BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20250706T005658EDT-0931B6BCAc@132.216.98.100
DTSTAMP:20250706T045658Z
DESCRIPTION:Arthur Chatton\, PhD\n\nIVADO postdoctoral fellow |\n	Faculté de
  Pharmacie\, Université de Montréal\n\nWHEN: Monday\, January 22\, 2024\, 
 from 4 to 5 p.m.\n\nWHERE: Hybrid| 2001 9IÖÆ×÷³§Ãâ·Ñ College\, Rm 1140 | Zoom\n
 \nNOTE: Arthur Chatton will be presenting in-person\n\nAbstract\n\nObtaini
 ng continuously updated predictions is a major challenge for personalized 
 medicine. In end-stage kidney diseases\, a major cause of morbidity and mo
 rtality worldwide\, dialysis is the standard therapy. However\, achieving 
 high blood-filtered volumes time after time and across patient populations
  requires clinical skills and readily accessible information and data. Nep
 hrologists and nurses must continually re-assess multiple parameters refre
 shed with each HDF session and consider time-varying clinical status chang
 es\, which is daunting in busy dialysis centres.\n\nDynamic prediction mod
 els provide predicted outcome values that can be updated over time for an 
 individual as new measurements become available. Previous approaches to pr
 ediction were mainly based on parametric models\, but there is a current t
 rend towards using more flexible machine learning approaches. Ensemble met
 hods leverage combinations of parametric regressions and machine learning 
 approaches into one final prediction.\n\nWe extend an ensemble method call
 ed super learner for (i) dynamically predicting a repeated continuous outc
 ome and (ii) optimizing the prediction for the patients the clinician face
 s up by combining approaches trained on the personal history of the patien
 t or on an external (i.e.\, 'historical') cohort. We also propose a new wa
 y to validate such personalized prediction models. We illustrate its perfo
 rmance by predicting the convection volume of patients undergoing hemodiaf
 iltration\, a specific dialysis technique\, in Montréal\, Canada.\n\nThe p
 ersonalized dynamic super learner outperformed its candidate learners with
  respect to median absolute error\, calibration-in-the-large\, discriminat
 ion\, and net benefit. We finally discuss the choices and challenges under
 lying its use and implementation.\n\nLearning Objectives\n\nBy the end of 
 this session\, attendees will:\n\n\n	Have a better understanding of the sup
 er learning framework\;\n	Become acquainted with dynamic prediction\;\n	Unde
 rstand the challenges of validating personalized prediction models.\n\n\nS
 peaker Bio\n\nArthur Chatton is a French biostatistician working on the cr
 ossroads of causal inference and prediction. His current interests focus m
 ainly on using machine learning approaches for causal inference\, either f
 or estimation purposes or identifiability checking. His work is supported 
 by an IVADO postdoctoral fellowship. He has an MSc and a PhD in Biostatist
 ics from the Université de Nantes\, France.\n
DTSTART:20240122T210000Z
DTEND:20240122T220000Z
SUMMARY:Personalized dynamic prediction in dialysis using a novel super lea
 rning framework
URL:/epi-biostat-occh/channels/event/personalized-dyna
 mic-prediction-dialysis-using-novel-super-learning-framework-353821
END:VEVENT
END:VCALENDAR