INTRODUCTION: On 2 March 2020, the first positive subject for the New Coronavirus was detected in the Municipality of Tortona, in the province of Alessandria, the third metropolitan area of Piedmont bordering Lombardy, the initial outbreak of the epidemic in Italy. METHODOLOGY: The study discussed in this thesis has, first of all, examined the monitoring system implemented by the ASL of Alessandria to manage the health emergency caused by COVID-19. Secondly, some prediction models were applied to the collected data to know in advance the spread of the virus from a space-time point of view in the Italian province of interest. The main epidemiological parameters were estimated (transmission rate β, recovery rate ɣ, basic reproduction rate R0 and final attack rate τ) for the entire province and for the Municipalities which, at the date of data extraction (21 April 2020), had a statistically significant sample of tested subjects. A SIR epidemiological model was also applied to the data for the prediction of the final number of infected, cured and deceased, in addition to a Machine Learning algorithm, the K-Nearest-Neighbors (KNN), thanks to which it was possible to predict in advance the positive or negative outcome of the swab based on four simple data: the Municipality of the subject, the day of the test request, the day of the test outcome and the age. RESULTS: A first significant result emerged from the survey is that 6.7% of swab requests were uploaded more than once to the IT platform: this fact highlighted some anomalies in the use of the information system by the operators in charge. From the analysis of the swabs carried out, it was then possible to obtain the average response time for obtaining a result (4,25 days). For the production chain including the laboratory of Alessandria, a performance index was also calculated for evaluating the efficiency of production resources and equipment (OEE), which recorded a not very encouraging value (17%). In fact, at the date of extraction of the data, a significant percentage of swabs was still not received (19.9%) or awaiting a result (29.2%). From an epidemiological point of view, a prevalence of infection equal to 40% (CI = 95%, percentage error = 7%) was obtained for the province of Alessandria, with some substantial differences between the Municipalities: Alessandria (866 observations, 41.45%), Tortona (670, 40.26%), Casale Monferrato (615, 32.84%), Novi Ligure (408, 25.49%), Acqui Terme (362, 34.53%), Ovada (221.51.58%). Finally, a strong correlation was shown between swab positives and age (mean age of infected subjects = 65.2 years, median age = 66 years). CONCLUSIONS: The analyses conducted have suggested that the lacking performances measured, with reference to the entire production chain, are mainly attributable to logistics. On the other hand, the study performed on tested subjects showed that the ASL of Alessandria was able to effectively identify the target population to be tested. From the results of the SIR model, however, it emerged that the so-called "phase 2" began, for the province of Alessandria, before the expected date of stopping the infections, consequently making the reopening in this specific geographical area more dangerous than in other Italian provinces. The results obtained also made it possible to state that the areas most affected by the epidemic were those of the most populous Municipalities in the province, as well as the areas neighbouring the regions of Lombardy and Liguria. Finally, a tool that proved to be very powerful was the KNN classification algorithm. The latter, if implemented with data on hospitalizations and deaths, will allow to control at the local health level (therefore with municipal granularity) the catchment area to be monitored in anticipation of a new wave of infections. This result would allow an ASL to draw up a screening plan sufficiently in advance, thus allowing the implementation of targeted prevention measures on the territory in the future.
INTRODUZIONE: Il 2 marzo 2020 è stato rilavato nel Comune di Tortona il primo soggetto positivo al Nuovo Coronavirus nella provincia di Alessandria, terza area metropolitana del Piemonte a confine con la Lombardia, focolaio iniziale dell’epidemia in Italia. METODOLOGIA: Lo studio discusso in questa tesi ha, innanzitutto, esaminato il sistema di monitoraggio implementato dall’ASL di Alessandria per la gestione dell’emergenza sanitaria da COVID-19. In secondo luogo, sono stati applicati ai dati raccolti alcuni modelli di predizione per conoscere in anticipo la diffusione del virus dal punto di vista spazio-temporale nella provincia italiana di interesse. Sono stati stimati i principali parametri epidemiologici (velocità di trasmissione β, tasso di recupero ɣ, tasso di riproduzione di base R0 e tasso finale d’attacco τ) per l’intera provincia e per i Comuni che, alla data di estrazione dei dati (21 aprile 2020), possedevano un campione statisticamente significativo di soggetti tamponati. Ai dati è stato, inoltre, applicato un modello epidemiologico SIR per la predizione del numero finale di infetti, guariti e deceduti, e un algoritmo di Machine Learning, il K-Nearest-Neighbors (KNN), grazie al quale è stato possibile prevedere in anticipo l’esito positivo o negativo del tampone sulla base di quattro semplici dati: il Comune del soggetto, la data di richiesta del test, la data dell’esito e l’età. RISULTATI: Un primo risultato significativo emerso dall’indagine è che il 6,7% delle richieste di tamponi sono state inserite più di una volta all’interno della piattaforma informatica: tale fatto ha messo in luce alcune anomalie nell’utilizzo del sistema informativo da parte degli operatori preposti. Dall’analisi dei tamponi effettuati, poi, è stato possibile ricavare il tempo medio di risposta per l’ottenimento di un esito (4,25 giorni). Per la catena produttiva comprendente il laboratorio di Alessandria è stato calcolato, inoltre, un indice di performance per la valutazione dell’efficienza delle risorse produttive e dell’impianto (OEE), che ha registrato un valore poco incoraggiante (17%). Infatti, alla data di estrazione dei dati, una percentuale rilevante di tamponi risultava ancora non pervenuta (19,9%) o in attesa di un esito (29,2%). Dal punto di vista epidemiologico, per la provincia di Alessandria è stata ricavata una prevalenza dell’infezione pari al 40% (IC = 95%, errore percentuale = 7%), con alcune sostanziali differenze tra Comune e Comune: Alessandria (866 osservazioni, 41.45%), Tortona (670, 40.26%), Casale Monferrato (615, 32.84%), Novi Ligure (408, 25.49%), Acqui Terme (362, 34.53%), Ovada (221, 51.58%). È stata, infine, dimostrata una forte correlazione tra i positivi al tampone e l’età (età media dei soggetti infetti = 65,2 anni, età mediana = 66 anni). CONCLUSIONI: Le analisi condotte hanno suggerito che le scarse prestazioni misurate, in riferimento all’intera catena produttiva, sono attribuibili principalmente alla logistica. D’altra parte, lo studio eseguito sui soggetti tamponati, ha dimostrato che l’ASL di Alessandria è riuscita a individuare in modo efficace la popolazione target da sottoporre a test. Dai risultati del modello SIR, è però emerso che la cosiddetta “fase 2” è iniziata, per la provincia di Alessandria, prima della prevista data di arresto dei contagi, rendendo di conseguenza la riapertura in questa specifica area geografica più pericolosa rispetto ad altre province italiane. I risultati ottenuti, inoltre, hanno permesso di affermare che le zone più colpite dall’epidemia sono state quelle dei Comuni più popolosi della provincia, nonché le aree a confine con la Lombardia e la Liguria. Infine, uno strumento che si è rivelato molto potente è stato l’algoritmo di classificazione KNN. Quest’ultimo, se implementato con dati sui ricoveri e i decessi, permetterà di controllare a livello della sanità locale (quindi con granularità municipale) il bacino d’utenza da sottoporre a sorveglianza in previsione di una nuova ondata di contagi. Tale risultato consentirebbe ad una ASL di stilare con sufficiente anticipo un piano di screening, permettendo in tal modo di attuare in futuro misure di prevenzione mirate sul territorio.
Modelli di predizione a supporto delle decisioni in sanità : l'epidemia da COVID-19 nella provincia di Alessandria
VAZZANA, DAVIDE
2019/2020
Abstract
INTRODUCTION: On 2 March 2020, the first positive subject for the New Coronavirus was detected in the Municipality of Tortona, in the province of Alessandria, the third metropolitan area of Piedmont bordering Lombardy, the initial outbreak of the epidemic in Italy. METHODOLOGY: The study discussed in this thesis has, first of all, examined the monitoring system implemented by the ASL of Alessandria to manage the health emergency caused by COVID-19. Secondly, some prediction models were applied to the collected data to know in advance the spread of the virus from a space-time point of view in the Italian province of interest. The main epidemiological parameters were estimated (transmission rate β, recovery rate ɣ, basic reproduction rate R0 and final attack rate τ) for the entire province and for the Municipalities which, at the date of data extraction (21 April 2020), had a statistically significant sample of tested subjects. A SIR epidemiological model was also applied to the data for the prediction of the final number of infected, cured and deceased, in addition to a Machine Learning algorithm, the K-Nearest-Neighbors (KNN), thanks to which it was possible to predict in advance the positive or negative outcome of the swab based on four simple data: the Municipality of the subject, the day of the test request, the day of the test outcome and the age. RESULTS: A first significant result emerged from the survey is that 6.7% of swab requests were uploaded more than once to the IT platform: this fact highlighted some anomalies in the use of the information system by the operators in charge. From the analysis of the swabs carried out, it was then possible to obtain the average response time for obtaining a result (4,25 days). For the production chain including the laboratory of Alessandria, a performance index was also calculated for evaluating the efficiency of production resources and equipment (OEE), which recorded a not very encouraging value (17%). In fact, at the date of extraction of the data, a significant percentage of swabs was still not received (19.9%) or awaiting a result (29.2%). From an epidemiological point of view, a prevalence of infection equal to 40% (CI = 95%, percentage error = 7%) was obtained for the province of Alessandria, with some substantial differences between the Municipalities: Alessandria (866 observations, 41.45%), Tortona (670, 40.26%), Casale Monferrato (615, 32.84%), Novi Ligure (408, 25.49%), Acqui Terme (362, 34.53%), Ovada (221.51.58%). Finally, a strong correlation was shown between swab positives and age (mean age of infected subjects = 65.2 years, median age = 66 years). CONCLUSIONS: The analyses conducted have suggested that the lacking performances measured, with reference to the entire production chain, are mainly attributable to logistics. On the other hand, the study performed on tested subjects showed that the ASL of Alessandria was able to effectively identify the target population to be tested. From the results of the SIR model, however, it emerged that the so-called "phase 2" began, for the province of Alessandria, before the expected date of stopping the infections, consequently making the reopening in this specific geographical area more dangerous than in other Italian provinces. The results obtained also made it possible to state that the areas most affected by the epidemic were those of the most populous Municipalities in the province, as well as the areas neighbouring the regions of Lombardy and Liguria. Finally, a tool that proved to be very powerful was the KNN classification algorithm. The latter, if implemented with data on hospitalizations and deaths, will allow to control at the local health level (therefore with municipal granularity) the catchment area to be monitored in anticipation of a new wave of infections. This result would allow an ASL to draw up a screening plan sufficiently in advance, thus allowing the implementation of targeted prevention measures on the territory in the future.| File | Dimensione | Formato | |
|---|---|---|---|
|
2020_10_Vazzana.PDF
non accessibile
Descrizione: Testo della tesi
Dimensione
3.59 MB
Formato
Adobe PDF
|
3.59 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/10589/166553