Last EPAFit recap of the 2023/24 academic year. This research group was created on the promises of transparency, robustness, and impactful research on physical activity and public health across lifespan, and we'll keep pushing them. Courses, projects, articles,... stay tuned!

Os esperamos en el V Congreso Internacional Healthy-Age: Envejecimiento Activo, Ejercicio y Salud. 🗓️ ¡Reserva la fecha! 14 y 15 de noviembre de 2024. ℹ️ Más info: healthy-age-meeting.blogspot.com/?m=1 Organizan CAFD UMU Facultad Educa-UAL CSD HealthyAgeNet Universidad Murcia Universidad de Almería

Do you want to know more about Bayesian statistics advantages? Take a peek to our research fellow Daniel Gallardo-Gómez content! #Bayesian #stats #data #science

From EPAFit, we are aware of the importance of data literacy in our society. It is critical to understand how data is collected, analysed, and interpreted. Great resource: youtube.com/watch?v=81vaSt…

I love spaghettis, but I like spaghettis plots better to visualise the variability in the effect estimate of the conditional expectation with posterior draws across a continuous covariate in our GAM Souped up! #Bayesian #stats #metaanalysis #data #science

📷 Nuestro compañero de AETSA, Daniel Gallardo, ha defendido su tesis “Evidence Synthesis Research: Applications on Physical Activity and Public Health” en Universidad de Sevilla 👉 Miguel Ángel Armengol, del Área de #BigData y Juan Carlos Rejón, de #AETSA, han sido miembros del Tribunal

Prednisone presented fewer adverse events in people with Crohn’s Disease (CD) than other treatments. Males had more associated adverse events than females; and age and BMI were determined as risk factors. Posterior predictive check showed good performance using poisson family.

Code to replicate these analyses in github.com/dgalgom/Treatm…

Minimally (clinically) important difference and minimal effective dose. If we have evidence of the minimal impact that must have a treatment, we can predict the exact dose required to reach this goal, and potentially, save money. #dose #response #modelling

Importance of individual-level treatment estimates when great inter-personal variability exists for regulatory decision-makers. In addition, how do you think is more appropriate: treat the time as discrete or continuous variable to visualise changes over time? 🤔

EPAFit was present during the 10th International Society for Physical Activity and Health Congress in Paris through our colleagues Jesús del Pozo-Cruz, Borja del Pozo Cruz, and Jaime López García. Well done team! ISPAH

It is all about this, people. Wonderful weekend in the GenAI Health Hackaton organised by the Hospital Clínic We generated an AI-based, ready-to-use information system for clinicians from free-text data, potentially saving time, money, and resources, optimising treatments.

(major) Sex differences in dose-response of VILPA (vigorous intermitent l/style #PhysicalActivity) with CVD events New UK Biobank study in British Journal of Sports Medicine (BJSM) bjsm.bmj.com/content/early/… 🙏Matthew Ahmadi Raaj Kishore Biswas Mark Hamer Jason Gill Angelo Sabag Cecilie Thøgersen Borja del Pozo Cruz

A powerful way to beautify your RMarkdown reports: embedding simple, visual Shiny R Apps. Example: Obesity Predictor Simulator. After model fitting (C5.0 algorithm), we predicted probabilities for BMI status. Video: probabilities flip from no food between meals to sometimes 🤔

📊 Our latest meta-analysis, incorporating data from over 12,000 participants across 7 countries, provides new insights into how changes in daily movement behaviors influence obesity risks across the lifespan. 🔗Full text: doi.org/10.1007/s40279… 👇Explore more in this video.

Our new review in cardio metabolic benefits of combining MVPA and muscle strengthening activities.

Our researcher Javier Ramos Munell led a study quantifying the tracking of Moderate-Vigorous Physical Activity (MVPA) across childhood and adolescence in a recent cohort from England. Results? Moderate tracking... but we could do more about it! More info in: 10.1016/j.jsams.2024.03.006

Confounding in observational data? Meet the parametric g-computation. Steps: 1. Fit a regression for outcomes 2. Create counterfactual datasets (treated vs. untreated, Fig1) 3. Predict outcomes, compare means = treatment effect (Fig2) Implementation in R using {marginaleffects}

Making patient experiences more clinically relevant via LLMs I used 🦙 3.2 in R for sentiment analysis on ICU’s Drug Reviews dataset for free-text classification, and perceived side effect extraction Revealing patterns combining objective and subjective insights can emerge!

Conformal prediction to adequately cover a specific share of out-of-sample observations. Key advantage: well-calibrated intervals regardless of the prediction model we use (even a misspecified one). Spoiler alert: 95% CIs don’t cover 95% of unseen observations! 🫢