Article

¿Odds ratio o razón de proporciones? Su utilización en estudios transversales

Gaceta Sanitaria (Impact Factor: 1.25). 01/2003; DOI: 10.1157/13043426
Source: DOAJ

ABSTRACT Antecedentes: En los estudios transversales las medidas de asociación clásicamente descritas son la razón de odds (odds ratio, OR) y la razón de prevalencias (prevalence ratio, PR). Algunos estudios epidemiológicos con diseño transversal expresan sus resultados en forma de OR, pero utilizan la definición de PR. El objetivo principal de este trabajo es describir y comparar diferentes métodos de cálculo de la PR discutidos en la bibliografía reciente en dos escenarios (prevalencia < 20% y prevalencia > 20%). Material y métodos: Se realizó una búsqueda bibliográfica para conocer las técnicas más utilizadas para la estimación de la PR. Los 4 procedimientos más empleados fueron: a) seguir obteniendo OR mediante regresión logística no condicional, pero utilizando su definición correcta de OR; b) utilizar una regresión de Breslow-Cox; c) utilizar un modelo lineal generalizado con la transformación logaritmo y familia binomial, y d) utilizar una fórmula de conversión de una OR, obtenida mediante regresión logística tradicional, a una PR. Se han replicado para cada uno de los dos escenarios (prevalencia < 20% y prevalencia > 20%) los modelos hallados utilizando datos reales de la Encuesta de Salud de Catalunya de 1994. Resultados: No se observan grandes diferencias entre las estimaciones ni entre los errores estándar obtenidos al utilizar una u otra técnica cuando la prevalencia es baja. Cuando la prevalencia es alta existen diferencias entre los estimadores y entre los intervalos de confianza, aunque todas las medidas mantienen la significación estadística. Conclusión: Todos los métodos propuestos tienen sus pros y sus contras, y debe ser el propio investigador/a quien escoja la técnica que mejor se adapte a sus datos y ser coherente a la hora de utilizar un estimador y su interpretación.

Full-text

Available from: Esteve Fernández, Apr 25, 2015
0 Followers
 · 
192 Views
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Objective To compare the magnitude of inequalities in the frequency of physical inactivity in Spain in 1987 and 2007, and assess whether the magnitude of inequality varies depending on the wealth of the area of residence. Design Descriptive cross-sectional study, national scope. Participants Data from the National Health Survey, 1987 and 2007, adult population between 25-64 years: 30,000 individuals (1987) and 29,478 (2006/7). Main measurements Main outcomes variable, leisure-time physical inactivity; exposure factor, educational level. An analysis was made of the prevalence and association using odds ratio (OR). Adjustment for socioeconomic variables: age, marital status, employment status, social class of head of household, and household income. Results Physical inactivity prevalence decreased in the two decades. There were more than three times more inactive males among those with primary education or less, compared to those with university studies. The magnitude of inequalities has increased in time: in a 45-64 years old man with primary school education: OR 2.43 (1.91-3.09) in 1987, to OR 2.77 (2.17-3.54) in 2007, adjusted for all socioeconomic variables. Conclusions The prevalence of physical inactivity decreased between 1987 and 2007, and the largest decreases were in individuals with university studies. The gap in the differences in prevalence and OR of leisure-time physical inactivity has increased over time. It's necessary to contribute, with health education strategies and equity promotion are needed to help reduce the inequalities in risk behaviors.
    Atención Primaria 12/2014; 46(10). DOI:10.1016/j.aprim.2014.02.010 · 0.89 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Objective To analyze the perception of the Spanish population of risk factors for cancer. Methods Data were extracted from the OncoBarometro 2010 survey. Multivariate logistic models were applied to analyze the perception of the population on the importance of various risk factors: smoking, alcohol, sun, food, weight, sexually transmitted diseases, family history, radiation exposure, exposure to toxic substances and air pollution. The answers were rated on a 0 to 10 scale and were converted to low (0-6) and high (7-10) categories. The measure of association used was the prevalence ratio (PR). Results The greatest importance was assigned to smoking (high importance: 83.1%), whereas the least importance was assigned to weight (26.5%). In general, the probability of perceiving risk factors as important was lower among men (PR sun: 0.87; PR sexually transmitted diseases: 0.78) and increased among people who received professional advice on cancer prevention (PR alcohol: 1.11; PR sun: 1.18; PR food; 1.31; PR weight: 1.92). In particular, knowledge of symptoms and extreme fear of cancer were associated with perceiving smoking as an important risk factor, whereas a high perceived vulnerability to cancer was associated with perceiving exposure to toxic substances, pollution and smoking as important risk factors. Conclusions Greater awareness is required of the association of cancer with overweight and sexually transmitted diseases. The recommendations given by health professionals on cancer prevention are key to increasing the population's awareness of risk factors for cancer.
    Gaceta Sanitaria 01/2013; DOI:10.1016/j.gaceta.2013.10.008 · 1.25 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Objectives: Identify strong predictors of smoke consumption by teenagers of high school Santiago of Cali. Materials and methods: Cross-sectional study with a sample of 286 adolescents enrolled and assets from 6 to 9 degree in school year 2005. From the prevalence of smoke consumption was estimated to differences with the independent variables with Ji2 and P values. The predictive factors were established with a logistic model not conditional upon assessment of confusion and colineality. Results: The prevalence of consumption was 40,5% [CI: 34,8-46,4]. In all three models, consumption of friends had an OR 2,31, 95%[CI: 1,06-5,04]; smoking if a friend is what gives was OR 7,45, 95% [CI 3,9-14,0] and exposure to cigarette smoke explain consumption with an OR 2,18 95% [CI:1,23-3,87]; Hosmer Lemeshow Ji2 Test =10,60 and P Value =0,2253. Conclusions: In this group of teenagers just as friends were possible predictors of the consumption of tobacco, although there was some consumption by parents was not statistically significant, if it was for exposure to smoke outside the home. There were no associations with alcohol or psychoactive substances.
    Salud Uninorte 06/2009; 25(1):56-72.