A Latent Class Approach to Measuring Adolescent Sexual Behavior

Introduction. Adolescent sexual risk behavior is multifaceted, characterized by behaviors such as number of partners, timing of first sex and condom use. Yet the majority of research on adolescent sexual behavior focuses on a single aspect of sexual behavior at a time. Latent class analysis (LCA) provides an opportunity to measure adolescent sexual risk behavior in a holistic manner by identifying latent subgroups of individuals based on multiple behaviors. LCA can reveal key patterns of co-occurring sexual behaviors, as well as predictors and consequences of the behavior patterns. We use LCA to identify classes of adolescents based on multiple sexual behaviors. We then apply a new, model-based approach to LCA with distal outcomes to examine associations between adolescent sexual behavior patterns and young adult physical, mental and social health outcomes.

Method. Classes were estimated using data from adolescents aged 16-18 at Wave II of the National Longitudinal Study of Adolescent Health. A broad set of indicators were used to identify class membership, including age at first intercourse, use of a condom at first sex, sex with a non-relationship partner, number of recent sexual partners, and timing of first oral sex. Outcomes included sexually transmitted infection in the past year, depressive symptoms, marriage and cohabitation using data from Wave IV, when participants were young adults aged 29-31.

Results. Fit statistics and interpretability suggested a model with five classes: Abstinent (39%), Oral Sex Only (10%), Low-risk Behavior (26%), Multi-partner/Normative Timing (12%) and Multi-partner/Early Timing (13%). Class membership was significantly associated with young adult STIs, depressive symptoms and partner violence (p<.05), and was marginally associated with body image (p=.07).  Gender differences in the associations emerged. For example, men’s risk of STIs increased with membership in more risky classes, yet women’s STI risk remained constant among the Low-risk, Multi-partner/Normative and Multi-partner/Early Timing classes. Also, membership in the Multi-partner/Early Timing class corresponded to the highest mean depressive symptoms, an association that was stronger for women than for men.

Conclusions. The majority of adolescents were in classes marked by relatively low-risk behaviors, although about a quarter were in classes with multiple or non-relationship partners. Young adult health outcomes differed by both gender and class membership, suggesting an interplay of behavior, partner characteristics and socialization may influence outcomes of sexual behavior.  These findings demonstrate the multidimensional nature of sexual behavior and how LCA can be used to understand sexual behavior and its consequences in a more holistic way.