Abstract: Accurate rainfall data are essential for hydrological forecasting and climate modeling. However, many developing regions, including Tunisia, struggle with significant data gaps in rainfall measurements, particularly from gauge stations. These missing data impair climate model validation and reduce forecasting accuracy across both spatial and temporal dimensions. To overcome these limitations, we conduct a comprehensive evaluation of novel deep learning (DL) architectures designed for imputing missing rainfall gauge data and generating monthly rainfall forecasts. Our framework systematically compares multiple DL approaches: Long Short-Term Memory (LSTM), a hybrid Bidirectional LSTM with a Transformer attention mechanism (BiLSTM-Transformer), and a pure Transformer model. Subsequently, we employ Principal Component Analysis (PCA), K-Means clustering, and quantile techniques to further refine DL model outputs. The processed data are then analyzed using Light Gradient Boosting Machine (LightGBM) to produce final results. Our rigorous evaluation across 47 Tunisian gauges covering 1983–2012 (70% training, 30% testing) demonstrates that the BiLSTM-Transformer hybrid delivers superior performance, achieving an 18.4% reduction in root mean squared errors (RMSE) compared to conventional interpolation methods (14.2 mm versus 17.4 mm monthly error) and improving R2 values by 0.15–0.23 across all test stations. The model shows particular strength in capturing Mediterranean rainfall patterns, correctly predicting 83% of extreme rainfall events (greater than 95th percentile). Furthermore, spatial graph networks boost performance at data-sparse stations by 12.7% through explicit modeling of topographic influences.

Abstract: For a Hutchinson iterated function system (IFS), a Banach contraction on a complete metric space, or a finite-metric dynamical system, a natural question is: at which resolution σ does the contraction's geometric structure (fractal attractor, basin of attraction, periodic part) become optimally visible? We answer this by introducing a scale-selection principle: define the observation scale σ_c := argmax_σ χ(σ), where χ(σ) = |dO(σ)/d log σ| is the susceptibility of a resolution-dependent observable, and prove that σ_c exists under explicit boundary-regularity hypotheses.The framework's main quantitative results are three theorems specialised to Banach contractions and IFS: (i) a geometric scaling identity σ_c = qL for affine Banach contractions with operator norm q and basin scale L, applying directly to Hutchinson IFS with σ_c ∼ q · diam(K⋆); (ii) a discrete Banach theorem on finite metric structures under uniform Lipschitz Lip_d(f) = q < 1, giving an exact collapse-time N⋆ = ⌈log(∆/d_min) / log(1/q)⌉; (iii) a spectral concentration theorem placing σ_c at the inverse log-spectral-gap of the transfer operator at fixed positive noise. A stability lemma for canonical normalisation under smooth windowing and a parametric Banach correspondence observation complete the technical core.The framework is stated explicitly in the non-expansive Lipschitz regime Lip_d(f) ≤ 1 on the metric side, and at fixed positive noise ε ∈ (0, 1) on the spectral side. A four-type classification of operations by injectivity structure organises the broader landscape; cross-domain empirical evidence anchored on a peer-reviewed NISQ-hardware measurement of σ_c is summarised. The middle-thirds Cantor set IFS appears as the principal worked example.

Abstract: The study applies a multiphase, multimethod research approach based on the participatory methodology. It integrates perspectives of professionals in the travel industry and academic experts with the aim to develop an integrated conceptual model of the AI and IoT influence on work, skills development, and job attractiveness in the industry. The research provides a comprehensive understanding of the ways in which digital technologies indirectly shape employment through changes in work processes and development of transferable digital and socio-emotional skills. The paper aimed to donate to redefining the perception of work in tourism and hospitality, emphasizing the sector not only as a career choice but also as a platform for the acquisition of skills relevant in other industries as well. The outcomes revealed that the employees’ aspirations to enter or stay in the industry are not directly influenced by AI and IoT technologies; rather, their effects are mediated through changes in work processes and, more importantly, through the development of skills. The study contributes theoretically by evolving and analytically confirming an incorporated theoretical model that connects technology implementation, work transformation, skills development, and employment outcomes. Practically, the results underscore the importance of human-centered implementation strategies, emphasizing training, communication, and employee inclusion to maximize the benefits of digital technologies.

Abstract: Background/ Objectives: Burn care remains a major clinical challenge, as they require effective wound management to promote healing, reduce pain, and prevent complications. This study aimed to describe the use of a polylactic acid membrane in patients with burns of different depths and to assess wound healing outcomes. Methods: A descriptive case series was conducted including six patients with burn injuries treated with a polylactic acid (PLA) membrane (Suprathel®) at a specialized burn unit. The study was performed according to CARE guidelines. Clinical assessment included burn depth evaluation, pain measurement using the Visual Analog Scale, and scar assessment using the Vancouver Scar Scale. The application of the PLA membrane and its clinical performance during the healing process was described. Results: Complete wound closure was achieved in all cases by re-epithelialization as the PLA adhered, integrated and became transparent, clearly visualizing the progression of healing. Time to re-epithelialization ranged from 14 to 35 days, and final Vancouver Scar Scale scores ranged from 1 to 3 points. No major wound-related complications or local infections were documented during the clinical follow-up period. Conclusions: The use of a PLA membrane in burn management represents a promising advance in wound care, as it was associated with wound closure and pain reduction in this case series, with potential benefits for the optimization of nursing clinical practice.

Abstract: Hero Ridge shale oil reservoirs are characterized by stacked pay boxes, strong vertical heterogeneity, rapid variations in lithology and in situ stress, and significant well-to-well interference during platform-scale three-dimensional development. Conventional fracturing design methods that focus mainly on single-well stimulation are insufficient to simultaneously address fracture propagation, reservoir contact and development economics. Taking the 1H platform and representative wells in the upper member of the Xiaganchaigou Formation (E₃², Boxes 5-6) as examples, this study establishes a workflow integrating reservoir-engineering dual-quality evaluation, single-well parameter optimization, platform-coordinated fracturing, dynamic pore-pressure-stress updating, and EUR-IRR response-surface analysis. Results show that Box 6 has better reservoir quality and fracability than Box 5, with average porosity, oil saturation and brittle-mineral content of 7.6%, 50.9% and 67.6%, respectively. Well 1H6-1, with a 1500 m lateral, penetrated Class I + II sweet spots for 90.6% of the horizontal interval, providing a geological basis for efficient volume stimulation. For conventional sweet-spot wells, the optimal single-well design includes eight clusters per stage, a pumping rate of 18 m³/min, a fluid intensity of 35 m³/m and a proppant intensity of 3.25 m³/m. For 200 m-spaced wells, the pumping rate and fluid intensity should be reduced to 16 m3/min and 32 m³/m, respectively, with 100 m³ of prepad gel to mitigate fracture overlap and stress interference. Further response-surface analysis based on actual EUR-IRR data shows that the highest EUR occurs at a lateral length of 4000 m and well spacing of 50 m (EUR = 566,261 m³), but the IRR is -27.1%. By contrast, the best IRR point is at a lateral length of 4000 m and well spacing of 600 m (IRR = 14.5%), with EUR of 377,500 m³. This demonstrates that the production-optimal and economics-optimal schemes are not coincident. The expanded pilot scheme has an after-tax IRR of 9.31%, after-tax NPV of RMB 131.38 million and payback period of 5.93 years. The results indicate that fracturing optimization in Hero Ridge should move from single-well engineering maximization to integrated decision-making that combines single-well design, platform coordination, lateral-length/well-spacing optimization and techno-economic evaluation.

Abstract: We propose a mechanically programmable nanoscale Chern valve based on an altermagnet–topologicalinsulator (AM–TI) heterostructure, where thin altermagnetic electrodes impose an anisotropic exchange mass on the surface states of a few-quintuple-layer topological-insulator channel. Periodic strain, delivered for example by integrated piezoelectric or surface-acoustic-wave actuators, modulates the inplane crystalline phase of the altermagnetic order and renormalizes the twofold and fourfold interfacial exchange harmonics through zeroth-order Bessel functions. This amplitude-selective renormalization produces re-entrant Chern plateaus, Hall and thermoelectric polarity inversions, and quantized adiabatic charge pumping with winding number changing from 0 to 2. For representative RuO₂/Bi₂Se₃ parameters, the induced gaps remain in the meV range, while MHz mechanical driving places the system deeply within the adiabatic regime. The predicted signatures are directly accessible in nanoscale Hall-bar geometries through the strain-amplitude dependence of transverse Hall response, gate-tracked thermoelectric Hall response, and the collapse of topological sectors near Bessel zeros. The proposed mechanism therefore provides a low-frequency, on-chip route to mechanically controlled topological transport in nano-spintronic AM–TI devices, without optical Floquet driving or net magnetization reversal.

Abstract: This study examined the wood properties, i.e. anatomical characteristics, chemical composition, physical and mechanical properties of seven-year-old plantation-grown B. microphyllum harvested from a research plot at the Forest Research Institute Malaysia. Microscopic analysis revealed diffuse-porous wood with very large solitary vessels, aliform to confluent parenchyma, medium-sized rays, and non-septate fibres. Fibre morphology showed a Runkel ratio below 1.0 and a slenderness ratio of 41.9, indicating favourable fibre flexibility and bonding potential. The absence of tyloses and silica suggests good treatability and machinability. Chemical analysis showed high holocellulose content (79.5–81.9%), α-cellulose (~44%), moderate lignin (22.6–23.9%), and low extractives (0.9–2.1%), indicating a substantial carbohydrate fraction with minimal non-structural compounds. Preliminary phytochemical screening detected flavonoids, tannins/polyphenols, and triterpenes/steroids as dominant constituents, supporting its traditional medicinal relevance. The wood density ranged from 441.4 to 606.8 kg m⁻³ (mean: 524.1 kg m⁻³), classifying the timber as light to moderately heavy. Shrinkage at 15% moisture content was 2.2% (tangential), 1.2% (radial), and 0.6% (longitudinal), giving a tangential-to-radial ratio of 1.6 and indicating moderate dimensional stability. Despite being harvested at only seven years of age, B. microphyllum exhibited mechanical properties comparable to or superior to several commonly planted fast-growing species, such as Eucalyptus nitens, rubberwood (Hevea brasiliensis), and batai (Paraserianthes falcataria). In particular, the bending and shear strengths were considerably higher than those reported for some older plantation timbers. These findings suggest that B. microphyllum has strong potential as a fast-growing plantation timber with favourable strength characteristics and other promising properties, making it a suitable candidate for structural and value-added wood applications.

Abstract: Background: To inform epidemic control, Rapid Test for Recent Infection (RTRI) assays, such as the Asanté HIV-1 Rapid Recency Assay (ARRA), have been developed to detect potential signals of increased HIV acquisition. However, ensuring the accuracy of these tests remains a challenge in resource-limited settings. While ARRA has been implemented for surveillance, there is a lack of documented experiences and lessons learned regarding quality assurance through Proficiency Testing (PT). This study examined Uganda's recent HIV infection PT program from 2020 through 2022 to highlight challenges and successes of implementation in resource-limited settings. Methods: We analyzed proficiency testing (PT) implementation for HIV recency testing in Uganda from 2020–2022. The study included biannual PT cycles (Cycle 1: Jan–Jun, Cycle 2: Jul–Dec) across 676 facilities in 133 districts. We assessed performance using pass rates (percentage of testers correctly identifying all samples in a PT panel) and response rates (proportion of testers submitting results within the stipulated timeframe out of the total number expected to participate). To evaluate sustainability, we longitudinally tracked a fixed cohort of the first 175 testers enrolled at the project's inception, representing diverse facility levels and cadres. Results: Analysis of six proficiency testing cycles from 2020-2022 revealed two key trends: a significant expansion in program participation and a concurrent longitudinal decline in performance among a consistent cohort. Overall, participation grew from 175 testers in Cycle 1, 2020, to 568 testers by Cycle 2, 2022. Among all participating facilities in each cycle, pass rates fluctuated, ranging from a high of 87.3% (Cycle 2, 2020) to a low of 54.9% (Cycle 2, 2021). A longitudinal analysis of the initial 175-testing-site cohort, however, revealed a significant inverse relationship between participation and performance. Non-response within this cohort increased drastically from 0% to 81.7% by early 2022 (p-value for trend <0.001). Among the diminishing subset of sites that continued to submit results, the pass rate showed a statistically significant declining trend, from 85.1% to 76.6% over the study period (p-value for trend = 0.012). Conclusion: This study identified three critical challenges: a steep rise in non-response, declining pass rates, and persistent performance gaps at lower-level health facilities. To address these gaps, we recommend individual tracking with digital feedback and targeted mentorship to re-engage staff and sustain competency at lower-level facilities.

Abstract: Complex percutaneous coronary intervention (PCI) represents a growing proportion of contemporary coronary revascularization, driven by aging populations, increasing comorbidity burden, and advances in interventional techniques. Complex PCI encompasses a spectrum of anatomically and procedurally challenging lesions, including left main disease, bifurcation lesions requiring two-stent strategies, chronic total occlusions, long stent lengths, severe calcification requiring atherectomy, and multivessel revascularization. Antithrombotic therapy, comprising antiplatelet and anticoagulant agents, is essential for preventing stent thrombosis and other ischemic events in both the early and long-term phases after PCI. While antithrombotic therapy mitigates ischemic risks associated with complex PCI, these patients frequently carry high-bleeding risk, thus making the choice of antithrombotic regimen challenging. Recent guideline recommendations emphasize balancing ischemic and bleeding risks rather than relying solely on procedural complexity. This review synthesizes contemporary evidence, guideline recommendations, and clinical considerations for antithrombotic therapy after complex PCI.

Abstract: Rel/SpoT family enzymes participate in controlling the cellular levels of the alarmone (p)ppGpp, thereby activating the stringent response and promoting survival under stress conditions. These proteins contain an N-terminal catalytic domain and a C-terminal regulatory domain. They catalyze both the synthesis of ppGpp/pppGpp from ATP and GDP/GTP and their hydrolysis to GDP/GTP and pyrophosphate. Here, we report the crystal structure of the N-terminal domain of Rel from Streptococcus equisimilis (Rel_Seq385) in complex with pppGpp at 3.2 Å resolution. The asymmetric unit contains a dimer with asymmetric ligation, in which pppGpp occupies only the synthetase site in one monomer, whereas it is observed in both the hydrolase and synthetase sites in the other. Molecular dynamics simulations supported this binding arrangement for the monomer with both sites occupied and revealed additional probable transient binding sites that may contribute to alarmone binding.

Abstract: Traditional food systems are increasingly threatened by industrialised agri-food production, which relies on standardised processes, economies of scale, and lower production costs. This transformation risks undermining not only the economic viability of artisanal producers but also the cultural heritage, local knowledge, pastoral practices, and territorial identities embedded in traditional foods. This study investigates whether consumers’ willingness to pay a premium for traditionally produced foods can help safeguard rural cultural heritage in a competitive PDO market. Focusing on an Italian cheese, the Fiore Sardo PDO, the research combines a Bertrand duopoly framework with the Theory of Planned Behaviour (TPB) to examine the relationships among market competition, consumer beliefs, and support for traditional production systems. Data from 1,640 Italian consumers were analysed using structural equation modelling. The results show that attitudes towards cultural preservation, social recognition of traditional production, and perceived support for rural shepherd communities significantly influence consumers’ willingness to purchase and pay higher prices for traditionally produced cheese. Consumers associate artisanal production not only with superior sensory quality and authenticity but also with the protection of cultural identity, traditional pastoral knowledge, and rural landscapes.

Abstract: Given the "dual-carbon" goals of China, research has been carried out on the impact of the digital economy on carbon emission intensity. Based on the panel data of the 288 cities in China from 2013 to 2022, a two-way fixed-effects model is employed in this paper to study how the digital economy affects carbon emission intensity at the level of cities and urban agglomerations. The results show that the development of the digital economy reduces the intensity of urban carbon emissions, and there is cross-regional spatial spillover at the level of urban agglomerations. According to the results of the mechanism test, there are two paths for "industry-technology" transmission: At the city level, the digital economy can reduce pollution by improving the structure and upgrades of the industrial system; At the level of urban agglomeration, it can strengthen green-technology innovation capabilities. According to the analysis of heterogeneity, polycentric agglomeration, optimisation-and-upgrading type agglomeration and coastal areas have relatively good carbon reduction effects. Based on the above, personalised regional policies will be formulated to promote the development of the digital economy in line with carbon reduction objectives.

Abstract: Corridor management—such as reliance on manual warning zone delineation and inconsistent boundaries—this paper proposes 3DSim-WZD, an automatic ground-level warning zone detection method based on 3D simulation data augmentation. Guided by "interpretable geometric priors combined with deep learning regression," the framework integrates four modules: parametric simulation generation, simulation-to-real transfer, boundary vertex regression, and voltage-level-based expansion. Specifically, parametric virtual scenes are constructed in Unity3D to automatically derive accurate vertex labels. The open-source Stable Diffusion framework, combined with ControlNet and LoRA, is employed for sim-to-real style transfer to reduce domain gaps. Furthermore, directional detection convolutional kernels are incorporated into the YOLO12m backbone to enhance sensitivity to transmission structures. Finally, safety clearance distances are mapped according to voltage levels for regulatory-compliant warning zones. Evaluated on a dataset of 5,000 simulated and 300 real samples, the method achieves a mIoU of 91.2% and an inference speed of 46.8 FPS, demonstrating significant potential for large-scale deployment.

Abstract: Background: The thermodynamic cost of local large language model (LLM)inference on consumer hardware is poorly characterised. Unlike data-centre deployments with hardware power monitors (NVML, RAPL), Apple Silicon unied-memory systems require alternative instrumentation strategies, and no Landauer-grounded framework for local inference energy has previously been validated. Methods: We deploy seven open-source LLMs (2.020.2 GB; 3.2B32.8B parameters, Q4_K_M quantisation) on a single Apple M5 MacBook Pro (32 GBunied memory, 25 GB Metal GPU VRAM) via Ollama v0.23.2, instrumentingthe system with a custom telemetry daemon (1.5 s polling; top, vm_stat, ioreg,ps). We apply the Kerimov-Alekberli (KA) information-geometric framework, which monitors KL divergence between consecutive output distributions relative to a Fisher Information Metric (FIM)-derived threshold (τ = 0.065), and compare energy consumption against an unoptimised baseline using a unied Pythoncode-generation benchmark. Energy estimates are grounded in Landauer's ther-modynamic lower bound E_min = k_BTln 2, scaled macroscopically by an empirical power-size model. Results: KA achieves a consistent 38 % energy reduction across all seven models, saving 59 mJ (llama3.2, 2.0 GB, 55.7 tok/s) to 32,841 mJ(qwen3:32b, 20.2 GB, 2.6 tok/s) per run. Measured power draw follows a linear modelP = 5.0 + 0.75 SGB W (R² = 0.97). Token eciency under KAranges from 1,321 tok/J (qwen3:32b) to 8,287 tok/J (llama3.2). The First-PassageTime (FPT) anomaly detector recorded 602 KL-divergence threshold exceed ancesacross 9,501 total inference tokens; the highest-energy model (qwen3:32b) regis-tered 562 anomalies and the greatest absolute saving. Conclusions: These results constitute the first empirical validation of a Landauer-grounded energy reduction mechanism in local LLM inference via an information-geometric output-distribution stabilisation framework, with extrapolated annual savings of 105.4 kJand 11.7 mg CO₂ per workstation.

Abstract: Transport infrastructure is commonly viewed as a key driver of development, alt-hough its actual contribution remains debated and appears to be dependent on geo-graphical and economic context. This study investigates the impact of transport infra-structure on regional economic growth in Romania, with a particular focus on spatial spillover effects Using panel data for Romanian regions over the period 2000–2024, the analysis applies spatial econometric techniques to capture both direct and indirect ef-fects of transport infrastructure and economic factors. A structured model selection procedure, based on Lagrange Multiplier tests and robust diagnostics, supports the use of the Spatial Autoregressive Model (SAR) as the preferred specification. The results reveal significant spatial dependence in regional economic performance, indicating that growth processes extend across regional boundaries. Nonetheless, the findings show that transport infrastructure does not exert a statistically significant direct effect on economic growth once spatial and structural factors are controlled. Instead, labor and private gross capital formation emerge as the primary drivers, generating both strong local impacts and substantial spillover effects. These results suggest that transport infrastructure acts mainly as an enabling factor rather than a standalone driver of growth, making the concept of “political mythification” of transport infra-structure effectiveness relevant in the Romanian context.

Abstract: The integration of generative artificial intelligence (AI) into agricultural extension services presents a transforma- tive opportunity to address the unique challenges faced by smallholder farmers, particularly in resource-constrainedsettings. While traditional extension services often struggle with scalability and personalized support, generative AI offers potential solutions through dynamic content generation, real-time decision-making assistance, and adaptive learning tools. This systematic literature review examines the efficacy of generative AI in enhancing agricultural extension services, focusing on its applications, benefits, and limitations for smallholder farmers. We synthesize existing research across multiple dimensions, including AI-driven farmer support, IoT-enabled monitoring, andclimate-smart agriculture, to identify gaps and trends in the current knowledge landscape. A rigorous methodol- ogy was employed to select and analyze relevant studies, ensuring a comprehensive evaluation of both theoreticalframeworks and practical implementations. The findings reveal that generative AI can significantly improve access to tailored agricultural advice, optimize resource allocation, and mitigate climate-related risks; however, challengessuch as digital literacy, infrastructure limitations, and ethical concerns remain critical barriers to widespread adop- tion. The review also highlights the disproportionate focus on high-income regions, underscoring the need for moreinclusive research in low-resource agricultural systems. By consolidating these insights, we provide actionable rec- ommendations for policymakers, researchers, and practitioners to harness generative AI’s potential while addressingits socio-technical constraints, thereby fostering equitable and sustainable agricultural development.

Abstract: This study investigates how intersection-related factors affect traffic crash severity through a comparative analysis of two major U.S. cities: Chicago and New York City (NYC). Using large-scale crash datasets, the analysis applies logistic regression and machine-learning methods to assess how intersections and temporal conditions influence injury outcomes. The results indicate that intersection-related crashes significantly increase the probability of injury in both cities, though the magnitude is substantially larger in Chicago. Nighttime conditions consistently elevate crash severity across both cities. Model evaluation using ROC curves suggests moderate predictive performance, indicating the influence of additional unobserved factors. A comparative modeling framework further reveals that the relationship between intersection-related factors and crash severity is context-dependent, varying across urban environments. These findings highlight the importance of developing location-specific traffic safety strategies and demonstrate the value of integrating statistical, machine-learning, and spatial analyses in crash severity research.

Abstract: Background: Periodontitis is a chronic inflammatory disease increasingly associated with systemic immune dysregulation and microbial imbalance beyond the oral cavity. Emerging evidence suggests that gut microbiome dysbiosis contributes to periodontal inflammation through the oral–gut microbial axis. Methods: This systematic review was conducted according to PRISMA guidelines using the PECOS framework. A comprehensive literature search was performed across PubMed, Embase, Scopus, Web of Science, Cochrane Library, CINAHL, and Google Scholar databases to identify studies evaluating the association between gut microbiota and periodontitis, including microbiome alterations, inflammatory pathways, and microbiome-modulating interventions. Results: The included studies demonstrated that patients with periodontitis frequently exhibit reduced gut microbial diversity, enrichment of pro-inflammatory taxa, impaired intestinal barrier function, and elevated inflammatory mediators including C-reactive protein, interleukin-6, and tumor necrosis factor-alpha. Several studies identified translocation of periodontal pathogens such as Porphyromonas gingivalis and Fusobacterium nucleatum to the gastrointestinal tract, supporting the existence of an oral–gut axis. Probiotics, prebiotics, synbiotics, and periodontal therapy showed potential benefits in improving periodontal parameters and restoring microbial homeostasis. Conclusions: Current evidence supports a significant relationship between gut microbiome dysbiosis and periodontitis through immune, inflammatory, and metabolic mechanisms. However, heterogeneity among studies and limited longitudinal evidence warrant further standardized clinical and mechanistic investigations to establish causality and optimize microbiome-targeted therapeutic strategies in periodontal disease.

Abstract: Sleep disorders affect a substantial proportion of hospitalized patients yet remain among the most systematically underdiagnosed conditions in acute care medicine, with up to 80% of moderate-to-severe cases carrying no formal diagnosis at the time of admission. At the same time, frailty—a state of heightened physiological vulnerability arising from cumulative multi-system biological decline—is present in 40–80% of inpatients and shares deep, bidirectional neurobiological pathways with sleep pathology through shared mechanisms of circadian dysregulation, hypothalamic-pituitary-adrenal axis activation, and chronic low-grade inflammation. Despite this convergence, no study has integrated validated, administratively computable frailty phenotyping with a machine learning framework specifically designed to predict inpatient sleep disorder diagnosis at the point of hospital admission. To address this gap, we developed and evaluated a suite of five binary classification models—XGBoost, Random Forest, LightGBM, CatBoost, and Decision Tree—using 9,682 balanced hospitalization episodes from the MIMIC-IV (version 2.2) database. The predictor set comprised 23 admission-time structured features across three domains: frailty and comorbidity burden, including the Hospital Frailty Risk Score (HFRS) derived from ICD-10 codes, the Elixhauser comorbidity index, prior admission history, and six binary disease flags (obesity, hypertension, type 2 diabetes, heart failure, COPD, and depression/anxiety); physiological and laboratory biomarkers from the first 24 hours of care, including minimum SpO₂, heart rate variability, hemoglobin, creatinine, albumin, and arterial blood gas parameters; and sociodemographic and administrative variables encompassing age, sex, ethnicity, insurance type, and admission acuity. Two binary outcomes were modeled independently: any sleep disorder diagnosis (ICD-10: G47.x) and insomnia specifically (ICD-10: G47.00). Model performance was assessed through five-fold stratified cross-validation and bootstrap confidence intervals (n = 1,000 iterations), with predictor importance quantified using SHapley Additive exPlanations (SHAP). XGBoost achieved the strongest aggregate performance across all evaluation metrics, attaining an area under the receiver operating characteristic curve (AUC) of 0.871 (95% CI: 0.856–0.887), accuracy of 79.6%, F1-score of 0.820, and sensitivity of 94.9%, correctly identifying 903 of 952 true positive cases in the held-out test set; all gradient boosting frameworks substantially outperformed the Decision Tree baseline (AUC 0.836). SHAP analysis identified the HFRS and Elixhauser index as the two dominant predictors, followed by depression/anxiety, obesity, hypertension, and minimum SpO₂—a pattern that is mechanistically consistent with established pathophysiological literature on frailty-associated sleep pathology. The well-calibrated probability outputs of the XGBoost model make it directly suitable for integration into clinical decision support systems, offering a deployable, interpretable screening tool for inpatient sleep disorder identification that requires no dedicated instrumentation beyond routine admission data.

Abstract: The paper seeks to find the direct and indirect association amongst capital structure and firm value among all the nonfinancial firms listed in PSX from (2014-2019). Secondary panel data was used to conduct analysis. Structural equation modelling technique in Stata was used to estimate the direct effects. MedSEM, a special package for Stata, was used to estimate the indirect effects. Results showed that capital structure had no direct effect on value of the firm, but financial distress mediated the association amongst capital structure and value of the firm. Substantial indirect effect clearly manifests the existence of indirect nature of association amongst capital structure and value of the firm.