Scalable Model Research Articles

Leveraging Knowledge Graphs for Enhanced Medical Reasoning in Personalized Medicine for Rare Diseases

Leveraging knowledge graphs offers a promising approach to enhancing medical reasoning in the personalized treatment of rare diseases. Knowledge graphs provide an effective framework for organizing and interpreting complex biomedical data, facilitating a more comprehensive approach to diagnosis, prognosis, and treatment of rare diseases. By integrating diverse data sources—such as genomic profiles, clinical histories, and pharmacological information—into a cohesive graph structure, the project aims to improve clinical outcomes by enabling patient-specific insights and treatment recommendations. Rare diseases pose unique challenges due to limited data and complex symptoms, making traditional diagnostic methods inadequate. Knowledge graphs have the potential to bridge these gaps by linking disparate datasets, allowing clinicians to identify biomarkers, make more accurate diagnoses, and formulate personalized treatment pathways. This project involves constructing a multi-layered knowledge graph, applying machine learning and reasoning algorithms for pattern detection, and validating its accuracy in real-world settings. Expected outcomes include improved diagnostic accuracy, the discovery of new biomarkers, and the development of a scalable knowledge graph framework adaptable for a wide range of diseases. This work aims to contribute to the advancement of personalized medicine, offering a model for AI-driven reasoning that supports more informed and timely clinical decisions in the treatment of rare diseases. By organizing complex biomedical data into interconnected graphs, clinicians can gain comprehensive insights into rare disease mechanisms, facilitating more accurate diagnoses and personalized treatment strategies. Integrating diverse data sources, such as genomic, clinical, and pharmacological information, the knowledge graph structure enables the identification of relationships and patterns that are often missed in traditional analyses. Machine learning algorithms further support this process by identifying potential biomarkers and recommending treatment pathways based on patient-specific data. Through real-world validation with clinical partners, the knowledge graph approach is expected to improve diagnostic accuracy and support early intervention strategies. Ultimately, this framework not only addresses the unique challenges of treating rare diseases but also establishes a scalable model that can be adapted for broader applications in personalized medicine.

Expanding access to addictions care: Implementation of a 24-hour healthcare provider support line in British Columbia, Canada

BackgroundMorbidity and mortality related to substance use have risen to catastrophic levels in North America, and treatment services are often difficult to access. In response, the province of British Columbia (BC), Canada, launched a province-wide addiction medicine support phone line that offers clinicians immediate access to phone consultation with an addictions medicine expert. The service operates 24/7 is accessible to any clinician in the province seeking assistance with an addiction-related question. We describe an evaluation of the reach and perceived impact of the service over its first two years.MethodsThe 24/7 Addiction Medicine Clinician Support Line was evaluated prospectively from June 2020 to April 2022. All provider-to-provider encounters were included. Data was collected from two primary sources: health provider demographic information collected at the time of consultation, and optional clinician surveys conducted after the consultation was complete. Descriptive data are presented as numerical values and percentages.ResultsOver the 22-month evaluation period, 1,279 consultations were requested by 631 distinct care providers across British Columbia. The service averaged 15 calls per week across the province, and 51.5% of calls were made outside of business hours. Physicians made the majority of calls to the service (n = 865, 67.6%), followed by nurse practitioners (n = 162, 12.7%). Among those who completed a follow-up survey (n = 258 calls, 20.2% total calls), 81.8% (n = 211) were “very” or “extremely” satisfied with the consultation. Of these respondents, 65.5% (n = 169) reported that the consultation led to the provision of better care for their patient, with 58.1% (n = 150) initiating a new prescription and 22.1% (n = 57) reporting expedited treatment for their patient. The consultation area of focus was most commonly opioid use (n = 417; 59.6%), followed by polysubstance use (n = 98; 14.0%).ConclusionsThe impact of the 24/7 Addiction Clinician Support Line was widespread, and the service increased accessibility to evidence-based addictions treatment across a range of care settings. Clinicians expressed a high degree of satisfaction with the service. To our knowledge, this province-wide program is the first of its kind in North America, offering a scalable and adaptable model to support access to evidence-based addictions care in under-resourced settings.

A CDSS based Integrated Pathway for Influenza Vaccination Across Primary and Secondary Care

Abstract Background In Italy, influenza vaccination coverage in people ≥ 65 years is unsatisfactory (56.7%), underlining the need to improve current vaccination strategies. The objectives of this study have been: to define a novel pathway to increase influenza vaccination adherence among ≥ 65 years population in the Lazio region; to provide a predictive estimate of the epidemiological and economic pathway’s potential impact. Methods A multidisciplinary working group (WG) featuring cross-sectoral expertise was created and WG periodic meetings were held to define the patient journey and its flow-chart representation. The pathway’s potential impact was assessed through epidemiological and economic indicators and scenario analyses. Results An integrated pathway across primary and secondary care was defined, based on the active patient in-Hospital recruitment and vaccination and enhanced by a Clinical Decision Support System relying on a digital algorithm to identify eligible patients. Assuming an increase of influenza vaccination coverage from the current rate of 60% (scenario 1) to 65% (scenario 2) in ≥ 65 years population in the Lazio region thanks to the pathway implementation, an increase of 8% in avoided influenza cases, influenza- or pneumonia-related hospitalizations and influenza-related outpatient visits was estimated with a relative increase in savings for hospitalizations and outpatients visits of up 2,367,310 euros. Setting the vaccination coverage at 70% (scenario 3), an increase of 16% in avoided influenza cases, hospitalizations and outpatient visits was estimated with a relative increase in savings for hospitalizations and outpatients’ visits of up to 4,833,259 euros. Conclusions Alongside offering a predictive estimate of the relevant pathway’s potential impact, both epidemiological and economic, this project, with its robust methodology, may serve as a scalable and transferable model for enhancing vaccination coverage at national and international level. Key messages • The proposed pathway, offering the option of receiving flu vaccination within the Hospital, supports the paradigm shift towards primary prevention pathways in secondary care settings. • Another relevant aspect of the integrated pathway is the adoption of an Artificial Intelligence tool to identify suitable patients and improve their recruitment and adherence to vaccination.

Expanding access to non-communicable disease screening in war-affected communities in Ukraine

Abstract Background Non-communicable diseases (NCDs) account for 91% of premature deaths in Ukraine. However, NCD screening lags among young people and across the entire population. The ongoing war has exacerbated barriers to NCD screening due to damaged infrastructure, understaffed medical facilities, and health system deterioration. Intervention/response To improve screening, diagnosis, and care, in March 2023 the PATH-led, USAID-funded Support TB Control Efforts in Ukraine (STBCEU) project established community level health access points (HAPs) in five conflict-affected communities. STBCEU procured screening tools (tannometers, glucometers, portable ECGs) for common NCDs, printed informational materials, trained 143 specialists, designed a reporting system, and facilitated monitoring visits to 77 HAPs. Results By the end of 2023, the project equipped 113 HAPs in the five communities, and 102 of them initiated NCD screening. From January-March 2024, 1,694 people under 40 years old were screened for NCDs. Among 1,488 blood pressure measurements, 143 cases had high readings (9.6 ±1.5%), 90 (62.9%) of which were first-time diagnoses, and 53 (37.1%) treatment monitoring. Glucose levels were measured 809 times, with elevated levels in 44 cases (5.4 ±1.5%), 30 (68.2%) of which were first-time diagnoses, and 14 (31.8%) treatment monitoring. Additionally, 1,144 mental health consultations were provided. Conclusions This initiative represented a first in Ukraine: per national requirements, young people have not historically been screened for NCDs. Expanding NCD screening in war-affected communities has demonstrated substantial early success, with significant detection of high blood pressure and elevated glucose levels among people under 40. This proactive, cost-effective intervention underscores the need for accessible, community-level health interventions in crisis settings and showcases a scalable model for integrating NCD screening into emergency health response frameworks. Key messages • The implementation of community-level NCD screening has shown early success in detecting high blood pressure and elevated glucose levels among people under 40. • Cost-effective NCD screening underscores the need for accessible, community health interventions, providing a scalable model for integrating NCD screening into emergency health response.

Boosting influenza vaccination rates through policy gamification: the ‘Let’s Control Flu’ project

Abstract Background Influenza remains a public health challenge globally, leading to substantial morbidity and mortality, particularly among vulnerable groups. Despite the benefits of vaccination, achieving the desired vaccination coverage rates (VCR) proves challenging across Europe. The ‘Let’s Control Flu’ (LCF) tool emerges as a new contribution, merging gamification with epidemiological data and public health policies to help increase influenza VCR. Methods The LCF tool, developed from an extensive literature review and validated by a National Advisory Board, initially showcased its capabilities in Sweden. Offering an interactive simulation, it allows stakeholders to explore the outcomes of implementing 13 targeted public health policies (PHP) on influenza VCR and health outcomes such as infection rates, hospitalizations, and mortality. This approach simplifies epidemiological complexities, paving the way for informed decision-making. Results Following its success in Sweden, the LCF tool was extended to Germany and Czechia, underlining the value of gamified tools in public health decision-making. The application across these new countries not only resulted in paths to improve VCR but also provided insights into the tool’s scalability and adaptability different landscapes. Conclusions The LCF project underscores the efficacy of integrating gamification into health policy decision-making, offering a replicable model for enhancing influenza vaccination uptake across Europe. With the tool now successfully applied in multiple countries, its potential for broader European and global application is evident. This initiative aligns with WHO’s objectives, spotlighting gamification as a relevant component in improving vaccination strategies and readiness for influenza and other preventable diseases, thereby reinforcing the importance of innovative approaches in public health policy development. Key messages • The LCF tool integrates gamification with public health strategies to help effectively raise influenza vaccination coverage. • By merging gamification and epidemiological insights, the LCF tool also serves as a scalable model for Europe, supporting WHO’s goals to improve vaccine strategies and preparedness.

Developing regional Diabetes Registries in Italy: pilot projects implementation within JACARDI

Abstract Issue/Problem In Italy a National Diabetes registry is planned but not yet set up at the Ministry; JACARDI is contributing to speed up the process so that critical gaps in health data availability can be filled. While some regional registries exist, the lack of cohesion and standardization hinders the tracking of prevalence, incidence, and sequelae, crucial for evidence-based policymaking amidst rising diabetes rates. Description of the Problem The JACARDI project, supported by EU funding and running for four years, engages 76 partners from 21 countries and includes 142 national-level projects. Among its focus areas is enhancing data availability, aiming to establish harmonized regional Diabetes registries in Italy to tackle this issue. These efforts aim to create a scalable model enabling data linkage, diabetes type differentiation, and socio-demographic integration. Collaboration with health authorities and societies, drawing insights from existing registries, informs the development process. Results A total of 11 regional Diabetes registries are currently in the design phase of implementation. They adhere to a clearly defined methodology, with established essential parameters and data linkage. Consensus on these parameters and linkage was reached through stakeholder discussions, drawing on past registry experiences. The primary goal of the registries is to monitor diabetes prevalence, stratify incidence by demographics, and track sequelae to inform targeted interventions and resource allocation. While scalability beyond Italy may encounter administrative challenges, standardized protocols facilitate transferability. Data Protection stands out as a significant implementation obstacle. Lessons Standardized data collection, stakeholder collaboration, and sustainability are vital in registry development. Leveraging data availability for surveillance demonstrates adaptability and scalability, offering valuable insights for enhancing diabetes management. Key messages • Standardized data collection and stakeholder collaboration are pivotal in establishing effective diabetes registries, fostering equity and improving health outcomes. • Leveraging data availability for surveillance offers a scalable and sustainable approach to diabetes management, facilitating informed policymaking and interventions.

Decentralized Machine Learning Framework for the Internet of Things: Enhancing Security, Privacy, and Efficiency in Cloud-Integrated Environments

The Internet of things (IoT) presents unique challenges for the deployment of machine learning (ML) models, particularly due to constraints on computational resources, the necessity for decentralized processing, and concerns regarding security and privacy in interconnected environments such as the Internet of cloud. In this paper, a novel decentralized ML framework is proposed for IoT environments characterized by wireless communication, dynamic data streams, and integration with cloud services. The framework integrates incremental learning algorithms with a robust decentralized model exchange protocol, ensuring that data privacy is preserved, while enabling IoT devices to participate in collaborative learning from distributed data across cloud networks. By incorporating a gossip-based communication protocol, the framework ensures energy-efficient, scalable, and secure model exchange, fostering effective knowledge sharing among devices, while addressing the potential security threats inherent in cloud-based IoT ecosystems. The framework’s performance was evaluated through simulations, demonstrating its ability to handle the complexities of real-time data processing in resource-constrained IoT environments, while also mitigating security and privacy risks within the Internet of cloud.

Iterative Development of a Clinical Decision Support Tool to Enhance Naloxone Co-Prescribing.

Background Opioid overdoses have contributed significantly to mortality in the United States. Despite long-standing recommendations from the Centers for Disease Control and Prevention to co-prescribe naloxone for patients receiving opioids who are at high risk of overdose, compliance with these guidelines has remained low. Objectives The objective of this study was to develop and evaluate a hospital-wide electronic health record (EHR)-based clinical decision support (CDS) tool designed to promote naloxone co-prescription for high-risk opioids. Methods We employed an iterative approach to develop a point-of-order, interruptive EHR alert as the primary intervention and assessed naloxone prescription rates, EHR efficiency metrics, and barriers to adoption. Data was obtained from our EHR's clinical data warehouse and analyzed using statistical process control and Chi-square analyses to assess statistically significant differences in prescribing rates during the intervention periods. Results The initial implementation phase of the intervention, spanning from April 2019 to May 2022, yielded a nearly 3-fold increase in the proportion of high-risk patients receiving naloxone, rising from 13.4% [95% CI, 12.9% - 13.8%] to 36.4% [95% CI, 35.2% - 37.5%; p = 1 x 10-38]. Enhancements to the CDS design and logic during the subsequent iteration's study period, June 2022 and December 2023, reduced the number of CDS triggers by more than 30-fold while simultaneously driving an additional increase in naloxone receipt to 42.7% [95% CI, 40.6% - 44.8%; p = 2 x 10-5]. The efficiency of the CDS demonstrated marked improvement, with prescribers accepting the naloxone co-prescription recommendation provided by the CDS in 41.1% of the encounters in version two, compared to 6.2% in version one (p = 6 x 10-9). Conclusion This study offers a sustainable and scalable model to address low rates of naloxone co-prescription and may also be used to target other opportunities for improving guideline-concordant prescribing practices.

Addressing Inpatient Hyponatremia Through Targeted Automatic E-consults: A Pilot Randomized Trial.

Hyponatremia is the most common electrolyte abnormality in hospitalized patients. Treatment of hyponatremia is associated with improved outcomes, but more than one in three cases of new onset hyponatremia is not corrected by the time of hospital discharge. Nephrologist input may improve the diagnosis and treatment of hyponatremia, but specialist resources are limited. Targeted automatic electronic consultations (TACos) may be one approach to provide expert nephrologist guidance to the workup and management of hyponatremia using a scalable model. Evaluate the feasibility and acceptability of a TACo intervention for hospitalized patients with hyponatremia. Single-site, parallel-group cluster randomized trial. Adult inpatients with hyponatremia on the hospital medicine service. A nephrologist conducted TACos on intervention patients, making diagnostic and therapeutic recommendations daily (if warranted) until discharge or resolution of hyponatremia. Measures of feasibility included the number of eligible participants, percentage receiving TACos, number of TACos per participant, and percentage of formal nephrology consults. Acceptability was assessed by a post-intervention survey. Clinical outcomes, including the percentage of hyponatremia cases that resolved by discharge, were also assessed. We identified 62 patients who met inclusion criteria: 38 in the intervention group and 24 in the control group. A nephrologist determined that 26 of 38 intervention patients (68%) would likely benefit from diagnostic and management recommendations; 67 TACos were performed (mean 2.6 per patient). Fourteen of 18 primary team physicians (78%) reported that the e-consults changed their management, and 15 of 18 (83%) wanted TACOs to continue. Resolution of hyponatremia, length of stay, 30-day readmissions, and costs were similar in the intervention and control groups. Inpatient TACos for hyponatremia were feasible and acceptable to primary teams, and frequently led to changes in diagnosis and management. Further studies are needed to determine the impact of the TACo model on clinical outcomes and cost-effectiveness.

Human Astrocytes Synchronize Neural Organoid Networks.

Biological neural networks exhibit synchronized activity within and across interconnected regions of the central nervous system. Understanding how these coordinated networks are established and maintained may reveal therapeutic targets for neurodegeneration and neuromodulation. Here, we tested the influence of astrocytes upon synchronous network activity using human pluripotent stem cell-derived bioengineered neural organoids. This study revealed that astrocytes significantly increase activity within individual organoids and across long distances among numerous rapidly merged organoids via influencing synapses and bioenergetics. Treatment of amyloid protein inhibited synchronous activity during neurodegeneration, yet this can be rescued by propagating activity from neighboring networks. Altogether, this study identifies critical contributions of human astrocytes to biological neural networks and delivers a rapid, reproducible, and scalable model to investigate long-range functional communication of the nervous system in healthy and disease states.

Flow induced acoustic characteristics analysis of propulsion pump by indirect acoustic variable method and FEM

The acoustic characteristics of water-jet propulsion pumps present significant challenges, primarily due to the interaction between flow and high-speed rotating impellers wall. Notably, such complicated turbulence induced acoustic sources by brings great difficulty to identify and control. This paper firstly analyzes the flow characteristics of the propulsion pump by using the scalable Hybrid RANS-LES model. Subsequently, an indirect acoustics variable method based on Lamb vector divergence is proposed to characterize acoustic sources caused by multiscale turbulence. The effectiveness and advantages of indirect acoustics variable method are evaluated through comparison with the acoustic finite element method and experimental results. The proposed indirect acoustics variable method demonstrates the ability to accurately quantify the contribution of turbulent flow to acoustic sources. Furthermore, the intensity of acoustic sources in different flow regions can be obtained through volume integrals, which can serve as optimization parameters in the initial hydraulic design of propulsion pumps. This study provides a novel approach for predicting turbulent noise and offers guidance for the design of low-noise propulsion systems.

Advanced supply chain optimization for emerging market healthcare systems

The complexities of pharmaceutical supply chains in emerging markets present significant challenges that directly impact healthcare outcomes, particularly in underserved regions. This research focuses on the optimization of supply chain systems for the healthcare sector, with an emphasis on enhancing pharmaceutical delivery. By leveraging data analytics and innovative procurement strategies, the study seeks to improve efficiency, reduce operational costs, and ensure the timely distribution of essential medications. Drawing on practical experiences at eMedic Technologies Africa, the research highlights the need for advanced models that integrate real-time data to optimize inventory management, forecast demand, and address bottlenecks in last-mile delivery. The study proposes a multi-faceted approach to addressing common challenges in emerging market supply chains, including unreliable infrastructure, fluctuating demand, and limited access to technology. By utilizing predictive analytics, the proposed models aim to improve accuracy in demand planning, reducing stockouts and overstock situations. Additionally, innovative procurement strategies, such as centralized procurement and strategic partnerships with local suppliers, are explored as key methods for enhancing cost-effectiveness and streamlining supply chain operations. A critical component of the research is the focus on last-mile delivery, which remains a significant hurdle in rural and underserved areas. The study advocates for the use of digital platforms and mobile technologies to enhance delivery tracking and coordination, ensuring that essential medications reach their destination efficiently. Furthermore, it emphasizes the importance of fostering collaboration between governments, non-governmental organizations, and private sector stakeholders to build resilient supply chain systems capable of adapting to local conditions. By providing actionable insights and scalable models for supply chain optimization, this research aims to contribute to improved healthcare delivery in emerging markets. The proposed strategies not only reduce costs and improve efficiency but also ensure a reliable and consistent supply of medications, ultimately leading to better health outcomes in regions most in need. Keywords: Supply Chain Optimization, Data-Driven Procurement, Last-Mile Delivery, Healthcare Efficiency, Emerging Markets

Progress in carbon dioxide capture, storage and monitoring in geological landform

Carbon Capture and Storage (CCS) is recognized as a potent strategy for managing the accumulation of human-generated CO2 in the atmosphere, helping to alleviate climate change’s effects. The CO2 gas is captured from the point source through methods such as pre-treating fossil fuels, oxy-fuel combustion, or post-combustion capture; thereafter; it is transported to a storage location and injected into geological formations. This article provides an overview of carbon dioxide capture and sequestration, focusing on its key principles, technologies, associated risks, and challenges. Direct Air Capture (DAC) and Scalable Modelling, Artificial intelligence (Al), Rapid Theoretical calculations SMART technologies are detailed as emerging and promising approaches to CO2 capture. Numerous pilot and commercial projects commissioned to manage carbon dioxide emissions are presented. Additionally, the paper explores approaches combining geological, geophysical, geochemical, and environmental monitoring techniques to ensure the secure and sustainable storage of CO2 underground. These are essential to address uncertainties, minimize risks, and build public confidence in CCS as a viable climate mitigation strategy. The successful deployment of these technologies on a global scale will require continued innovation, particularly in the areas of monitoring, risk management, and public engagement. Emerging technologies such as AI and SMART systems could play a crucial role in enhancing the efficiency and safety of CCS operations. However, the integration of these advancements with existing infrastructure and regulatory frameworks remains a challenge. Ultimately, a multi-disciplinary approach, combining technological, economic, and regulatory perspectives, will be vital to realizing the full potential of CCS in combating climate change.

Decoding Spatial Tissue Architecture: A Scalable Bayesian Topic Model for Multiplexed Imaging Analysis.

Recent progress in multiplexed tissue imaging is advancing the study of tumor microenvironments to enhance our understanding of treatment response and disease progression. Cellular neighborhood analysis is a popular computational approach for these complex image data. Despite its popularity, there are significant challenges, including high computational demands that limit feasibility for large-scale applications and the lack of a principled strategy for integrative analysis across images. This absence hampers the precise and consistent identification of spatial features and tracking of their dynamics over disease progression. To overcome these challenges, we introduce SpaTopic, a spatial topic model designed to decode high-level spatial architecture across multiplexed tissue images. This algorithm integrates both cell type and spatial information within a topic modelling framework, originally developed for natural language processing and adapted for computer vision. Spatial information is incorporated into the flexible design of documents, representing densely overlapping regions in images. The model employs an efficient collapsed Gibbs sampling algorithm for both statistical and computational inference. We benchmarked the performance against five state-of-the-art algorithms through various case studies using different single-cell spatial transcriptomic and proteomic imaging platforms across different tissue types. Our findings demonstrate that SpaTopic consistently identifies biologically and clinically significant spatial "topics" such as tertiary lymphoid structures (TLSs) and tracks dynamic changes in spatial features over disease progression. Its computational efficiency and broad applicability across various molecular imaging platforms will enhance the analysis of large-scale tissue imaging datasets.

Workforce Optimization for Quick Commerce Dark Stores: Talent Sourcing and Workforce Modelling Strategies

Quick commerce (Q-commerce) has transformed the retail and e-commerce sectors by enabling ultra-fast delivery of goods through hyper-local logistics and dark stores. The operational success of this model hinges on a structured workforce strategy that supports efficient scaling and agility. This paper explores innovative workforce optimization techniques that focus on organizational design, talent sourcing, and workforce modelling, referencing industry leaders like Flipkart, Zepto, Instamart, and Blinkit. The study identifies the vital roles within dark stores and the approaches to talent acquisition, emphasizing localized recruitment, employee retention, and technology integration. It further presents best practices for scalable workforce modelling to accommodate seasonal demands, high attrition rates, and hyper-growth environments, ensuring sustainable development in the Q-commerce industry. Keywords: Quick Commerce, Workforce Strategy, Talent Sourcing, Dark Stores, Organizational Design, Hyper-Local Delivery, Workforce Modelling

Scalable incident detection via natural language processing and probabilistic language models

Post marketing safety surveillance depends in part on the ability to detect concerning clinical events at scale. Spontaneous reporting might be an effective component of safety surveillance, but it requires awareness and understanding among healthcare professionals to achieve its potential. Reliance on readily available structured data such as diagnostic codes risks under-coding and imprecision. Clinical textual data might bridge these gaps, and natural language processing (NLP) has been shown to aid in scalable phenotyping across healthcare records in multiple clinical domains. In this study, we developed and validated a novel incident phenotyping approach using unstructured clinical textual data agnostic to Electronic Health Record (EHR) and note type. It’s based on a published, validated approach (PheRe) used to ascertain social determinants of health and suicidality across entire healthcare records. To demonstrate generalizability, we validated this approach on two separate phenotypes that share common challenges with respect to accurate ascertainment: (1) suicide attempt; (2) sleep-related behaviors. With samples of 89,428 records and 35,863 records for suicide attempt and sleep-related behaviors, respectively, we conducted silver standard (diagnostic coding) and gold standard (manual chart review) validation. We showed Area Under the Precision-Recall Curve of ~ 0.77 (95% CI 0.75–0.78) for suicide attempt and AUPR ~ 0.31 (95% CI 0.28–0.34) for sleep-related behaviors. We also evaluated performance by coded race and demonstrated differences in performance by race differed across phenotypes. Scalable phenotyping models, like most healthcare AI, require algorithmovigilance and debiasing prior to implementation.

A new deep learning-based model for reconstructing high-quality NDVI time-series data in heavily cloudy areas: fusion of Sentinel 1 and 2 data

ABSTRACT Reconstructing high-quality Normalized Difference Vegetation Index time series data is essential for ecological and agricultural applications but remains challenging in heavily cloudy areas. Fusing Sentinel SAR and optical data with deep learning could be helpful but is also challenging for stable models due to unstable SAR-NDVI relationships caused by imaging mechanism differences and environmental complexities. In this study, we developed a new Bidirectional Recurrent Imputation for Optical-SAR fusion (BRIOS) model to reconstruct high-quality Sentinel-2 NDVI time series data. BRIOS designs a two-layer recurrent architecture that integrates the autocorrelation of discrete, cloud-free NDVI observations into the model for a more stable SAR-NDVI relationship. Evaluating BRIOS against three baseline methods (GF-SG spatiotemporal fusion, Harmonic regression interpolation, and MCNN-Seq deep learning) across three full Sentinel-2 tiles in reconstructing 8-day NDVI time series, BRIOS consistently outperformed in scenarios of either random or continuously missing data, as evidenced by lower RMSE values (e.g. 0.075 for BRIOS vs. 0.108 for GF-SG vs. 0.143 for Harmonic regression vs. 0.303 for MCNN-Seq), better Edge index, and high linear correlation coefficients (R values up to 0.97). Further ablation experiments revealed that deep integration of NDVI autocorrelation features and SAR temporal change patterns has improved the stability and generalization of BRIOS. Discussions on the model's scalability across various cloud sizes and training dataset sizes affirm its practicality for broad-scale application in vegetation monitoring under challenging cloudy conditions.

Transitioning to rural practice together: a rural fellowship model (in 6 Ps).

Maintaining a robust healthcare workforce in underserved rural communities continues to be a challenge. To better meet healthcare needs in rural areas, training programs must develop innovative ways to foster transition to, and integration into, these communities. Mountain Area Health Education Center designed and implemented a 12-month post-residency Rural Fellowship program to enhance placement, transition, and retention in rural North Carolina. Utilizing a '6 Ps' framework, the program targeted physicians and pharmacists completing residency with the purpose of recruiting and supporting their transition into the first year of rural practice. To better understand Rural Fellows' experiences and the immediate impact of their Fellowship year, we conducted a semi-structured interview using a narrative technique and evaluated retention rates over time. Interviews with the eight participants, which included Fellowship alumni and current Fellows, demonstrated the impact and influence of the key curricular '6 Ps' framework. An early retention rate of 100% and a long-term retention rate of 87%, combined with expressed clarity of curricular knowledge, skills, and attitudes related to the '6 Ps', demonstrate the potential and effectiveness of this Rural Fellowship model. Participants indicated the Rural Fellowship experience supports the transition to rural practice communities and expands their clinical skills. The Rural Fellowship program demonstrates an effective model to support early career healthcare providers as they begin practice in rural communities in western North Carolina through academic opportunities, personal growth, and professional development. Implementation of this model has demonstrated the success of a rural retention model over a 6-year period. This model has the potential to target an array of clinical providers and disciplines. We started with family medicine and have expanded to psychiatry, obstetrics, pharmacy, and nursing. This study demonstrated that this model supports clinical providers during the critical transition period from residency to practice. Targeting the most important stage of one's medical training, the commencement of professional practice, this is a scalable model for other rural-based health professions education sites where rural recruitment and retention remain a problem.

Eco-Friendliness and Cost-Effectiveness of Reusable Sanitary Pads for Rural Women: A Model for Menstrual Hygiene Management in Low Resource Settings

Background Menstrual Hygiene Management (MHM) is fundamental to women’s health, well-being, and gender equality. The study aimed to assess the cost-effectiveness and acceptability of reusable sanitary pads among rural women in Pakistan, proposing a scalable model for improving MHM practices in low-income countries. Methods A cross-sectional study was conducted using Multi- Stage Cluster sampling to accomplish a sample size of 340 childbearing-age women from the rural areas of Sindh, Pakistan. The data was collected using a standard questionnaire regarding the affordability and accessibility of reusable versus disposable sanitary pads; further information was also gathered from Market surveys for MHM products. The biodegradability of the pads was assessed using a soil burial test. The data was analyzed using SPSS version 27. Results Our research found a significant preference for reusable sanitary pads. The biodegradability tests demonstrated that reusable pads had a superior degradation rate (1.88%) to disposable pads (1.59%) over four months. Economic assessments showed a low utilization of commercial pads (12.5%) due to affordability issues, with 87.5% using reusable cloths. Despite 70% awareness of menstrual hygiene, 48% of women showed interest in creating sanitary solutions, underscoring the feasibility of implementing locally sourced, cost-effective products across similar socio-economic landscapes. Conclusion The present study underscores the universal applicability of affordable, sustainable MHM solutions in rural communities of low-income countries, advocating for broader implementation of such models to address global menstrual hygiene challenges. Bangladesh Journal of Medical Science Vol. 23 No. 04 October’24 Page : 1030-1037

Efficient Bayesian Policy Reuse With a Scalable Observation Model in Deep Reinforcement Learning.

Bayesian policy reuse (BPR) is a general policy transfer framework for selecting a source policy from an offline library by inferring the task belief based on some observation signals and a trained observation model. In this article, we propose an improved BPR method to achieve more efficient policy transfer in deep reinforcement learning (DRL). First, most BPR algorithms use the episodic return as the observation signal that contains limited information and cannot be obtained until the end of an episode. Instead, we employ the state transition sample, which is informative and instantaneous, as the observation signal for faster and more accurate task inference. Second, BPR algorithms usually require numerous samples to estimate the probability distribution of the tabular-based observation model, which may be expensive and even infeasible to learn and maintain, especially when using the state transition sample as the signal. Hence, we propose a scalable observation model based on fitting state transition functions of source tasks from only a small number of samples, which can generalize to any signals observed in the target task. Moreover, we extend the offline-mode BPR to the continual learning setting by expanding the scalable observation model in a plug-and-play fashion, which can avoid negative transfer when faced with new unknown tasks. Experimental results show that our method can consistently facilitate faster and more efficient policy transfer.