Income-based disparities in health care utilisation under universal health coverage in Brazil, 2002–2003

Abstract

Since Brazil's adoption of universal health care in 1988, the country's health care system has consisted of a mix of private providers and free public providers. We test whether income-based disparities in medical visits and medications remain in Brazil despite universal coverage using a nationally representative sample of over 48,000 households. Additional income is associated with less public sector utilisation and more private sector utilisation, both using simple correlations and regressions controlling for household characteristics and local area fixed effects. Importantly, the increase in private care use is greater than the drop in public care use. Also, income and unmet medical needs are negatively associated. These results suggest that access limitations remain for low-income households despite the availability of free public care.

Keywords:

• universal health care
• Brazil
• family health programme
• disparities
• income

Notes

1. Because SSS reimburses funds to the Ministry of Health when their patients use public facilities and because there is a 27.5% health deduction of the income tax for private expenditures, some researchers consider SSS part of SUS.

2. An alternative survey, the Pesquisa Nacional por Amostra de Domicilios 2008, contains more detailed health information, but it is not suitable for our purposes since it does not allow for separate counts of medical visits (or medicines) based on whether the provision is public or private. Additionally, the PNAD's health system utilisation data have a reference period of only two weeks.

3. For example, surveyed individuals were asked: did you have the need of any medication that you could not get because you did not have the money to pay for it? If yes, the interviewer entered the medication and coded ‘restriction’ under payment method. Underreporting of actual needs is a likely limitation.

4. As robustness checks we also defined public visits in two alternative ways: (1) any visit whose payment method was coded as donation, and (2) any donation whose source of provision was coded as public and whose provider location was coded as SUS. The results were virtually identical and are available upon request. This robustness reflects the fact that virtually all free visits in Brazil are publicly provided by the SUS.

5. Occasionally, medical visits reported by the POF do not fall into either our private or public classifications, so our classification of ‘all’ visits does not include every single POF visit. However, these ‘other’ visits are rare and may in many cases simply reflect reporting error. We therefore elect not to include them when computing ‘all visits’. This exclusion is of little consequence for the sample means, correlations or regression estimates, and results using the broader classification are available upon request. A similar caveat applies to the following discussion of medications.

6. We do not use the POF sampling weights since some of the Stata modules used in the analysis (specifically, the zero-truncated Poisson regressions) do not support them. In unreported regressions (available upon request), we verified that the results from the regressions for which sampling weights are supported remain similar if they are used. This is not surprising, as the POF is nationally representative and does not aim to oversample any groups.

7. The 2008–2009 POF questionnaire on individual expenditures contains the variable payment method but not source of provision or provider location. Hence, publicly provided medicines were identified only through the code that corresponds to donation as payment method, introducing a degree of noise in the variable. We nonetheless obtain similar results as with the 2002–2003 data.