This paper tests for the existence of adverse selection in the Brazilian individual health insurance market in 2003. The testing approach adapts that conceived by Chiappori and Salanié (Eur Econ Rev 41, 943–950, 1997; J Polit Econ 108, 56–78, 2000). After controlling for sex, age, income, number of dependents, occupational groups and schooling levels, the evidence favors adverse selection as indicated by a positive correlation between the coverage of the contract and occurrence of illnesses (as approximated by hospitalization) was not strong. The consideration of complex sampling in the probit estimations led to empirical evidence that does not indicate the presence of adverse selection, but which highlighted some interesting features of the relationship between the selected variables.