Predictive artificial intelligence (AI) offers an opportunity to improve clinical practice and patient outcomes but risks perpetuating biases if fairness is inadequately addressed. However, the definition of fairness remains unclear. We conducted a scoping review to identify and critically appraise fairness metrics in clinical predictive AI models. We defined a fairness metric as a metric quantifying whether a model discriminates (societally) against individuals or groups defined by sensitive attributes. We searched five databases for literature published during 2014-24, screened 820 records, included 42 studies, and extracted 63 fairness metrics. The search was limited to studies published in English. These metrics, which were classified by performance dependency, model output level, and base performance metric, revealed a fragmented landscape in the field of clinical predictive AI, with inadequate clinical validation and over-reliance on threshold-dependent metrics. 19 metrics, including only one metric for clinical use, were explicitly developed for health care. Our findings highlight conceptual challenges in defining and quantifying fairness and identify gaps in uncertainty quantification, intersectionality, and real-world applicability. Therefore, future works on clinical predictive AI models should prioritise clinically meaningful metrics.

The COVID-19 pandemic disproportionately affected vulnerable populations, including individuals with rare diseases (RDs) and, in the general population, those from ethnic minority backgrounds. However, the intersectional risk, and how these vulnerabilities combine, is poorly understood. A comprehensive baseline map of RD prevalence by granular ethnicity is required to investigate the pandemic’s true impact on these complex patient groups. This study had two aims: (1) to generate the national scale prevalence estimates for 406 rare diseases stratified by 19 ethnicity groupings; and (2) to describe the burden of recorded COVID-19 infection across these distinct populations. We conducted a cross-sectional study within the National Health Service (NHS) England Secure Data Environment, accessed via the BHF Data Science Centre’s CVD-COVID-UK/COVID-IMPACT Consortium, linking primary care, hospital, and mortality records for individuals alive on 31 July 2023. We calculated age- and sex-standardised prevalence for 406 RD and calculated COVID-19 infection rates for the overall cohort, for each RD, and by the six-group Office for National Statistics (ONS) and the 19-group NHS ethnicity classifications. We observed a higher burden of COVID-19 infection in the RD cohort (1.7% of the total population) compared to the non-RD population (32.8% vs. 29.0%). The burden varied significantly by ethnicity (e.g., 27.7% in the Pakistani 23.8% in the Black African groups, vs. 33.7% in the White British group) and by specific RD (24.6%–41.2% among the top 25 diseases). This highlights the importance of this ethnicity-stratified prevalence map, which revealed significant underlying ethnic disparities in the RDs themselves. For example, the Pakistani population had markedly elevated odds for certain metabolic disorders (e.g., medium-chain acyl-CoA dehydrogenase deficiency, OR=4.46[4.07–4.90]), and Black Caribbean individuals showed increased odds of autoimmune conditions (e.g., discoid lupus erythematosus, OR=3.34[3.06–3.64]). The burden of the COVID-19 pandemic disproportionately affected individuals with rare diseases. However, the patterning of this risk by ethnicity is complex and runs contrary to general population trends, likely reflecting the deep-seated ethnic disparities in the prevalence of specific RDs. Our foundational map of 406 rare diseases by granular ethnicity is essential for understanding these factors and identifying which specific patient-ethnic subgroups face the greatest intersectional risk.

Background: Climate-exacerbated flooding triggers global public health crises, costing lives, livelihoods, and jeopardising fragile healthcare systems. Despite well-established disaster management frameworks, the timely communication of rapid, locally actionable information remains lacking—leaving vulnerable populations at risk. We developed and operationally validated an open-access monitoring system across six South Asian nations, reporting the public health impact of the 2025 floods and executing live daily nowcasting during the July 2026 monsoon emergency to better understand who is in harms’ way, democratise anticipatory disaster intelligence, and support healthcare resilience. Methods: We engineered a multi-stakeholder co-designed ensemble pipeline for flood-health intelligence by fusing observations from six synthetic aperture radar and optical satellites with hydrological, demographic, agricultural, transportation, and infrastructure datasets. We estimated cumulative exposure for children and adults, alongside functional disruption to healthcare facilities, built-up areas, road networks, and cropland across remote, rural, and urban areas of Pakistan, India, Bangladesh, Bhutan, Nepal, and Sri Lanka. The 2025 analysis was benchmarked against publicly-available official surveillance reports and independent field-validation. The 2026 deployment is ongoing live, externally field-validated against 8 ground-sensor stations along the Indus River system. Findings: Our analysis revealed a critical gap in current surveillance. In 2025, continuous monitoring identified 257.67 million exposed individuals (29.1% in rural areas) including 89.3 million children under 18 (34.6%). A five-fold increase (72.19 million additional people) was found over the 18.97 million captured in official episode-level reports for Pakistan, Bangladesh, and Sri Lanka. Over 460, 000 km² of agricultural land was exposed (18.5 % to >2m deep water), having significant implications for the region’s predominantly agrarian economy.Across publicly unmonitored regions and flood episodes, an estimated 237 million flood-exposed people were left in a data vacuum, completely missing from publicly-available mapping. Functionally, 22,818 low-lying healthcare facilities were inundated (3,041 functionally disrupted), severing last-mile care for 46.79 million people, including 28.54 million exposed to deep-water flooding (>2m). In each country, access to at least 5% healthcare facilities was disrupted. Our rapid monitoring of the July–August 2026 crisis has already reported 11,905 km² of inundation, 92,140 km of submerged roads, 5.19 million exposed people including 1.6 million children (30.8%), and disruption to 1,900 basic health units, community clinics, pharmacies, and hospitals. Interpretation: This first region-wide, multi-year study demonstrates that current global disaster response relies on severely incomplete or inaccessible data, effectively disenfranchising the world's most underserved and climate-vulnerable populations from healthcare when most needed. Rapid, reliable, and scalable precision intelligence is not a technical luxury but a necessary and achievable instrument for health equity and climate justice. To counter 21st-century global health threats, we recommend: (1) prioritising anticipatory planning for children and vulnerable communities via national action plans and community toolkits, (2) democratising disaster data through digital public infrastructure to inform e

Malaria, childhood acute respiratory infection, and child undernutrition together account for over two million deaths annually in children under five, with the burden concentrated in low and middle-income countries where climate variability modulates transmission, exposure, and nutritional outcomes. Routine health surveillance in these settings remains sparse and reactive. Satellite-derived representations of the Earth's surface offer a scalable, low-cost complement to traditional covariates, yet their utility as predictors of population health outcomes is poorly characterised. We summarise findings from three studies evaluating AlphaEarth Foundations 64-dimensional satellite embeddings as predictors of population health outcomes, focusing on vulnerable populations. The studies span infectious disease (malaria, respiratory infection) and stunting. In each study, embeddings provide predictive value at sufficient spatial granularity: (i) malaria prediction across Nigeria shows consistent per-region R^2 gains; (ii) childhood acute respiratory infection prediction across 11 DHS countries increases pooled R^2 from 0.157 to 0.206 across three tree-based estimators; (iii) stunting prediction across 35 countries is neutral at country level due to collinearity with fixed effects. The stunting case is currently limited by lack of DHS cluster-level coordinates, which is the next key experiment.

We previously identified genetic correlation between pairs of musculoskeletal (MSK) and respiratory conditions. Strategies to prevent or delay their onset remain underexplored in the context of multimorbidity. This study investigated whether MSK–respiratory disease pairs show evidence of potential causal relationships, identified modifiable risk factors, and quantified intervention windows to prevent progression to multimorbidity. We examined combinations of one respiratory condition (asthma, COPD) and one MSK condition [rheumatoid arthritis (RA), osteoarthritis (OA), polymyalgia rheumatica (PMR), psoriasis]. Two-sample Mendelian randomisation (MR) evaluated potential causal relationships in both directions. Linked electronic health records from CPRD (N = 11,042,985; age ≥ 40 years) were used to assess longitudinal disease trajectories, prognostic consequences, and mediation by potentially modifiable or treatable factors. We found evidence for bidirectional relationships between COPD and RA/OA (ORs 1.10–1.19) and between asthma and RA/OA (ORs 1.03–1.14). COPD genetic liability also increased PMR risk (OR 1.14, 95

Objectives We aimed to assess the risk of incident autoimmune and inflammatory conditions during the post-acute period of COVID-19.Design Descriptive network cohort study.Setting Electronic health records from the UK and Dutch primary care, Norwegian linked health registry, hospital records of specialist centres in Spain, France and Korea and healthcare claims from Estonia and the USA.Participants We followed individuals between September 2020 and the latest available data from day 91 after a SARS-CoV-2 negative test (comparator) or a COVID-19 record (exposed patients, ie assessing patients during the post-acute phase). We further established a reinfection cohort (any further COVID-19 record among the exposed patients). We followed patients until an outcome, end of study period, death, day 365 or an infection (comparator only) or reinfection (exposed patients only).Main outcome measures We assessed postural orthostatic tachycardia syndrome (POTS) diagnoses/symptoms, myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) diagnoses/symptoms, multi-inflammatory syndrome (MIS) and several autoimmune diseases (rheumatoid arthritis (RA), juvenile idiopathic arthritis (JIA), systemic lupus erythematosus (SLE), inflammatory bowel disease (IBD) and type 1 diabetes mellitus (T1DM)).Meta-analysed crude incidence rate ratios (IRRs) of outcomes after COVID-19 versus negative testing and after reinfection versus a previous COVID-19 record yield the ratios of respective absolute risks of each assessed outcome. We performed subgroup analyses by age, sex and predominant variant periods.Results We included 2 521 812 individuals with a first COVID-19 record, 4 233 145 with a first negative test and 135 551 with a reinfection. Age and sex were largely comparable between exposure groups with a shorter follow-up for the reinfection cohorts. After COVID-19 compared with test-negative patients and equally after reinfection compared with previous COVID-19 patients, we did not observe increased rates for all outcomes and all subgroup analyses. Counts of MIS and JIA were too small for meta-analyses.Conclusions In our descriptive meta-analyses of crude IRRs among databases from various countries and settings, we did not observe increased rates of incident POTS, ME/CFS, RA, IBD, SLE and T1DM in COVID-19 versus test-negative or reinfection versus COVID-19 during the first 9 months of the post-acute phase of COVID-19 or reinfection (>90 days postinfection until month 12). Since causal interpretation cannot be made from this study, further causal research is warranted.

Pakistan experienced an unusually severe flood season between June and December 2025, with cascading impacts on population, infrastructure, and agriculture. Existing operational flood products (e.g., UNOSAT) provide valuable episode-level snapshots but rarely deliver spatially and temporally continuous inundation maps at near-real-time latency within the country. We present a multi-sensor, ensemble-based remote-sensing framework for continuous flood nowcasting in Pakistan that integrates Sentinel-1 SAR, Harmonized Landsat-Sentinel (HLS L30 and S30), MODIS, and VIIRS observations on a harmonized grid in Google Earth Engine. The framework employs a tiered nowcasting ensemble that prioritizes higher-resolution sensors (Sentinel-1 and HLS) and falls back to MODIS and VIIRS when necessary, preserving daily continuity of flood extent at each sensor's native resolution. Applied to the 2025 monsoon period, the system generates near-real-time, spatially consistent inundation maps across Pakistan. As a nowcasting case study, we track the super-flood of 26 August-7 September 2025 day by day, demonstrating the framework's ability to capture the evolving flood footprint in near real time and extend beyond the temporal limits of episodic mapping products. Validation against GloFAS discharge anomalies and precipitation datasets (CHIRPS v3.0, MSWEP) shows strong agreement with observed hydrometeorological conditions. By integrating nowcast outputs with exposure layers (WorldPop, ESA WorldCover, Giga-HOTOSM), the framework enables rapid estimation of affected populations, cropland, and critical infrastructure, supporting timely disaster response and resilience planning in South Asia.

Abstract Background/Aims Detection of people with psoriasis at increased risk for developing psoriatic arthritis (PsA) is essential to aid early diagnosis and treatment. However, there are limitations of the existing clinical prediction models: none have been externally validated and most are based on studies with modest sample sizes. We aimed to develop and externally validate a multivariable prediction model to predict the risk of developing PsA in adults newly diagnosed with psoriasis in primary care. Methods A retrospective observational cohort study using primary care electronic health record data was performed from 2000-2024. The exposure cohort was adults with an incident diagnosis of psoriasis. Outcome was a diagnosis of PsA. Time at risk for the model was five years. The Clinical Practice Research Datalink (CPRD) GOLD was used for developing and internally validating the model. The Health Improvement Network (THIN) databases from UK, France, Spain, Italy, Romania, and Belgium were used for external validation. Extreme Gradient Boosting was used to build the model. Results Table 1 summarises the results from model development and external validation. The discrimination in the train and test datasets in CPRD GOLD showed that the model performed well above the threshold. There was good calibration for the train dataset in CPRD GOLD. However, there was poor calibration in the test dataset: the model underpredicted the outcome. There was also poor calibration in the THIN databases used for external validation. The predictors with the highest discriminatory power were age and visit occurrence to the GP in the one year prior to their psoriasis diagnosis. Conclusion This is the first clinical prediction model for predicting PsA developed using clinical markers available in primary care that has been externally validated. The model performs well in the dataset it was developed in (similar to other published models). However, it did not perform well when externally validated. It could be that further granular data not available in primary care data and other markers (e.g., genetic) are needed to predict PsA. The Health initiatives in Psoriasis and PsOriatic arthritis ConsoRTium European States (HIPPOCRATES) consortium is aiming to address this through the HIPPOCRATES Prospective Observational Study (HPOS). Disclosure A. Vivekanantham: None. M. Pineda Moncusi: None. E. Burn: None. S. Khalid: None. D. Prieto Alhambra: Consultancies; DPA has provided consultancy services, with fees paid to the University, for UCB Biopharma. Honoraria; DPA is a member of the Board of the EHDEN Foundation and has received grants from the European Medicines Agency and the Innovative Medicines Initiative. Grants/research support; DPA’s department has received grants from Amgen, Chiesi-Taylor, Gilead, Lilly, Janssen, Novartis, and UCB Biopharma. Other; Janssen has funded or supported training programmes organised by DPA’s department. L.C. Coates: Consultancies; LCC worked as a paid consultant for AbbVie, Amgen, Boehringer Ingelheim, Bristol Myers Squibb, Celgene, Eli Lilly, Gilead, Galapagos, Janssen, Novartis, Pfizer and UCB. Grants/research support; LCC received grants/research support from AbbVie, Amgen, Celgene, Eli Lilly, Novartis and Pfizer. Other; LCC has been paid as a speaker for AbbVie, Amgen, Biogen, Celgene, Eli Lilly, Galapagos, Gilead, Janssen, Medac, Novartis, Pfizer and UCB.

Brick kilns are a major source of air pollution and forced labor in South Asia, yet large-scale monitoring remains limited by sparse and outdated ground data. We study brick kiln detection at scale using high-resolution satellite imagery and curate a multi city zoom-20 (0.149 meters per pixel) resolution dataset comprising over 1.3 million image tiles across five regions in South and Central Asia. We propose ClimateGraph, a region-adaptive graph-based model that captures spatial and directional structure in kiln layouts, and evaluate it against established graph learning baselines. In parallel, we assess a remote sensing based detection pipeline and benchmark it against recent foundation models for satellite imagery. Our results highlight complementary strengths across graph, foundation, and remote sensing approaches, providing practical guidance for scalable brick kiln monitoring from satellite imagery.

An increased risk of COVID-19 mortality risk among certain ethnic groups is well-reported, however data on ethnic disparities in COVID-19-related cardiovascular disease (CVD) are lacking. We estimated age-standardised incidence rates and adjusted hazard ratios for 28-day mortality and 30-day CVD by sex for individual ethnicity groups from England and Wales, using linked health and administrative data. We studied 6-level census-based ethnicity group classification, 10-level classification (only for Wales), and 19-level classification as well as any ethnicity sub-groups comprising >1000 individuals each (only for England). COVID-19 28-day mortality and 30-day CVD risk was increased in most non-White ethnic groups in England, and Asian population in Wales, between 23rd January 2020 and 1st April 2022. English data show mortality decreased during the Omicron variant’s dominance, whilst CVD risk [95% confidence interval] remained elevated for certain ethnic groups when compared to White populations (January-April 2022): by 120% [28-280%] in White and Asian men and 58% [32-90%] in Pakistan men, as compared to White British men; and by 75% [13-172%] in Bangladeshi women, 55% [19-102%] in Caribbean women, and 82% [31-153%] in Any Other Ethnic Group women, as compared to White British women. Ethnically diverse populations in the UK remained disproportionately affected by CVD throughout and beyond the COVID-19 pandemic.

Using the UK Clinical Practice Research Datalink, our cohort study matched 237,297 individuals with hearing loss (HL) to 829,431 without HL. The study found an 8–10

Heat waves are a major public health challenge, yet the link between heat-related illness (HRI) and regional climate and geography is underexplored. We examined HRI and excess sun exposure incidence rates (IR) [95% confidence interval (CI) per 100,000 person-years], and their correlation with regional maximum temperatures across 9 US climatic zones 33,603,572 individuals were followed from 2017 to 2022. We observed 10,652 individuals with HRI diagnosis (median age: 49 years, 62.3% male). Seasonal peaks occurred during summer: highest overall IR (130.97 [119.93–142.75]) was recorded in July 2019, highest regional IR was reported in the South (186.04 [117.93–279.15]) during 2020. Strongest correlations between monthly maximum temperature and incidence of HRI were observed in the West (Pearson Correlation Coefficient (cor) = 0.854) and Southwest (cor = 0.832). In contrast, we observed 131,204 individuals with excess sun exposure (predominantly older adults [median age: 67 years], 52.3% female, 30% with history of cancer). Overall IR for sun exposure peaked in March 2021 (664.31 [644.84–684.21]) and lacked a consistent seasonal pattern. Sun exposure exhibited weaker correlations with regional temperatures, even in high-temperature regions like the West (cor = 0.305). These data indicate regional variations in HRI. With distinct at-risk groups for HRI and sun exposure, targeted regional interventions may be beneficial, such as heat safety protocols to reduce HRI risk and sun protection campaigns for older adults to mitigate sun exposure risk.

The relationships between air pollution, genetic susceptibility, and COVID-19-related outcomes, as well as the potential interplays between air pollution and genetic susceptibility, remain largely unexplored. The Cox proportional hazards model was used to assess associations between long-term exposure to air pollutants and the risk of COVID-19 outcomes (infection, hospitalization, and death) in a COVID-19-naive cohort (n = 458,396). Additionally, associations between air pollutants and the risk of COVID-19 severity (hospitalization and death) were evaluated in a COVID-19 infection cohort (n = 110,216). Furthermore, this study investigated the role of host genetic susceptibility in the relationships between exposure to air pollutants and the development of COVID-19-related outcomes. Long-term exposure to air pollutants was significantly associated with an increased risk of COVID-19-related outcomes in the COVID-19 naive cohort. Similarly, in COVID-19 infection cohort, hazard ratios (HRs) for COVID-19 hospital admission were 1.23 (1.19, 1.27) for PM2.5 and 1.22 (1.17, 1.26) for PM10, whereas HRs for COVID-19 death were 1.28 (1.18, 1.39) for PM2.5 and 1.25 (1.16, 1.36) for PM10. Notably, significant interactions were found between PM2.5/PM10 and genetic susceptibility in COVID-19 death. In COVID-19 infection cohort, participants with both high genetic risk and high air pollutants exposure had 1.86- to 1.97-fold and 1.91- to 2.14-fold higher risk of COVID-19 hospitalization and death compared to those with both low genetic risk and low air pollutants exposure. Exposure to air pollution is significantly associated with an increased burden of severe COVID-19, and air pollution-gene interactions may play a crucial role in the development of COVID-19-related outcomes.

Transfer learning enables the reuse of models trained on large datasets, reducing data collection, computation time, and costs. While widely used in computer vision, its application to models based on electronic health records (EHRs) remains limited. This study evaluates whether fine-tuning an EHR-based model from one country to another outperforms training a model from scratch. EHR from the SIDIAP (Spain) and CPRD (UK) databases were used, defining a cohort in each country of individuals aged 65+ followed between 2010 and 2019. A prediction model was trained and validated internally for each country to predict 1-year mortality, then externally validated and fine-tuned with the other country's population (recalibrated model). The models were based on ARIADNEhr, a previously validated architecture. Performance metrics, decision curve analysis, and attention maps were compared. Participants included 1,456,052 from SIDIAP and 1,507,736 from CPRD, with similar demographics. Performance on the external cohort varied between -10.9% and +39.5%. Fine-tuning consistently improved external performance (1.8%-15.5%), enhanced model calibration and clinical utility, and maintained key contributing variables. However, the fine-tuned models did not reach the performance of the country-specific models, showing a performance drop between 14% and 20%. Fine-tuning may be useful in other fields but still insufficient for tabular EHR-based prediction models in health applications.

Digital health technology tools (DHTTs) have the potential to transform health care delivery by enabling new forms of participatory and personalized care that fit into patients’ daily lives. However, realizing this potential requires careful navigation of numerous challenges. This viewpoint presents the authors’ experiences and perspectives on the development and implementation of DHTTs, addressing both established practices and controversial topics. This article offers a practical guide organized into 10 recommendations derived from a multidisciplinary lecture series and associated workshop discussions on “Digital Health and Digital Biomarkers” held at the University of Luxembourg in 2023-2024. Key messages include the need to understand specific health care challenges, form interdisciplinary teams, incorporate patient feedback, select appropriate measurement technologies, ensure data integration and interoperability, apply advanced data science techniques, use scalable designs and open standards, comply with regulatory requirements, and maintain continuous evaluation and improvement. While the guide highlights essential practices, it also addresses contentious issues such as balancing innovation with regulatory compliance, addressing ethical concerns in artificial intelligence adoption, managing privacy versus the need for comprehensive data integration and open science, and managing the financial sustainability of DHTTs. The authors argue that digital health’s greatest potential lies in its ability to provide participatory and personalized care, but this requires a delicate balance between technological advances and ethical, legal, and social implications. Overall, this workshop-derived viewpoint aims to help health care professionals, engineers, developers, and researchers not only adopt best practices but also address and resolve the controversial aspects inherent in the development of DHTTs.

Background: Hypertension and type 2 diabetes (T2D) are two of the most frequently co-occurring long-term conditions, but their shared mechanisms are not fully understood, often being attributed to adiposity pathways. Here, we aimed to identify shared genetic mechanisms independent of adiposity. Methods: We performed genome-wide association study meta-analyses of T2D and, separately, hypertension. We investigated the bidirectional causal relationship using Mendelian randomisation and quantified genetic correlation before and after accounting for common modifiable risk factors. We then applied a Bayesian GWAS approach to re-estimate SNP-disease effects after accounting for the causal genetic effects of adiposity-related traits. Colocalisation analysis identified shared causal genetic variants, and we investigated the biological pathways involved. Results: We observed a bidirectional causal relationship, and substantial genetic correlation between the two traits (rg = 0.48, 95%CI 0.45-0.52), which persisted after accounting for the genetic contributions of BMI, waist-hip ratio (WHR), and triglycerides (rg = 0.29, 95%CI 0.24-0.34). This indicated shared mechanisms beyond those captured by standard measures of adiposity. We found 98 genetic loci containing variants significantly associated with both hypertension and T2D; colocalisation analysis identified 37 that contained specific shared causal variants. Of these, eight remained statistically significant after adjusting for genetic measures of adiposity, and four were identified only after removing the causal effect of adiposity measures. Shared variants include an allele within PCSK7 associated with risk of both T2D and hypertension and with circulating PCSK7 protein levels, as well as a variant in the 3′ untranslated region of ZNF101 , within the TM6SF2 locus, likely reflecting regulatory variation affecting hepatic lipid metabolism and cardiometabolic traits.

The UK Health Data Research Alliance presents five recommendations for improving data collection for inclusive health research.