Throughout the world, it is estimated that one person in seven thousand is genetically affected by at least one type of Growth Hormone Deficiency. This medical condition can be treated through the use of growth hormone replacement using recombinant human growth hormone. Merck is one the main worldwide producer of this treatment, with the Saizen® medication. This company also produces medical devices that can be used by patients to perform daily injection of r-hGH (recombinant Growth Hormone). These devices are connected and transmit to Merck a set of data about Saizen injections. One key aspect of a successful r-hGH therapy holds in the patient’s ability in following the treatment plan (treatment adherence). In case of non-adherence, the treatment loses its efficacy. The main objective of this work is to determine the factors that influence high adherence using the growth hormone-related data obtained by Merck’s connected devices. This thesis presents a new approach based on machine learning to analyze patient data and obtain new insights about the factors that influence patient adherence. This thesis first presents a review of the machine learning technologies used, and a review of Merck in-house data managing processes. Then, it presents the dataset preparation, the choice of the features to be analyzed, the model training and the model evaluation for the XGBoost, Lasso, and Snowflake ML methods. Next, a clustering analysis of the patient is performed to discover other factors that impact adherence. Finally, the different results of the analyses are compared and discussed to determine the points on which Merck can act to better support the patients in their treatment. This work exhibits that the factors that most influence the probability that a patient remains adherent in the long term are: the consistency in administering the drug at the same hour across the months, the use of the latest version of Merck auto-injectors, as well as socio-cultural factors such as the country of origin or the clinic responsible for the treatment.
Si stima che, a livello mondiale, una persona su settemila sia geneticamente affetta da almeno un tipo di carenza di ormone della crescita. Questa condizione clinica può essere trattata mediante l’uso di ormone della crescita umano ricombinante. Merck è uno dei principali produttori mondiali di questo trattamento, con il farmaco Saizen®. Questa azienda produce inoltre dispositivi medici che possono essere utilizzati dai pazienti per eseguire iniezioni giornaliere di r-hGH (ormone della crescita umano ricombinante). Questi dispositivi sono connessi e trasmettono a Merck dati sulle iniezioni di Saizen® somministrate. Un aspetto fondamentale di una terapia r-hGH di successo consiste nella capacità del paziente nel seguire il trattamento (aderenza al trattamento). In caso di non aderenza, il trattamento perde la sua efficacia. L’obiettivo principale di questo lavoro è determinare i fattori che influenzano un’alta aderenza utilizzando i dati sull’uso dell’ormone della crescita ottenuti dai dispositivi connessi di Merck. In questa tesi si analizza un approccio innovativo che utilizza il machine learning per analizzare i dati dei pazienti e ottenere nuove informazioni sui fattori che influenzano l’aderenza. Questa tesi presenta innanzitutto una revisione delle tecnologie di machine learning utilizzate e una revisione dei processi di gestione dei dati interni di Merck. In seguito, descrive la preparazione del dataset e la scelta dei parametri da analizzare, l’allenamento e la valutazione del modello per i metodi XGBoost, Lasso e Snowflake ML. Viene poi eseguita una caratterizzazione dei pazienti per identificare altri fattori che influenzano l’aderenza. Infine, i diversi risultati delle analisi vengono confrontati e discussi per determinare i punti su cui Merck può intervenire per supportare meglio i pazienti nel loro trattamento. Questo lavoro dimostra che i fattori che influenzano maggiormente la probabilità che un paziente rimanga aderente al trattamento con r-hGH sul lungo periodo sono: la costanza nell’amministrare il farmaco alla stessa ora ogni mesi, l’uso dei più recenti auto-iniettori di Merck, oltre a fattori socio-culturali quali il paese di origine o la clinica responsabile del trattamento.
Detection of the factors impacting patients' adherence to growth hormone treatment using machine learning
SIMONIN, MÉLISSANDE SYLVIE GAËLLE
2023/2024
Abstract
Throughout the world, it is estimated that one person in seven thousand is genetically affected by at least one type of Growth Hormone Deficiency. This medical condition can be treated through the use of growth hormone replacement using recombinant human growth hormone. Merck is one the main worldwide producer of this treatment, with the Saizen® medication. This company also produces medical devices that can be used by patients to perform daily injection of r-hGH (recombinant Growth Hormone). These devices are connected and transmit to Merck a set of data about Saizen injections. One key aspect of a successful r-hGH therapy holds in the patient’s ability in following the treatment plan (treatment adherence). In case of non-adherence, the treatment loses its efficacy. The main objective of this work is to determine the factors that influence high adherence using the growth hormone-related data obtained by Merck’s connected devices. This thesis presents a new approach based on machine learning to analyze patient data and obtain new insights about the factors that influence patient adherence. This thesis first presents a review of the machine learning technologies used, and a review of Merck in-house data managing processes. Then, it presents the dataset preparation, the choice of the features to be analyzed, the model training and the model evaluation for the XGBoost, Lasso, and Snowflake ML methods. Next, a clustering analysis of the patient is performed to discover other factors that impact adherence. Finally, the different results of the analyses are compared and discussed to determine the points on which Merck can act to better support the patients in their treatment. This work exhibits that the factors that most influence the probability that a patient remains adherent in the long term are: the consistency in administering the drug at the same hour across the months, the use of the latest version of Merck auto-injectors, as well as socio-cultural factors such as the country of origin or the clinic responsible for the treatment.File | Dimensione | Formato | |
---|---|---|---|
2024_07_SIMONIN.pdf
non accessibile
Descrizione: Text of the thesis
Dimensione
7.54 MB
Formato
Adobe PDF
|
7.54 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/10589/223871