Abstract

Background/Objectives: Adolescent pregnancy remains a global public health challenge associated with adverse outcomes for both mothers and newborns. This study aimed to use two instruments and several multivariate techniques to identify a minimal set of variables that reproduces the instrument-based classification of adolescent pregnancy risk, while remaining parsimonious with the original instruments. Methods: A cross-sectional, quantitative methodological study was conducted among Colombian schoolgirls, comprising 160 adolescents in the face-validity phase and 319 in the risk-estimation and modeling phase. The IRENE and REND instruments, originally developed to assess the risk of adolescent pregnancy, underwent cultural adaptation, face validity and content validity. Subsequently, the instruments were administered to estimate the risk. Finally, Factor Analysis and Categorical Principal Component Analysis were applied as exploratory dimensionality reduction techniques to identify the most relevant variables for retention, thereby preserving the parsimony of the original versions. Results: Overall, 80.3% of participants were classified as not at risk, while the remainder were classified as at risk by one or both instruments. The results suggest that variables such as who the adolescent lives with, the age at which the adolescent had their first complete sexual intercourse, and how the adolescent would respond to an unplanned pregnancy are factors associated with the instrument-based risk classification. Subsequently, dimensionality reduction and logistic regression analyses identified a small subset of variables that can be used to reproduce the instrument-based classification of adolescent pregnancy risk. Conclusions: The reduced set of variables reproduced the instrument-based risk classification with high internal accuracy. Because the predictors and the outcome derive from the same instruments, these findings reflect internal reproduction of the instrument classification rather than prediction of an observed pregnancy, and they require external validation. Overall, the reduced and optimized set of variables from the IRENE and REND instruments offers a parsimonious approach that reproduces the instrument-based risk classification and that may support rapid screening once validated in independent adolescent samples. The main limitations are the use of a single institution, the non-probabilistic convenience sample, and the absence of an observed pregnancy outcome.

Keywords