Magdalena Katharina Wekenborg

Vagally‐mediated heart rate variability (vmHRV) is widely proposed as a biomarker for physical and mental health, but its utility hinges on psychometric robustness, particularly temporal stability and consistency across measurement contexts. Prior reliability studies have been constrained by small samples, short retest intervals, and analytic approaches that cannot disentangle stable individual differences from situational fluctuations and measurement error. The present study examined the temporal and transsituational stability of vmHRV over a 1‐year interval in a large, population‐based sample ( N = 175) from the Dresden Burnout Study. Short‐term inter‐beat‐interval (IBI) recordings were obtained at two annual assessments under two conditions: recumbent recovery following venipuncture and seated rest. Four cardiac and HRV‐derived parameters, the established vmHRV indices root mean square of successive differences (RMSSD), percentage of successive RR intervals differing by more than 50 ms (pNN50), and high‐frequency power (HF power), complemented by mean RR interval (RR) as a general cardiac timing parameter, were analyzed using single‐trait–multistate–method‐minus‐one [STMSM(M–1)] models within the latent state–trait (LST) framework, with recovery specified as the reference condition and resting modeled as a condition‐specific method factor. Reliability ranged from 0.650 to 0.903, and trait specificity ranged from 0.375 to 0.773. RR and RMSSD showed the most consistent trait‐like profiles across occasions and conditions. pNN50 showed the highest individual trait‐specificity estimate, but also greater heterogeneity across occasions, whereas HF power showed comparatively lower trait specificity and greater occasion‐specific variance. Method specificity was small (0.034–0.087), indicating that stable condition‐specific deviations from recovery accounted for a limited proportion of variance. These findings support the interpretation of vmHRV, particularly RMSSD and pNN50, as mainly trait‐like indicators of cardiac vagal function, while also showing that single short‐term assessments contain meaningful occasion‐related variance. Standardized assessment conditions and repeated measurements remain essential for biomarker applications of vmHRV.

The visual environment is a fundamental yet unquantified determinant of mental health. While the concept of the environmental exposome is well-established, current methods rely on coarse geospatial proxies or biased self-reports, failing to capture the first-person visual context of daily life. Here we show that coupling ecological momentary assessment with vision-language models can quantify the semantic richness of human visual experience. Across 2,674 participant-generated photographs, model-derived estimates of “greenness” robustly predicted momentary affect and chronic stress, consistent with established benchmarks. We then developed a semi-autonomous large language model-based pipeline that mined over seven million scientific publications to extract nearly 1,000 environmental features empirically linked to mental health. When applied to real-world imagery, up to 33% of model-extracted context ratings significantly correlated with affect and stress.

Background: Previous evidence suggest that chatbots based on generative artificial intelligence (AI) are increasingly used in clinical practice, but little evidence exists on the specific conditions: What tools are used and for which purpose? Who provides access? What happens to patient data? Existing evidence is fragmented, with surveys constrained to single specialties, career stages, or practice types, limited by small samples, or conducted by AI vendors. We therefore aimed to characterize the usage among German physicians across sectors, specialties, and career stages, and the institutional conditions under which this use occurs.Methods: We conducted a nationwide, anonymous, cross-sectional online survey of German physicians between December 2025 and August 2026, recruited by convenience sampling through professional societies and direct contact with healthcare providers. Findings: After excluding incomplete responses and non-physicians, n = 3229 were analyzed. Participant group proportions resembled the national physician statistics except for an over representation of hospital-based physicians (67% vs 52%). 64% of respondents reported using AI chatbots regularly in clinical practice. Among chatbot users, use was more common for clinical decision support (77%) than for documentation assistance (44%). The most frequently used tools were general-purpose chatbots, most commonly ChatGPT (38% of tools mentioned). The most mentioned tool specifically designed for clinical use was OpenEvidence with 11%. Among chatbot users, only 6% had access to an institutionally provided solution; most relied on private self-paid accounts (27%) or free offerings (53%). Only a minority of participants reported that their workplace provided mandatory (9%) or voluntary (6%) training on chatbot use, and only 40% knew of applicable institutional guidelines.Interpretation: Our survey suggests that chatbot use occurs predominantly outside institutionally sanctioned channels, with little formal training or guidance. The findings indicate an urgent need for implementation strategies, certified tools, and training to ensure safe and effective use. Funding: No external funding; the study was conducted using institutional resources of Else Kröner Fresenius Center for Digital Health, TUD Dresden and Institute for Digitization in Medicine, Medical Center University of Freiburg.

The last years have seen an acceleration in the development and uptake of artificial intelligence (AI) systems by “early adopter” hospitals, caught between the pressures to “perform” and “transform” in a struggling health care system. This transformation has raised concerns among health care providers as their voices and location-specific workflows have often been overlooked, resulting in technologies that fail to integrate meaningfully into routine care and worsen rather than improve care processes. How can positive AI implementation be carried out in health care, aligned with European values? Based on a perspective that spans all stakeholders, we have created EURAID (European Responsible AI Development), a practical, human-centric framework for AI development and implementation based on agreed goals and values. We illustrate this approach through the co-development of a narrow-purpose “in-house” AI system, designed to help bridge the AI implementation gap in real-world clinical settings. This example is then expanded to address the broader challenges associated with complex, multiagent AI systems. By portraying all key stakeholders across the AI development life cycle and highlighting their roles and contributions within the process, real use cases, and methods for achieving iterative consensus, we offer a unique practical approach for safe and fast progress in hospital digital transformation in the AI age.

Stress‐related chronic exhaustion can be predicted longitudinally by reduced basic vagal tone (i.e., vagally‐mediated heart rate variability [vmHRV]). However, little is known about the relationship between phasic vmHRV and momentary exhaustion in daily life. To examine this relationship, this preregistered study used ecological momentary assessment (EMA) in a sample of N = 151 healthy participants (age = 22.17 years [SD = 4.98 years]; 14.57% male) for three consecutive weekdays. Exploratorily, we examined if individuals with higher chronic exhaustion would show different patterns of phasic vmHRV when perceiving acute stress. We analysed data on momentary (emotional, cognitive, physical) exhaustion, perceived acute stress, ambulatory ECG data and adjusted for relevant covariates (e.g., age, gender, and momentary movement acceleration) using multi‐level analyses. After adjusting for preregistered covariates, phasic vmHRV showed a positive association with momentary emotional and cognitive exhaustion, but not with momentary physical exhaustion. Our exploratory analyses revealed that individuals with higher levels of chronic exhaustion did not show the expected negative association between situationally perceived acute stress and phasic vmHRV, whereas those with lover levels did. These findings indicate that momentary exhaustion is associated with increased phasic vmHRV in daily life. Combined with our exploratory results that chronic exhaustion modulates vagal withdrawal under perceived acute stress, this study offers important directions for future research into the link between stress‐related exhaustion and autonomic changes.Study Registration: The study and analysis plan were preregistered at OSF (DOI: 10.17605/OSF.IO/T2C4X).

Safe and agile experimentation spaces are essential for developing AI-enabled medical devices and digital health innovations. “Living Labs” offer a collaborative environment for testing technologies in near-real-world settings, capturing user perspectives and outcomes, often missed in traditional testing. However, the flexibility of Living Labs often clashes with the European Union’s rigid regulatory frameworks for medical devices that were not designed for digital technologies. We examine this intersection, showing how flexibility in evaluation and patient interaction can coexist with safety guardrails. Our approach integrates technology readiness, preemptive planning, and iterative modifications to bridge innovation and regulation, fostering adaptive health care technology development.

Objective: Feeling comfortable and safe has been discussed to foster health and well-being. However, the pathways to better health are complex, involving both behavioral and physiological routes. Methods: In this study, we examined the role of safety perception for cardiac health by (1) examining associations with baseline heart rate variability (HRV; Study 1) and (2) evaluating a novel measure of autonomic cardiac flexibility in daily life, namely increases in HRV independent of metabolic demands (ImdHRVi; Study 2). Results: Study 1 (N = 76) found evidence for a positive association between vagally mediated HRV and the Neuroception of Psychological Safety scale (Morton L, Cogan N, Kolacz J, et al. “A new measure of feeling safe: developing psychometric properties of the Neuroception of Psychological Safety Scale (NPSS)”: Correction. Psychol Trauma. 2022; https://doi.org/10.1037/tra0001374), thus suggesting a link between safety and cardiac vagal regulation. In Study 2, a sample of N = 245 adult volunteers participated in a four-day-ambulatory assessment measuring HRV and bodily movement. A regression was calculated between HRV and bodily movement for 12 h of the first recording day, which was then used to calculate minute-by-minute ImdHRVi (beyond those predicted by bodily movement) in the following days. It turned out that safety perception predicted more episodes of ImdHRVi in everyday life, even after controlling for several confounds. Conclusions: Findings suggest that feeling safe and everyday life cardiac autonomic regulation are interrelated, thus possibly contributing to adaptive adjustment and health.

Objective: This study investigates the mediating role of rumination in the relationship between work-related stress and sleep quality. Sleep quality is a critical component of mental health and wellbeing. Work-related stress is a potential risk factor for poor sleep quality, and this research focuses on how the constructs of rumination—affective rumination, detachment, and problem-solving rumination—may mediate this association. Methods: Data from the Dresden Burnout Study were analyzed cross-sectionally and longitudinally over three timepoints. The sample included 360 full- and parttime employees. Work-related stress was assessed using the effort–reward Imbalance Questionnaire, sleep quality using the Pittsburgh Sleep Quality Index, and rumination using the Work-Related Rumination Questionnaire. Results: The study found that work-related stress was positively correlated with reduced sleep quality, both cross-sectionally and longitudinally. In addition, work-related stress predicted increased levels of affective rumination and detachment but not of problem-solving rumination. Affective rumination and detachment partially mediated the relationship between work-related stress and sleep quality, suggesting that individuals who experience higher levels of work-related stress are more likely to engage in ruminative thoughts about their work experiences, which, in turn, leads to reduced sleep quality. Conclusion: These findings highlight the importance of considering cognitive processes such as rumination when addressing sleep disturbance related to work stress. Future research should include objective measures of sleep quality and consider shorter intervals to capture immediate changes in stress and sleep patterns. By addressing these factors, healthcare professionals can better support those suffering from sleep disturbance due to work-related stress. Ultimately, this may have a positive impact on their overall wellbeing and productivity at work.