Alireza Vafaei Sadr

Stroke survivors face considerable risk for mental health complications including depression, anxiety, and mental distress. Despite evidence linking mental health to stroke, frequent mental distress (FMD), a broader public health marker of mental health burden, remains underexamined in stroke populations. This study aims to find the prevalence and associated factors of FMD among U.S. stroke survivors. Analysis was expanded to compare FMD with individuals who have a history of myocardial infarction (MI) and healthy adults. We conducted a cross-sectional analysis using Behavioral Risk Factor Surveillance System data from 2016 to 2023. FMD was defined in two different ways, with the standard threshold of ≥14 days of poor mental health in the past 30 days, and an alternative ≥6-day threshold. Weighted prevalence estimates and multivariate logistic regressions examined associations between social and demographic factors and FMD. Comparative analyses assessed differences in FMD between stroke and MI survivors with the healthy adult population as reference. FMD was most prevalent among stroke survivors who were younger adults (18–64) at 32.5%, females at 26.0%, and Hispanics at 28.7%. Geographically, the highest FMD rates occurred in West Virginia (29.0%) and Tennessee (27.6%), concentrated in the Stroke Belt and Appalachian regions. Stroke survivors had significantly greater odds of experiencing FMD than healthy adults in both unadjusted (4.36) and adjusted regression models (1.96). MI survivors also had greater odds of FMD compared to healthy adults (3.02 unadjusted, 1.69 adjusted), but FMD was less likely when compared to stroke survivors. Our findings suggest that FMD is more prevalent in younger adults, females, Hispanic individuals, and those residing in the Stroke belt and Appalachian states. FMD is also more prevalent among stroke survivors compared to individuals with MI or healthy adults. Future research is necessary for targeted mental health interventions in these specific subgroups.

Background: Artificial intelligence (AI) can improve stroke imaging workflows, but its computational carbon footprint remains poorly quantified. We estimated carbon dioxide equivalent (CO ₂ eq) emissions from AI use in US stroke management and evaluated carbon‐aware mitigation. Methods: We combined TriNetX neuroimaging utilization data with Global Burden of Disease stroke burden estimates for 2018 to 2019. Imaging utilization rates were aligned by state, year, stroke type, sex, and age group. We modeled 2 AI scenarios: Minimal AI, defined as one 2‐dimensional classification model and one 2‐dimensional segmentation model; and Ideal AI, defined as two 2‐dimensional classification models, two 3‐dimensional segmentation models, and one 3‐dimensional classification model. Model energy use was converted to CO ₂ eq using state‐level grid carbon intensity from Electricity Maps. Results: Ideal AI generated 16 375.09 metric tons CO ₂ eq/year (95% CI, 16 340.83–16 409.36), compared with 7692.70 metric tons CO ₂ eq/year (95% CI, 7676.08–7709.32) for Minimal AI. Computed tomography angiography and magnetic resonance imaging were the largest modality contributors. Per‐capita emissions varied substantially by state, reflecting local grid composition. Relocating AI‐related processing to cleaner‐energy states reduced modeled Ideal AI emissions to 7408.68 metric tons CO ₂ eq/year (95% CI, 7386.99–7430.38), a 54.75% reduction. Conclusions: AI‐assisted stroke imaging has a measurable carbon footprint that depends on model intensity, imaging modality, and grid carbon intensity. Carbon‐aware routing of computation offers an immediate mitigation strategy while preserving clinical AI use.

Background: Lack of physical activity is a major risk factor for cardiovascular diseases. Wearables and smartphones offer potential advantages in cost, accessibility, and scale to change physical activity. This systematic review aims to determine if interventions comprising smartphones and wearables are effective in enhancing physical activity among individuals with cardiovascular disease. Methods: PubMed, Cochrane, and Web of Science databases were searched for randomized controlled trials, published since January 2000, on wearables and smartphone applications to enhance physical activity among patients with cardiovascular disease. Nonrandomized trials and interventions without a smartphone/wearable component were excluded. Risk of bias was assessed by the Cochrane Collaboration tool. Meta‐analyses were performed to assess the pooled effect on steps per day, distance, oxygen consumption, and (moderate‐to‐vigorous) physical activity in minutes per day. Results: Fourteen randomized controlled trials were included. Interventions comprising smartphones and wearables resulted in a mean difference of 1097.4 (95% CI, 409.2–1785.6; P =0.0018) steps per day, and 3.9 (95% CI, 0.2–7.6; P =0.0413) minutes of moderate‐vigorous physical activity per day compared with control groups. Conclusion: This meta‐analysis showed that interventions comprising smartphones and wearable devices are effective at increasing physical activity among patients with cardiovascular disease. Wearables and smartphones could provide accessible and tailored interventions to enhance physical activity.

Background: Machine learning models for stroke mortality prediction typically treat each time horizon independently and use flat tabular features that ignore the relational structure of electronic health records (EHRs). In this pilot study, we leveraged graph-based machine learning models to predict post stroke all-cause-mortality across three different time horizons. Methods: We developed Stroke Temporal Heterogeneous Graph (StrokeTHG), a heterogeneous graph neural network model for simultaneous multi-horizon stroke mortality prediction (30-day, 90-day, 1-year) using EHR data from Penn State Health System. The model encodes various relations among EHR entities (e.g., patient, diagnosis, comorbidity) and temporal encoding of admission time to better predict stroke mortality. We compared our proposed approach against various baseline methods, including Logistic Regression, Random Forest, and XGBoost. We also performed ablation and subgroup analyses, evaluated the quality of learned graph embeddings, and assessed the importance of different edge types in the graph. Results: We included 4,144 stroke patients (mean age 69.2 years; 54.3% men), of whom 3,332 (80.4%) survived their stroke after one year. 30-day, 90-day, and 1-year mortality rates were 9.7%, 13.7%, and 19.6%, respectively. Our proposed approach, StrokeTHG, achieved AUROC of 0.872, 0.878, and 0.837 across horizons, outperforming all tabular baselines. At ≥75% specificity, the model identified 5–10 percentage points more mortality cases than the best baseline at each horizon. Subgroup analysis demonstrated consistent performance across sex subgroups and the largest discriminative gains in the Age 65–80 stratum. Edge-type ablation identified phenotype–patient and admission–patient edges in the constructed EHR graph as the most influential relational edges for mortality prediction. StrokeTHG embeddings outperformed all graph and matrix factorization baselines under an identical downstream classifier, confirming that performance gains stem from representation quality rather than classifier capacity. Conclusions: StrokeTHG demonstrates that heterogeneous graph representations of EHR data provide a consistent improvement over flat tabular models for multi-horizon stroke mortality prediction, with particular advantage at clinically actionable sensitivity thresholds and novel multi-horizon monotonic prediction capability. This methodological framework may be adaptable to other EHR-based clinical research studies seeking to leverage heterogeneous relational structures for predictive modeling.

Background: Artificial intelligence (AI) is increasingly being used in healthcare. Despite its promise, physicians and trainees remain cautiously optimistic. This systematic review and meta-analysis aimed to assess knowledge and attitudes toward AI and to provide recommendations for AI buy-in by physicians. Methods: Searches of PubMed-OVID-IEEE-Scopus, and Web-of-Science for studies in 2013–2024 identified 11,437 records. One-hundred-and-fifteen met inclusion criteria. Fifty-three studies reported quantitative data on physicians’/trainees’ knowledge and were included in the meta-analysis. Results: Our meta-analysis estimated that only 19.6% of physicians and trainees have high overall AI knowledge, while 36.3% have low knowledge. Fifty-five studies evaluated the depth of AI knowledge. These studies consistently concluded that most physicians or trainees possess only moderate conceptual knowledge of AI, and their technical knowledge is usually limited. Qualitative evaluations also highlighted that a high level of conceptual AI knowledge is associated with greater receptiveness to AI implementation in medicine. We identified five major barriers to translating AI to the bedside with physician buy-in. Conclusion: Although physicians and trainees are generally receptive to AI, many barriers hinder adoption. To address them, we recommend establishing standardized AI education and workforce training, involving clinicians early in AI design, clarifying legal and regulatory issues, leveraging insights from clinical decision support system implementation to reduce workflow challenges, and integrating patient-centered communication principles to enhance trust and transparency.

Background: Lack of physical activity is a major risk factor for Cardiovascular diseases (CVDs). Wearables and smartphones offer potential advantages in cost, accessibility, and scale to change physical activity This systematic review aims to determine if interventions comprising smartphones and wearables are effective in enhancing physical activity among individuals with CVD. Methods: PubMed, Cochrane, and Web of Science databases were searched for RCTs, published since January 2000, on wearables and smartphone applications to enhance physical activity among CVD patients. Non-randomized trials, interventions without a smartphone/wearable component were excluded. Risk of bias was assessed by the Cochrane Collaboration tool. Meta-analyses were performed to assess the pooled effect on steps per day, distance, VO2, and (moderate-to-vigorous) physical activity in minutes per day. Results: Fourteen RCTs were included. Interventions comprising smartphones and wearables resulted in a mean difference of 1097.4 (95% CI = [409.2-1785.6]; p=0.0018) steps per day, and 3.9 (95% CI = [0.2-7.6]; p=0.0413) minutes of moderate-vigorous physical activity per day compared to control groups. Conclusion: This meta-analysis showed that interventions comprising smartphones and wearable devices are effective at increasing physical activity among CVD patients. Wearables and smartphones could provide accessible and tailored interventions to enhance physical activity.

Background: Socioeconomic status plays a crucial role in determining health outcomes, particularly the age at which individuals experience their first ischemic stroke. This study investigates the relationship between ZIP‐code level median income and the age at the first ischemic stroke, considering various demographic and clinical characteristics. Methods: We identified patients >18 years old from the 2019 National Inpatient Sample database with a primary diagnosis of ischemic stroke. Multivariate linear regression adjusted for sex, race, patient location, and comorbidities was used to estimate the impact of income on the age at the first ischemic stroke. Results: Among 94 074 patients, representing a national estimate of 470 370, the mean age at first stroke was 70.1±13.9 years, with 49.5% women. In the lowest (<$47 999) area‐level income group, the average age of the first stroke was 4.3 years lower than the highest ($82 000+) income group (68.1±13.8 versus 72.4±13.8). After adjusting for sociodemographic factors and comorbidities, the highest‐income individuals were 2.1 years older (95% CI, 2.3–2.5) at first ischemic stroke compared with those in the lowest‐income category. Area‐level income disparities also affected stroke age across racial groups and sexes. Specifically, first stroke occurred 2.6, 0.9, and 0.9 years earlier for White, Black, and Hispanic individuals, respectively. Women experienced a 2.3‐year delay compared with a 1.9‐year delay for men. Conclusion: Lower income levels are associated with the first incidence of stroke at a younger age, especially among White individuals and women, highlighting socioeconomic status‐related disparities in stroke incidence. Further studies to investigate the underlying factors are warranted.

Predicting paroxysmal atrial fibrillation (PAF) is challenging due to its transient nature. Existing methods often rely solely on electrocardiogram (ECG) waveforms or Electronic Health Record (EHR)-based clinical risk factors. We hypothesized that explicitly balancing the contributions of these heterogeneous data sources could improve prediction accuracy. We developed a Transformer-based deep learning model that integrates 12-lead ECG signals and 47 structured EHR variables from 189 patients with cryptogenic stroke, including 49 with PAF. By systematically varying the relative contributions of ECG and EHR data, we identified an optimal ratio for prediction. Best performance (accuracy: 0.70, sensitivity: 0.72, specificity: 0.87, Area Under Curve - Receiver Operating Characteristics (AUROC): 0.65, Area Under the Precision-Recall Curve (AUPRC): 0.43) was achieved using a 5-fold cross-validation when EHR data contributed one-third and ECG data two-thirds of the model’s input. This multimodal approach outperformed unimodal models, improving accuracy by 35% over EHR-only and 5% over ECG-only methods. Our results support the value of combining ECG and structured EHR information to improve accuracy and sensitivity in this pilot cohort, motivating validation in larger studies.

Sepsis is a leading cause of mortality in the United States, with over 350,000 deaths annually, yet the contribution of neighborhood‐level social determinants of health (SDoH) remains underexplored. We conducted a retrospective analysis of 4.4 million hospitalized patients with sepsis, identified using ICD‐10 codes, leveraging de‐identified data from the TriNetX Diamond Network. Neighborhood‐level variables included median income, housing cost burden, broadband access, park proximity, racial/ethnic diversity, and the Area Deprivation Index (ADI). Mortality was the primary outcome, assessed using generalized linear models and Random Forest analysis. Lower neighborhood income, higher ADI scores, limited park access, and lack of broadband connectivity were consistently associated with increased sepsis mortality. Unexpectedly, greater racial/ethnic diversity correlated with lower mortality, potentially reflecting regional differences in healthcare access and socioeconomic context. In multivariate models, several associations observed in univariate analyses were attenuated, highlighting the complex interplay of SDoH. Random Forest modeling identified park access as the most important predictor, suggesting its utility as a proxy for broader neighborhood resources. Neighborhood‐level SDoH significantly influence sepsis outcomes and should be integrated into prognostic models to enhance risk stratification. Targeted interventions addressing social vulnerabilities, expanding access to green spaces, and reducing disparities in healthcare resources may help mitigate the burden of sepsis mortality across diverse populations.

Background: Although physical activity correlates with reduced depressive symptoms among survivors of stroke, how frequency, intensity, and duration of physical activity relate to mental health is unknown. This study aims to find the association of physical activity doses recommended in 2011, 2014, and 2021 American Heart Association/American Stroke Association guidelines with mental health burden (number of days of poor mental health). Analysis was expanded to find the optimal dose linked to mental health burden. Methods: We used data from the 2011 to 2019 Behavioral Risk Factor Surveillance System, a nationally representative survey. The mental health burden between guideline‐adherent and nonadherent groups was compared. Propensity score matching was used to balance the 2 groups with respect to sociodemographic factors and comorbidities. The relative difference in mental health burden between the groups is reported. The impact of duration, frequency, and intensity was analyzed using nonlinear regression, adjusting for potential confounders. We further performed a comparative analysis among survivors of myocardial infarction. Results: Survivors of stroke who adhered to 2011, 2014, and 2021 guidelines had 0.6, 0.8, and 0.3 fewer days of poor mental health, with relative differences of 11.1%, 14.8%, and 5.8%, respectively. Among survivors of stroke adhering to 2014 guidelines (3–4 sessions/week, 40 minutes/session), younger (18–64) individuals showed greater relative differences in mental health burden at 15.7%. For those adhering to 2021 guidelines (4 moderate 10‐minute or 2 vigorous 20‐minute sessions), older (≥65) individuals had higher relative differences at 11.1%. Regression analysis showed that the optimal dose is 4 to 6 weekly sessions of moderate physical activity, each lasting 45 minutes. Conclusions: The findings suggest that physical activity doses recommended in 2011, 2014, and 2021 American Heart Association/American Stroke Association guidelines are associated with lower mental health burden. Furthermore, the amount of physical activity associated with lower mental health may vary by subpopulations (eg, age). Randomized control trials are needed to validate the optimal dose.