38978 research outputs found
Sort by
Evaluation of LLM-generated narratives based on patient-reported outcome measures
No full textPatient-reported outcome measures (PROMs) are instruments, often validated questionnaires, that capture patients’ experiences of symptoms, functioning, and quality of life. While useful, patients often receive little or no personalized feedback based on their PROMs responses, limiting their potential to support understanding, reflection, and management of their health. One promising way to provide such feedback is through patient narratives, which can translate structured questionnaire data into coherent stories that may enhance interpretability and emotional resonance (Boomstra et al., 2024).
In practice, however, providing personalized narratives for each individual PROMs profile is not feasible at scale due to time and resource constraints. Given recent advances in large language models (LLMs) and their potential of generating natural-language narratives from structured numeric input, the goal of this work is to examine if LLMs can generate personalized patient narratives that both accurately represent PROMs data and are of sufficient quality to be useful and meaningful.
The present study therefore investigates whether and how LLMs can generate narratives based on PROM data, focusing on the accuracy and quality of the generated texts. This study aims to provide evidence on the feasibility of using LLM-generated narratives as personalized PROMs feedback
An Eight-Parameter Assessment Framework for Tectonic Stress Evolution and Major Earthquake Probability Forecasting
No full textSeismo Framework v2.0.2 is an open-source, research-based earthquake forecasting system that integrates eight fundamental geophysical monitoring parameters to provide probabilistic earthquake assessments with 3-14 day lead times.
**Version 2.0.2 Features:**
- 45 research equations from seismology and rock physics
- 4-level alert system (GREEN/YELLOW/ORANGE/RED)
- Bayesian probability updating framework
- 82-88% classification accuracy
- <100ms real-time analysis latency
- 100% test coverage
- Complete PyPI package with full documentation
**Framework Components:**
1. Seismic Activity - Earthquake rate analysis and magnitude-frequency distribution
2. Crustal Deformation - GPS, InSAR, and strainmeter measurements
3. Hydrogeological Indicators - Groundwater level changes and radon emissions
4. Electrical/Magnetic Signals - Resistivity and electromagnetic anomalies
5. Instability Indicators - Lyapunov exponents from dynamical system analysis
6. Tectonic Stress State - Coulomb stress transfer calculations
7. Rock Properties - Seismic velocity variations and attenuation
8. Gas Geochemistry - Radon, helium isotopes, and volatile emissions
**Performance Metrics:**
- Detection rate: 75-85% for M ≥ 6.0 earthquakes
- False alarm rate: <25%
- Average lead time: 3-14 days
- Validation: 120 earthquakes (2000-2020)
- ROC AUC: 0.876 ± 0.021
**Case Studies:**
- 2011 Tōhoku Earthquake (M9.0): 7-day warning capability
- 2016 Kumamoto Earthquakes (M7.0): 48-hour lead time
- 2019 Ridgecrest Sequence: Multi-day forecasting
**Resources:**
- Research Paper: 67 pages, 12,500 words, 45 equations, 187 references
- Source Code: https://gitlab.com/gitdeeper3/seismo
- PyPI Package: https://pypi.org/project/seismo-framework/2.0.2/
- Zenodo DOI: 10.5281/zenodo.18563973
- Website: https://seismo.netlify.app
- Dashboard: https://seismo.netlify.app/dashboard
**Implementation:**
- Python 3.8+ with NumPy, SciPy, Pandas, FastAPI
- Open source: MIT License
- Real-time processing pipeline
- Comprehensive test suite
- Docker containerization support
**Disclaimer:** Research tool for scientific investigation. Not for public earthquake warnings without proper regional validation
Core Self-Evaluation and School Disengagement in Early Adolescence: A Three-Wave Longitudinal Network Analysis
No full textClinicians’ perspectives of Depict VR, a virtual-reality application for sharing experiences about mental distress between young people and their trusted confidante
No full textThese data are anonymised transcripts of interviews of 14 clinicians who tested Depict VR, a virtual-reality application for sharing experiences about mental distress between young people and their trusted confidante. The dataset also includes the coding of the transcripts for a Thematic Analysis that was performed by the researcher
Implementing Martha’s Rule in Paediatric Care
No full textThis study explores how Martha’s Rule is implemented and experienced in paediatric hospital care, with a focus on the daily check-in component (component 1).
Martha’s Rule is intended to help patients and families raise concerns early if they think someone is getting worse. The study examines what helps or gets in the way of children and young people (CYP), parents and carers, and healthcare staff engaging with daily check-ins, and how implementation varies across NHS Trusts.
The project uses a mixed-methods design with qualitative emphasis, including non-participant observations, semi-structured interviews, and surveys across multiple wards and NHS Trusts in England.
Findings will be used to propose practical improvements to make daily check-ins more accessible, equitable, and usable in paediatric settings. Outputs will include a PhD thesis, peer-reviewed publications, conference presentations, and plain-language summaries
Refractory and Difficult-to-Treat Axial Spondyloarthritis: Systematic Review and Meta-analysis of Prevalence and Determinants
No full text
Referring Effectively to a Group of Objects
No full textIn communication, we often need to refer to multiple objects. We have several possibilities: enumerating the objects one by one or referring to a shared feature. This is the first study to systematically compare different ways of referring to groups of objects in terms of response times, accuracy, and memory encoding. In a within-subjects design, participants listened to instructions either using a shared feature or enumerating coordinates of individual objects We find that there is no one way of referring that is universally more effective, but that the optimal way of referring depends on the type of common feature and the number of objects. When the shared feature is prominent, such as color, and multiple objects need to be identified, referring to it results in shorter identification times and fewer errors. In contrast, when the shared feature is less salient, such as pattern, referring by location tends to be better
Anxiety/stress symptoms uniquely predict greater negative global metacognitive bias in post-9/11 veterans
No full textObjective: Negative metacognitive bias, underestimating one’s abilities, is consistently linked to psychopathology, yet prior work has often collapsed anxiety/stress and depression or examined depression alone. We tested the unique associations of depression, anxiety/stress, and PTSD with metacognitive bias in post-9/11 veterans. Methods: Veterans from the Translational Research Center for TBI and Stress Disorders (TRACTS, N=601; 90% male; M_age=34.31) completed DASS-21 (depression, anxiety/stress), CAPS-IV (PTSD), WHODAS-II Understanding/Communicating (self-reported cognition), and an objective cognition composite (assessment of executive function, memory, attention). Bias was computed as self-report minus objective cognition. A subsample (n=239) repeated testing ~2 years later. Results: At time 1, more negative metacognitive bias was associated with greater anxiety/stress (r=−.41), depressive (r=−.37), and PTSD symptoms (r=−.31) (all ps<.001). In a simultaneous model, anxiety/stress (β=−.29 p<.001) and depressive symptoms (β=−.12, p=.045) explained unique variance, though PTSD symptoms did not (β=−.03, p=.524). Longitudinally, changes in bias were uniquely predicted by symptom changes in anxiety/stress (β=−.33, p<.001) and PTSD (β=−.16, p=.001), but not depression (β=−.10, p=.137). Conclusions: Across cross-sectional and longitudinal models, anxiety/stress emerged as the most consistent correlate of metacognitive bias, with weaker contributions from depression and PTSD. These findings highlight the importance of assessing the self-report vs. objective cognition gap, and the need to further understand the temporal relationship between anxiety/stress and metacognitive bias
