Zack van Allen
PhD, MBA

This preprint examines whether LLM-generated feedback on resident scholarly projects can approach expert human feedback in quality, usefulness, and consistency. The study evaluates AI-generated reports in blinded comparisons, focusing on whether structured prompting and evaluation workflows can produce feedback that is educationally credible while also reducing reviewer workload in medical training contexts.

This study uses longitudinal data from the Canadian Longitudinal Study on Aging to prospectively classify functional dependence using machine learning. The analysis compares predictive performance across models and identifies a parsimonious set of variables associated with later dependence, showing how predictive analytics may support earlier identification of older adults at risk for decline in everyday functioning.

Using eight waves of longitudinal data, this study examined whether physical activity before stroke is associated with the long-term evolution of functional limitations after stroke compared with matched adults without stroke. The findings showed that the beneficial effect of prestroke physical activity on basic activities of daily living was stronger in people with stroke, supporting physical activity as both a preventive and prognostic factor in functional recovery.

This study applies temporal network analysis to multiple health behaviours and pandemic-related protective behaviours using longitudinal Canadian data from the iCARE study. It shows that behaviours such as physical activity, healthy eating, vaping, alcohol use, and other protective actions are interconnected over time, supporting the view that health behaviours are better understood as dynamic systems rather than isolated outcomes.

Drawing on baseline and follow-up data from more than 40,000 participants in the Canadian Longitudinal Study on Aging, this paper compares co-occurrence-based and co-variation-based approaches to understanding multiple health behaviours. Seven behavioural clusters were identified, but individual behaviours proved to be stronger predictors of later health outcomes, suggesting that clustering is most useful for targeting subgroups whereas correlations are more useful for understanding behaviour systems.

This protocol outlines a program of analyses using Canadian Longitudinal Study on Aging baseline data to examine how health and everyday-life behaviours cluster together. It details the planned use of hierarchical cluster analysis, multinomial models, and network community detection to study links between behavioural groupings, health, life satisfaction, and healthcare use, laying the methodological foundation for later empirical work.

This paper tests whether acting more extraverted in everyday life can improve well-being across three week-long interventions. Participants completed brief daily behavioural exercises, and the results showed reliable boosts in positive affect, some gains in authenticity, and a more nuanced picture of when acting counter to dispositional style feels beneficial versus effortful in daily life.

Using a patient-oriented qualitative approach with immigrants to Canada from Pakistani, Chinese, and French-speaking African and Caribbean communities, this study identified barriers and enablers to diabetic retinopathy screening attendance. Shared issues included awareness, language, cost, wait times, appointment logistics, and social support, while some barriers were group-specific, informing the design of linguistically and culturally tailored interventions.

This experiment tested the trait-state isomorphism hypothesis in the domain of openness to experience by assigning participants to five days of exercises designed either to activate openness-related thoughts and behaviours or to serve as a control task. Results supported the hypothesis for positive affect but not for creative thinking or personal growth, suggesting that some benefits associated with trait openness may be more malleable than others.