Context: Measures of contraceptive use at one point in time do not account for its changing nature. A measure that addresses the pattern of method use over time may better predict the cumulative risk of unintended pregnancy. Methods: Women at risk of unintended pregnancy were selected from the 1995 National Survey of Family Growth, and their contraceptive use patterns were compared across age-groups. Survival analysis was used to validate women's long-term use pattern as an indicator of pregnancy risk, and multivariate regression analyses were used to explore potential covariates of current patterns of contraceptive use. Results: More than two-thirds of women aged 15-19 report long-term uninterrupted contraceptive use, but they are more likely to report sporadic use and less likely to report uninterrupted use of a very effective method than are women aged 25-34. Compared with women aged 25-34, women aged 20-24 have higher rates of sporadic use and lower rates of effective uninterrupted use. Among teenagers, nonusers are 12 times as likely as uninterrupted effective users to experience an unintended pregnancy within 12 months at risk. Women in less stable relationships, those having more infrequent intercourse and women who have recently experienced nonvoluntary intercourse for the first time are more likely than others to have a high-risk contraceptive pattern. Women aged 17 and younger whose current partner is more than three years older are significantly less likely to practice contraception than are their peers whose partner is closer in age. Conclusions: Long-term contraceptive use pattern is a valid predictor of unintended pregnancy risk. Policies aimed at reducing unintended pregnancies should target women who do not practice contraception and those who are sporadic users. Women in unstable relationships, those having infrequent sex and women who experience sexual coercion need access to methods, such as emergency contraception, that can be used sporadically or after unprotected intercourse.