Article Text

Download PDFPDF
P89 Structural equation modelling with latent variables to study the relationship between health status, access to health care and socieconomic status in Chile
  1. VM Pedrero1,
  2. M Oyarte1,
  3. B Cabieses1,
  4. P Zitko2
  1. 1Facultad de Medicina, Universidad del Desarrollo, Santiago, Chile
  2. 2Unidad de Estudios, Complejo Asistencial Barros Luco Trudeau, Santiago, Chile


Background Social inequalities in health are a complex and multidimensional phenomena. There is a great deal of evidence about the relationship between the perception of people’s health, their social position within society and access to healthcare services. However, many of these studies do not consider that these variables are not directly observed (i.e. latent variables) and are collected with a variable degree of error. Structural equation modelling (SEM) with latent variables is a multivariate method that allows researchers to isolate the error of measurement and test of mediation and moderation relations that are useful in this context. This study aimed at demonstrating the applicability of SEM for the understanding of social inequalities in health using data from an anonymous national representative survey in Chile.

Methods The national socioeconomic characterisation survey (CASEN) corresponds to one of the main tools for decision-making for social policies in Chile. The main objective of CASEN is to know the socioeconomic situation of households in Chile (demographic aspects, income, health, housing and work) and to assess the impact of social policies in Chile. Using data from the household heads in CASEN 2013 (n = 66.677), a SEM model was fitted to evaluate the relationship between latent variables: people’s health state (10 variables), access to healthcare services (4 variables) and socioeconomic position (9 variables). Afterwards, the model was adjusted using control variables not included as latent like educational level, employment, rurality, age and sex.

Results The SEM adjusted appropriately to the data (RMESA = 0.02; CFI = 0.97; TLI = 0.97). In this model the people’s health status is more influenced by access to healthcare (standarized coeficient (βstd) = 0.32, p < 0.01) that the household income (βstd = 0.07; p < 0.01). Income in turn influences significantly on access to healthcare (βstd = 0.17; p < 0.01). This situation suggests a significant mediation of income on population self-reported health status. The direct effect of educational attainment on health is not significant (βstd = 0.02; p > 0.05); however, the (indirect) mediated effect (income, access) and the total effect (direct and indirect) was large and significant (βstd = 0.11; p < 0.01). The lack of influence of the geographical area on the global access to health services is also striking (βstd = 0.04; p > 0.05).

Conclusion This study confirmed a strong relationship between income, health status and access to healthcare in Chile. Also showed the importance of the effects of mediation in these relationships. The use of SEM is a useful method to explore these phenomena.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.