Abstract
OBJECTIVES: To identify and analyze the patterns of change in Oral Health-Related Quality of Life (OHRQoL) from late adolescence to early adulthood (ages 17 to 23) using a machine learning algorithm. METHODS: This longitudinal trajectory study used data from the Iowa Fluoride Study (IFS). Participants were recruited at birth from eight Iowa hospitals between 1992 and 1995. OHRQoL was assessed at ages 17, 19, and 23 using three validated questionnaires: the Child Perceptions Questionnaire (CPQ(11-14)), Global Oral Health Rating (GOHR), and Visual Analog Scale of Quality of Life (VisQoL). Of the 437 individuals assessed at age 17, 402 were re-assessed at age 19 and 367 at age 23 (91% retention rate). The K-Means for Longitudinal data (KmL) algorithm was applied to identify distinct trajectory groups for each measure. Associations between trajectory group membership and sociodemographic variables were examined using logistic regression. All analyses were performed in R (version 4.1.3). RESULTS: Two distinct trajectory groups were identified for each OHRQoL measure, representing consistently better versus persistently worsening oral health experiences. For the CPQ(11-14), 84.8% of participants were in the "Favorable" group and 15.2% in the "Unfavorable" group. GOHR classified 57.2% as "Favorable" and 42.8% as "Unfavorable," while VisQoL showed 67.5% and 32.5%, respectively. Despite differing proportions, all instruments reflected similar directional trends. Higher socioeconomic status was associated with favorable trajectory group membership (p < 0.05). CONCLUSION: Most participants followed a favorable OHRQoL trajectory and were from higher socioeconomic backgrounds. These findings highlight the value of longitudinal, multi-measure approaches in identifying at-risk subgroups.