Communications medicine最新文献

Background: To accurately detect individuals' mental health issues using artificial intelligence and self-report scales, it is crucial to recognize how personal characteristics can affect the detection. This study focuses on the role of alexithymia-a condition where individuals struggle to recognize and articulate emotions and symptoms-in the detection of depression. We aimed to determine whether deep learning models could enhance the accuracy of depression detection in people with alexithymia compared to self-report scales.

Methods: We analyzed data from 194 patients with major depressive disorder and 105 community controls, employing eight large language models (LLMs) trained on transcript text from clinician-administered structured interviews using the Hamilton Depression Rating Scale (HAMD).

Results: Here we show that generalized logistic regression analysis indicates a positive relationship between alexithymia and depression. Using the HAMD as the gold standard, individuals with alexithymia show poorer performance on the self-reported Hospital Anxiety and Depression Scale-Depression Subscale (HADS-D) in identifying depression (b = -0.37, p = .002). Four of the eight LLMs (AUCs=0.87-0.89) significantly outperform the HADS-D (AUC = 0.79) in depression detection (p <0.05). Subgroup analysis demonstrates that while LLMs achieve AUCs ranging from 0.79 to 0.96, the HADS-D only reaches an AUC of 0.35 for individuals with alexithymia.

Conclusions: Our findings reveal that LLMs can potentially outperform self-report scales in detecting depression, particularly in those with alexithymia. These results highlight the importance of considering patient characteristics, such as alexithymia, when detecting depression. Deep learning analyses can enhance the accuracy of clinical assessments for depression and potentially for other mental health disorders.

Background: The skeletal muscle hypothesis refers to a vicious cycle of successive deterioration of left ventricular function, skeletal muscle remodeling, and functional capacity in patients with heart failure. Despite extensive research, the regulatory mechanisms and their associations with clinical status and prognosis are still largely unclear.

Methods: To identify mechanisms and characterize underlying processes involved in the disease pathophysiology, we performed RNA sequencing and network analysis using human skeletal muscle samples from 58 patients with severe symptomatic heart failure. A co-expression network with communities involved in established biological processes within human skeletal muscle was identified and validated in two independent cohorts.

Results: Here, we show network communities associated with mitochondrial beta-oxidation, extracellular matrix remodeling, oxidative phosphorylation, and contractile elements with lower expression in heart failure patients than in age-matched controls. Based on the strong correlation with clinical features and prognosis, extracellular matrix remodeling, mitochondrial beta-oxidation, and p53 signalling communities are identified as key underlying processes. The former two communities are highly enriched with genes regulated by physical (in)activity, i.e., bed rest and exercise, and associated weakly with prognosis. Community related to p53 signalling, with CDKN1A as a key regulator, is increased in heart failure patients relative to age-matched controls and associated with worse prognosis.

Conclusion: The current work differentiates previously proposed factors underlying heart failure-induced skeletal muscle dysfunction, emphasizing the p53 signalling community and importance of biological age in this process. The distinct association with clinical status and prognosis furthermore supports pathophysiological significance and clinical potential of this community.

Background: Immigrants in Sweden, particularly those from low- and middle-income countries, had higher risks of COVID-19 mortality and morbidity compared to the Swedish-born. However, prior studies have not quantified the contribution of the differential distribution of health and social determinants to the increased risks.

Methods: We used total population registers from Sweden to investigate disparities in COVID-19 hospitalization between five groups of immigrants and Swedish-born, using a cohort 577911 working-age adults (18-65 years) living in Stockholm during the first two waves of the COVID-19 pandemic. Applying a decomposition analysis, we quantified the relative contribution of age, sex, income, education, occupation type, residential area, and pre-existing medical conditions to these disparities.

Results: Our study shows that immigrants have higher risks of hospitalization compared to Swedish-born, and that the investigated factors accounted for these disparities to varying degrees across immigrant groups. For the most affected immigrant groups (from Africa and Middle East), the examined factors together account for only a minor part of the disparities (21% and 18% for Wave 1; 16% and 11% for Wave 2), with occupation type and residential area contributing substantially.

Conclusions: Common observable social determinants of health account for a moderate share of the overall disparities in COVID-19 hospitalizations between Swedish-born individuals and immigrant from the most affected regions of origin.

Background: Shortages in mental healthcare lead to long periods of inadequate support for many patients. While digital interventions offer a scalable solution to this unmet clinical need, patient engagement remains a key challenge. Generative artificial intelligence (genAI) presents an opportunity to deliver highly engaging, personalized mental health treatment at scale.

Methods: In a pre-registered (ClinicalTrials.gov: NCT06459128, 10 June 2024), parallel, 2-arm, unblinded, randomized controlled trial (N = 540), we evaluate whether a genAI-enabled cognitive behavioral therapy (CBT) app enhances engagement or symptom reduction compared with digital CBT workbooks. Eligible participants are adults residing in the United States with elevated self-reported symptoms of anxiety (GAD-7 ≥ 7) or depression (PHQ-9 ≥ 9), recruited online. After an online baseline assessment, participants are automatically randomly allocated (3:2) to receive either the genAI-enabled app or a digital workbook, both self-guided over six weeks. Primary outcomes are: 1) engagement frequency and duration, and 2) change in anxiety (GAD-7) and depression (PHQ-9) symptom severity. Secondary outcomes include adverse events and functional impairment. The study is unblinded to participants and researchers due to the nature of the digital interventions.

Results: A total of 540 participants are recruited and randomized to each group (intervention: n = 322, active control: n = 218). Nine participants from the control group are excluded from analysis due to protocol deviations. Over six weeks, the genAI solution (n = 322) increases engagement frequency (2.4×) and duration (3.8×) compared to digital workbooks (n = 209), with moderate to large effect sizes. We observe comparable outcomes for anxiety (GAD-7) and depression (PHQ-9) with no differences in adverse events. Moreover, exploratory analyses suggest that participants who choose to engage with clinical personalization features powered by genAI experience stronger anxiety symptom reduction and improved overall wellbeing.

Conclusions: Our findings suggest that, in self-directed usage, tailored genAI-enabled therapy safely enhances user engagement above and beyond static materials, without showing an overall enhancement in anxiety or depression symptom reduction.

Background: The Reproductive Mental health of Ontario Virtual Intervention Network (MOVIN) aims to improve perinatal depression care across a large Canadian health jurisdiction. It involves a web-based platform, care coordinator for personalized treatment planning, and psychiatrist to liaise with primary care clinicians, and provide direct consultation as needed. This was a pilot randomized controlled trial (RCT) of MOVIN.

Methods: Participants aged ≥18 years, pregnant or within 12 months postpartum and with Edinburgh Postnatal Depression Scale (EPDS) score >12 recruited from across Ontario, Canada, were randomized 1:1 to MOVIN for 24 weeks or a control condition. The primary outcome was feasibility, inclusive of recruitment, acceptability, and research protocol follow-up. Depression symptom remission was a main secondary outcome for the pilot trial. EPDS scores and remission (EPDS ≤ 12) were compared between groups.

Results: Of 101 participants (n = 48 MOVIN; n = 53 control), 80% completed 24-week follow-up. Participant views of MOVIN were very positive (high acceptability) and multiple opportunities for protocol adjustment in a larger future study were identified. At 24 weeks post-randomization, EPDS scores were lower in MOVIN vs. controls, with a mean difference adjusted for baseline score of -2.32 (95% Confidence Interval, CI -4.23 to -0.42). At the same time point, 75.0% of the MOVIN group vs. 51.1% of controls were in remission (chi-square=4.83, p = 0.03).

Conclusions: With high feasibility, including in recruitment, acceptability, and research protocol adherence, and preliminary suggestion of efficacy, the results of this study support proceeding to a large-scale RCT of MOVIN to definitively evaluate its efficacy at a larger scale.

Background: Neural recordings capture crucial pathophysiological processes along the dementia continuum. However, cross-center variability in recording techniques and paradigms limit their generalizability and diagnostic power, preventing clinical use. We here propose a computational approach enabling cross-center classification even in the presence of completely different clinical pipelines.

Methods: We leveraged a digital twin model to derive digital biomarkers linking neurodegeneration mechanisms to alterations in neural activity across multiple recording modalities. We tested the generalizability of digital biomarkers through cross-center classification of Mild Cognitive Impairment (MCI) and healthy subjects in two independent clinics. The two datasets presented different recording techniques (EEG and MEG), preprocessing modalities, recruitment criteria and diagnostic guidelines. Digital biomarkers derived from one clinic were tested for classifying patients in the other clinic and vice versa employing a transfer learning approach.

Results: Digital biomarkers outperform standard biomarkers in the MCI vs healthy classification in both separate datasets (83% vs 58% for EEG dataset and 75% vs 68% for MEG dataset). Moreover, they achieve accurate and consistent cross-center classification (77-78% accuracy), while standard biomarkers perform poorly in the generalization attempt (56-65%). Additionally, digital biomarkers reliably predict global cognitive status across clinics across both datasets ( p < 0.01), while standard biomarkers present no correlation.

Conclusions: Digital biomarkers generalize across recording techniques and datasets, enabling a cross-modal and cross-center classification of a patient's condition. These biomarkers offer a robust measure of patient-specific neurodegeneration, mapping neural recordings anomalies into a common framework of underlying structural alterations. The vast differences between the two datasets support the applicability of this approach also in the presence of high inter-center variability.

Background: Overweight and obesity are rising globally, with Ghana experiencing significant increases among women over the past two decades, raising public health concerns. This study aimed to identify and quantify the key drivers of overweight and obesity among women of reproductive age in Ghana, analysing how these factors have contributed to prevalence changes over time.

Methods: Data from the 2003, 2008, 2014, and 2022 Ghana Demographic and Health Surveys were analysed using binary logistic regression to assess associations with factors such as age, wealth, and education. Multivariate decomposition analysis quantified the contributions of these factors to the observed increases in overweight and obesity prevalence over time.

Results: Here we show overweight and obesity among Ghanaian women rise significantly, reaching 43% in 2022. Key drivers of change in overweight and obesity include wealth, education, urban residence, age, and region. Women in the wealthiest quintile have three times the odds of overweight (aOR: 3.07 [2.02-4.67]) and over six times the odds of obesity (aOR: 6.73 [3.80-11.91]) compared to the poorest quintile. Decomposition analysis shows that 22.5% of the increase in prevalence was due to changes in population characteristics, such as marital and educational status.

Conclusions: Our findings reveal that socio-demographic changes in society, beyond individual behavioural factors, drive the rising overweight and obesity prevalence among Ghanaian women of childbearing age. These findings highlight the dynamic factors influencing weight outcomes and the need for tailored strategies addressing the diverse and evolving determinants of overweight and obesity in Ghanaian women.

Background: Real-world studies based on electronic health records often require manual chart review to derive patients' clinical phenotypes, a labor-intensive task with limited scalability. Here, we developed and compared computable phenotyping based on rules using the spaCy framework and a Large Language Model (LLM), GPT-4, for sub-phenotyping of patients with Crohn's disease, considering age at diagnosis and disease behavior.

Methods: For our rule-based approach, we leveraged the spaCy framework and for the LLM-based approach, we used the GPT-4 model. The underlying data included 49,572 clinical notes and 2204 radiology reports from 584 Crohn's disease patients. A test set of 280 clinical texts was labeled at sentence-level, in addition to patient-level ground truth data. The algorithms were evaluated based on their recall, precision, specificity values, and F1 scores.

Results: Overall, we observe similar or better performance using GPT-4 compared to the rules. On a note-level, the F1 score is at least 0.90 for disease behavior and 0.82 for age at diagnosis, and on patient level at least 0.66 for disease behavior and 0.71 for age at diagnosis.

Conclusions: To our knowledge, this is the first study to explore computable phenotyping algorithms based on clinical narrative text for these complex tasks, where prior inter-annotator agreements ranged from 0.54 to 0.98. There is no statistical evidence for a difference to the performance of human experts on this task. Our findings underline the potential of LLMs for computable phenotyping and may support large-scale cohort analyses from electronic health records and streamline chart review processes in the future.

Background: Physical activity plays an important role in preventing chronic diseases, but most studies rely on self-reported or short-term data that fail to capture habitual behavior. This study utilizes Fitbit data to investigate the relationship between physical activity and various chronic diseases.

Methods: We analyzed data from 22,019 participants in the All of Us Research Program who shared at least six months of Fitbit activity data linked with electronic health records. Various physical activity patterns were evaluated using Cox proportional hazards and logistic regression models, adjusting for age, sex, and body mass index (BMI). To test robustness, sensitivity analyses were conducted using obesity defined by BMI, applying a two-year exclusion window for outcome diagnoses to mitigate potential reverse causation, and incorporating lifestyle covariates (smoking and alcohol use) under a simplified directed acyclic graph (DAG) framework to address residual confounding.

Results: Here, we show that higher physical activity levels are associated with lower risks of multiple chronic conditions. Higher daily step counts were negatively associated with obesity and type 2 diabetes, while greater elevation gains and longer vigorous activity are associated with lower risks of conditions such as morbid obesity, obstructive sleep apnea, and major depressive disorder. All sensitivity analyses yield consistent results, supporting the robustness of findings against reverse causation and lifestyle confounding.

Conclusions: Higher physical activity and lower sedentary time may help prevent diverse chronic diseases. These findings demonstrate the potential of large-scale wearable data to inform personalized prevention and population health strategies.

Background: Approximately half of all high-grade serous ovarian carcinomas (HGSCs) have a therapeutically targetable defect in homologous recombination (HR) DNA repair. While there are genomic and transcriptomic methods, developed for other cancers, to identify HR deficient (HRD) samples, there are no gene expression-based tools to predict HR status in HGSC specifically. We have built a HGSC-specific model to predict HR status using gene expression.

Methods: We separated The Cancer Genome Atlas (TCGA) cohort of HGSCs into training (n = 288) and testing (n = 73) sets and labelled each case as HRD or HR proficient (HRP) based on the clinical standard for classification. Using the training set, we performed differential gene expression analysis between HRD and HRP cases. The 2604 significantly differentially expressed genes were used to train a penalised logistic regression model.

Results: IdentifiHR uses the expression of 209 genes to predict HR status in HGSC. These genes preserve the genomic damage signal, capturing known regions of HR-specific copy number alteration which impact gene expression. IdentifiHR is 85% accurate in the TCGA test set and 86% accurate in an independent cohort of 99 samples, taken from primary tumours, ascites and normal fallopian tubes. Further, IdentifiHR is 84% accurate in pseudobulked single-cell HGSC sequencing from 37 patients and outperforms existing expression-based methods to predict HR status, being BRCAness, MutliscaleHRD and expHRD.

Conclusions: IdentifiHR is an accurate model to predict HR status in HGSC. It is available as an open source R package, empowering researchers to robustly classify HR status when only transcriptomic sequencing data is available.