Nature, Published online: 26 March 2026; doi:10.1038/d41586-026-00867-4

Data gathered during a 1,200-kilometre polar trek show where PFAS levels are the highest.

Background: Early detection of Alzheimer disease (AD) is essential for timely intervention; yet, diagnostic performance varies widely across modalities and datasets. Recent multimodal artificial intelligence (AI) models have made significant progress, but the evidence base remains fragmented due to heterogeneous datasets, modeling frameworks, and reporting quality. Objective: This systematic review aimed to analyze studies on multimodal AI models for AD diagnosis, prognosis, and risk prediction over 5 years. We evaluated dataset characteristics, modality combinations, modeling strategies, performance metrics, and methodological limitations. We further discuss real-world implications and translational pathways. Methods: Following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 guidelines, we systematically searched PubMed, IEEE Xplore, Scopus, ACM Digital Library, Cochrane, and arXiv, with the final datasets last searched on November 15, 2025. Studies applying multimodal machine learning or deep learning to AD, mild cognitive impairment, and dementia outcomes were included, whereas studies using a single modality or lacking sufficient methodological detail were excluded. QUADAS-2 (Revised Quality Assessment of Diagnostic Accuracy Studies tool) assessed risk of bias. Extracted performance results were synthesized across 4 major multimodal dataset families. Results: A total of 66 studies met the inclusion criteria. Across datasets, multimodal models consistently outperformed single-modal baselines. Alzheimer’s Disease Neuroimaging Initiative–based diagnosis achieved an average accuracy of 92.5% (SD 3.8%), while mild cognitive impairment–conversion models achieved an average area under the curve (AUC) of 0.922 (SD 0.045), and several fusion architectures reported AUCs above 0.95. In contrast, UK Biobank risk-prediction studies reported an average AUC of 0.84 (SD 0.056), and this reflects performance in large, population-based datasets. DementiaBank speech-language studies achieved an average AUC of 0.813 (SD 0.042), and cross-lingual AD detection achieved an accuracy of 77% (SD 6.5%). Self-collected multimodal datasets demonstrated average accuracies around 96% (SD 2.4%), but their generalizability is limited due to small sample sizes and single-center designs. Conclusions: This systematic review demonstrates that multimodal AI models consistently outperform single-modal models for AD diagnosis, prognosis, and risk prediction by integrating complementary biological, clinical, and behavioral information. Unlike prior reviews, this review provides a unified synthesis across heterogeneous clinical, imaging, genetic, and linguistic datasets, enabling cross-domain comparison of modeling strategies and performance. However, the generalizability of reported performance was limited due to substantial heterogeneity in dataset composition, outcome definitions, and validation, and prevalent risks of bias. By evaluating these factors, this review clarifies where current evidence is robust and where caution is warranted. The findings highlight the need for standardized multimodal benchmarks, transparent evaluation protocols, and clinically grounded model design to enable reliable real-world deployment. Overall, this work advances the field by framing multimodal AI not only as a performance-driven tool but also as a translational framework for equitable, interpretable, and scalable AD diagnosis. Trial Registration: PROSPERO CRD420251241895;

Background: Increasing suicidal thoughts and behaviors (STB) among adolescents raise social concerns and have a well-recognized association with sexual abuse (SA). However, research regarding the mechanisms explaining the association between SA and STB remains limited. Objective: This study aims to examine the chained mediating effects of negative life events (NLE) and internet addiction (IA) between SA and STB among adolescents in China. Methods: This cross-sectional study used data from the Science Database of the People Mental Health survey conducted between March 2013 and December 2022 by the National Population Health Data Center of the National Research Institute for Family Planning. Through stratified sampling, 20,893 adolescents were recruited from 16 Chinese provinces. After excluding samples with missing relevant variables, 10,664 (55.89%; aged 16-17.9 y; n=5826, 54.63% women) adolescents were included in the final analysis. STB was the outcome variable, with NLE and IA as mediators, all assessed via a questionnaire that was uniformly administered by trained investigators in school settings. The Pearson χ test was used to analyze the association between SA and STB. Using a combination of multiple linear regression and bootstrap testing, the study constructed a chain mediation model to explore how SA influences STB in adolescents through NLE and IA. Results: The scores for SA, NLE, IA, and STB were 1.330 (SD 1.714), 51.960 (SD 23.822), 34.88 (SD 13.852), and 0.690 (SD 1.396), respectively. Multiple linear regression analysis indicated SA was associated with NLE (β=2.382, 95% CI 2.112‐2.653;

Background: Preliminary research has suggested that internet use data could offer digital signals of early disease and has the potential to facilitate early detection and improve patient outcomes. However, there are significant challenges in linking individual-level internet use data with health outcomes. One key aspect is that the public might not be willing to share data for research or that selective data sharing might create bias in datasets and increase inequalities. Objective: Our study aimed to investigate the willingness of the public to share their internet use data for medical research and to identify key criteria that affect willingness to share. Methods: We conducted a web-based, cross-sectional online survey with 2390 UK adults with and without a history of cancer, heart disease, and depression using quota sampling. Participants were randomly assigned to explore willingness to share different types of internet use data for 1 of 3 health conditions (cancer, heart disease, and depression) and for provision of a pictorial example of internet use data. Logistic regression analysis (α=.05) for each condition was used to determine key factors of willingness to share, including sociodemographics and attitudes toward sharing. Open-ended comments regarding facilitators of sharing and concerns were analyzed thematically. Results: Willingness to share internet use data was high across conditions (74%‐77%, 95% CI 70.5%‐80.3%), especially for health app data (73%‐76%, 95% CI 69.8%‐79.1%). The pictorial example of browsing history did not affect willingness to share. For all conditions, factors consistently associated with willingness to share were perceived benefits (odds ratios [ORs] 5.692‐8.850; all

Background: The targeted application and design of digital health interventions (DHIs) require an understanding of usage determinants. Usage includes uptake (initial use) and frequency (extent of use), but it is unclear whether both components are driven by the same determinants. Objective: This study aimed to examine the determinants of uptake and frequency of use and assess whether they differ. Methods: The investigated DHI was a web portal provided in an intervention for improving disease-related self-management. This study is a secondary analysis of intervention group data from a parallel-group randomized controlled trial. Eligibility criteria were being an adult and being diagnosed with type 2 diabetes and/or coronary heart disease. Sociodemographic, psychological, and health-related variables were examined as determinants. Determinants were analyzed using simple and multiple regression models. Uptake was analyzed using logistic regression, and frequency was analyzed using negative binomial regression with robust SEs. Frequency was analyzed for those who used the DHI at least once. Except for sociodemographic variables, all other variables were standardized to a range from 0 to 1. For simple regression, inflation of the α error due to multiple testing was controlled via the approach of Benjamini and Hochberg, and for multiple regression, it was controlled via the significance of the complete multiple regression model. Results: Of 462 intervention group members, 199 (43.1%) used the web portal at least once. After controlling for inflation of the α error, simple regression for uptake yielded significant effects for higher education (B=0.56, 95% CI 0.18-0.95; =.004), openness (B=1.08, 95% CI 0.33-1.83; =.005), intention regarding physical activity (B=2.28, 95% CI 1.30-3.26;

Background: Embedding models are critical components of Retrieval Augmented Generation (RAG) systems for retrieving and searching unstructured medical data. However, existing models are predominantly trained on publicly available English datasets, limiting their effectiveness in non-English health care settings. More importantly, these models lack training on real-world clinical documents, leading to inaccurate context retrieval when integrated into RAG systems for health care applications. This gap is particularly pronounced in specialized medical documentation containing domain-specific terminology, abbreviations, and nuanced clinical language. Objective: This retrospective study aimed to develop and validate embedding models specifically trained on real-world clinical documents from multiple medical specialties to improve medical information retrieval (IR) and RAG system performance in both German and English language contexts. Methods: We fine-tuned embedding models, so-called sentence transformers, using the multilingual-e5-large architecture as a foundation. Training data consisted of approximately 11 million question-answer pairs synthetically generated from 400,000 diverse clinical documents from a large German tertiary hospital, spanning 163,840 patients and 282,728 clinical cases between 2018 and 2023. The large language model generated medically relevant questions and corresponding answers for each document. The dataset was additionally pseudonymized and translated into English to aim for broader applicability. Models were evaluated in 2 distinct scenarios: IR using questions with multiple relevant passages, and RAG system performance in both cross-patient and patient-centered contexts. Results: In the IR evaluation, the fine-tuned miracle model achieved a mAP@100 of 0.27, outperforming the multilingual-e5-large baseline (0.14) and state-of-the-art models such as bge-m3 (0.11). In the RAG evaluation, the model demonstrated robust performance comparable with the baseline in the constrained patient-centered scenario (BERTScore F1 0.781 vs 0.778) and showed moderate improvements in the unconstrained cross-patient setting (BLEURT 0.56 vs 0.53). Notably, the model trained on pseudonymized data achieved comparable retrieval performance (mAP@100 0.25) and the highest scores for patient-centered contextual precision (0.93). Performance gains were robust in the German dataset, while the translated English model demonstrated promising results as a proof of concept for cross-lingual transfer. Conclusions: By leveraging a comprehensive real-world dataset spanning multiple medical specialties and using large language models for synthetic question generation, we successfully created and validated domain-specific embedding models. These models can improve medical IR in large-scale search spaces and perform competitively in constrained RAG applications. By publishing the models trained on pseudonymized data, other health care institutions can integrate or adapt these embedding models to their needs. This work establishes a reproducible framework for developing domain-specific clinical embedding models, with the potential to improve data retrieval in medical settings.

Background: Stroke is a leading cause of long-term upper limb disability, severely impacting patients’ independence and quality of life. Robot-assisted therapy (RAT) has emerged as a promising, high-intensity rehabilitation alternative. However, conclusions from existing systematic reviews on its efficacy are inconsistent and often lack a holistic framework, limiting their use for guiding personalized clinical decisions. Objective: This study aims to systematically synthesize recent evidence on RAT for upper limb rehabilitation after stroke. Guided by the International Classification of Functioning, Disability and Health framework, it moves beyond singular outcomes to provide a multidimensional evaluation across body function, activity, and participation levels. The review aims to provide stratified guidance for clinical decision-making based on patient- and intervention-specific characteristics, thereby supporting evidence-based practice and informing future research. Methods: This study included systematic reviews and meta-analyses published from January 1, 2019, to December 26, 2025, comparing RAT with conventional therapy for upper limb rehabilitation after stroke. Overall, 6 databases, including PubMed, Web of Science, and Embase, were searched. Two reviewers (XZ and LZ) independently performed study selection, data extraction, and quality assessment using the AMSTAR 2 tool. The synthesis integrated outcome measures and subgroup analyses derived from the included studies. Results: This umbrella review included 21 meta-analyses encompassing 535 randomized controlled trials and 27,598 patients across acute, subacute, and chronic stroke stages. According to AMSTAR 2, 17 reviews were high quality, 3 moderate, and 1 critically low. The synthesis demonstrated that RAT was superior in improving upper limb motor function, but no statistically significant advantages were observed in activities of daily living compared to conventional therapy. Subgroup analyses revealed that treatment effects were influenced by stroke stage, upper limb motor impairment level, and robot type. Conclusions: RAT is an effective intervention for improving upper limb motor function after stroke. However, its benefits are primarily observed at the level of body function, with limited evidence for long-term maintenance. The current evidence is constrained by significant outcome heterogeneity and methodological limitations inherent to umbrella reviews. Future research should validate these findings in broader clinical practice, focus on translating functional gains into sustained improvements in daily activities and participation, and include cost-effectiveness evaluations. Trial Registration: PROSPERO CRD42024497183; https://www.crd.york.ac.uk/PROSPERO/view/CRD42024497183

Scientists have discovered that losing a key protein in small cell lung cancer triggers inflammation that actually helps tumors grow and spread. Even more surprising, it pushes cancer cells into a more aggressive, neuron-like state linked to relapse.

In a remarkable deep-sea breakthrough, researchers have discovered 24 new species of amphipods in the Pacific’s Clarion-Clipperton Zone—including a rare, entirely new superfamily. The findings reveal previously unknown branches of life and push the boundaries of how deep these creatures are known to live.

Flower nectar often contains small amounts of alcohol, meaning pollinators like hummingbirds are drinking it all day long. Despite consuming human-equivalent amounts, they show no signs of intoxication—suggesting a surprising evolutionary tolerance.

A major discovery reveals that metformin works not just in the body, but in the brain. By switching off a key protein and activating specific neurons, the drug lowers blood sugar through a previously hidden pathway, opening new doors for diabetes treatment.

A star you can see with the naked eye has kept astronomers guessing for decades with its unusually powerful X-rays. Now, thanks to highly precise observations from Japan’s XRISM space telescope, scientists have finally uncovered the source: a hidden white dwarf companion pulling in material and generating extreme heat. This discovery not only solves a 50-year-old mystery surrounding Gamma Cassiopeiae, but also confirms the existence of a long-predicted type of binary star system.

Astronomers have narrowed down the cosmic search for life, identifying fewer than 50 rocky planets among thousands of known exoplanets that may have the right conditions to support life. Using new data from ESA’s Gaia mission and NASA archives, researchers pinpointed worlds in the “habitable zone,” where temperatures could allow liquid water to exist. Some of the most intriguing targets include nearby systems like TRAPPIST-1 and Proxima Centauri, offering tantalizing possibilities just dozens of light-years away.

Nearly half of Americans don’t know that processed meat increases colorectal cancer risk, according to a new poll. But once they learn the connection, most support warning labels—suggesting people want clearer information. Experts warn that awareness is lagging even among healthcare providers. The good news: diets rich in plant foods and fiber, along with healthy habits, can dramatically lower risk.

Balance problems in aging and Parkinson’s may come from the body working too hard, not too little. Scientists found that the brain and muscles become overactive during even minor disturbances, yet this actually weakens balance recovery. At the same time, muscles can stiffen against each other, making movement less stable. This unexpected pattern could help predict who is more likely to fall.

Researchers have visualized atoms in motion just before a radiation-driven decay process occurs, revealing a surprisingly dynamic scene. Instead of remaining fixed, the atoms roam and rearrange, directly influencing how and when the decay unfolds. This “atomic movie” shows that structure and motion play a central role in radiation damage mechanisms. The findings could improve our understanding of how harmful radiation affects biological matter.

Honey bees don’t just perform their famous waggle dance to share directions, they actually adjust how well they dance depending on who’s watching. Researchers found that when fewer bees pay attention, the dancer becomes less precise as it moves around trying to attract an audience. This means the dance is not simply a fixed message about food location, but a flexible performance shaped by social feedback.

New fathers appear to have fewer mental health diagnoses during pregnancy and the early months after birth. But that early stability does not last. About a year later, depression and stress-related disorders increase significantly, surprising researchers. The findings suggest that the emotional toll of fatherhood builds over time rather than hitting immediately.

A new neural implant is so small it can rest on a grain of salt, yet it can track and wirelessly transmit brain activity for over a year. It’s powered by laser light that safely passes through tissue and communicates using tiny infrared signals. This ultra-miniature device could transform how scientists study the brain without invasive wiring.