Análisis predictivo de pacientes No-Show mediante un modelo multivariable basado en fuentes de información heterogéneas
Cargando...
Archivos
Fecha
2018
Profesor/a Guía
Facultad/escuela
Idioma
es
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad Andrés Bello
Nombre de Curso
Licencia CC
Licencia CC
Resumen
El gasto en Salud como porcentaje del Producto Interno Bruto (PIB) de los países latinoamericanos es menor en comparación con los países desarrollados de América y Europa. Sólo Cuba (11,8%) y Costa Rica (10,1%) tienen un gasto total en salud que se aproxima al promedio de los países desarrollados. Los países que son parte de la OCDE presentan un promedio de 3.2 médicos por cada mil habitantes, sin embargo, Chile se encuentra por debajo de esta cifra con 1.7 médicos por cada mil habitantes. Este déficit de profesionales de la salud genera una serie de complicaciones en diversos ámbitos, como el aumento de la demanda del servicio y, por consiguiente, la existencia de largas listas de espera para obtener atención médica. El sistema de salud de Chile es un sistema mixto, es decir, está compuesto por un sistema público y un sistema privado. En el sistema público, los establecimientos de salud están clasificados de acuerdo a su nivel de complejidad (primaria, secundaria y terciaria). La atención primaria de salud (APS) tiene como objetivo principal brindar a los pacientes una atención ambulatoria. La APS es la puerta de entrada al sistema de salud pública, es debido a esto que posee una alta cobertura a nivel nacional con alrededor de dos mil centros de atención y cerca de 11 millones de consultas médicas al año.
La atención médica de especialidad es uno de los recursos más escasos, costosos y de difícil acceso del sistema público de salud. Aspectos económicos, sociales, culturales y geográficos son variables fundamentales a la hora de acceder a este tipo de atención, es debido a esto que se puede detectar una problemática asociada al alto nivel de ausentismo a las horas médicas de especialidad, también conocido como pacientes “que no se presentan” o “no-show patient” (NPS) lo que genera pérdidas de entre un 8% y un 20% de las consultas médicas solicitadas.
Diversos estudios nacionales e internacionales han abordado esta problemática destacando la importancia de identificar los factores relacionados con la ocurrencia de un paciente No Show. Debido a esto, se realizó una selección de variables influyentes utilizando criterios de evaluación referentes a la relevancia, acceso y procesamiento de la información. Se construyó una base de datos para almacenar esta información y acceder a ella de forma eficiente. Adicionalmente se realizó una selección preliminar de algoritmos que pudiesen resultar útiles para el diseño de un modelo predictivo, lo que dio paso a la construcción de cinco modelos: Modelo de regresión logística multinomial, modelo de agrupación K-means, Naive Bayes, Árboles de decisión y modelo de clasificación supervisado KNN. De estos se logró determinar que los modelos Naive Bayes y Regresión Logística Multinomial son los que responden de mejor forma a las pruebas tanto con datos simulados como con datos reales. Estos resultados nos permiten dilucidar de mejor forma la real influencia de cada variable frente a un posible caso de NSP y con esto, generar medidas de mitigación para la reducción de las extensas listas de espera en los centros de atención primaria de salud.
Health spending as a percentage of the Gross Domestic Product (GDP) of Latin American countries is lower to the countries of North America and Europe. Only Cuba (11.8%) and Costa Rica (10.1%) have a total health expenditure that approximates to the average these countries. The countries that are part of the OECD have an average of 3.2 per thousand inhabitants, however, Chile is below this figure with 1.7 doctors per thousand inhabitants. This deficit of health professionals presents a series of complications in several areas, such as the increase in demand for health services, consequently, the existence of long waiting lists to obtain medical attention. The health system of Chile is a mixed system, composed of a public system and a private system. In the public system, health facilities are classified according to their level of complexity (primary, secondary and tertiary). Primary health care (PHC) aims to provide to patients with outpatient care. The PHC is the entry point to the public health system; it has a high coverage nationwide with around two thousand care centers and about 11 million medical consultations per year. Medical specialties are the most scarce, expensive and difficult to access resources of the public health system. Economic, social, cultural and geographical aspects are fundamental variables that determinate the access to this type of care. For this reason, it is important to pay attention to the absenteeism associated to the hours of medical specialties. A patient that lose a medical consultation is known as No-Show Patient (NSP), which goes from 8 to 20% of the requested medical consultations. Several national and international studies have addressed this problem by identifying the factors related to the occurrence of a No-Show Patient. This work has considered these studies to present a selection of variables. This selection considered the evaluation criteria, the relevance, access, and feasibility of processing of the information. A database was built to store this information and access it efficiently. In addition, a preliminary selection of algorithms that could be useful for the design of a predictive model was made. Five analysis models were considered for this purpose: multinomial logistic regression model, K-group model means, Naive Bayes, decision trees and KNN supervised classification model. In this way, the search results of the Naive Bayes and Multinomial Logistic Regression models are those that had better respond to the tests with both, the simulated and the real data. These results allow to better determinate the real influence of each variable against a possible case of NSP and with this, generate measures to mitigate the reduction of the long waiting lists in primary health care centers.
Health spending as a percentage of the Gross Domestic Product (GDP) of Latin American countries is lower to the countries of North America and Europe. Only Cuba (11.8%) and Costa Rica (10.1%) have a total health expenditure that approximates to the average these countries. The countries that are part of the OECD have an average of 3.2 per thousand inhabitants, however, Chile is below this figure with 1.7 doctors per thousand inhabitants. This deficit of health professionals presents a series of complications in several areas, such as the increase in demand for health services, consequently, the existence of long waiting lists to obtain medical attention. The health system of Chile is a mixed system, composed of a public system and a private system. In the public system, health facilities are classified according to their level of complexity (primary, secondary and tertiary). Primary health care (PHC) aims to provide to patients with outpatient care. The PHC is the entry point to the public health system; it has a high coverage nationwide with around two thousand care centers and about 11 million medical consultations per year. Medical specialties are the most scarce, expensive and difficult to access resources of the public health system. Economic, social, cultural and geographical aspects are fundamental variables that determinate the access to this type of care. For this reason, it is important to pay attention to the absenteeism associated to the hours of medical specialties. A patient that lose a medical consultation is known as No-Show Patient (NSP), which goes from 8 to 20% of the requested medical consultations. Several national and international studies have addressed this problem by identifying the factors related to the occurrence of a No-Show Patient. This work has considered these studies to present a selection of variables. This selection considered the evaluation criteria, the relevance, access, and feasibility of processing of the information. A database was built to store this information and access it efficiently. In addition, a preliminary selection of algorithms that could be useful for the design of a predictive model was made. Five analysis models were considered for this purpose: multinomial logistic regression model, K-group model means, Naive Bayes, decision trees and KNN supervised classification model. In this way, the search results of the Naive Bayes and Multinomial Logistic Regression models are those that had better respond to the tests with both, the simulated and the real data. These results allow to better determinate the real influence of each variable against a possible case of NSP and with this, generate measures to mitigate the reduction of the long waiting lists in primary health care centers.
Notas
Tesis (Ingeniero Bioinformático)
Este trabajo de Tesis se realizó en el Centro de Bioinformática y Biología Integrativa (CBIB) de la Universidad Andrés Bello y fue financiado por el Proyecto Fondecyt AMoDDI n° 11130583.
Este trabajo de Tesis se realizó en el Centro de Bioinformática y Biología Integrativa (CBIB) de la Universidad Andrés Bello y fue financiado por el Proyecto Fondecyt AMoDDI n° 11130583.
Palabras clave
Salud Pública, Atención al Paciente, Procesamiento de Datos, Chile