For over two decades, the Italian healthcare system has been facing increasingly complex challenges that compromise its long-term sustainability. Among the most significant issues are persistent cuts to public funding; the progressive demographic shift towards an aging population and the consequent increase in chronic diseases; and a significant delay in digital transformation of public infrastructures. These factors collectively contribute to increasing the pressure on the National Health System (SSN), limiting its capacity to respond adequately and effectively to the growing and diverse needs of users. This dissertation aims to explore and analyze the application of Computer Vision (CV) and Machine Learning (ML) within the context of Italian public healthcare. Through a series of case studies and practical applications implemented in various healthcare settings nationwide, the present work demonstrates the innovative potential of these enabling technologies in enhancing operational efficiency, augmenting user-perceived quality, and optimizing resource management. The first case study examines the application of CV in the automated recognition of nasopharyngeal swabs used in COVID-19 self-testing. This research demonstrates how the implementation of an automatic outcome classification system, integrated with the Electronic Health Record (FSE), can contribute not only to tracking new infections and containing viral spread but also to alleviating the workload of healthcare and administrative personnel during critical phases of a pandemic. Subsequently, the application of ML techniques for predicting the Length of Stay (LOS) in Emergency Departments is investigated. Through the analysis of data collected during the triage phase, the predictive models developed in this study facilitate more effective management of patient flow in emergency care departments, enabling the implementation of targeted and proactive interventions. A further application of ML that has been explored involves the prediction of LOS in hospital wards. The implemented models exhibit notable accuracy in estimating the duration of patient hospitalization, favouring the timely identification of potential “bed blockers”, more precise scheduling of discharges and admissions, and optimal use of available beds. Finally, a distributed variant of a bed management system, designated as Distributed Electronic Bed Management System (DEBMS), is presented. The system, specifically designed for Bed Managers and Hospital Administrations, collects and aggregates information from primary hospital management flows, processing them through predictive models to provide valuable insights that support operational decision-making. The DEBMS facilitates real-time sharing of statistics among nearby hospitals, and enables distributed management of patient admission requests, fostering more effective synergy among hospital structures at the provincial level and improving coordination and response capacity to healthcare needs. Digital transformation represents a strategic opportunity to reform healthcare systems, modernize their infrastructures, and enhance their resilience in the face of recent challenges, such as the 2020 pandemic, and future exigencies. Through the analysis of concrete cases and the discussion of related practical implications, this work offers an important contribution to the debate on how the judicious adoption of Artificial Intelligence (AI) in Italian healthcare settings can lead to substantial evolutions in the current landscape, while maintaining an unwavering focus on the quality of care provided and the sustainability of a public healthcare system internationally recognized for its excellence.

Da oltre due decenni, il sistema sanitario italiano si trova ad affrontare situazioni di crescente complessità, tali da comprometterne la sostenibilità a lungo termine. Tra le problematiche più rilevanti vi sono i tagli ai finanziamenti pubblici; l’invecchiamento progressivo della popolazione e il conseguente aumento delle patologie croniche; nonché un significativo ritardo nella transizione digitale delle infrastrutture pubbliche. Questi fattori contribuiscono ad aumentare la pressione sul Sistema Sanitario Nazionale (SSN), limitandone la capacità di rispondere in modo adeguato ed efficace alle crescenti e diversificate esigenze degli utenti. La presente tesi si propone di esplorare e analizzare l’applicazione della Computer Vision (CV) e del Machine Learning (ML) nel contesto della sanità pubblica italiana. Attraverso una serie di casi di studio e applicazioni pratiche realizzate in diverse realtà sanitarie del Paese, il lavoro dimostra il potenziale di queste tecnologie abilitanti nel migliorare l’efficienza operativa, aumentare la qualità percepita dagli utenti e ottimizzare la gestione delle risorse. Il primo caso di studio riguarda l’applicazione della CV nel riconoscimento automatizzato di tamponi nasofaringei utilizzati nell’autotesting del COVID-19. Viene dimostrato come l’implementazione di un sistema di classificazione automatica degli esiti possa contribuire non solo al tracciamento delle nuove infezioni e al contenimento della diffusione del virus, ma anche a ridurre il carico di lavoro del personale sanitario e amministrativo durante le fasi critiche di una pandemia. Successivamente, viene esplorata l’applicazione di tecniche di ML per la previsione della durata della permanenza (Length of Stay, LOS) nei reparti di Pronto Soccorso. Analizzando dati raccolti durante la fase di triage, i modelli predittivi sviluppati consentono una gestione più efficace del flusso di pazienti, promuovendo l’attuazione di interventi mirati e proattivi. Un’ulteriore applicazione del ML riguarda la previsione della LOS nei reparti ospedalieri. I modelli implementati mostrano un’accuratezza soddisfacente nella stima della durata del ricovero dei pazienti, facilitando l’individuazione tempestiva di potenziali “bed blockers”, una programmazione più precisa di dimissioni e ingressi, e un utilizzo ottimale dei posti letto disponibili. Infine, viene illustrata una variante distribuita di un sistema di gestione informatizzata dei posti letto (Distributed Electronic Bed Management System, DEBMS), rivolta ai Bed Manager e alle Direzioni ospedaliere. Tale sistema raccoglie e integra informazioni dai principali flussi di gestione ospedaliera, elaborandole attraverso modelli predittivi per fornire indicazioni di supporto ai processi decisionali. Il DEBMS facilita la condivisione di statistiche in tempo reale tra gli ospedali di prossimità e consente una gestione distribuita delle richieste di presa in carico dei pazienti, incoraggiando una sinergia più incisiva tra le strutture ospedaliere a livello provinciale e migliorando la capacità di coordinamento e risposta alle esigenze sanitarie. La trasformazione digitale rappresenta un’opportunità strategica per riformare i sistemi sanitari, modernizzarne le infrastrutture e renderli più resilienti di fronte alle sfide future. Attraverso l’analisi di casi concreti e la discussione delle relative implicazioni pratiche, il presente lavoro offre un contributo importante al dibattito su come l’adozione ponderata dell’Intelligenza Artificiale nelle realtà sanitarie italiane possa aprire a evoluzioni sostanziali nel panorama attuale, pur mantenendo al centro la qualità delle cure erogate e la sostenibilità di un sistema sanitario pubblico riconosciuto come uno tra i migliori a livello globale.

Trasformazione Digitale e Apprendimento Automatico Applicati alla Sanità Pubblica / Paolo Perliti Scorzoni , 2025 Apr 07. 37. ciclo, Anno Accademico 2023/2024.

Trasformazione Digitale e Apprendimento Automatico Applicati alla Sanità Pubblica

PERLITI SCORZONI, PAOLO
2025

Abstract

For over two decades, the Italian healthcare system has been facing increasingly complex challenges that compromise its long-term sustainability. Among the most significant issues are persistent cuts to public funding; the progressive demographic shift towards an aging population and the consequent increase in chronic diseases; and a significant delay in digital transformation of public infrastructures. These factors collectively contribute to increasing the pressure on the National Health System (SSN), limiting its capacity to respond adequately and effectively to the growing and diverse needs of users. This dissertation aims to explore and analyze the application of Computer Vision (CV) and Machine Learning (ML) within the context of Italian public healthcare. Through a series of case studies and practical applications implemented in various healthcare settings nationwide, the present work demonstrates the innovative potential of these enabling technologies in enhancing operational efficiency, augmenting user-perceived quality, and optimizing resource management. The first case study examines the application of CV in the automated recognition of nasopharyngeal swabs used in COVID-19 self-testing. This research demonstrates how the implementation of an automatic outcome classification system, integrated with the Electronic Health Record (FSE), can contribute not only to tracking new infections and containing viral spread but also to alleviating the workload of healthcare and administrative personnel during critical phases of a pandemic. Subsequently, the application of ML techniques for predicting the Length of Stay (LOS) in Emergency Departments is investigated. Through the analysis of data collected during the triage phase, the predictive models developed in this study facilitate more effective management of patient flow in emergency care departments, enabling the implementation of targeted and proactive interventions. A further application of ML that has been explored involves the prediction of LOS in hospital wards. The implemented models exhibit notable accuracy in estimating the duration of patient hospitalization, favouring the timely identification of potential “bed blockers”, more precise scheduling of discharges and admissions, and optimal use of available beds. Finally, a distributed variant of a bed management system, designated as Distributed Electronic Bed Management System (DEBMS), is presented. The system, specifically designed for Bed Managers and Hospital Administrations, collects and aggregates information from primary hospital management flows, processing them through predictive models to provide valuable insights that support operational decision-making. The DEBMS facilitates real-time sharing of statistics among nearby hospitals, and enables distributed management of patient admission requests, fostering more effective synergy among hospital structures at the provincial level and improving coordination and response capacity to healthcare needs. Digital transformation represents a strategic opportunity to reform healthcare systems, modernize their infrastructures, and enhance their resilience in the face of recent challenges, such as the 2020 pandemic, and future exigencies. Through the analysis of concrete cases and the discussion of related practical implications, this work offers an important contribution to the debate on how the judicious adoption of Artificial Intelligence (AI) in Italian healthcare settings can lead to substantial evolutions in the current landscape, while maintaining an unwavering focus on the quality of care provided and the sustainability of a public healthcare system internationally recognized for its excellence.
Digital Transformation and Machine Learning Applied to Public Healthcare
7-apr-2025
GRANA, Costantino
BOLELLI, FEDERICO
File in questo prodotto:
File Dimensione Formato  
Perliti_PhD_Thesis_final.pdf

Open access

Descrizione: Tesi definitiva Perliti Scorzoni Paolo
Tipologia: Tesi di dottorato
Dimensione 14.38 MB
Formato Adobe PDF
14.38 MB Adobe PDF Visualizza/Apri
Pubblicazioni consigliate

Licenza Creative Commons
I metadati presenti in IRIS UNIMORE sono rilasciati con licenza Creative Commons CC0 1.0 Universal, mentre i file delle pubblicazioni sono rilasciati con licenza Attribuzione 4.0 Internazionale (CC BY 4.0), salvo diversa indicazione.
In caso di violazione di copyright, contattare Supporto Iris

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11380/1376157
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact