Modelización de las relaciones entre las variables implicadas en el tratamiento del Trastorno por Consumo de SustanciasReal-world evidence mediante estudios observacionales retrospectivos

  1. Dacosta Sánchez, Daniel
Supervised by:
  1. Oscar Martín Lozano Rojas Director
  2. Fermín Fernández Calderón Director
  3. María Carmen Díaz Batanero Director

Defence university: Universidad de Huelva

Fecha de defensa: 22 September 2023

Type: Thesis


This doctoral dissertation focuses on modeling the relationships between the variables involved in treating Substance Use Disorders. To achieve this, the nosology of the Substance Use Disorder, its prevalence, its treatment, the origin and evolution of clinical records are firstly described. Then, the variables that the specialized literature identifies as relevant for the treatment outcome, and the conceptual modeling of the relationships between these variables are detailed in depth. Once this theoretical framework is established in the introductory chapters, four specific objectives are proposed: (i) to determine how dual pathology affects treatment retention or dropout among patients treated in therapeutic communities; (ii) to analyze the utility of adherence and retention indicators as quantitative or dichotomous predictors of therapeutic success among Substance Use Disorder patients; iii) to identify different profiles of therapeutic progress on Cannabis Use Disorder patients, based on adherence and abstinence indicators, to know the relationship between these profiles with baseline and outcome variables; iv) to analyze which variables are more predictive of the therapeutic success and relapse of Cocaine Use Disorder patients according to the Texas Christian University Treatment Process Model. To contrast these objectives, a retrospective observational design was applied, using the data contained in the electronic health records of the Information System of the Andalusian Plan on Drugs and Addictions. The target population consists of 96,770 patients who started treatment in the net of Andalusian addiction care centers between 2015 and 2019. The main results have shown that: Patients with diagnoses of polydrug use and personality disorders in Cluster B have lower retention and a higher probability of dropping out than patients with other types of dual pathology. - Retention and adherence to therapeutic appointments present greater predictive capacity for therapeutic success when defined as quantitative variables such as months in treatment and proportion of attendance to scheduled appointments, respectively. Using retention as a dichotomous variable with a threshold of three months explains therapeutic success better than using a threshold of six or more months. - Progress profiles are identified and related to patients' baseline characteristics at treatment entry and to their discharge and post-treatment outcomes. The highest abstinence/highest adherence profile shows greater probability of therapeutic success and lower readmission rates, requiring more time in treatment. The lowest abstinence/lowest adherence profile shows a higher risk of treatment dropout. - Regarding the empirical modeling of relationships between variables involved in treatment, limited predictive capacity is observed for the baseline attributes of the patients on the adherence and retention variables. Adherence and time in treatment consistently predict the type of discharge, and time in treatment consistently predicts readmission 24 months after starting treatment. In addition, discharge type consistently predicts readmission. Adequate modeling of the therapeutic process using electronic clinical records requires epistemological positions that support the correct use of such data. Evidence provided in the present research has been grouped under categories of information in a new model that ease treatment planning and therapeutic outcomes assessment. The baseline attributes of patients and the treatment program variables can guide the therapeutic process in the medium and long term, but the focus should be on the relationship between process and outcome categories to achieve or predict therapeutic success.