Search Results (41,811)

Airborne fungal contamination poses an occupational health risk in environments where engineering-controls are essential for maintaining indoor air quality. However, evidence integrating assessments of engineering-control readiness and airborne fungal contamination in forensic pathology facilities remains limited, particularly in low- and middle-income countries. This study assessed engineering-control readiness, indoor environmental conditions, and airborne fungal contamination in three public forensic pathology facilities in Gauteng Province, South Africa, and explored factors associated with airborne fungal contamination. A cross-sectional occupational hygiene study was conducted in three public forensic pathology facilities. Engineering-control readiness was assessed using structured walkthroughs and summarised using a study-specific Engineering Control Readiness Index (ECRI). Indoor temperature, relative humidity, carbon dioxide (CO₂), and airborne fungal concentrations were measured concurrently. Duplicate fungal air samples were collected at predefined locations over three consecutive sampling days. For the primary regression analysis, duplicate fungal measurements were averaged within each physical sampling location and sampling day, yielding 54 indoor location-day observations from 18 physical indoor sampling locations. Multivariable log-linear regression included facility, functional work area, area-specific ventilation and environmental-control deficiency, temperature, relative humidity and CO₂, with cluster-robust inference accounting for repeated measurements within physical sampling locations. Overall ECRI scores ranged from 50.1% to 53.9%, with ventilation and environmental control representing the weakest domain. Airborne fungal concentrations differed significantly between facilities, with FPF G recording the highest indoor median concentration (740 CFU/m³). Cladosporium spp. and Penicillium spp. predominated. After adjustment for facility, functional work area and environmental conditions and accounting for repeated sampling, engineering-control deficiency was not independently associated with fungal concentration (adjusted change per 10-percentage-point increase: −16.3%; 95% CI: −70.9% to 140.9%; p = 0.727). Substantial engineering-control deficiencies and airborne fungal contamination were identified; however, the independent contribution of engineering-control deficiency could not be estimated with adequate precision in this three-facility study. Larger longitudinal studies incorporating objective ventilation-performance measurements are required to clarify the contribution of specific engineering controls to occupational bioaerosol exposure. Full article

Background: This study aimed to compare grazing behavior scores between adolescents with essential hypertension (HTN) and controls. It also investigated the associations of grazing behavior with obesity, emotion regulation difficulties (ERD), problematic smartphone use, and screen time. Methods: This cross-sectional case–control study enrolled 60 adolescents with essential HTN and 60 controls selected via frequency matching based on age and sex, all aged 12–18 years. The diagnosis of essential HTN was confirmed using 24-h ambulatory blood pressure monitoring (ABPM). Grazing behavior, ERD, problematic smartphone use, screen time and body mass index (BMI) were assessed in all participants. Group comparisons and associations between grazing behavior and the study variables were analyzed. Results: The essential HTN and control groups were comparable with respect to age and sex distribution (p > 0.05). The mean BMI in the essential HTN group was 31.18 ± 6.81 kg/m², which was significantly higher than that in the control group (23.35 ± 4.47 kg/m²) (p < 0.001). A significant positive correlation was found between total Grazing Questionnaire (GQ) scores and BMI (r = 0.252; p = 0.006), ERD (r = 0.187; p = 0.041), and problematic smartphone use (r = 0.315; p < 0.001). A statistically significant difference was observed between the essential HTN and control groups in total GQ scores [F(1, 117) = 4.85, p = 0.030, η²p = 0.040]. After adjustment for BMI, participants with hepatic steatosis had significantly higher total GQ scores (p = 0.019, η²p = 0.138), Grazing Behaviors subscale scores (p = 0.046, η²p = 0.107), and Uncontrollability subscale scores (p = 0.025, η²p = 0.123) than those without hepatic steatosis. Conclusions: Adolescents with essential HTN had higher grazing behavior scores than controls after adjustment for BMI. Grazing behavior was also associated with ERD and problematic smartphone use across the study sample. The findings regarding grazing behavior and hepatic steatosis warrant cautious interpretation, as the subgroup analysis was exploratory and included a relatively small number of participants. Full article

Background/Objectives: Schatzker type is assigned early, usually from the anteroposterior (AP) radiograph. A single benchmark accuracy cannot say whether a model read the fracture, the anatomy around it, the annotation, or how the archive was assembled. We ran four inexpensive controls to separate those contributions. Methods: We benchmarked a ResNet-50 on PlaTiF, a 2026 public release built for artificial-intelligence research that pairs 421 AP knee radiographs from 186 patients with expert Schatzker labels and per-image tibial segmentations. Evaluation used stratified group five-fold cross-validation grouped by patient, five seeds and balanced accuracy. Inputs were cropped to the expert tibial segmentation shipped with the dataset, an oracle localisation unavailable at deployment. Four controls ran on identical folds: a regression given no pixel content; ablation of the tibial pixels with its complement; a regression on the expert mask alone; and an augmentation audit for label-erasing invariances. Results: Among the 128 fracture patients the network reached 0.345 ± 0.030 six-class balanced accuracy, +0.168 over a non-anatomical baseline fitted on the same folds and the same labels (95% CI +0.106 to +0.230, p = 0.002). Recall was graded: 0.72 for Schatzker VI, 0.11 for V and 0.04 for IV, the last two below chance (0.167). Erasing the tibial pixels left 0.257 ± 0.025, read on its own as the target bone being unused; its complement, the tibia with everything else removed, reached 0.367 ± 0.012, and the whole radiograph, which carries both, only 0.297 ± 0.032 (+0.071 for the tibia alone, 95% CI +0.031 to +0.111, p = 0.008). A regression on the expert mask alone reached 0.213 ± 0.034 and was not distinguishable from the erased model. On fracture versus no classifiable fracture the network reached 0.833 ± 0.028 against 0.814 ± 0.016 for a model given no pixels (p = 0.264), and a coronal computed tomography section accompanied 126 of 128 fracture patients but 24 of 58 others (p = 2.9 × 10⁻¹⁹). Conclusions: Each headline number admitted an explanation other than the fracture in the target bone. An ablation reported without its complement misstated where the signal lay, and an augmentation audit overturned our own explanation for the failure of type IV. Controls of this kind cost minutes, and this study illustrates why they can be informative when a benchmark is built on a retrospective clinical archive. Full article

Background: Diabetes distress is a key mental health dimension of type 2 diabetes (T2D) and a growing concern as populations age, yet its determinants in older adults remain insufficiently characterized. The independent contributions of demographic and metabolic factors to severe distress and their implications for multidisciplinary screening across the lifespan remain unclear. Methods: In this cross-sectional study, we assessed 453 adults with T2D using the 17-item Diabetes Distress Scale (DDS-17). We categorized participants as experiencing low (<2), moderate (2–2.9), or severe (≥3) distress. We compared categories using the Kruskal–Wallis H test and chi-square tests. Logistic regression models were constructed with severe distress (DDS ≥ 3) as the outcome, adjusting for sex, educational level, and HbA1c. Results: Severe distress was identified in 19 participants (4.2%). Age differed significantly across DDS categories (Kruskal–Wallis H test, p = 0.003), with the severe group being significantly older (72.95 ± 4.38 years). Emotional burden was the dominant subscale across all groups. In multivariable logistic regression, increasing age (aOR = 1.163 per year; 95% CI: 1.055–1.282; p = 0.002) and lower educational level (high school vs. basic: aOR = 0.187; p = 0.020) were independently associated with severe distress, whereas metabolic variables were not independently associated with severe distress. ROC analysis identified age ≥ 70 years as an exploratory, sample-derived cut-off (AUC = 0.726, 95% CI 0.602–0.827; sensitivity 84%, specificity 55%). Results were robust in sensitivity analyses that excluded patients with diabetes duration < 5 years. Conclusions: Severe diabetes distress is more strongly associated with demographic vulnerability than with metabolic burden. Age ≥ 70 years and lower educational attainment may help identify older adults at increased risk of diabetes-related distress, supporting targeted psychological screening strategies in this lifespan stage. Full article

Background/Objectives: Impaired masticatory function may compromise nutritional status in older adults wearing removable dentures, but its associations with nutritional risk indices remain unclear. This study aimed to examine associations between masticatory performance and risk defined by the Geriatric Nutritional Risk Index (GNRI), Prognostic Nutritional Index (PNI), and Controlling Nutritional Status (CONUT) score. Methods: This cross-sectional study included 179 outpatients aged ≥60 years wearing removable dentures. Masticatory performance was quantified by glucose concentration after chewing a gummy jelly. GNRI-defined risk was the prespecified primary outcome; PNI- and CONUT-defined risk were secondary. Logistic regression used Model 1 adjusted for age, sex, and comorbidity and an exploratory Model 2 with additional oral, dietary, and socioeconomic covariates. Results: Mean age was 74.1 ± 6.9 years, and 60.9% were female. Nutritional risk was identified in 21.8%, 48.6%, and 31.3% using the GNRI, PNI, and CONUT, respectively. Higher masticatory performance, indicated by higher glucose concentration, was associated with lower odds of GNRI-defined nutritional risk in both the primary parsimonious model (OR per 10 mg/dL increase, 0.89; 95% CI, 0.82–0.95; p = 0.001) and the exploratory additionally adjusted model (OR, 0.89; 95% CI, 0.82–0.96; p = 0.002). Associations with PNI- and CONUT-defined risk were not consistent across models. Conclusions: Lower masticatory performance was consistently associated with higher GNRI-defined risk, whereas comparable associations were not consistently observed for PNI- or CONUT-defined risk. Objective assessment of masticatory performance may provide clinically relevant information regarding GNRI-defined nutritional risk in older removable denture wearers. Full article

Background: Diabetes mellitus is a chronic metabolic disorder often accompanied by psychological distress, including stress, anxiety, and depression. Such emotional symptoms can influence self-care behaviours and metabolic control. In Saudi Arabia, cultural and religious practices may mitigate the effects of psychological distress on glycaemic regulation. Limited evidence exists regarding the simultaneous assessment of depression, anxiety, and stress in relation to objective glycaemic indicators such as HbA1c among adults with diabetes in Saudi Arabia. This study aimed to determine the association between psychological distress and glycaemic control among adults with diabetes, while exploring the influence of demographic and biochemical factors. Methods: A cross-sectional study was conducted from 21 January 2026 to 19 April 2026 among 546 adults diagnosed with diabetes in multiple healthcare settings in Riyadh, Saudi Arabia. Psychological distress was assessed using the DASS-21 questionnaire, while glycaemic control was measured using HbA1c, categorised as controlled (<7%) or uncontrolled (≥7%). Demographic, anthropometric, and biochemical parameters including BMI, blood pressure, and lipid profile were collected from medical records and standardized assessments. Associations were examined using chi-square tests and multiple linear regression analysis was performed including psychological distress scores, BMI, blood pressure measurements, and lipid profile parameters to identify factors associated with HbA1c variation. Results Psychological distress was prevalent; however, stress (6.2–10.7%), anxiety, and depression were significantly associated with variation in HbA1c levels. In contrast, nationality, total cholesterol, LDL, triglycerides, and systolic blood pressure were significantly associated with HbA1c. Multiple regression showed that stress (B = 0.09, p = 0.015), systolic blood pressure (B = 0.146, p = 0.006), and total cholesterol (B = 0.229, p < 0.001) were independent predictors, collectively explaining 14.8% of HbA1c variance. Multiple regression analysis identified stress, systolic blood pressure, and total cholesterol as significant factors associated with variation in HbA1c levels. Conclusions: While psychological distress is common among adults with diabetes, its direct impact on HbA1c may be attenuated in populations with strong religious and cultural coping practices. Conversely, metabolic factors, particularly total cholesterol and systolic blood pressure, alongside stress, showed stronger associations with HbA1c levels in this study population. These findings underscore the importance of integrating physiological, psychological, and culturally contextualised strategies in diabetes management to optimize outcomes. Full article

Attention-Deficit/Hyperactivity Disorder (ADHD) presents substantial challenges in higher education, including elevated mind wandering, difficulties sustaining attention during academic reading, and diminished self-regulatory capacity. Yet, many students with ADHD achieve academic outcomes comparable to their neurotypical (NT) peers, suggesting that protective and compensatory mechanisms may help decouple attentional vulnerability from distal academic achievement. This study examined whether artificial intelligence (AI) tools may function as compensatory support for college students with ADHD and whether resilience is differentially associated with academic success in this population. A sample of 195 engineering students (ADHD = 84, NT = 111) completed self-report measures of trait mind wandering, boredom proneness, self-efficacy, resilience, return-to-focus ability, AI usage and motivation, alongside Raven’s Progressive Matrices, an academic reading task with attention probes and comprehension questions, and end-of-semester course grades. Students with ADHD reported significantly greater AI usage than NT students (d = 0.32), despite no significant group differences in cognitive ability or course grades. Hierarchical regression showed that once attentional traits were included, diagnostic group membership no longer accounted for AI usage, indicating that attentional vulnerability, particularly elevated mind wandering, rather than diagnosis per se, may better explain AI usage. Self-efficacy was positively associated with AI usage among students with ADHD but not among NT students, though this group difference was not statistically significant; the pattern is therefore reported as tentative rather than as an established compensatory mechanism. Resilience was lower among students with ADHD than among NT students but was not associated with course grade in either group, and neither resilience nor AI usage was independently associated with course grade once cognitive ability and attentional traits were accounted for. These findings suggest that AI use may reflect compensatory academic adaptation to attentional vulnerability in ADHD. More broadly, the results highlight how attentional vulnerability, self-efficacy, AI-based compensatory strategies, and resilience jointly shape academic outcomes in ADHD and raise questions about how equitable, AI-supported learning environments might be designed in higher education. Because the design is cross-sectional, the present data do not establish that AI use improves academic achievement. Full article

Background: Cyberbullying victimization is associated with recent poor mental health among adolescents, but the roles of digital and offline contextual indicators require careful interpretation. Objective: Guided primarily by Stress Process Theory and secondarily by bounded routine activity concepts, this study examined cross-sectional associations among traditional school bullying, cyberbullying, social media use, physical activity frequency, and recent poor mental health among U.S. high school students. Methods: Complete-case 2023 Youth Risk Behavior Survey data for U.S. high school students (N = 10,295; 49.5% female; 42.7% White, 21.9% Hispanic/Latino, 11.2% Black or African American, and 24.2% all other races) were analyzed in two non-survey-weighted, model-based indirect-association specifications using PROCESS Model 6. Both models adjusted for sex, grade, sexual identity, race/ethnicity, and the alternate bullying indicator. Results: Traditional bullying and cyberbullying were each positively associated with more frequent recent poor mental health after mutual adjustment (β = 0.113 and β = 0.111, respectively). No indirect associations were statistically supported in the traditional bullying model. In the cyberbullying model, the indirect association through social media use was statistically supported but small, and the sequential indirect association through social media use and physical activity frequency was statistically supported but extremely small. Conclusions: Findings support a cautious, cross-sectional interpretation in which bullying victimization is treated as peer-related strain and social media use and physical activity frequency are contextual indicators rather than causal mediators. Full article

The war in Ukraine, beginning in February 2022, disrupted continuous medical care for Ukrainian childhood and adolescent cancer patients (UCC), many of whom were relocated to pediatric cancer centers worldwide, including Switzerland. In this Brief Report, we describe their health-related quality of life (HRQoL) after arrival. In a multicenter, cross-sectional survey across five Swiss pediatric oncology centers, we included patients ≤18 years at diagnosis who arrived after 24 February 2022 and were undergoing active treatment. HRQoL was assessed by the PedsQL™ 4.0 Generic Core Scales (self- and parent-reports). Fourteen of 23 eligible families (61%) participated. HRQoL declined with age, particularly in physical functioning, while psychosocial functioning remained relatively stable but lower among adolescents; parent scores closely matched self-reports. Scores were referenced against a published cohort of healthy Ukrainian children and previously reported pediatric cancer populations. Given the small sample and lack of a matched control group, the findings cannot separate the effects of displacement, cancer, and treatment and are hypothesis-generating. They suggest a potential role for age-tailored supportive care in displaced children with cancer and call for larger, controlled studies. Full article

Background: Population ageing, multimorbidity, and chronic conditions have increased the demand for person-centred care in Primary Health Care. Current stratification systems rely on diagnoses or resource utilization and fail to capture the multidimensional nature of care complexity, which also depends on functional, social, environmental factors. Objective: To develop a HealthCare Complexity Index using routinely collected electronic health record (EHR) data and the Person-Centred Care Knowledge Model. Methodology: A retrospective, cross-sectional observational study was conducted in Primary Health Care in the Community of Madrid (Spain). Deductive Care Methodology was used to identify functional assessment variables associated with four care dimensions: vulnerability, risk, etiology, and signs/symptoms. A multidisciplinary expert panel selected and validated the variables through iterative consensus rounds. A weighted scoring algorithm, the Care Complexity Index, was designed to quantify care complexity on a 0–100 scale. Relationships between the selected variables and NANDA-I nursing diagnoses were established and validated by an expert in standardized nursing languages. Results: 27 assessment variables were identified and classified into the 4 dimensions of the care model. These variables were integrated into a weighted complexity index with five severity levels ranging from low to critical complexity. Furthermore, 48 NANDA-I nursing diagnoses were linked to the selected variables. Conclusions: The HealthCare Complexity Index provides a standardized, person-centred approach for estimating care complexity using routinely available (EHR) data. It may support early identification of individuals with complex care needs, facilitate individualized care planning, and strengthen clinical decision-making. Future studies should validate its predictive performance and implementation in other clinical practice. Full article

Obesity is a widespread condition strongly linked to increased cardiovascular risk, and heart rate (HR) responses to physical activity are widely used physiological indicators of this relationship. Although the role of fat mass- and obesity-associated (FTO) gene variants in obesity risk is well established, their potential associations with HR changes during and after physical activity (independent of adiposity) have not yet been fully elucidated. This study examined whether common FTO polymorphisms (rs9939609, rs1121980, rs1558902, and rs9941349) and an exploratory FTO multi-variant score (MVS_FTO), representing regional variation across the high-LD FTO locus, are associated with HR responses to a brief submaximal exercise stimulus. A population-based cohort of Hungarian adults (n = 660) completed the YMCA 3-min step test, with HR measured at rest, at first-minute post-exercise (HR_aft), and at 5- (HR_5min) and 10-min recovery (HR_10min). Acute HR response (ΔHR) and percent of predicted maximal HR (HR_max%) were calculated. Linear regression analyses showed no association between any FTO variant and resting HR (all p > 0.05). In contrast, all four candidate SNPs demonstrated significant negative associations with HR_aft (β ≈ −3.6 to −4.2, all p ≤ 0.005), ΔHR (β ≈ −3.7 to −3.9, p ≤ 0.005), and HR_max% (β ≈ −2.5, p = 0.001). The MVS_FTO revealed even more pronounced and consistent associations: higher MVS_FTO burden was associated with lower HR_aft (β = −4.19, p = 0.001), HR_5min (β = −2.06, p = 0.004), HR_10min (β = −1.15, p = 0.009), ΔHR (β = −4.04, p = 0.001), and HR_max% (β = −2.47, p = 0.001). Exploratory four-way mediation decompositions and saturated structural equation models indicated that these inverse associations remained statistically robust after multivariable adjustment for measured BMI and leisure-time physical activity (all direct paths p ≤ 0.001). Furthermore, multi-trait latent cluster analysis identified distinct phenotypic profiles consistent with these findings, showing that individuals with high MVS_FTO burden and moderate physical activity exhibit blunted exercise HR reactivity across recovery phases. Rather than establishing biological causality or clinical cardiovascular benefit, these observational findings demonstrate that regional FTO locus variation captured by the multi-variant score is statistically associated with inter-individual variability in short-term post-exercise HR reactivity, providing hypothesis-generating insights that warrant validation in prospective cohorts and interventional exercise trials. Full article

This study examined the relationship between perceived internal coaching, toxic leadership, and perceived job performance, seeking to determine whether toxic leadership moderates the relationship between perceived internal coaching and performance, as well as whether the managerial/leadership role moderated that same relationship. A quantitative, correlational, and cross-sectional design was adopted, based on 313 valid survey responses. The results showed that perceived internal coaching (seen to be an intermediate level of leadership and management) was positively and significantly associated with perceived performance, supporting the hypothesis that internal coaching practices can contribute to professional effectiveness through reflection, feedback, the clarification of objectives, and behavioral adjustment. Hence, internal coaching is viewed as being essential. Toxic leadership did not have a statistically significant direct effect on performance; however, it significantly moderates the relationship between perceived internal coaching and performance, indicating that higher levels of toxic leadership attenuate the positive relationship between these variables (but perhaps less than thought initially). On the other hand, the managerial/leadership role did not significantly moderate the relationship between perceived internal coaching and performance. These results suggest that perceived internal coaching constitutes a relevant developmental resource for perceived professional performance, although relational and contextual dynamics may condition its impact, such as whether the culture is more individualistic or collectivist. The presence of toxic leadership appears to weaken the potential positive effect of perceived internal coaching, reinforcing the need for broader organizational interventions, including ethical leadership practices, accountability mechanisms, and psychologically safe work environments. Full article

Environmental regulations and stricter emission limits are driving the development of advanced fuel supply systems. Precise fuel metering under varying engine loads has become critical, with modern strategies using multiple injections of very short duration. However, injector behavior, particularly nonlinear flow characteristics, is not fully understood. This study presents an experimental analysis of the flow characteristics Q = f (t_inj) and opening dynamics of five low-pressure gas injectors with different valve system designs. The tests were conducted for injection times t_inj = 0–20 ms. For t_inj > 2.5 ms, the characteristics were very well described by a linear model (R² > 0.995), whereas for t_inj < 2.5 ms, there was a clear deviation from the linear relationship between flow rate and injection time. Analysis of the electrical signals, outlet pressure, and body vibrations made it possible to identify the mechanistic sources of the observed nonlinearity. It was demonstrated that the initial lack of flow results from an electromechanical delay associated with the rise in current and the electromagnetic force required to overcome the spring force, friction, and inertia of the valve element. The subsequent movement of the valve contributing factors a dynamic change in the flow cross-sectional area and, consequently, a nonlinear change in flow rate. Additionally, the change in the position of the valve element affects the inductance of the coil and the nature of the electromagnetic force. Near the maximum lift, the element bounces off the stop, causing a momentary change in its position and a local decrease in flow rate. Only after the valve element’s motion stabilizes does the flow transition to a nearly linear relationship. The response times of the injectors ranged from 0.60 to 1.30 ms, and the times to reach full opening ranged from 1.08 to 2.14 ms, corresponding, respectively, to the onset and the transition to the steady-state region of the characteristic curve. The results indicate that the nonlinearity of the short-time portion of the characteristic has a mechanistic, electromechanical nature and results from the coupling of electromagnetic phenomena, the motion of the valve element, and the varying flow cross-section. This means that accurately modeling it requires taking into account the actual dynamics of valve-opening, particularly in the case of strategies that use short and repeated injection pulses. These findings highlight a significant limitation in fuel dosing precision and emphasize the need to incorporate nonlinear injector models or dynamic corrections in ECU control algorithms—an essential step for further reducing exhaust emissions. Full article

The increasing demand for thermal protection materials in new energy vehicle (NEV) systems has stimulated the development of lightweight coatings capable of simultaneously providing thermal insulation and fire resistance. In this study, a lightweight ceramic-like inorganic coating was prepared by incorporating hollow glass microspheres (HGMs), expanded vermiculite, and ceramic fibers into a sodium silicate matrix. The expanded vermiculite/ceramic-fiber ratio and coating thickness were varied to evaluate their effects on thermal protection. Among the investigated formulations, Formulation C showed the lowest mean thermal conductivity, with a value of 0.0649 ± 0.0004 W m⁻¹ K⁻¹ for the 1-mm-thick C-1 coating. The apparent density of the representative Formulation C coating was 0.54 ± 0.02 g cm⁻³ (n = 3). During representative single-specimen direct butane-flame exposure tests, the spatially averaged backside temperature of the 2-mm-thick C-2 specimen was 147.1 °C after 600 s, compared with 420.1 °C for the uncoated aluminum plate, corresponding to a temperature difference of 273.0 °C. Cross-sectional SEM showed non-uniform structural evolution after flame exposure, with more pronounced morphological changes near the flame-facing region and greater retention of microspherical structures in the inner region. The representative Formulation C coating also retained more than 80% of its mass at 800 °C in both nitrogen and air. These results indicate that the combined presence of HGMs, expanded vermiculite, ceramic fibers, and the sodium silicate matrix provides effective resistance to heat transfer under high-temperature exposure. The coating therefore offers a lightweight inorganic approach for high-temperature thermal protection. Full article

Introduction: Voice-based digital biomarkers have emerged as a promising, non-invasive approach for the early detection and monitoring of neurodegenerative disorders, particularly Parkinson’s disease (PD). Although voice recordings can be acquired easily using mobile health technologies, their integration into routine clinical practice remains limited due to challenges related to model interpretability and clinical validation. This study aimed to investigate the diagnostic potential of voice recordings acquired through the TALIA smartphone and web-based platform and to improve model transparency using explainable artificial intelligence (XAI) techniques. Methods: Voice recordings from participants with PD and control group (CG) were acquired over a six-month period using the TALIA digital-health platform at Vito Fazzi Hospital in Lecce, Italy. The data were evaluated cross-sectionally at the recording level. Sustained vowel (/a/) phonations were preprocessed and transformed into spectrogram images. A transfer-learning framework based on the pre-trained VGGish convolutional neural network (VGGish-CNN) was developed to classify PD and CG voice recordings. To enhance interpretability, explainable artificial intelligence (XAI) methods, including Local Interpretable Model-Agnostic Explanations (LIME) and Occlusion Sensitivity, were integrated to identify the spectro-temporal regions contributing most strongly to model predictions. Results: The proposed VGGish-CNN framework demonstrated excellent classification performance in distinguishing PD from CG. The model achieved an accuracy of 0.93, precision of 0.91, recall of 0.95, F1-score of 0.93, loss of 0.16, and an area under the receiver operating characteristic curve (AUC) of 0.987. XAI analyses provided qualitative, sample-specific visualizations of the spectro-temporal regions contributing to individual model predictions, thereby improving the transparency and interpretability of the deep-learning model. Conclusions: The findings demonstrate that a transfer-learning approach based on VGGish-CNN could achieve promising recording-level classification performance in distinguishing voice recordings from participants with PD and CG within this pilot dataset. Furthermore, XAI techniques provide qualitative insight into the model’s decision-making process. These preliminary findings support further investigation of voice-based approaches for PD screening in larger cohorts using participant-level and external validation. Full article