Horizontal inequity in the use and access to health care in Uruguay

This index does not control for potential endogeneity between the needs for health services and the medical care received, which may be due, for example, to the contemporary measurement of both variables; neither between income and the use of health services, due to both the simultaneity in the measurement and the existence of omitted need-variables that affect the use of services and may be correlated with income [3, 4]. Only works that use longitudinal data overcome these problems

Standard errors are calculated through bootstrapping with 400 replications (StataCorp, 2014).

The ENS was based on a randomized, stratified, and multi-stage sample design (three stages in Montevideo and four stages in the rest of the country), The sampling frame was the 2011 Census. More details in “First Report of National Health Survey (2016)” in https://​www.​gub.​uy/​ministerio-salud-publica/​datos-y-estadisticas/​datos/​encuesta-nacional-de-salud

Continuous Household Survey is a national survey with a long tradition of confiability, performed by the National Statistics Institute. Income information from this Survey is using to calculate official statistics about poverty and income inequality.

As a robustness check, another strategy for imputation of income was carried out. The method performs multiple stochastic imputations, using information from the variable itself as well as the relationship between different ENS variables. This method assumes that the probability that the data is missing does not depend on unobservable characteristics (they are “missing at random”). The missing data was imputed 10 times to give variability of the imputation. An algorithm was used based on the so-called chained equations (Stata software), which performs a sequence of univariate imputations with specific models for each variable (“Full 6 Conditional Specifications”). Therefore, a multivariate imputation is carried out, using the conditional density of the observed data for each variable. The results do not vary significantly from those presented in this paper and are available by request.

The five original options of health status were grouped into three categories due to the low frequency of some of these options.

Diseases considered were renal failure, heart problems, hyper or hypothyroidism, respiratory diseases (emphysema or asthma), bone-muscle diseases (arthritis / osteoarthritis, rheumatism, osteoporosis, tendinitis and spinal problems); biological risk factors considered were overweight or obesity, hypertension, high cholesterol and diabetes.

Limitations considered were: blindness, difficulties to see even with glasses, deafness, hearing difficulties even using hearing aids, speech difficulties, inability or difficulty to move that requires the permanent use of a wheelchair, impossibility or difficulty to move that requires the use of a cane or crutches permanently, limitations for using hands and arms, limitations for moving outside the house or using means of transport, limitations for moving inside the house, mental limitations that make learning and application of knowledge and task development difficult; mental limitations that make it difficult to interact with others.

The ENS allows for a partial approximation to these behaviors, since it does not release information necessary to establish the risk levels. In this sense, the binary variables are defined: alcohol, which takes the value 1 if the person consumed alcohol three or more times per week, or if he got drunk several times, in the last 30 days; sedentary, which takes the value 1 if the person declares that he is sitting for most of the day and at the same time does not usually perform at least 10 minutes of physical activity; smoke, takes the value 1 if the person currently smokes every day. As for eating habits, the ENS collects the frequency of habitual consumption of a set of foods (not the amount consumed), allowing to specify the binary variable unhealthy food that takes the value 1 if the person consumes fruits or vegetables 1 or 2 times per week (or less), if eats snacks (chips, cookies), “fast” food or drinks with sugar 3 times per week (or more).

Double coverage refers to people who declare have public and private health coverage.

More details in “First Report of National Health Survey (2016)” in https://​www.​gub.​uy/​ministerio-salud-publica/​datos-y-estadisticas/​datos/​encuesta-nacional-de-salud

The results are available by request.

In O'Donnell et al. [12] is an exhaustive detail of the definitions of equity in health and its implementation through the use of data from household surveys in Stata.

There is a debate about the use of health care variables as dependent variables since it is binary variables. Although non-linear models are recommended in these cases, there is also evidence of similar results using linear and non-linear models [1]. We calculate the linear approximation to the nonlinear model for all estimates since this allows the decomposition analysis.

This index does not control for potential endogeneity between the needs for health services and the medical care received, which may be due, for example, to the contemporary measurement of both variables; neither between income and the use of health services, due to both the simultaneity in the measurement and the existence of omitted need-variables that affect the use of services and may be correlated with income [3, 4]. Only works that use longitudinal data overcome these problems

Standard errors are calculated through bootstrapping with 400 replications (StataCorp, 2014).