TRADITIONALLY cross-sectional studies examining the prevalence of human immunodeficiency virus (HIV) and sexually transmitted diseases (STDs) among sex workers have used workplace as a variable to proxy socioeconomic status.1–3 As a result, when developing a survey instrument for a multicenter repeated cross-sectional study of sex workers in Spain, we included a similar closed question (i.e., In the past year where have you worked for the majority of the time?).4,5 However, during the course of the study, a number of the interviewers expressed concerns with this question. For example, it was believed that a number of women who worked from the street were actually better off than colleagues working in hotels. Therefore, in the final two (of five) surveys, we added a new closed question: What is the minimum amount you earned for your service? This therefore provided the opportunity to test the hypothesis that income has greater classificatory power in predicting HIV and other STDs than workplace. Methods of recruitment, interview techniques, and analysis are reported in detail elsewhere.4 In total, information on 1,668 sex workers working in 29 provinces in Spain were collected as part of a cross-sectional study conducted by 22 collaborating center in five 2-month periods between November 1989 and January 1991.5 Data were collected on sociodemographic information, sexual behavior, and health status. These included information on reproductive and sexual histories and possible known risk factors for HIV infection. Data on the 555 sex workers that were interviewed in the final two sections of the study were selected (i.e., 33.3% of the total sample). Logistic regressions were computed with HIV, syphilis, and hepatitis as the dependent variables and workplace or income as the independent variables, while controlling for intravenous drug use (IDU). To examine the ability of the various models in classifying individuals correctly into either a positive or negative serological result two parameters were assessed: (1) the sensitivity is the proportion of true-positives that are correctly identified; and (2) the specificity is the proportion of truenegatives that are correctly identified.6 The independent variable that has the highest rate of sensitivity for a given model would therefore demonstrate the greater ability in classifying a serological-positive women. As illustrated in Table 1, there is no apparent difference in the classificatory power of income over workplace. In the HIV model controlling for IDU, sensitivity was 76.32 % with the inclusion of the workplace variable and 71.05 % for income. In other words both models correctly predicted approximately three quarters of HIV-positive women. Likewise, for hepatitis B only 25.56 % of cases were correctly predicted by workplace, and only 30.00 % by income. Furthermore, as highlighted by the 0.00/100.00 sensitivity/specificity rate, neither workplace nor income possessed the classificatory power to predict past or present serological markers for syphilis.TABLE 1: The Sensitivity and Specificity of Workplace and Income in Predicting HIV, Hepatitis B, and Syphilis Status (While Controlling for Intravenous Drug Use)On the basis of the data presented, the hypothesis that income has greater classificatory power in predicting HIV and other STDs than workplace has to be rejected: Both variables demonstrate approximately the same predictive ability. Furthermore, the low sensitivity indicates that workplace and income (and IDU) are poor determinants of hepatitis B and syphilis. One possible explanation for the similar predicative ability of workplace and income is that the variables themselves are associated. Indeed this is the case: There is a highly significant association between income and workplace (Pearson chi-square, χp2 = 130.75, which is significant at p = 0.0001 level), and this association is linear (Mantel-Haenszel test for linear association, χMH2 = 6.51, which is significant at p = 0.05). Moreover, 71.6 % of women who worked from the street earned between 2,001 to 5,000 pesetas (≈$16 to $40) per sexual service, compared with 33.3% for hotel sex workers; whereas only 1.5% of street sex workers earned between 10,001 to 25,000 pesetas (≈$80 to $200), in comparison with 24.2% of women who worked from hotels. Nevertheless, it is still important to consider the observations made by the interviewers during this study, namely, that the socioeconomic status of sex workers is both difficult to measure and heterogeneous. Therefore, although a negative result such as this may question the validity of including the income variable in the final two stages of the study, we would argue that the qualitative feedback of the interviewers made this justifiable. Furthermore, we have been able to assess scientifically this observation, and in this case we have demonstrated it to be false.