Statistical techniques for estimating causal effects in biomedical research

Causal inference methods are statistical techniques used to analyse the causal effect of a treatment/exposure on an outcome. Their use is increasing in the last decade, especially in the framework of observational studies where the no randomization of the treatment/exposure may lead to confounding b...

Descripción completa

Detalles Bibliográficos
Autor: Coscia Requena, Claudia
Tipo de recurso: tesis doctoral
Fecha de publicación:2022
País:España
Institución:Universidad Complutense de Madrid (UCM)
Repositorio:Docta Complutense
Idioma:inglés
OAI Identifier:oai:docta.ucm.es:20.500.14352/3739
Acceso en línea:https://hdl.handle.net/20.500.14352/3739
Access Level:acceso abierto
Palabra clave:519.2:61(043.2)
Medical statistics
Estadística médica
Estadística
1209 Estadística
Descripción
Sumario:Causal inference methods are statistical techniques used to analyse the causal effect of a treatment/exposure on an outcome. Their use is increasing in the last decade, especially in the framework of observational studies where the no randomization of the treatment/exposure may lead to confounding bias. These methods present great advantages versus classic regression models due to their capability of reducing and controlling for confounding bias.This thesis begins with the use of known techniques applied in real clinical scenarios, second, a lack of developed statistical methods to estimate causal effects in complex epidemiological scenarios is noted. These findings support the main objective of this thesis, which is the development of causal inference methods to better understand and diagnose clinical and epidemiological outcomes. A comparison between the Propensity Score and classic regression models was made using an Intensive Care Unit database where it was shown that, in presence of confounding bias, Propensity Score performed better. Moreover, based on a systematic review and metaanalysis, causal estimates from Propensity Score and Randomized Controlled Trials were compared. It was observed that similar estimations were obtained in both approaches...