Data-driven mathematical models represent nowadays valuable tools to provide a wide support in many clinical practices thanks to their capability to describe relationships among variables and pattern within the data, exploit historical data to predict future events and simulate experiments several times and in different scenarios. The application of mathematical modeling in the healthcare field can be particularly beneficial when considering chronic diseases, which nowadays constitute the leading cause of death and disability worldwide, as well as an economic burden for the public health. Therefore, a current interest for public health organizations is to design strategies aimed at reducing the impact of the disease on the healthcare systems, by focusing on the 3 main directions of action for chronic diseases management, namely prevention, prognosis and therapy. Mathematical models are valuable resources to achieve such purposes, allowing, for example, to identify major risk factors for a disease development, highlight different patterns of disease progression and subjects phenotype, and test the effectiveness of treatments or therapy on virtual subjects without any risk for real patients. Being long-lasting conditions, chronic diseases intrinsically generate clinical data over time, also called longitudinal data, usually collected in subsequent medical appointments to monitor the patient’s status under natural or treated conditions. Longitudinal data constitute a precious resource for developing mathematical models, which provide a dynamic picture of the disease evolution from different aspects (symptoms, treatments, lab tests, etc) allowing, for example, to investigate how relationships among the observed variables change over time. However, their employment in developing mathematical models leads to various challenges which need to be appropriately addressed, mainly due to their temporal sparsity and heterogeneous nature, and thus is still limited in the literature. The availability of models developed on dynamic data and capable to reproduce the temporal evolution of a disease or a clinical state under natural or treated conditions also enables the realization of the so-called in silico clinical trials (ISCTs). ISCTs, allowing to simulate a large number of virtual subjects with different characteristics and for very long time with reduced costs, are particularly helpful when implementing a clinical trial is ethically or logistically impossible due to time, patients safety, available/required resources and other limitations which can hinder the realization of meaningful studies. As an example, ISCTs allow to run multiple experiments on the same virtual subjects by testing the effect of different treatments while maintaining the same surrounding conditions (including patients’ physiology and behaviour), scenario very difficult to recreate in real life. The aim of this thesis is to develop innovative mathematical modeling approaches taking advantage of dynamic data and, as a complement to real data, simulated data aimed at supporting the main objectives of chronic diseases management, namely prevention, prognosis and therapy optimization. Three different case studies are addressed, one for each level of action on chronic diseases, concerning different clinical contexts and for which ad hoc methodological approaches are proposed.
Modelli matematici data-driven rappresentano preziosi strumenti volti a fornire un grande supporto in diverse procedure cliniche grazie alla loro capacità di descrivere relazioni tra le variabili e pattern risconstrabili nei dati, sfruttare dati cronologici per predire eventi futuri, e simulare esperimenti diverse volte e in diversi scenari. L'applicazione di modelli matematici nell'ambito sanitario può essere particolarmente utile quando si considerano malattie croniche, le quali ad oggi costituiscono la principale causa di morte e disabilità mondiale, oltre a rappresentare un enorme peso economico per la sanità pubblica. Di conseguenza, un interesse attuale per le organizzazioni di sanità pubblica è di progettare strategie volte a ridurre l'impatto delle malattie croniche sul sistema sanitario, focalizzandosi su 3 principali direzioni di azione, ovvero prevenzione, prognosi e terapia. I modelli matematici rappresentano una preziosa risorsa per conseguire questi obbiettivi, permettondo, ad esempio, di identificare i principali fattori di rischio per lo sviluppo di una malattia, evidenziare diversi pattern di progressione della malattia e fenotipi di popolazione, così come testare l'efficiacia di trattamenti o di una terapia su soggetti virtuali senza alcun rischio per i pazienti reali. Essendo patologie di lunga durata, le malattie croniche generano dati clinici nel tempo, chiamati anche dati longitudinali, comunemente raccolti durante successive visite mediche per monitorare lo stato del paziente in condizioni normali o sotto trattamento. I dati longitudinali costituiscono una preziosa risorsa per lo sviluppo di modelli matematici, costituendo una fotografia dinamica dell'evoluzione della malattia da diversi aspetti (sintomi, trattamenti, esami clinici, etc) e quindi permettendo, ad esempio, di investigare come le relazioni tra le variabili osservate cambiano nel tempo. Tuttavia, il loro utilizzo nello sviluppo di modelli matematici, al momento ancora limitato in letteratura, genera diverse sfide metodologiche che, principalmente dovute alla sparsità temporale e natura eterogenea dei dati, devono essere adeguatamente affrontate. La disponibilità di modelli sviluppati su dati dinamici e capaci di riprodurre l'evoluzione temporale di una malattia o di uno stato clinico in condizioni normali o sotto trattamento consente inoltre di realizzare i cosiddetti trial clinici simulati. I trial clinici simulati, permettendo di simulare un grande numero di soggetti virtuali con diverse caratteristiche e per una lunga durata a costi ridotti, sono particolarmente utili quando la realizzazione di trial clinici reali è eticamente o logisticamente impossibile a causa del tempo richiesto, sicurezza dei pazienti, disponibilità di risorse e altre limitazioni che possono ostacolare la realizzazione di studi significativi. Per riportare un esempio, i trial clinici simulati permettono di eseguire diversi esperimenti sugli stessi soggetti virtuali testando l'effetto di diversi trattamenti in condizioni riproducibili (inclusi la fisiologia e il comportamento del paziente), scenario molto difficile da ricreare in condizioni reali. L'obbiettivo di questa tesi riguarda lo sviluppo di approcci innovativi basati su modelli matematici traendo vantaggio da dati dinamici e, a complemento di dati reali, dati simulati, con lo scopo di supportare i principali obbiettivi della gestione delle malattie croniche, ovvero prevenzione, prognosi e ottimizzazione del trattamento. In questa tesi sono affrontati tre diversi casi di studio, uno per ciascun livello di azione sulla malattie croniche, riguardanti diversi contesti clinici e per cui sono proposti approcci metodologici realizzati ad hoc per il problema in questione.
Mathematical modelling approaches to support the prevention, prognosis and treatment optimization of chronic diseases / Roversi, Chiara. - (2023 Mar 17).
Mathematical modelling approaches to support the prevention, prognosis and treatment optimization of chronic diseases
ROVERSI, CHIARA
2023
Abstract
Data-driven mathematical models represent nowadays valuable tools to provide a wide support in many clinical practices thanks to their capability to describe relationships among variables and pattern within the data, exploit historical data to predict future events and simulate experiments several times and in different scenarios. The application of mathematical modeling in the healthcare field can be particularly beneficial when considering chronic diseases, which nowadays constitute the leading cause of death and disability worldwide, as well as an economic burden for the public health. Therefore, a current interest for public health organizations is to design strategies aimed at reducing the impact of the disease on the healthcare systems, by focusing on the 3 main directions of action for chronic diseases management, namely prevention, prognosis and therapy. Mathematical models are valuable resources to achieve such purposes, allowing, for example, to identify major risk factors for a disease development, highlight different patterns of disease progression and subjects phenotype, and test the effectiveness of treatments or therapy on virtual subjects without any risk for real patients. Being long-lasting conditions, chronic diseases intrinsically generate clinical data over time, also called longitudinal data, usually collected in subsequent medical appointments to monitor the patient’s status under natural or treated conditions. Longitudinal data constitute a precious resource for developing mathematical models, which provide a dynamic picture of the disease evolution from different aspects (symptoms, treatments, lab tests, etc) allowing, for example, to investigate how relationships among the observed variables change over time. However, their employment in developing mathematical models leads to various challenges which need to be appropriately addressed, mainly due to their temporal sparsity and heterogeneous nature, and thus is still limited in the literature. The availability of models developed on dynamic data and capable to reproduce the temporal evolution of a disease or a clinical state under natural or treated conditions also enables the realization of the so-called in silico clinical trials (ISCTs). ISCTs, allowing to simulate a large number of virtual subjects with different characteristics and for very long time with reduced costs, are particularly helpful when implementing a clinical trial is ethically or logistically impossible due to time, patients safety, available/required resources and other limitations which can hinder the realization of meaningful studies. As an example, ISCTs allow to run multiple experiments on the same virtual subjects by testing the effect of different treatments while maintaining the same surrounding conditions (including patients’ physiology and behaviour), scenario very difficult to recreate in real life. The aim of this thesis is to develop innovative mathematical modeling approaches taking advantage of dynamic data and, as a complement to real data, simulated data aimed at supporting the main objectives of chronic diseases management, namely prevention, prognosis and therapy optimization. Three different case studies are addressed, one for each level of action on chronic diseases, concerning different clinical contexts and for which ad hoc methodological approaches are proposed.File | Dimensione | Formato | |
---|---|---|---|
tesi_definitiva_Chiara_Roversi.pdf
embargo fino al 16/03/2026
Descrizione: tesi_definitiva_Chiara_Roversi
Tipologia:
Tesi di dottorato
Dimensione
6.52 MB
Formato
Adobe PDF
|
6.52 MB | Adobe PDF | Visualizza/Apri Richiedi una copia |
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.