Purpose: Although vaginal douching is associated with several adverse outcomes, the reasons why women douche have not been studied prospectively. Methods: Non-pregnant (N = 3620) women aged 15 to 44 years presenting for routine care at 12 clinics in Birmingham, Alabama, participated in a longitudinal study of vaginal flora (1999-2003). Participants were assessed quarterly for 1 year. The authors applied conditional logistic regression in a case-crossover analysis to determine the individual-level factors that vary between a woman's douching and non-douching intervals. Findings were compared to a population-level analysis utilizing generalized estimating equations. Results: Thirty percent of participants douched in every interval; 28% douched in some but not all intervals. The case-crossover analysis indicated a woman was more likely to douche when reporting "fishy" vaginal odor (odds ratio [OR]:2.74; 95% confidence interval [CI]: 1.55, 1.84), vaginal irritation (OR: 1.52; 95% CI: 1.10, 2.11), summer month (OR: 1.37, 95% CI: 1.13, 1.67), or increase in number of sex partners (≥3, OR: 2.42, 95% CI: 1.11, 5.26). Bacterial vaginosis/trichomoniasis treatment (OR: 0.72, 95% CL: 0.59, 0.89) and absent menses (OR: 0.37, 95% CI: 0.28, 0.50) were negatively associated with douching. These ORs were farther from the null than comparable population-level estimates. Conclusions: Programs targeting these predictors may decrease the untoward sequelae associated with douching. Furthermore, a case-crossover analysis applied to prospective studies can provide insights into time-varying factors. © 2008 Elsevier Inc. All rights reserved.