Objective:To explore reliable and efficient sampling methods and corresponding formulae for quantitative sensitive question survey on three－stage random sampling. To provide scientific data for the prevention and control of high risk AIDS population. Methods:We derived statistical formulae from mathematical statistics theory;investigated MSMs in Beijing for instance;used SAS programming to simulate investigation and analysis of 100 samples through Monte Carlo simulation and accessed validity and reliability of methods and formulae studied in this article. Results:We derived statistical formulae of the overall mean estimator and its variance on quantitative sensitive questions under stratified three－stage sampling. The means of ages of the first occurrence of MSM,the monthly number of different sexual partners and the monthly number of sex were 21.96 years old,2.80 people,4.85 times,respectively,among MSM in Beijing,and their standard errors were 0.127 years old,0.096 people,0.559 times,respectively. All of the population mean of 95% confidence interval obtained from 100 analog samples contains population mean. Conclusion:For high－risk sexual behavior among MSM with AIDS,its prevention and control should be strengthened. Monte Carlo computer simulation results show that stratified three－stage sampling method and corresponding formule are feasiblae and could be applied in the sensitive issue of large－scale sampling in the future.