This Dissertation explores the interplay between data and decision-making in healthcare, focusing on two key dimensions: data representativeness and analytical scope. As healthcare systems increasingly rely on data-driven methods to face newer and complex challenges, it becomes essential to critically appraise how different approaches can influence decision-making. Structured as a collection of papers, most of which were developed in collaboration with a healthcare provider, this Dissertation examines how data can be used to inform both policy and clinical practice, addressing methodological challenges that may arise when the available information is limited. The nine studies included in this Dissertation lay at the intersection of operations research and data science, and address fundamental questions such as: how is it possible to construct robust solutions when data on uncertain parameters is scarce? How can data science tools be used to design high-level healthcare strategies? In what ways can machine learning enhance clinical decision-making? When extensive data on a particular problem are available, how to best exploit this knowledge? Is it possible to assess the potential impact of a machine learning approach even before developing it? The works included in this Dissertation employ several techniques, covering a wide range of approaches, from text mining and regression/classification models to simulation, (robust) (mixed) integer linear programming and (mat)heuristics. Two novel methodologies to address uncertain parameters are also proposed herein. Besides, these studies span all the levels of decision-making — strategical, tactical and operational — addressing managerial topics such as primary care network planning and blood donation governance, and clinical applications such as surgical planning and scheduling. All the reported case studies illustrate how data-driven operations research can improve efficiency, enhance patient outcome, and ultimately support decision-makers in navigating uncertainty. This Dissertation offers insights for both researchers and practitioners, emphasizing the applicability and impact of analytical tools in real-world settings.
Questa Tesi esplora le interazioni tra dati e processi decisionali in ambito sanitario, concentrandosi su due dimensioni chiave: la rappresentatività dei dati stessi e la finalità analitica. Poiché i sistemi sanitari si affidano sempre più a metodi data-driven per affrontare sfide nuove e complesse, diventa essenziale valutare in maniera critica come i diversi approcci possono influenzare le decisioni. Strutturata come una collezione ragionata di articoli, molti dei quali sviluppati con enti sanitari, questa Tesi esamina come i dati possano essere utilizzati per guidare sia le politiche sanitarie che la pratica clinica, considerando le sfide che possono sorgere quando le informazioni disponibili sono limitate. I nove lavori inclusi in questa Tesi si collocano all'intersezione tra la ricerca operativa e la scienza dei dati, affrontando domande cruciali quali: come è possibile ottenere soluzioni robuste quando le informazioni sui parametri incerti sono scarse? Come si possono usare strumenti data-driven per progettare strategie sanitarie? In che modo possono le tecniche di machine learning migliorare i processi decisionali nella pratica clinica? Quando sono disponibili molti dati su un particolare problema, come si può sfruttare al meglio questa conoscenza? È possibile valutare il potenziale impatto di un approccio data-driven ancor prima di svilupparlo? I lavori inclusi in questa Tesi usano un'ampia gamma di tecniche, dagli approcci di text mining e modelli di regressione/classificazione alla simulazione, la programmazione lineare intera (mista) (robusta) e le (mat)euristiche. Due metodi innovativi per affrontare l'incertezza nei parametri sono inoltre qui proposti. Questi studi abbracciano tutti i livelli del processo decisionale — strategico, tattico e operativo — affrontando problemi gestionali come la pianificazione delle reti di assistenza primaria e la governance del sistema-sangue, e applicazioni cliniche come la pianificazione e la programmazione chirurgica. Tutti i casi di studio riportati illustrano come la ricerca operativa data-driven può incrementare l'efficienza, migliorare il benessere dei pazienti e, in ultima analisi, supportare la gestione dell'incertezza. Questa Tesi offre spunti sia ai ricercatori che ai professionisti, sottolineando l'applicabilità e l'impatto degli strumenti analitici in contesti reali.
On data and decision-making: perspectives from healthcare applications considering representativeness and analytical scope
Doneda, Martina
2024/2025
Abstract
This Dissertation explores the interplay between data and decision-making in healthcare, focusing on two key dimensions: data representativeness and analytical scope. As healthcare systems increasingly rely on data-driven methods to face newer and complex challenges, it becomes essential to critically appraise how different approaches can influence decision-making. Structured as a collection of papers, most of which were developed in collaboration with a healthcare provider, this Dissertation examines how data can be used to inform both policy and clinical practice, addressing methodological challenges that may arise when the available information is limited. The nine studies included in this Dissertation lay at the intersection of operations research and data science, and address fundamental questions such as: how is it possible to construct robust solutions when data on uncertain parameters is scarce? How can data science tools be used to design high-level healthcare strategies? In what ways can machine learning enhance clinical decision-making? When extensive data on a particular problem are available, how to best exploit this knowledge? Is it possible to assess the potential impact of a machine learning approach even before developing it? The works included in this Dissertation employ several techniques, covering a wide range of approaches, from text mining and regression/classification models to simulation, (robust) (mixed) integer linear programming and (mat)heuristics. Two novel methodologies to address uncertain parameters are also proposed herein. Besides, these studies span all the levels of decision-making — strategical, tactical and operational — addressing managerial topics such as primary care network planning and blood donation governance, and clinical applications such as surgical planning and scheduling. All the reported case studies illustrate how data-driven operations research can improve efficiency, enhance patient outcome, and ultimately support decision-makers in navigating uncertainty. This Dissertation offers insights for both researchers and practitioners, emphasizing the applicability and impact of analytical tools in real-world settings.| File | Dimensione | Formato | |
|---|---|---|---|
|
M. Doneda - Thesis (versione deposito, cap 6 escluso per copyright).pdf
accessibile in internet per tutti
Descrizione: Versione definitiva della tesi. Il Capitolo 6, corrispondente ad un articolo già pubblicato (DOI: 10.3280/PDS2023-002004) è stato estromesso dalla versione online accessibile della tesi per accordi in merito al copyright con la casa editrice Franco Angeli.
Dimensione
8.74 MB
Formato
Adobe PDF
|
8.74 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/241777