministry of Education of the Republic of Korea and the NRF (No. 2024S1A5B5A16021673; Hwayoung Park). This study received no specific grants from funding agencies in the public, commercial, or non-profit sectors. The funding sources had no role in the study design; collection, analysis, and interpretation of the data; or in writing the manuscript. The authors also thank Editage (www.editage.co.kr) for English language editing. AUTHOR’S CONTRIBUTIONS JH, CY, HP, BK, HC, MK, and SC conceived and designed the study. JH, HP, BK, HC, MK, and SC recruited the participants. JH, CY, HP, BK, HC, MK, and SC acquired the data. JH, CY, HP, BK, HC, MK, and SC analyzed and interpreted the data. JH, CY, HP, BK, HC, MK, and SC drafted the article. All authors read and approved the final version of the manuscript submitted. COMPETING INTERESTS The authors declare no competing interests. The code for training and testing the ML models was written in Python 3.8 using PyTorch 1.9.1 and Torchvision 0.10.1. Data management and feature processing scripts were written in Python 3.8 using Pandas 1.3.3 and NumPy 1.21.2. The datasets supporting this study's findings, as well as the code used for the analysis, are available from the corresponding author upon reasonable request. DATA AVAILABILITY STATEMENT The datasets supporting the findings of this study are available from the corresponding author upon reasonable request. References Fayers, P. M. & Machin, D. Quality of Life: the Assessment, Analysis and Interpretation of Patient-Reported Outcomes (John Wiley & Sons, 2013). Haraldstad, K. et al. A systematic review of quality of life research in medicine and health sciences. Qual. Life Res. 28 , 2641-2650 (2019). Batista, P. S. P. & Pereira, A. Quality of life in patients with neurodegenerative diseases. J. Neurol. Neurosci. 7 , 74 (2016). Nance, M. A. Comprehensive care in Huntington's disease: a physician's perspective. Brain Res. Bull. 72 , 175-178 (2007). Alexander, G. E., DeLong, M. R. & Strick, P. L. Parallel organization of functionally segregated circuits linking basal ganglia and cortex. Annu. Rev. Neurosci. 9 , 357-381 (1986). Takakusaki, K., Tomita, N. & Yano, M. Substrates for normal gait and pathophysiology of gait disturbances with respect to the basal ganglia dysfunction. J. Neurol. 255 (Supplement 4), 19-29 (2008). Tosin, M. H., Goetz, C. G. & Stebbins, G. T. Patient with Parkinson disease and care partner perceptions of key domains affecting health-related quality of life: systematic review. Neurology 102 , e208028 (2024). Opara, J. A., Brola, W., Leonardi, M. & Błaszczyk, B. Quality of life in Parkinson’s disease. J. Med. Life 5 , 375–381 (2012). Soh, S. E., Morris, M. E. & McGinley, J. L. Determinants of health-related quality of life in Parkinson’s disease: a systematic review. Parkinsonism Relat. Disord. 17 , 1-9 (2011). Zhao, N. et al. Quality of life in Parkinson's disease: A systematic review and meta‐analysis of comparative studies. CNS Neurosci. Ther. 27 , 270-279 (2021). Bock, M. A., Brown, E. G., Zhang, L. & Tanner, C. Association of motor and nonmotor symptoms with health-related quality of life in a large online cohort of people with Parkinson disease. Neurology 98 , e2194–e2203 (2022). Hagell, P. & Nilsson, M. H. The 39-item Parkinson's disease Questionnaire (PDQ-39): is it a unidimensional construct? Ther. Adv. Neurol. Disord. 2 , 205-214 (2009). Sanchez-Luengos, I. et al . Predictors of health-related quality of life in Parkinson’s disease: the impact of overlap between health-related quality of life and clinical measures. Qual. Life Res. 31 , 3241-3252 (2022). Boonstra, T. A., Van Der Kooij, H., Munneke, M. & Bloem, B. R. Gait disorders and balance disturbances in Parkinson's disease: clinical update and pathophysiology. Curr. Opin. Neurol. 21 , 461-471 (2008). Horak, F. B. & Mancini, M. Objective biomarkers of balance and gait for Parkinson's disease using body‐worn sensors. Mov. Disord. 28 , 1544-1551 (2013). Gonçalves, H. R., Branquinho, A., Pinto, J., Rodrigues, A. M. & Santos, C. P. Digital biomarkers of mobility and quality of life in Parkinson's disease based on a wearable motion analysis LAB. Comput. Methods Programs Biomed. 244 , 107967 (2024). Martinez‐Martin, P. What is quality of life and how do we measure it? Relevance to Parkinson's disease and movement disorders. Mov. Disord. 32 , 382-392 (2017). Kazemi, K. et al . Impact of physical activity on quality of life during pregnancy: A causal ML approach in 46th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) 2024 , (1-6) (2024). Jannani, A., Sael, N. & Benabbou, F. Artificial intelligence for quality of life study: A systematic literature review. IEEE Access 12 , 62059–62088 (2024). Borsboom, D. A network theory of mental disorders. World Psychiatry 16 , 5-13 (2017). Schönenberg, A. et al. Using network analysis to explore the validity and influential items of the Parkinson’s disease Questionnaire-39. Sci. Rep. 13 , 7221 (2023). Curtze, C., Nutt, J. G., Carlson-Kuhta, P., Mancini, M. & Horak, F. B. Objective gait and balance impairments relate to balance confidence and perceived mobility in people with Parkinson disease. Phys. Ther. 96 , 1734-1743 (2016). Bange, M. et al. Deficient interhemispheric connectivity underlies movement irregularities in Parkinson’s disease. J. Parkinsons Dis. 12 , 381-395 (2022). Karmakar, C., Khandoker, A., Begg, R. & Palaniswami, M. Understanding ageing effects using complexity analysis of foot–ground clearance during walking. Comput. Methods Biomech. Biomed. Eng. 16 , 554-564 (2013). Yentes, J. M. et al . The appropriate use of approximate entropy and sample entropy with short data sets. Ann. Biomed. Eng. 41 , 349-365 (2013). Lukšys, D. & Griškevičius, J. Application of nonlinear analysis for the assessment of gait in patients with Parkinson’s disease . Technol. Health Care 30 , 201-208 (2022). Bouça-Machado, R. et al. Gait kinematic parameters in Parkinson’s disease: a systematic review. J. Parkinsons Dis. 10 , 843-853 (2020). Dissanayaka, N. N. N. W. et al. The clinical spectrum of anxiety in Parkinson's disease. Mov. Disord. 29 , 967-975 (2014). Ge, Y. et al. Correlation between motor function and health-related quality of life in early to mid-stage patients with Parkinson disease: a cross-sectional observational study. Front. Aging Neurosci. 16 , 1399285 (2024). Ahlskog, J. E. & Muenter, M. D. Frequency of levodopa‐related dyskinesias and motor fluctuations as estimated from the cumulative literature. Mov. Disord. 16 , 448-458 (2001). Dissanayaka, N. N. et al. Phenomenology of atypical anxiety disorders in Parkinson's disease: a systematic review. Am. J. Geriatr. Psychiatry 30 , 1026-1050 (2022). Bloem, B. R., Grimbergen, Y. A., Cramer, M., Willemsen, M. & Zwinderman, A. H. Prospective assessment of falls in Parkinson's disease. J. Neurol. 248 , 950-958 (2001). Chuquilín-Arista, F., Álvarez-Avellón, T. & Menéndez-González, M. Prevalence of depression and anxiety in Parkinson disease and impact on quality of life: a community-based study in Spain. J. Geriatr. Psychiatry Neurol. 33 , 207-213 (2020). Scalzo, P. L., Flores, C. R., Marques, J. R., Robini, S. C. O. & Teixeira, A. L. Impact of changes in balance and walking capacity on the quality of life in patients with Parkinson's disease. Arq. Neuro Psiquiatr. 70 , 119-124 (2012). Bowman, T. et al. Predictors of mobility domain of health-related quality of life after rehabilitation in Parkinson’s disease: A pilot study. Arch. Physiother. 8 , 10 (2018). Canning, C. G., Allen, N. E., Dean, C. M., Goh, L. & Fung, V. S. C. Home-based treadmill training for individuals with Parkinson’s disease: a randomized controlled pilot trial. Clin. Rehabil. 26 , 817-826 (2012). Franchignoni, F., Martignoni, E., Ferriero, G. & Pasetti, C. Balance and fear of falling in Parkinson's disease . Parkinsonism Relat. Disord. 11 , 427-433 (2005). Grimbergen, Y. A. M., Schrag, A., Mazibrada, G., Borm, G. F. & Bloem, B. R. Impact of falls and fear of falling on health-related quality of life in patients with Parkinson's disease. J. Parkinsons Dis. 3 , 409-413 (2013). Hanna, K. K. & Cronin-Golomb, A. Impact of anxiety on quality of life in Parkinson′s disease. Parkinsons Dis. 2012 , 640707 (2012). Haertner, L. et al. Effect of fear of falling on turning performance in Parkinson’s disease in the lab and at home. Front. Aging Neurosci. 10 , 78 (2018). Shah, V. V. et al. Digital biomarkers of mobility in Parkinson’s disease during daily living. J. Parkinsons Dis. 10 , 1099-1111 (2020). Lo Buono, V. L. et al. Anxiety, depression, and quality of life in Parkinson’s disease: the implications of multidisciplinary treatment. Neural Regen. Res. 16 , 587-590 (2021). Hughes, A. J., Ben-Shlomo, Y., Daniel, S. E. & Lees, A. J. UK Parkinson’s Disease Society Brain Bank clinical diagnostic criteria. J. Neurol. Neurosurg. Psychiatry 55 , e4 (1992). Park, H. et al. Classification of Parkinson’s disease with freezing of gait based on 360 turning analysis using 36 kinematic features. J. Neuroeng. Rehabil. 18 , 1-18 (2021). Choi, H. et al. Convolutional neural network based detection of early stage Parkinson’s disease using the six minute walk test. Sci. Rep. 14 , 22648 (2024). Tibshirani, R. Regression shrinkage and selection via the lasso. J. R. Stat. Soc. B 58 , 267-288 (1996). Hsieh, C. P., Chen, Y. T., Beh, W. K. & Wu, A. Y. A. Feature selection framework for XGBoost based on electrodermal activity in stress detection in IEEE International Workshop on Signal Processing Systems (SiPS) 330-335 (IEEE, 2019). Abhisheka, B., Biswas, S. K. & Purkayastha, B. Infusing weighted average ensemble diversity for advanced breast cancer detection. Int. J. Imaging Syst. Technol. 34 , e23146 (2024). Friedman, J., Hastie, T. & Tibshirani, R. Sparse inverse covariance estimation with the graphical lasso. Biostatistics 9 , 432-441 (2008). Igual, L. & Seguí, S. Network analysis in Introduction to Data Science: A Python Approach to Concepts, Techniques and Applications 151-174 (Springer, 2024). Supplementary File Supplementary File is not available with this version. Additional Declarations No competing interests reported. Supplementary Files SupplementaryTables.docx Cite Share Download PDF Status: Published Journal Publication published 21 Oct, 2025 Read the published version in Scientific Reports → Version 1 posted Editorial decision: Revision requested 28 Jul, 2025 Reviews received at journal 26 Jul, 2025 Reviewers agreed at journal 10 Jul, 2025 Reviews received at journal 27 Jun, 2025 Reviewers agreed at journal 27 Jun, 2025 Reviewers invited by journal 12 Jun, 2025 Editor assigned by journal 06 Jun, 2025 Editor invited by journal 29 May, 2025 Submission checks completed at journal 27 May, 2025 First submitted to journal 27 May, 2025 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-6755283","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":470229249,"identity":"9d0dac3e-90db-460c-b5fb-12e98e17ef1d","order_by":0,"name":"Juseon Hwang","email":"","orcid":"","institution":"The Graduate School of Dong-A University","correspondingAuthor":false,"prefix":"","firstName":"Juseon","middleName":"","lastName":"Hwang","suffix":""},{"id":470229250,"identity":"5d86a618-fdd3-4ed2-acb1-5671a5261921","order_by":1,"name":"Changhong Youm","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAt0lEQVRIiWNgGAWjYDACCTBpw9gA4SYQrSWNdC2HSdDCP7vH8DFvznnZ/hkJjB9+MKTlE7bkzhljY95tt41n3EhgluxhyLFsIKTFQCLHTBqoJbHhRgKDNANDhQFBW6BaziXOB9rymxQtBxI33EhgA9qSQ1iLxI20YsO525KNN5552GbZY5BGWAv/jOSND95us5Oddzz58I0fFcmEtTAwcMAUgaKGGA0MDOwPiFI2CkbBKBgFIxgAAMSZOXu0+WMMAAAAAElFTkSuQmCC","orcid":"","institution":"The Graduate School of Dong-A University","correspondingAuthor":true,"prefix":"","firstName":"Changhong","middleName":"","lastName":"Youm","suffix":""},{"id":470229251,"identity":"3f9bfdb1-df8e-484a-a7ea-699d257f8328","order_by":2,"name":"Hwayoung Park","email":"","orcid":"","institution":"Dong-A University","correspondingAuthor":false,"prefix":"","firstName":"Hwayoung","middleName":"","lastName":"Park","suffix":""},{"id":470229252,"identity":"a96f1a40-4638-406b-b653-0c93af947f00","order_by":3,"name":"Bohyun Kim","email":"","orcid":"","institution":"Dong-A University","correspondingAuthor":false,"prefix":"","firstName":"Bohyun","middleName":"","lastName":"Kim","suffix":""},{"id":470229253,"identity":"e8b109b8-169a-4fd5-a07e-baa3cd79340c","order_by":4,"name":"Hyejin Choi","email":"","orcid":"","institution":"The Graduate School of Dong-A University","correspondingAuthor":false,"prefix":"","firstName":"Hyejin","middleName":"","lastName":"Choi","suffix":""},{"id":470229254,"identity":"0f4b8774-8c97-4230-a7fb-a1f3dd31dbee","order_by":5,"name":"Minsoo Kim","email":"","orcid":"","institution":"The Graduate School of Dong-A University","correspondingAuthor":false,"prefix":"","firstName":"Minsoo","middleName":"","lastName":"Kim","suffix":""},{"id":470229255,"identity":"ae4018e0-0c13-4d59-84d4-5da24d3cf9fc","order_by":6,"name":"Sang-Myung Cheon","email":"","orcid":"","institution":"Dong-A University","correspondingAuthor":false,"prefix":"","firstName":"Sang-Myung","middleName":"","lastName":"Cheon","suffix":""}],"badges":[],"createdAt":"2025-05-27 04:38:29","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6755283/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6755283/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41598-025-20656-9","type":"published","date":"2025-10-21T16:16:16+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":84761133,"identity":"6f23603b-7513-4f2e-b16b-731fc8aec0b9","added_by":"auto","created_at":"2025-06-17 06:03:52","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":415934,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFlow diagram for participants.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-6755283/v1/7755d7e6ee4ee179b5e0eeb7.png"},{"id":84761135,"identity":"623c6a64-72ed-44e6-a0e6-a9845b3aa3e0","added_by":"auto","created_at":"2025-06-17 06:03:52","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1627257,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePermutation feature importance with the weighted ensemble method.\u003c/strong\u003e PDQ-39: Parkinson’s disease questionnaire-39; BAI: Beck anxiety inventory; FESK: Falls efficacy scale-Korea; NMSS: Non-motor Symptoms Scale; NFOG-Q: New Freezing of Gait Questionnaire; H\u0026amp;Y: Hoehn \u0026amp; yahr scale; UPDRS: Unified Parkinson’s Disease Rating Scale; FW: forward walking; Gyr: gyroscope; BDI: Beck depression inventory; TurnFL: Left turning at faster speed; LELB: Left elbow; SampEn: Sample entropy; SPPB: Short Physical Performance Battery; TurnFR: Right turning at faster speed; SLR: Right stride length; TurnPR: Right turning at preferred speed; LANK: Left ankle; Acc: Acceleration; CLL: Left contralateral phase; BW: Backward walking; DSR: Right double support phase; SEE: Self-efficacy for exercise; TurnPL: Left turning at preferred speed; RELB: Right elbow; RANK: Right ankle; PDSS2: Parkinson’s Disease Sleep Scale 2nd version; Max Angular Jerk: Max angular velocity jerk; 6MWT: 6-Minute Walk Test; Rms: Root mean square; FSS: Fatigue Severity Scale; ESS: Epworth Sleepiness Scale; CLR: Right contralateral phase; ADL: Activity of daily living.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-6755283/v1/5e321948457e13ea04fb356c.png"},{"id":84761956,"identity":"267e3612-a9db-4b38-a435-3b851f29424b","added_by":"auto","created_at":"2025-06-17 06:11:52","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":761797,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eNetworking plot\u003c/strong\u003e. PDQ-39: Parkinson’s disease questionnaire-39; BAI: Beck anxiety inventory; FESK: Falls efficacy scale-Korea: FW: forward walking; SampEn: Sample entropy; Acc: Acceleration; RANK: right ankle; H\u0026amp;Y: Hoehn \u0026amp; yahr scale; TurnFL: Left turning at faster speed; Gyr: gyroscope; ADL: Activity of daily living; UPDRS: Unified Parkinson’s disease rating scale; TurnPR: Right turning at preferred speed; LANK: Left ankle; FESK: Fall efficacy scale-korean; SEE: Self-efficacy for exercise; BW: Backward walking; SPPB: Short Physical Performance Battery; DSL: Left double support phase; NFOGQ: New freezing of gait questionnaire; 6MWT: 6-minute walk test; BDI: Beck depression inventory.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-6755283/v1/b31f85096b9e55348ec74c61.png"},{"id":84761138,"identity":"9888d2a2-3b76-473f-b293-daf1115eeba9","added_by":"auto","created_at":"2025-06-17 06:03:52","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1292092,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSchematics of the proposed approach for predicting HRQoL.\u003c/strong\u003e ML: Machine learning; LASSO: Least absolute shrinkage and selection operator; XGBoost: Extreme gradient boosting; W\u003csub\u003e1\u003c/sub\u003e: LASSO corresponding weights; \u0026nbsp;\u0026nbsp;W\u003csub\u003e2\u003c/sub\u003e: XGBoost corresponding weights; HRQoL: Health related quality of life; PD: people with PD; PDQ-39: Parkinson’s disease questionnaire-39.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-6755283/v1/6e0d8c8ad930cea0e6494365.png"},{"id":94490558,"identity":"99f94d42-ed0c-49ec-92c0-17b64004f230","added_by":"auto","created_at":"2025-10-27 17:12:05","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":5687447,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6755283/v1/6bb7b8f0-ee7e-48c3-9c7d-6f8e32f1cb6c.pdf"},{"id":84761955,"identity":"17c7ac45-99ad-48c9-b9e4-8fe2b515d8fb","added_by":"auto","created_at":"2025-06-17 06:11:52","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":33942,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTables.docx","url":"https://assets-eu.researchsquare.com/files/rs-6755283/v1/e2f94388ee586b5c33b70085.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Multidimensional Predictors of Health-Related Quality of Life in Parkinson’s Disease Using Ensemble Learning and Network Analysis","fulltext":[{"header":"INTRODUCTION","content":"\u003cp\u003eAs healthcare systems increasingly prioritize patient-centered care, quality of life (QoL) has become a critical metric for evaluating health outcomes [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. This shift has expanded the focus of medical research beyond traditional treatment outcomes to include the broader impacts of disease on individuals' daily lives [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Research on QoL has primarily focused on chronic illnesses\u0026mdash;particularly cancer\u0026mdash;where long-term survivors often continue to face persistent challenges even after completing treatment [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Similarly, neurodegenerative disorders such as Huntington\u0026rsquo;s and Alzheimer\u0026rsquo;s diseases impose compounded physical and psychological burdens that severely compromise patients\u0026rsquo; QoL [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Managing these conditions remains challenging due to their multifaceted symptomology and the lack of effective disease-modifying therapies [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eParkinson\u0026rsquo;s disease (PD) is a progressive neurodegenerative disorder caused by the loss of dopaminergic neurons in the substantia nigra [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. It is characterized by motor symptoms\u0026mdash;including tremor, bradykinesia, rigidity, postural instability, and gait dysfunction\u0026mdash;and nonmotor symptoms, such as neuropsychiatric, autonomic, sensory, and sleep disturbances. These symptoms profoundly impact health-related QoL (HRQoL) and progressively deteriorate over time [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Consequently, improving HRQoL is a key objective in PD management [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eNumerous studies have explored the predictors of HRQoL in PD, identifying motor and nonmotor contributors [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Depression, anxiety, apathy, cognitive decline, falls, and motor impairment (Unified Parkinson\u0026rsquo;s Disease Rating Scale [UPDRS]-II scores) are all associated with poor HRQoL [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. However, reliance on the Parkinson\u0026rsquo;s Disease Questionnaire (PDQ)-39 summary index rather than its subscales can obscure the complexity and multidimensionality of HRQoL [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Identifying domain-specific predictors may better guide targeted interventions. Previous studies have attempted to do so, revealing domain-specific associations (such as fatigue predicting emotional well-being, UPDRS-III predicting activities of daily living (ADL), and communication) [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]; however, most previous studies relied heavily on subjective assessments and are prone to recall bias and clinical misinterpretation, especially regarding motor symptoms [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eGait and balance impairments are major determinants of reduced QoL in patients with PD but are often inadequately captured in brief clinical exams [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Integrating sensor-based gait analysis offers a promising, objective complement to clinical evaluations. Recent studies show that as QoL declines, distinctive changes in gait\u0026mdash;such as reduced pace, increased variability, and asymmetry\u0026mdash;become more pronounced [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e], indicating that sensor data may be valuable for continuous, real-world monitoring and intervention.\u003c/p\u003e \u003cp\u003eDespite these insights, most research has focused on associated factors (correlates) rather than determinant factors. The associated factors do not show a clear causal link and are less consistent than the determinant factors [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. Understanding how different symptoms contribute\u0026mdash;directly or indirectly\u0026mdash;to HRQoL is essential, and traditional statistical methods often fall short of revealing these complex interdependencies [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThe emergence of machine learning (ML) techniques offers new opportunities to model and predict HRQoL using diverse data sources, including clinical assessments and sensor-based metrics [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. In particular, network analysis has emerged as a powerful approach to visualize and quantify symptom interactions [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e], allowing for a more comprehensive understanding of symptom interrelations and potential intervention targets [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eTherefore, in this study, we aimed to identify key symptom predictors of overall HRQoL and its subdomains in individuals with PD using ML approaches and examine the structure of symptom-HRQoL networks to reveal interdependencies that could inform more precise, domain-specific interventions.\u003c/p\u003e"},{"header":"RESULTS","content":"\u003cdiv id=\"Sec3\"\u003e\n \u003ch2\u003eParticipant characteristics\u003c/h2\u003e\n \u003cp\u003eThe cohort selection process is illustrated in Fig. 1. Of the 102 individuals with PD, 101 completed the HRQoL assessments and were included in the analysis. Table 1 summarizes the cohort\u0026rsquo;s demographic and clinical characteristics (n\u0026thinsp;=\u0026thinsp;101).\u0026nbsp;\u003c/p\u003e\n \u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 1\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003eDemographic and clinical characteristics of people with Parkinson\u0026rsquo;s disease (n\u0026thinsp;=\u0026thinsp;101)\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eCharacteristics\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eMean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD or n (%)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e47 (46.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e54 (53.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAge (years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e68.07\u0026thinsp;\u0026plusmn;\u0026thinsp;7.23\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHeight (cm)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e160.48\u0026thinsp;\u0026plusmn;\u0026thinsp;8.39\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBody weight (kg)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e63.00\u0026thinsp;\u0026plusmn;\u0026thinsp;10.63\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBMI (kg/m\u003csup\u003e2\u003c/sup\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e24.38\u0026thinsp;\u0026plusmn;\u0026thinsp;3.08\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eK-MMSE (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e27.77\u0026thinsp;\u0026plusmn;\u0026thinsp;2.09\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMoCA (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e25.88\u0026thinsp;\u0026plusmn;\u0026thinsp;2.97\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHoehn and Yahr scale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eStage 1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e28 (27.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eStage 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e50 (49.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eStage 3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e23 (22.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUPDRS total (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e54.20\u0026thinsp;\u0026plusmn;\u0026thinsp;23.06\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUPDRS part Ⅰ (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e10.19\u0026thinsp;\u0026plusmn;\u0026thinsp;5.15\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUPDRS part Ⅱ (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e13.02\u0026thinsp;\u0026plusmn;\u0026thinsp;7.14\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUPDRS part Ⅲ (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e29.26\u0026thinsp;\u0026plusmn;\u0026thinsp;15.19\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUPDRS part Ⅳ (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.73\u0026thinsp;\u0026plusmn;\u0026thinsp;2.68\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eL-Dopa equivalent dose (mg/day)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e546.90\u0026thinsp;\u0026plusmn;\u0026thinsp;291.04\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTreatment duration (years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.74\u0026thinsp;\u0026plusmn;\u0026thinsp;4.26\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSymptom duration (years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5.69\u0026thinsp;\u0026plusmn;\u0026thinsp;4.37\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBDI (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e14.65\u0026thinsp;\u0026plusmn;\u0026thinsp;8.95\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBAI (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e30.67\u0026thinsp;\u0026plusmn;\u0026thinsp;7.62\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePDSS2 (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e15.56\u0026thinsp;\u0026plusmn;\u0026thinsp;10.66\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eESS (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6.21\u0026thinsp;\u0026plusmn;\u0026thinsp;4.17\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFSS (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e32.58\u0026thinsp;\u0026plusmn;\u0026thinsp;14.01\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFall\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e80 (79.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFESK (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e25.91\u0026thinsp;\u0026plusmn;\u0026thinsp;10.54\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSEE (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e62.73\u0026thinsp;\u0026plusmn;\u0026thinsp;21.26\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNFOGQ (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.68\u0026thinsp;\u0026plusmn;\u0026thinsp;7.12\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNMSS (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e44.62\u0026thinsp;\u0026plusmn;\u0026thinsp;25.26\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePDQ39_total (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e34.94\u0026thinsp;\u0026plusmn;\u0026thinsp;18.17\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePDQ39_Mobility (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e9.23\u0026thinsp;\u0026plusmn;\u0026thinsp;7.38\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePDQ39_ Activities of daily living (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6.70\u0026thinsp;\u0026plusmn;\u0026thinsp;4.89\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePDQ39_Emotional (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5.50\u0026thinsp;\u0026plusmn;\u0026thinsp;5.12\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePDQ39_Stigma (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.59\u0026thinsp;\u0026plusmn;\u0026thinsp;4.06\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePDQ39_Socialsupport (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.71\u0026thinsp;\u0026plusmn;\u0026thinsp;1.48\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePDQ39_Cognition (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.62\u0026thinsp;\u0026plusmn;\u0026thinsp;2.75\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePDQ39_Communication (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2.24\u0026thinsp;\u0026plusmn;\u0026thinsp;2.16\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePDQ39_Bodilydiscomfort (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2.35\u0026thinsp;\u0026plusmn;\u0026thinsp;2.68\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGrip_L (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e25.86\u0026thinsp;\u0026plusmn;\u0026thinsp;7.89\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGrip_R (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e26.65\u0026thinsp;\u0026plusmn;\u0026thinsp;7.87\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFTSST (seconds)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e9.62\u0026thinsp;\u0026plusmn;\u0026thinsp;3.91\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6MWT distance (m)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e389.82\u0026thinsp;\u0026plusmn;\u0026thinsp;101.20\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSPPB (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e10.96\u0026thinsp;\u0026plusmn;\u0026thinsp;1.69\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMini-BEST (scores)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e23.03\u0026thinsp;\u0026plusmn;\u0026thinsp;3.49\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNQ\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e66.04\u0026thinsp;\u0026plusmn;\u0026thinsp;11.09\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIPAQ\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2792.29\u0026thinsp;\u0026plusmn;\u0026thinsp;4912.22\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003ctfoot\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\"\u003eAll the data are indicated as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation for continuous variables and counts and percentages for categorical data; BMI: Body mass index; K-MMSE: Korean mini-mental state examination; MoCA: Montreal cognitive assessment; UPDRS: Unified Parkinson\u0026rsquo;s disease rating score; BDI: Beck depression inventory; BAI: Beck anxiety inventory; PDSS2: Parkinson\u0026rsquo;s disease sleep scale 2nd version; ESS: Epworth sleepiness scale; FSS: Fatigue severity scale; FESK: Fall efficacy scale-Korean; SEE: Self-efficacy for exercise; NFOG-Q: New freezing of gait questionnaire; NMSS: Nonmotor symptoms scale; PDQ39: the 39-item Parkinson\u0026rsquo;s disease questionnaire; FTSST: five times sit-to-stand test; 6MWT: 6-minute walk test; SPPB: Short physical performance battery; Mini-BEST: Mini balance evaluation systems test; NQ: Nutrition quotient; IPAQ: international physical activity questionnaire.\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tfoot\u003e\n \u003c/table\u003e\n \u003cp\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003ePerformance evaluation of machine learning models\u003c/h3\u003e\n\u003cp\u003eThe model\u0026apos;s performance evaluation included reporting the average performance across five folds for each model. The performance of each model and a weighted average for each model using the coefficient of determination (R\u003csup\u003e2\u003c/sup\u003e) matrix are summarized in Table 2.\u0026nbsp;\u003c/p\u003e\n\u003ctable id=\"Tab3\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 2\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003eModel evaluation metrics and their corresponding weights.\u0026nbsp;\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eDomain\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eML models\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eMSE\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eR\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eWeights\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003ePDQ-39 total\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLASSO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.54\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.42\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.60\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eXGBoost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.67\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.29\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.40\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eMobility\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLASSO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.63\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.51\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.44\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eXGBoost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.45\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.66\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.56\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eADL\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLASSO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.40\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.51\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.44\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eXGBoost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.45\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.66\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.56\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eEmotional well-being\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLASSO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.83\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.33\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.60\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eXGBoost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.97\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.22\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.40\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003c/p\u003e\n\u003cp\u003ePDQ-39: Parkinson\u0026rsquo;s disease questionnaire-39; ADL: Activity of daily living; ML: Machine learning; LASSO: Least absolute shrinkage and selection operator; XGBoost: Extreme gradient boosting.\u003c/p\u003e\n\u003ch3\u003ePredictors of overall HRQoL\u003c/h3\u003e\n\u003cp\u003eSupplementary Table S1 presents each feature importance derived from Least Absolute Shrinkage and Selection Operator (LASSO) regression and Extreme Gradient Boosting (XGBoost) algorithms. Figure 2 illustrates the average permutation importance of 41 selected features across models. After feature reduction using linear regression, the most significant predictors of the Parkinson\u0026rsquo;s disease questionnaire-39 (PDQ-39) total score were Beck Anxiety Inventory (BAI) (\u0026beta;\u0026thinsp;=\u0026thinsp;0.341, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), Fall Efficacy Scale-Korean (FES-K) (\u0026beta;\u0026thinsp;=\u0026thinsp;0.446, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), Hoehn and Yahr (H\u0026amp;Y) (\u0026beta;\u0026thinsp;=\u0026thinsp;0.264, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), treatment duration (\u0026beta;\u0026thinsp;=\u0026thinsp;0.145, p\u0026thinsp;=\u0026thinsp;0.023), Left turning at faster speed (TurnFL)_posterior superior iliac spine (PSI) sample entropy for gyroscope (SampEnGyr) (\u0026beta;\u0026thinsp;=\u0026thinsp;0.170, p\u0026thinsp;=\u0026thinsp;0.007), forward walking (FW)_right ankle (RANK) sample entropy for acceleration (SampEnAcc) (\u0026beta;\u0026thinsp;=\u0026thinsp;0.190, p\u0026thinsp;=\u0026thinsp;0.008), FW_PSI SampEnAcc (\u0026beta;\u0026thinsp;=\u0026thinsp;0.147, p\u0026thinsp;=\u0026thinsp;0.028), and FW_RANK maximum jerk value (Max Jerk) (\u0026beta;\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.194, p\u0026thinsp;=\u0026thinsp;0.004) (Table 3).\u003c/p\u003e\n\u003ctable id=\"Tab4\" border=\"1\" class=\"fr-table-selection-hover\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 3\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003eAssociation between features and PDQ-39 and its domain using linear regression analysis\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eVariables\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e\u0026Beta; (SE)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003et-value\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e95% CI for OR\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ep-value\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eR\u003csub\u003eN\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"8\"\u003e\n \u003cp\u003ePDQ-39 total\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBAI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.341 (0.062)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5.505\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.218\u0026ndash;0.464\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" rowspan=\"8\"\u003e\n \u003cp\u003e0.657\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFESK\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.446 (0.068)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6.521\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.310\u0026ndash;0.582\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFW_RANK_SampEnAcc\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.190 (0.070)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2.724\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.052\u0026ndash;0.329\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.008\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eH\u0026amp;Y\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.264 (0.065)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.032\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.134\u0026ndash;0.394\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTurnFL_PSI_SampEnGyr\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.170 (0.062)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2.762\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.048\u0026ndash;0.293\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.007\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFW_RANK_MaxJerk\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026minus;0.194 (0.065)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026minus;2.974\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026minus;0.324 \u0026ndash; \u0026minus;0.065\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.004\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTreatment_duration\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.145 (0.063)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2.304\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.020\u0026ndash;0.269\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.023\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFW_PSI_SampEnAcc\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.147 (0.066)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2.240\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.017\u0026ndash;0.277\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.028\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"8\"\u003e\n \u003cp\u003eADL\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUPDRS_total\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.409 (0.076)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5.377\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.067\u0026ndash;2.732\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" rowspan=\"8\"\u003e\n \u003cp\u003e0.610\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTurnPR_LANK_MaxAcc\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.175 (0.071)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2.481\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.082\u0026ndash;0.341\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.015\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFESK\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.375 (0.078)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.784\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.219\u0026ndash;0.530\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSEE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.224 (0.068)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.309\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.090\u0026ndash;0.359\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBW_RANK_SampEnAcc\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.186 (0.067)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2.789\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.053\u0026ndash;0.318\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.006\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSPPB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026minus;0.236 (0.073)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026minus;3.206\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026minus;0.381 \u0026ndash; \u0026minus;0.090\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.002\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTurnFL_T10_SampEnGyr\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.182 (0.068)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2.685\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.047\u0026ndash;0.317\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.009\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFW_DSL\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026minus;0.177(0.073)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026minus;2.403\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026minus;0.322 \u0026ndash; \u0026minus;0.031\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.018\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"7\"\u003e\n \u003cp\u003eMobility\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFESK\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.443 (0.068)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6.542\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.309\u0026ndash;0.578\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" rowspan=\"7\"\u003e\n \u003cp\u003e0.687\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUPDRS_total\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.259 (0.076)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.397\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.108\u0026ndash;0.411\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNFOGQ\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.156 (0.065)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2.386\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.026\u0026ndash;0.285\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.019\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFW_RANK_MaxJerk\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026minus;0.219 (0.060)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026minus;3.645\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026minus;0.338 \u0026ndash; \u0026minus; 0.100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFW_LANK_SampEnAcc\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.237 (0.062)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.820\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.114\u0026ndash;0.360\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6MWT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026minus;0.182 (0.064)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026minus;2.868\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026minus;0.309 \u0026ndash; \u0026minus;0.056\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.005\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBAI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.146 (0.059)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2.460\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.028\u0026ndash;0.264\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.016\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"4\"\u003e\n \u003cp\u003eEmotional well-being\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBAI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.363 (0.090)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.025\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.184\u0026ndash;0.542\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" rowspan=\"4\"\u003e\n \u003cp\u003e0.408\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBDI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.323 (0.090)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.570\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.143\u0026ndash;0.502\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTurnPL_LANK_MaxGyr\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.254 (0.081)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.125\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.093\u0026ndash;0.415\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.002\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6MWT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026minus;0.167 (0.082)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026minus;2.044\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026minus;0.329 \u0026ndash; \u0026minus;0.005\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.044\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003ctfoot\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"7\"\u003ePDQ-39: Parkinson\u0026rsquo;s disease questionnaire-39; ADL: Activity of daily living; OR: Odds ratio; B: Logistic regression coefficient; SE: Standard error; 95% CI: 95% confidence interval; \u003cstrong\u003eR\u003c/strong\u003e\u003csub\u003e\u003cstrong\u003eN\u003c/strong\u003e\u003c/sub\u003e\u003csup\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003c/sup\u003e is the fit statistic for the Nagelkerke model; BAI: Beck anxiety inventory; FESK: Falls efficacy scale-Korea: FW: forward walking; SampEn: Sample entropy; Acc: Acceleration; RANK: right ankle; H\u0026amp;Y: Hoehn \u0026amp; yahr scale; TurnFL: Left turning at faster speed; Gyr: gyroscope; UPDRS: Unified Parkinson\u0026rsquo;s disease rating scale; TurnPR: Right turning at preferred speed; LANK: Left ankle; FESK: Fall efficacy scale-korean; SEE: Self-efficacy for exercise; BW: Backward walking; SPPB: Short Physical Performance Battery; DSL: Left double support phase; NFOGQ: New freezing of gait questionnaire; 6MWT: 6-minute walk test; BDI: Beck depression inventory, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tfoot\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003c/p\u003e\n\u003cp\u003eThe network structure between these predictors and the PDQ-39 total score is depicted in Fig.\u0026nbsp;3. All predictors were associated with the PDQ-39 total score, with the strongest connection observed between FES-K and PDQ-39 (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.446). Additional key associations included BAI (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.341), H\u0026amp;Y (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.264), FW_RANK Max Jerk (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.194), FW_RANK SampEnAcc (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.190), TurnFL_PSI SampEnGyr (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.170), FW_PSI SampEnAcc (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.147), and treatment duration (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.145). Within the network of predictor variables, FES-K was linked to H\u0026amp;Y (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.204) and FW_RANK SampEnAcc (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.116). FW_RANK SampEnAcc was associated with FW_PSI SampEnAcc (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.184) and FW_RANK Max Jerk (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.184) and was also connected to TurnFL_PSI SampEnGyr (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.103).\u003c/p\u003e\n\u003ch3\u003ePredictors of subdimensions\u003c/h3\u003e\n\u003cp\u003eThe permutation importance of the variables identified through LASSO regression and XGBoost algorithms for PDQ-39 subdimensions is presented in Supplementary Table S1.\u003c/p\u003e\n\u003cp\u003eFor the mobility dimension, Fig.\u0026nbsp;2 illustrates the average permutation importance of 43 selected features across models. Linear regression analyses identified the most significant predictors of mobility as FES-K (\u0026beta;\u0026thinsp;=\u0026thinsp;0.438, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), UPDRS total score (\u0026beta;\u0026thinsp;=\u0026thinsp;0.157, p\u0026thinsp;=\u0026thinsp;0.001), New freezing of gait questionnaire (NFOGQ) (\u0026beta;\u0026thinsp;=\u0026thinsp;0.138, p\u0026thinsp;=\u0026thinsp;0.019), FW_RANK Max Jerk (\u0026beta;\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.229, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), FW_left ankle (LANK) SampEnAcc (\u0026beta;\u0026thinsp;=\u0026thinsp;0.243, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), 6-Minute Walk Test (6MWT) (\u0026beta;\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.159, p\u0026thinsp;=\u0026thinsp;0.005), and BAI (\u0026beta;\u0026thinsp;=\u0026thinsp;0.145, p\u0026thinsp;=\u0026thinsp;0.016) (Table\u0026nbsp;3). In the network analysis, all predictors were associated with mobility, with the strongest connection observed between FES-K and mobility (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.438). Additional key associations included FW_LANK SampEnAcc (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.243), FW_RANK Max Jerk (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.229), 6MWT (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.159), UPDRS total score (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.157), BAI (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.145), and NFOGQ (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.138). Within the network of predictor variables, the UPDRS total score was linked to H\u0026amp;Y (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.471), FES-K (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.215), NFOGQ (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.182), 6MWT (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.179), and BAI (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.114). FES-K was associated with NFOGQ (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.127), while 6MWT was connected to H\u0026amp;Y (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.153). In addition, FW_RANK Max Jerk was linked to FW_LANK SampEnAcc (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.163).\u003c/p\u003e\n\u003cp\u003eFor the ADL dimension, Fig.\u0026nbsp;2 illustrates the average permutation importance of 28 selected features across models. Linear regression analyses identified the most significant predictors of ADL as UPDRS total score (\u0026beta;\u0026thinsp;=\u0026thinsp;0.409, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), right turning at preferred speed (TurnPR)_LANK maximum acceleration value (MaxAcc) (\u0026beta;\u0026thinsp;=\u0026thinsp;0.175, p\u0026thinsp;=\u0026thinsp;0.015), FES-K (\u0026beta;\u0026thinsp;=\u0026thinsp;0.375, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), Self-Efficacy for Exercise (SEE) (\u0026beta;\u0026thinsp;=\u0026thinsp;0.224, p\u0026thinsp;=\u0026thinsp;0.001), backward walking (BW)_RANK SampEnAcc (\u0026beta;\u0026thinsp;=\u0026thinsp;0.186, p\u0026thinsp;=\u0026thinsp;0.006), Short Physical Performance Battery (SPPB) (\u0026beta;\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.236, p\u0026thinsp;=\u0026thinsp;0.002), TurnFL_T10 SampEnGyr (\u0026beta;\u0026thinsp;=\u0026thinsp;0.182, p\u0026thinsp;=\u0026thinsp;0.009), and FW_left double support phase (DSL) (\u0026beta;\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.177, p\u0026thinsp;=\u0026thinsp;0.018) (Table\u0026nbsp;3). In the network analysis, all predictors were associated with ADL, with the strongest connection observed between UPDRS total score and ADL (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.409). Additional key associations included FES-K (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.375), SEE (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.224), SPPB (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.236), TurnPR_LANK MaxAcc (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.175), BW_RANK SampEnAcc (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.186), TurnFL_T10 SampEnGyr (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.182), and FW_Double Support (L) (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.177). Within the network of predictor variables, the UPDRS total score was linked to FES-K (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.319) and SPPB (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.204). TurnPR_LANK MaxAcc was associated with TurnFL_T10 SampEnGyr (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.223) and FW_DSL (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.143). FES-K was further connected to SEE (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.235) and BW_RANK SampEnAcc (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.138). In addition, SPPB and FW_DSL exhibited a strong association (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.288).\u003c/p\u003e\n\u003cp\u003eFor the emotional dimension, Fig.\u0026nbsp;2 illustrates the average permutation importance of 34 selected features across models. Linear regression analyses identified the most significant predictors of emotional well-being dimension as BAI (\u0026beta;\u0026thinsp;=\u0026thinsp;0.363, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), Beck Depression Inventory (BDI) (\u0026beta;\u0026thinsp;=\u0026thinsp;0.323, p\u0026thinsp;=\u0026thinsp;0.001), left turning at preferred speed (TurnPL)_LANK maximum gyroscope value (MaxGyr) (\u0026beta;\u0026thinsp;=\u0026thinsp;0.254, p\u0026thinsp;=\u0026thinsp;0.002), and 6MWT (\u0026beta;\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.167, p\u0026thinsp;=\u0026thinsp;0.044) (Table\u0026nbsp;3). In the network analysis, all predictors were associated with the emotional domain, with the strongest connection observed between BAI and emotional well-being (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.363). Additional key associations included BDI (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.323), TurnPL_LANK MaxGyr (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.254), and 6MWT (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.167). Within the network of predictor variables, BAI exhibited a strong association with BDI (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.429). In addition, TurnPL_LANK MaxGyr was connected to 6MWT (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.228).\u003c/p\u003e"},{"header":"DISCUSSION","content":"\u003cp\u003eThis study identified non-motor and motor symptoms as key predictors of HRQoL in individuals with PD, with an overall explanatory power of 65.7%. Of the non-motor variables, anxiety (BAI), fear of falling (FOF; FESK), H\u0026amp;Y stage, and treatment duration emerged as significant predictors. Motor predictors included sample entropy of acceleration, angular velocity, and maximum jerk. Specifically, TurnFL_PSI SampEnGyr (β = 0.170), FW_RANK SampEnAcc (β = 0.190), and FW_PSI SampEnAcc (β = 0.147) were positively associated with higher PDQ-39 scores, indicating poorer HRQoL. While higher maximal jerk in FW was associated with better HRQoL in this study, previous studies show that individuals with PD typically exhibit reduced jerk, which may contribute to lower perceived mobility [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. This finding suggests that as ankle acceleration and angular velocity are more irregular, which reflected in increased sample entropy, HRQoL become declines [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eA healthy gait is marked by flexibility and adaptability, reflected in complex, fractal-like stride patterns. Sample entropy, derived from approximate entropy, quantifies movement regularity and is an effective marker of gait dysfunction [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. In PD, degeneration of motor control leads to increased entropy and loss of complexity in gait, signifying impaired adaptability [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. This study highlights the relevance of PSI and ankle-based metrics in capturing real-world gait characteristics [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. Depression has traditionally been cited as the strongest HRQoL predictor in PD [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]; however, this study emphasized the importance of anxiety, FOF, treatment duration, and H\u0026amp;Y stage [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. These results align with the known “wearing-off” effect of long-term dopaminergic therapy, which contributes to worsening motor and non-motor symptoms and reduced QoL [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. Treatment fluctuations often provoke anxiety, particularly fear of falling [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIn this study, our results underscore FOF (FES-K) as a dominant predictor of overall HRQoL, consistent with earlier research [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. FOF—a lack of confidence in safely performing daily tasks—leads to activity restriction and social isolation, worsening QoL [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. Network analysis revealed notable associations between FOF and both the H\u0026amp;Y stage (r = 0.204) and SampEnAcc (r = 0.116). As PD progresses, declining postural stability and heightened FOF reduce mobility and independence. FOF, although often underrecognized, plays a critical role in limiting function by diminishing confidence in walking ability [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. Thus, interventions targeting gait irregularities and FOF may positively impact HRQoL.\u003c/p\u003e \u003cp\u003eThe mobility dimension of PDQ-39 was predicted by BAI (β = 0.145), FESK (β = 0.438), UPDRS total score (β = 0.157), NFOGQ (β = 0.138), 6MWT distance (β = -0.159), FW_RANK Max Jerk (β = -0.229), and FW_LANK SampEnAcc (β = 0.234), with an explanatory power of 68.7%. These findings are consistent with previous research linking mobility-related QoL to anxiety [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e], FOF [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e], motor severity [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e], FOG [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e], and walking performance [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThese results suggest that reduced walking ability and gait instability strongly influence perceptions of mobility and QoL [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]. Notably, NFOGQ, a measure of FOG, was closely linked to FOF in the network analysis, indicating their joint role in diminishing functional independence. FOG-related motor instability, paired with psychological fear, may cause individuals to avoid walking and limit social interaction, contributing to physical deconditioning and reduced HRQoL [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThe ADL dimension was predicted by UPDRS total score (β = 0.409), TurnPR_LANK MaxAcc (β = 0.175), FESK (β = 0.375), SEE (β = -0.224), BW_RANK SampEnAcc (β = 0.186), SPPB (β = -0.236), TurnFL_T10 SampEnGyr (β = 0.182), and FW DSL (β = -0.177), explaining 60.1% of the variance. Reduced walking capacity and balance—reflected in prolonged double support time and irregular gait patterns—negatively influenced self-perceptions of functional ability [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]. Network analysis highlighted UPDRS as a central node, linking disease severity with reduced physical performance and increased FOF. As PD progresses, motor decline, functional impairment, and fear of movement interact to restrict independence in daily activities [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]. These results emphasize the need for interventions that address motor control and non-motor symptoms to improve ADL-related QoL.\u003c/p\u003e \u003cp\u003eThe emotional well-being domain was predicted by BAI (β = 0.363), BDI (β = 0.323), TurnPL_LANK MaxGyr (β = 0.254), and 6MWT (β = -0.167), with an explanatory power of 40.8%. Consistent with the findings of previous studies, anxiety had a stronger impact than motor severity or disease duration [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. Network analysis revealed a strong correlation between anxiety and depression, which are often comorbid in PD and associated with embarrassment about visible symptoms and resulting social withdrawal [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]. In addition, reduced 6MWT distance reflects impaired aerobic capacity and mobility [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e], which may contribute to emotional distress and diminished self-efficacy. Our findings also show that higher maximal angular velocity during turning—measured using gyroscopes—was linked to lower HRQoL. Individuals with PD often reduce turning speed to maintain balance, and such compensatory behavior may stem from FOF in public, further impacting emotional well-being [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e]. These findings suggest that emotional well-being interventions should go beyond treating anxiety and depression and incorporate physical strategies, such as improving aerobic capacity and promoting safer turning techniques [\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThis study has some limitations. First, its cross-sectional design precludes causal inference or longitudinal analysis of HRQoL changes over time. Future longitudinal studies are needed to determine whether modifying key predictors improves HRQoL. Second, all assessments were conducted during the ON-medication state. Evaluating participants during OFF-medication states may yield deeper insights into the relationship between motor function and QoL [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. Third, despite adjusting the hyperparameter in the LASSO model, the performance for certain subdomains of PDQ-39, including stigma, communication, bodily discomfort, cognition, and social support, remained negative. This result may be attributable to the relatively weak correlation between these subdomains and the PDQ-39 total score (see Supplementary Table S2 for correlations between the PDQ-39 total score and its subdimensions) and may explain the lack of association with variables. Finally, this study is based on a single-center Korean cohort, which may limit the generalizability of our results depending on the clinical and cultural backgrounds. Therefore, future studies should include a larger validation multiethnic cohort [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIn conclusion, this study identified key predictors of HRQoL and its domains in individuals with PD, integrating clinical, psychological, and sensor-based motor data through machine learning and network analysis. Anxiety, fear of falling, and sample entropy of acceleration emerged as central predictors of overall HRQoL and domain-specific outcomes. These findings underscore the importance of combining objective motor assessments with psychological and clinical evaluations to inform targeted interventions. Addressing motor irregularities and psychological barriers may be essential for improving QoL in individuals living with PD.\u003c/p\u003e "},{"header":"METHODS","content":"\u003ch2\u003eParticipants\u003c/h2\u003e\u003cp\u003eThis study included 101 individuals with PD. Participants were selected based on the following criteria: (a) a clinician-confirmed diagnosis of idiopathic PD according to the UK Parkinson’s Disease Society Brain Bank Clinical Diagnostic Criteria [\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e], (b) age between 40 and 85 years, and (c) H\u0026amp;Y stages 1–3, indicating independent walking and mobility without assistive devices. Individuals were excluded if they had a history of cardiovascular, musculoskeletal, vestibular, or other neurological disorders, required mobility-assistive devices, or had pharmacotherapy-refractory dyskinesia. All participants were in the ON medication state during experimental protocols, ensuring they experienced the full effects of their medication. All study procedures involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards. The study protocol and supplementary information files were approved by the Institutional Review Board of Dong-A University Hospital (IRB number: DAUHIRB-22-089) (see ethics approval letter in the supplementary file). All patients were informed of the experimental procedures and provided written informed consent before data collection. The study is registered with the Clinical Research Information Service in the Republic of Korea (KCT0009353).\u003c/p\u003e\u003ch3\u003eExperimental protocol\u003c/h3\u003e\u003cp\u003e Ethical consent was collected after participants’ recruitment, and all participants underwent physical assessments and clinician-administered evaluations of their clinical characteristics. Cognitive function was assessed using the Korean Mini-Mental State Examination (K-MMSE) and the Montreal Cognitive Assessment (MoCA). Gait and non-motor symptoms were evaluated using the NFOGQ and the Non-Motor Symptoms Scale in PD (NMSS), respectively. Fall-related concerns were measured with the FES-K, while exercise self-efficacy was assessed using the SEE. Psychological and sleep disturbances were evaluated using the BDI, BAI, Parkinson’s Disease Sleep Scale 2nd version (PDSS-2), ESS, and Fatigue Severity Scale (FSS). Disease severity and motor symptoms were assessed using the UPDRS, which includes Parts I (non-motor aspects of daily living), II (motor aspects of daily living), III (motor examination), and IV (motor complications), as well as the H\u0026amp;Y scale. They also responded to lifestyle questionnaires, including nutrition quotient (NQ) and the international physical activity questionnaire (IPAQ).\u003c/p\u003e\u003cp\u003eParticipants’ physical characteristics, including height, body weight, shoulder offset, elbow width, wrist width, hand thickness, leg length, knee width, and ankle width measurements, were measured in the second session. Participants completed a 5-min warm-up, including stretching exercises, before beginning the experiment. We used nine infrared cameras (Vicon MX-T10, Oxford Metrics, UK) and wearable sensors with a stretchable belt (Xsens DOT, Movella Technologies, Enschede, Netherlands) for this experiment.\u003c/p\u003e\u003cp\u003eParticipants were instructed to walk at a comfortable and fast pace. The walking tasks included (a) forward walking (FW), (b) backward walking (BW) at a self-selected speed, and (c) 360° turns in the right (Turn_PSR, Turn_FSR) and left (Turn_PSL, Turn_FSL) directions at self-selected and maximum speeds [\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e]. Each measurement was repeated three times, and the mean value was calculated. Finally, participants underwent postural and functional assessments, including the Short Physical Performance Battery (SPPB), Mini Balance Evaluation Systems Test (Mini-BEST), grip strength measurement, the Five Times Sit-to-Stand Test (FTSST), and the 6-Minute Walk Test (6MWT).\u003c/p\u003e\u003ch2\u003eData analysis\u003c/h2\u003e\u003cp\u003eThe physical model was constructed using 39 round reflective markers based on the Plug-in Gait Full Body Model (Vicon Motion Systems Ltd., Oxford, UK), a modified version of the Helen Hayes Marker Set. A global coordinate system was established, with the positive X-axis to the right, the positive Y-axis facing anteriorly, and the Z-axis defined as the cross-product between the X-axis and Y-axis, with the positive Z-axis facing superiorly. Data was recorded at a sampling rate of 100 Hz and filtered using a fourth-order Butterworth low-pass filter at 10 Hz [\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eSpatiotemporal gait variables were extracted from three-dimensional marker data and included walking speed (m/s), stride length (m), single-support phase (%), double-support phase (%), and contralateral temporal coordination. Left–right contralateral temporal coordination (CLL) was defined as the average difference between the left elbow and right knee, and right–left contralateral temporal coordination (CLR) used as that between the right elbow and left knee.\u003c/p\u003e\u003cp\u003eIn addition, wearable sensors were attached to six anatomical landmarks, including the left and right upper arms (LELB, RELB) (5 cm above the lateral humeral epicondyle), left and right lateral malleolus (LANK, RANK), 10th thoracic spine (T10), and lumbar spine (center of the left and right posterior superior iliac spine [PSI]) [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e]. The raw data obtained from the wearable sensor using inertial measurement units (IMU) was recorded at a sampling rate of 60 Hz in each axis (x-axis: vertical, y-axis: mediolateral, and z-axis: anteroposterior).\u003c/p\u003e\u003cp\u003eExtracted gait features from IMU sensors included maximum jerk (Maxjerk) and angular velocity jerk (Maxangveljerk); mean acceleration (Meanacc) and gyroscope values (Meangyr); maximum acceleration (Maxacc) and gyroscope values (Maxgyr); root mean square (RMS) acceleration (RMSacc) and gyroscope values (RMSgyr); and sample entropy for acceleration (SampEnacc) and gyroscope (SampEngyr) signals. Supplementary Table S3 provides details on the estimation of these metrics.\u003c/p\u003e\u003ch2\u003ePre-processing\u003c/h2\u003e\u003cp\u003eThe dataset was pre-processed to identify its structure and any inconsistencies. Pre-processing involved two main steps. First, we used k-nearest neighbor (k-NN) imputation to address the missing values, which were estimated and replaced using a k-NN imputer with five neighbors and leveraged the average of the nearest data points to ensure informed imputations. Lastly, this dataset was normalized with z-score normalization, which enabled the comparability of different features by consistent scaling.\u003c/p\u003e\u003ch2\u003eStatistical analysis\u003c/h2\u003e\u003cp\u003eAll data were tested for normality using the Shapiro–Wilk test, and statistical analyses were conducted using parametric and nonparametric methods. Demographic and clinical characteristics were summarized using mean and standard deviation (SD) and counts and percentages for categorical data. Spearman’s rho correlation analysis was performed to identify subdomains of the PDQ-39 that were significantly associated with the total PDQ-39 score. Statistical analyses were performed using SPSS 21.0 (IBM Corp., Armonk, NY, USA), and data preprocessing and analysis were conducted using MATLAB R2017b (MathWorks, Natick, MA, USA) and Python (version 3.10) with Pandas, NumPy, Scikit-learn.\u003c/p\u003e\u003ch2\u003ePrediction models\u003c/h2\u003e\u003ch2\u003eLeast Absolute Shrinkage and Selection Operator\u003c/h2\u003e\u003cp\u003eThe Least Absolute Shrinkage and Selection Operator (LASSO) algorithm was employed to analyze and identify variables influencing the total PDQ-39 score and its dimensions. LASSO is a linear regression technique that incorporates a penalty term based on the L1 norm to constrain model coefficients, effectively setting some coefficients to zero and enabling automatic feature selection [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e]. The optimization objective of LASSO is expressed as follows:\u003c/p\u003e\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:\\text{arg}min\\sum\\:_{n=1}^{N}{\\left({\\mathcal{Y}}^{n}-\\sum\\:_{q=1}^{m}{\\beta\\:}_{q}{\\mathcal{x}}_{q}^{n}\\right)}^{2}+\\:\\lambda\\:\\sum\\:_{q=1}^{m}\\left|{\\beta\\:}_{q}\\right|$$\u003c/div\u003e\u003c/div\u003e\u003cp\u003e \u003cspan class=\"InlineEquation\"\u003e \u003cspan class=\"mathinline\"\u003e\\(\\:{\\mathcal{Y}}^{n}\\)\u003c/span\u003e \u003c/span\u003e represents the class label for the \u003cem\u003en\u003c/em\u003e-th trial, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:N\\)\u003c/span\u003e\u003c/span\u003e is the total number of trials, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\mathcal{x}}_{q}^{n}\\)\u003c/span\u003e\u003c/span\u003e denoted the \u003cem\u003eq\u003c/em\u003e-th clinical factor of the \u003cem\u003en\u003c/em\u003e-th sample, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\beta\\:}_{q}\\)\u003c/span\u003e\u003c/span\u003e is the regression coefficient of the \u003cem\u003eq\u003c/em\u003e-th factor, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\lambda\\:\\)\u003c/span\u003e\u003c/span\u003e is the regularization parameter, where a larger \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\lambda\\:\\)\u003c/span\u003e\u003c/span\u003e results in a sparser model, setting more \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\beta\\:}_{q}\\)\u003c/span\u003e\u003c/span\u003e values to zero.\u003c/p\u003e\u003cp\u003eWe applied a 5-fold cross-validation approach to optimize the LASSO regularization parameter (alpha), which was selected based on the highest mean R² across the five folds to assess model selection stability. The data were split into training (80%) and test (20%) sets, ensuring model evaluation was performed on unseen data. After identifying the optimal alpha, the model was retrained using the entire training set, and the final evaluation was performed on the independent test set (20% of the data). The final model was evaluated using mean squared error (MSE) and coefficient of determination (R\u003csup\u003e2\u003c/sup\u003e), and the final selected variables were defined as those with non-zero regression coefficients.\u003c/p\u003e\u003ch2\u003eExtreme Gradient Boosting\u003c/h2\u003e\u003cp\u003eThe Extreme Gradient Boosting (XGBoost) classification algorithm is a high-performance boosting technique that minimizes the loss function through iterative model addition. XGBoost sequentially adds models to the ensemble in each iteration. The algorithm generates feature importance scores based on the gain score during the training process [\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e]. The model's output is described as follows:\u003c/p\u003e\u003cdiv id=\"Equb\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equb\" name=\"EquationSource\"\u003e\n$$\\:{\\widehat{\\mathcal{Y}}}_{i}=\\sum\\:_{k=1}^{K}{f}_{k}\\left({\\mathcal{x}}_{\\mathcal{i}}\\right),{f}_{k}\\in\\:F,\\:\\in\\:F,\\:k=1$$\u003c/div\u003e\u003c/div\u003e\u003cp\u003eWhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:k\\)\u003c/span\u003e\u003c/span\u003e represents the total number of predictors, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{f}_{k}\\)\u003c/span\u003e\u003c/span\u003e is the function corresponding to the \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:k\\)\u003c/span\u003e\u003c/span\u003e-th predictor in the functional space \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:F\\)\u003c/span\u003e\u003c/span\u003e. We used the training feature \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\mathcal{x}}_{\\mathcal{i}}\\)\u003c/span\u003e\u003c/span\u003e to predict the target variable \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\mathcal{Y}}_{i}\\)\u003c/span\u003e\u003c/span\u003e. We ranked the variables in order of importance and selected the variables that covered at least 80% of the cumulative score.\u003c/p\u003e\u003cp\u003eThe dataset was split into training (80%) and testing (20%) sets. Hyperparameter tuning was performed using the grid search method with 3-fold cross-validation to identify the optimal parameters. In particular, the number of estimators (50, 100, 200), learning rate (0.01, 0.1, 0.2), and maximum tree depth (3, 5, 7) were tuned. In addition, the generalization performance of the optimized model was evaluated using 5-fold cross-validation on the entire dataset. The final model was evaluated using MSE and R\u003csup\u003e2\u003c/sup\u003e.\u003c/p\u003e\u003ch2\u003eWeighted Average Ensemble Method\u003c/h2\u003e\u003cp\u003eThe weighted average ensemble method enhances prediction accuracy by assigning greater importance to models with better performance. To generate the final prediction, a weighted average of each model's output was calculated using the following Equation, where the predictions of all models were combined with their corresponding weights [\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e]. The final feature importance was obtained by summing the feature importance values from each model.\u003c/p\u003e\u003cdiv id=\"Equc\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equc\" name=\"EquationSource\"\u003e\n$$\\:{P}_{final}=\\frac{{\\mathcal{W}}_{1}\\bullet\\:\\:{P}_{1}+{\\mathcal{W}}_{2}\\bullet\\:\\:{P}_{2}+{\\mathcal{W}}_{3}\\bullet\\:\\:{P}_{3}+\\dots\\:+{\\mathcal{W}}_{N}\\bullet\\:\\:{P}_{N}}{{\\mathcal{W}}_{1}+{\\mathcal{W}}_{2}+{\\mathcal{W}}_{3}+\\dots\\:+{\\mathcal{W}}_{N}}$$\u003c/div\u003e\u003c/div\u003e\u003cp\u003eIn the ensemble classifier, \u003cem\u003eN\u003c/em\u003e models generate predictions for a given input, where \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{P}_{1}\\)\u003c/span\u003e\u003c/span\u003e, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{P}_{2}\\)\u003c/span\u003e\u003c/span\u003e, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{P}_{3}\\)\u003c/span\u003e\u003c/span\u003e, …, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{P}_{N}\\)\u003c/span\u003e\u003c/span\u003e represent the predictions from Models 1, 2, 3, and so on. The predictions from all models were combined with their corresponding weights, denoted as \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\mathcal{W}}_{1}\\)\u003c/span\u003e\u003c/span\u003e, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\mathcal{W}}_{2}\\)\u003c/span\u003e\u003c/span\u003e, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\mathcal{W}}_{3}\\)\u003c/span\u003e\u003c/span\u003e,…, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\mathcal{W}}_{N}\\)\u003c/span\u003e\u003c/span\u003e, to create the weighted average ensemble classifier. Finally, the optimal variables for evaluating the determinants of HRQoL in individuals with PD were identified using a forward stepwise linear regression.\u003c/p\u003e\u003ch2\u003eNetwork analysis\u003c/h2\u003e\u003cp\u003eNetwork analysis was employed to explore the associations between selected variables and HRQoL based on partial correlations. We applied the graphical lasso model [\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e] to estimate the partial correlations between variables and shrink the less relevant edges to zero. Therefore, this network structure represents the regularized partial correlations, highlighting the most relevant relationships while minimizing irrelevant connections. The nodes with the darkest color correspond to the raw scores of the PDQ-39 total score and its eight subscales. In addition, all nodes were positioned using the spring_layout of the Fruchterman-Reingold force-directed algorithm based on the strength of the connections between them [\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e]. Edge thickness reflects the strength of the relationships, using regularized partial correlations between nodes (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eACKNOWLEDGMENTS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors thank all the participants of this study. This work was supported by a Dong-A University Research Fund and grants from the National Research Foundation of Korea (NRF) funded by the Korean government (MSIT) (No. 2022R1A2C100933711; Changhong Youm), the Basic Science Research Program through the NRF funded by the Ministry of Education (No. 2022R1A6A3A0108756411; Hwayoung Park), and the Ministry of Education of the Republic of Korea and the NRF (No. 2024S1A5B5A16021673; Hwayoung Park). This study received no specific grants from funding agencies in the public, commercial, or non-profit sectors. The funding sources had no role in the study design; collection, analysis, and interpretation of the data; or in writing the manuscript. The authors also thank Editage (www.editage.co.kr) for English language editing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAUTHOR\u0026rsquo;S CONTRIBUTIONS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eJH, CY, HP, BK, HC, MK, and SC conceived and designed the study. JH, HP, BK, HC, MK, and SC recruited the participants. JH, CY, HP, BK, HC, MK, and SC acquired the data. JH, CY, HP, BK, HC, MK, and SC analyzed and interpreted the data. JH, CY, HP, BK, HC, MK, and SC drafted the article. All authors read and approved the final version of the manuscript submitted.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCOMPETING INTERESTS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests. The code for training and testing the ML models was written in Python 3.8 using PyTorch 1.9.1 and Torchvision 0.10.1. Data management and feature processing scripts were written in Python 3.8 using Pandas 1.3.3 and NumPy 1.21.2. The datasets supporting this study\u0026apos;s findings, as well as the code used for the analysis, are available from the corresponding author upon reasonable request.\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eDATA AVAILABILITY STATEMENT\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets supporting the findings of this study are available from the corresponding author upon reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eFayers, P. M. \u0026amp; Machin, D. Quality of Life: the Assessment, Analysis and Interpretation of Patient-Reported Outcomes (John Wiley \u0026amp; Sons, 2013).\u003c/li\u003e\n\u003cli\u003eHaraldstad, K. et al. A systematic review of quality of life research in medicine and health sciences. \u003cem\u003eQual. Life Res.\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 2641-2650 (2019).\u003c/li\u003e\n\u003cli\u003eBatista, P. S. P. \u0026amp; Pereira, A. Quality of life in patients with neurodegenerative diseases. \u003cem\u003eJ. Neurol. Neurosci.\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 74 (2016).\u003c/li\u003e\n\u003cli\u003eNance, M. A. Comprehensive care in Huntington\u0026apos;s disease: a physician\u0026apos;s perspective. \u003cem\u003eBrain Res. Bull.\u003c/em\u003e \u003cstrong\u003e72\u003c/strong\u003e, 175-178 (2007).\u003c/li\u003e\n\u003cli\u003eAlexander, G. E., DeLong, M. R. \u0026amp; Strick, P. L. Parallel organization of functionally segregated circuits linking basal ganglia and cortex. \u003cem\u003eAnnu. Rev. Neurosci.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 357-381 (1986).\u003c/li\u003e\n\u003cli\u003eTakakusaki, K., Tomita, N. \u0026amp; Yano, M. Substrates for normal gait and pathophysiology of gait disturbances with respect to the basal ganglia dysfunction. \u003cem\u003eJ. Neurol.\u003c/em\u003e \u003cstrong\u003e255\u003c/strong\u003e (Supplement 4), 19-29 (2008).\u003c/li\u003e\n\u003cli\u003eTosin, M. H., Goetz, C. G. \u0026amp; Stebbins, G. T. Patient with Parkinson disease and care partner perceptions of key domains affecting health-related quality of life: systematic review. \u003cem\u003eNeurology\u003c/em\u003e \u003cstrong\u003e102\u003c/strong\u003e, e208028 (2024).\u003c/li\u003e\n\u003cli\u003eOpara, J. A., Brola, W., Leonardi, M. \u0026amp; Błaszczyk, B. Quality of life in Parkinson\u0026rsquo;s disease. \u003cem\u003eJ. Med. Life\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 375\u0026ndash;381 (2012).\u003c/li\u003e\n\u003cli\u003eSoh, S. E., Morris, M. E. \u0026amp; McGinley, J. L. Determinants of health-related quality of life in Parkinson\u0026rsquo;s disease: a systematic review. \u003cem\u003eParkinsonism Relat. Disord.\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 1-9 (2011).\u003c/li\u003e\n\u003cli\u003eZhao, N. et al. Quality of life in Parkinson\u0026apos;s disease: A systematic review and meta‐analysis of comparative studies. \u003cem\u003eCNS Neurosci. Ther.\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 270-279 (2021).\u003c/li\u003e\n\u003cli\u003eBock, M. A., Brown, E. G., Zhang, L. \u0026amp; Tanner, C. Association of motor and nonmotor symptoms with health-related quality of life in a large online cohort of people with Parkinson disease. \u003cem\u003eNeurology\u003c/em\u003e \u003cstrong\u003e98\u003c/strong\u003e, e2194\u0026ndash;e2203 (2022).\u003c/li\u003e\n\u003cli\u003eHagell, P. \u0026amp; Nilsson, M. H. The 39-item Parkinson\u0026apos;s disease Questionnaire (PDQ-39): is it a unidimensional construct? \u003cem\u003eTher. Adv. Neurol. Disord.\u003c/em\u003e \u003cstrong\u003e2\u003c/strong\u003e, 205-214 (2009).\u003c/li\u003e\n\u003cli\u003eSanchez-Luengos, I. et al\u003cem\u003e.\u003c/em\u003e Predictors of health-related quality of life in Parkinson\u0026rsquo;s disease: the impact of overlap between health-related quality of life and clinical measures. \u003cem\u003eQual. Life Res.\u003c/em\u003e \u003cstrong\u003e31\u003c/strong\u003e, 3241-3252 (2022).\u003c/li\u003e\n\u003cli\u003eBoonstra, T. A., Van Der Kooij, H., Munneke, M. \u0026amp; Bloem, B. R. Gait disorders and balance disturbances in Parkinson\u0026apos;s disease: clinical update and pathophysiology. \u003cem\u003eCurr. Opin. Neurol.\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 461-471 (2008).\u003c/li\u003e\n\u003cli\u003eHorak, F. B. \u0026amp; Mancini, M. Objective biomarkers of balance and gait for Parkinson\u0026apos;s disease using body‐worn sensors. \u003cem\u003eMov. Disord.\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 1544-1551 (2013).\u003c/li\u003e\n\u003cli\u003eGon\u0026ccedil;alves, H. R., Branquinho, A., Pinto, J., Rodrigues, A. M. \u0026amp; Santos, C. P. Digital biomarkers of mobility and quality of life in Parkinson\u0026apos;s disease based on a wearable motion analysis LAB. \u003cem\u003eComput. Methods Programs Biomed.\u003c/em\u003e \u003cstrong\u003e244\u003c/strong\u003e, 107967 (2024).\u003c/li\u003e\n\u003cli\u003eMartinez‐Martin, P. What is quality of life and how do we measure it? Relevance to Parkinson\u0026apos;s disease and movement disorders. \u003cem\u003eMov. Disord.\u003c/em\u003e \u003cstrong\u003e32\u003c/strong\u003e, 382-392 (2017).\u003c/li\u003e\n\u003cli\u003eKazemi, K. \u003cem\u003eet al\u003c/em\u003e. Impact of physical activity on quality of life during pregnancy: A causal ML approach in 46th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) \u003cstrong\u003e2024\u003c/strong\u003e, (1-6) (2024).\u003c/li\u003e\n\u003cli\u003eJannani, A., Sael, N. \u0026amp; Benabbou, F. Artificial intelligence for quality of life study: A systematic literature review. \u003cem\u003eIEEE Access\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 62059\u0026ndash;62088 (2024).\u003c/li\u003e\n\u003cli\u003eBorsboom, D. A network theory of mental disorders. \u003cem\u003eWorld Psychiatry\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 5-13 (2017).\u003c/li\u003e\n\u003cli\u003eSch\u0026ouml;nenberg, A. et al. Using network analysis to explore the validity and influential items of the Parkinson\u0026rsquo;s disease Questionnaire-39. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 7221 (2023).\u003c/li\u003e\n\u003cli\u003eCurtze, C., Nutt, J. G., Carlson-Kuhta, P., Mancini, M. \u0026amp; Horak, F. B. Objective gait and balance impairments relate to balance confidence and perceived mobility in people with Parkinson disease. \u003cem\u003ePhys. Ther.\u003c/em\u003e \u003cstrong\u003e96\u003c/strong\u003e, 1734-1743 (2016).\u003c/li\u003e\n\u003cli\u003eBange, M. et al. Deficient interhemispheric connectivity underlies movement irregularities in Parkinson\u0026rsquo;s disease. \u003cem\u003eJ. Parkinsons Dis.\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 381-395 (2022).\u003c/li\u003e\n\u003cli\u003eKarmakar, C., Khandoker, A., Begg, R. \u0026amp; Palaniswami, M. Understanding ageing effects using complexity analysis of foot\u0026ndash;ground clearance during walking. \u003cem\u003eComput. Methods Biomech. Biomed. Eng.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 554-564 (2013).\u003c/li\u003e\n\u003cli\u003eYentes, J. M. et al\u003cem\u003e.\u003c/em\u003e The appropriate use of approximate entropy and sample entropy with short data sets. \u003cem\u003eAnn. Biomed. Eng.\u003c/em\u003e \u003cstrong\u003e41\u003c/strong\u003e, 349-365 (2013).\u003c/li\u003e\n\u003cli\u003eLuk\u0026scaron;ys, D. \u0026amp; Gri\u0026scaron;kevičius, J. Application of nonlinear analysis for the assessment of gait in patients with Parkinson\u0026rsquo;s disease\u003cem\u003e. Technol. Health Care\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 201-208 (2022).\u003c/li\u003e\n\u003cli\u003eBou\u0026ccedil;a-Machado, R. et al. Gait kinematic parameters in Parkinson\u0026rsquo;s disease: a systematic review. \u003cem\u003eJ. Parkinsons Dis.\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 843-853 (2020).\u003c/li\u003e\n\u003cli\u003eDissanayaka, N. N. N. W. et al. The clinical spectrum of anxiety in Parkinson\u0026apos;s disease. \u003cem\u003eMov. Disord.\u003c/em\u003e \u003cstrong\u003e29\u003c/strong\u003e, 967-975 (2014).\u003c/li\u003e\n\u003cli\u003eGe, Y. et al. Correlation between motor function and health-related quality of life in early to mid-stage patients with Parkinson disease: a cross-sectional observational study. \u003cem\u003eFront. Aging Neurosci.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 1399285 (2024).\u003c/li\u003e\n\u003cli\u003eAhlskog, J. E. \u0026amp; Muenter, M. D. Frequency of levodopa‐related dyskinesias and motor fluctuations as estimated from the cumulative literature. \u003cem\u003eMov. Disord.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 448-458 (2001).\u003c/li\u003e\n\u003cli\u003eDissanayaka, N. N. et al. Phenomenology of atypical anxiety disorders in Parkinson\u0026apos;s disease: a systematic review. \u003cem\u003eAm. J. Geriatr. Psychiatry\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 1026-1050 (2022).\u003c/li\u003e\n\u003cli\u003eBloem, B. R., Grimbergen, Y. A., Cramer, M., Willemsen, M. \u0026amp; Zwinderman, A. H. Prospective assessment of falls in Parkinson\u0026apos;s disease. \u003cem\u003eJ. Neurol.\u003c/em\u003e \u003cstrong\u003e248\u003c/strong\u003e, 950-958 (2001).\u003c/li\u003e\n\u003cli\u003eChuquil\u0026iacute;n-Arista, F., \u0026Aacute;lvarez-Avell\u0026oacute;n, T. \u0026amp; Men\u0026eacute;ndez-Gonz\u0026aacute;lez, M. Prevalence of depression and anxiety in Parkinson disease and impact on quality of life: a community-based study in Spain. \u003cem\u003eJ. Geriatr. Psychiatry Neurol.\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 207-213 (2020).\u003c/li\u003e\n\u003cli\u003eScalzo, P. L., Flores, C. R., Marques, J. R., Robini, S. C. O. \u0026amp; Teixeira, A. L. Impact of changes in balance and walking capacity on the quality of life in patients with Parkinson\u0026apos;s disease. \u003cem\u003eArq. Neuro Psiquiatr.\u003c/em\u003e \u003cstrong\u003e70\u003c/strong\u003e, 119-124 (2012).\u003c/li\u003e\n\u003cli\u003eBowman, T. et al. Predictors of mobility domain of health-related quality of life after rehabilitation in Parkinson\u0026rsquo;s disease: A pilot study. \u003cem\u003eArch. Physiother.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 10 (2018).\u003c/li\u003e\n\u003cli\u003eCanning, C. G., Allen, N. E., Dean, C. M., Goh, L. \u0026amp; Fung, V. S. C. Home-based treadmill training for individuals with Parkinson\u0026rsquo;s disease: a randomized controlled pilot trial. \u003cem\u003eClin. Rehabil.\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 817-826 (2012).\u003c/li\u003e\n\u003cli\u003eFranchignoni, F., Martignoni, E., Ferriero, G. \u0026amp; Pasetti, C. Balance and fear of falling in Parkinson\u0026apos;s disease\u003cem\u003e. Parkinsonism Relat. \u003c/em\u003e\u003cem\u003eDisord.\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 427-433 (2005).\u003c/li\u003e\n\u003cli\u003eGrimbergen, Y. A. M., Schrag, A., Mazibrada, G., Borm, G. F. \u0026amp; Bloem, B. R. Impact of falls and fear of falling on health-related quality of life in patients with Parkinson\u0026apos;s disease. \u003cem\u003eJ. Parkinsons Dis.\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 409-413 (2013).\u003c/li\u003e\n\u003cli\u003eHanna, K. K. \u0026amp; Cronin-Golomb, A. Impact of anxiety on quality of life in Parkinson\u0026prime;s disease. \u003cem\u003eParkinsons Dis.\u003c/em\u003e \u003cstrong\u003e2012\u003c/strong\u003e, 640707 (2012).\u003c/li\u003e\n\u003cli\u003eHaertner, L. et al. Effect of fear of falling on turning performance in Parkinson\u0026rsquo;s disease in the lab and at home. \u003cem\u003eFront. Aging Neurosci.\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 78 (2018).\u003c/li\u003e\n\u003cli\u003eShah, V. V. et al. Digital biomarkers of mobility in Parkinson\u0026rsquo;s disease during daily living. \u003cem\u003eJ. Parkinsons Dis.\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 1099-1111 (2020).\u003c/li\u003e\n\u003cli\u003eLo Buono, V. L. et al. Anxiety, depression, and quality of life in Parkinson\u0026rsquo;s disease: the implications of multidisciplinary treatment. \u003cem\u003eNeural Regen. Res.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 587-590 (2021).\u003c/li\u003e\n\u003cli\u003eHughes, A. J., Ben-Shlomo, Y., Daniel, S. E. \u0026amp; Lees, A. J. UK Parkinson\u0026rsquo;s Disease Society Brain Bank clinical diagnostic criteria. \u003cem\u003eJ. Neurol. Neurosurg. Psychiatry\u003c/em\u003e \u003cstrong\u003e55\u003c/strong\u003e, e4 (1992).\u003c/li\u003e\n\u003cli\u003ePark, H. et al. Classification of Parkinson\u0026rsquo;s disease with freezing of gait based on 360 turning analysis using 36 kinematic features. \u003cem\u003eJ. Neuroeng. Rehabil.\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 1-18 (2021).\u003c/li\u003e\n\u003cli\u003eChoi, H. et al. Convolutional neural network based detection of early stage Parkinson\u0026rsquo;s disease using the six minute walk test. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 22648 (2024).\u003c/li\u003e\n\u003cli\u003eTibshirani, R. Regression shrinkage and selection via the lasso. \u003cem\u003eJ. R. Stat. Soc. B\u003c/em\u003e \u003cstrong\u003e58\u003c/strong\u003e, 267-288 (1996).\u003c/li\u003e\n\u003cli\u003eHsieh, C. P., Chen, Y. T., Beh, W. K. \u0026amp; Wu, A. Y. A. Feature selection framework for XGBoost based on electrodermal activity in stress detection in\u003cem\u003e IEEE International Workshop on Signal Processing Systems (SiPS)\u003c/em\u003e 330-335 (IEEE, 2019). \u003c/li\u003e\n\u003cli\u003eAbhisheka, B., Biswas, S. K. \u0026amp; Purkayastha, B. Infusing weighted average ensemble diversity for advanced breast cancer detection. \u003cem\u003eInt. J. Imaging Syst. Technol.\u003c/em\u003e \u003cstrong\u003e34\u003c/strong\u003e, e23146 (2024).\u003c/li\u003e\n\u003cli\u003eFriedman, J., Hastie, T. \u0026amp; Tibshirani, R. Sparse inverse covariance estimation with the graphical lasso. \u003cem\u003eBiostatistics\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 432-441 (2008).\u003c/li\u003e\n\u003cli\u003eIgual, L. \u0026amp; Segu\u0026iacute;, S. Network analysis in \u003cem\u003eIntroduction to Data Science: A Python Approach to Concepts, Techniques and Applications \u003c/em\u003e151-174 (Springer, 2024).\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Supplementary File","content":"\u003cp\u003eSupplementary File is not available with this version.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Parkinson’s disease, HRQoL, gait, motor symptom, nonmotor symptom, machine learning","lastPublishedDoi":"10.21203/rs.3.rs-6755283/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6755283/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eParkinson\u0026rsquo;s disease (PD) causes motor, nonmotor, and mental health challenges that significantly impact health-related quality of life (HRQoL), and most previous studies relied on subjective assessments. Several factors are associated with poor HRQoL; however, the causal relationships remain unclear. Therefore, we aimed to identify key symptom predictors of HRQoL and its subdomains in PD and examine the structure of their interaction networks. We assessed 101 individuals with PD in the ON medication state. HRQoL was measured using a weighted ensemble of the Least Absolute Shrinkage and Selection Operator and Extreme Gradient Boosting for feature selection, followed by stepwise multivariate linear regression. Network analysis was conducted to explore variable interrelations. The HRQoL total score was predicted by the Beck Anxiety Inventory (BAI), Fall Efficacy Scale-Korean (FES-K), Hoehn and Yahr scale, treatment duration, maximum jerk, and sample entropy, with an explanatory power of 65.7%. Regarding subdimensions, anxiety, fear of falling, and sample entropy of acceleration were key determinants. This study identifies key motor and nonmotor predictors of HRQoL in PD and reveals domain-specific networks. These findings may inform targeted interventions and clinical decision-making to improve HRQoL outcomes in people with PD.\u003c/p\u003e","manuscriptTitle":"Multidimensional Predictors of Health-Related Quality of Life in Parkinson’s Disease Using Ensemble Learning and Network Analysis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-06-17 06:03:47","doi":"10.21203/rs.3.rs-6755283/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-07-28T11:37:26+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-07-26T16:01:24+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"295627319669158409607312292138171143214","date":"2025-07-10T17:34:47+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-06-27T11:19:31+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"4465025703979306908932151511624244666","date":"2025-06-27T08:36:42+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-06-12T04:39:14+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-06-06T04:37:34+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-05-29T11:13:35+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-05-27T05:00:36+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2025-05-27T04:31:46+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"38fa7293-76e0-424b-8ec8-49b4201e4631","owner":[],"postedDate":"June 17th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":49931271,"name":"Health sciences/Biomarkers"},{"id":49931272,"name":"Health sciences/Health care/Quality of life"}],"tags":[],"updatedAt":"2025-10-27T16:39:56+00:00","versionOfRecord":{"articleIdentity":"rs-6755283","link":"https://doi.org/10.1038/s41598-025-20656-9","journal":{"identity":"scientific-reports","isVorOnly":false,"title":"Scientific Reports"},"publishedOn":"2025-10-21 16:16:16","publishedOnDateReadable":"October 21st, 2025"},"versionCreatedAt":"2025-06-17 06:03:47","video":"","vorDoi":"10.1038/s41598-025-20656-9","vorDoiUrl":"https://doi.org/10.1038/s41598-025-20656-9","workflowStages":[]},"version":"v1","identity":"rs-6755283","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6755283","identity":"rs-6755283","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}