This thesis aims to explore predictive models for the risk of developing Post-Traumatic Stress Disorder (PTSD), with a particular focus on recent machine learning methodologies, as well as traditional statistical models such as logistic regression. Through a review of the scientific literature, pre- and peri-traumatic risk factors were analyzed, including age, gender, and peritraumatic dissociation. The thesis also focused on multivariable prediction models, including neurobiological ones. The findings highlight that the use of complex algorithms (Random Forest, SVM, XGBoost) allows for greater accuracy in risk prediction, opening new perspectives for the prevention and treatment of PTSD. However, study methodology and potential bias limit the generalizability of the conclusions. The thesis emphasizes the need for integrated and externally validated approaches for the effective identification of vulnerable individuals and the personalization of care in clinical settings.
La tesi si propone di esplorare i modelli predittivi del rischio di sviluppo del Disturbo da Stress Post-traumatico (PTSD). Pone particolare attenzione alle recenti metodologie di machine learning, così come a modelli statistici tradizionali quali la regressione logistica. Attraverso la revisione della letteratura scientifica, sono stati analizzati i fattori di rischio pre- e peri-traumatici, tra cui età, genere e dissociazione peri-traumatica. La tesi si è concentrata inoltre sui modelli di predizione multivariabili, come quelli neurobiologici. I risultati evidenziano che l'utilizzo di algoritmi complessi (Random Forest, SVM, XGBoost) consente maggiore accuratezza nella predizione del rischio, aprendo nuove prospettive nella prevenzione e nel trattamento del PTSD. Tuttavia, la metodologia degli studi, insieme al rischio di bias, limitano l'università delle conclusioni. La tesi sottolinea la necessità di approcci integrati e validati esternamente, per un'efficace identificazione dei soggetti vulnerabili e una personalizzazione delle cure in ambito clinico.
Predizione del rischio nel PTSD
CAPECE, DILETTA
2024/2025
Abstract
This thesis aims to explore predictive models for the risk of developing Post-Traumatic Stress Disorder (PTSD), with a particular focus on recent machine learning methodologies, as well as traditional statistical models such as logistic regression. Through a review of the scientific literature, pre- and peri-traumatic risk factors were analyzed, including age, gender, and peritraumatic dissociation. The thesis also focused on multivariable prediction models, including neurobiological ones. The findings highlight that the use of complex algorithms (Random Forest, SVM, XGBoost) allows for greater accuracy in risk prediction, opening new perspectives for the prevention and treatment of PTSD. However, study methodology and potential bias limit the generalizability of the conclusions. The thesis emphasizes the need for integrated and externally validated approaches for the effective identification of vulnerable individuals and the personalization of care in clinical settings.| File | Dimensione | Formato | |
|---|---|---|---|
|
Capece_Diletta.pdf
Accesso riservato
Dimensione
457.59 kB
Formato
Adobe PDF
|
457.59 kB | Adobe PDF |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/90814