The psychological wellbeing of adolescents, including their relationships with their parents and friends, is highly affected by romantic relationship conflict. At-risk youth - who are involved with the juvenile justice system and are especially vulnerable to aggression, defiance, and behavioral problems - experience relationship conflict at rates that exceed those of youth in the community. The pervasive usage of mobile technology by adolescents provides unprecedented data on their social interactions, communication patterns, and romantic relationship conflicts; as well as an opportunity to develop timely interventions to improve their wellbeing. This study investigates the potential of utilizing text messaging metadata from conversations with parental figures, friends, and other social connections for scalable early identification of moments of disagreement in romantic relationships in non-incarcerated, justice-involved youth. We utilize changepoint detection on text message metadata across various relationships, including the number of messages, length of messages, and response time, to identify changes in communication patterns across hours, days, and weeks, to estimate the day of a romantic disagreement. When comparing our predicted disagreement times to the current standard of intervention delivery, at random, we found statistically significant improvements in predicting disagreements with romantic partners. Our findings demonstrate that messaging metadata can effectively identify moments of disagreement in adolescents, and can be used for the delivery of just-in-time interventions for high-risk adolescents.