Background
After dramatic declines in teen births and pregnancies from 1991 to 2005, teen birth rates in the United States increased in 2006 and in 2007. We examined behavioral determinants of these trends and the likely direction of future trends.

Methods
Pregnancy risk was estimated based on recent sexual activity, method of contraception used, and method-specific contraceptive efficacy, using data from young women on the national Youth Risk Behavior Survey (N~125,000). Weighted logistic and linear regression were used to test for linear and quadratic (curved) trends over time.

Results
Between 1991 and 2007, behavioral risk for pregnancy declined, with all of the decline occurring between 1991 and 2003. Improvements in contraceptive use from 1991 to 2003 were found in condom use, nonuse, and use of withdrawal. Recent sexual activity (past 3 months) was unchanged over the entire period, except among black students. Quadratic changes were found in pregnancy risk for black teens and in condom use among all teens and black teens, suggesting that trends had reversed or flattened out. While no change was found for any behavior between 2003 and 2007, pregnancy risk among sexually active teens demonstrated a borderline increase (p=.06) and small non-significant declines were seen for specific contraceptive methods. Pregnancy risk estimated from behavioral data correlated well with actual changes in teen pregnancy rates (1991-2004) and birth rates (1991-2006).

Discussion
After improvement in the 1990s and early 2000s, trends in behavioral risk for pregnancy appear to have stalled or even reversed among certain groups since 2003. These behavioral trends are consistent with the 2006 and 2007 increases in the teen birth rate. They may well portend further increases in 2008.