Communication is fundamental for dementia care. The trouble source repair (TSR) framework can identify strategies that facilitate or impede communication in dyadic interactions. A secondary analysis of videos (N = 221) from a clinical trial of a family caregiver telehealth intervention was analyzed using sequential behavioral coding of communication behaviors and breakdowns for 53 caregiver and person with dementia dyads. Coded data from 3,642 30-s observations were analyzed using penalized regression for feature selection followed by Bayesian mixed-effects modeling to identify communication strategies associated with communication breakdown and repair. Breakdown (coded as 0) was associated with caregivers changing topic (median = -11.45, 95% credibility interval [CrI; -24.34, -4.37]), ignoring (median = -11.49, 95% CrI [-24.49, -4.72]), giving commands (median = -10.74, 95% CrI [-24.22, -3.38]), and taking over the task (median = -4.06, 95% CrI [-7.28, -1.77]). Successful repair of breakdown was associated with verbalizing understanding (median = 0.46, 95% CrI [0.09, 0.86]), tag questions, (median = 2.4, 95% CrI [0.33, 5.35]), and silence (median = 0.78, 95% CrI [0.42, 1.15]) and negatively associated with ignoring and changing topic (median = -3.63, 95% CrI [-4.81, -2.57] and -2.51 [-3.78, -1.33], respectively). The TSR was effective in identifying specific communication strategies to avoid (changing topic, ignoring, commands, and taking over the task) and to use to repair breakdown (verbalize understanding, tag questions, and silence). Future research is needed to test these strategies and explore the potential effects of dementia stage, diagnosis, and dyad characteristics in additional samples. Behavioral coding provides evidence of communication best practices as a basis for family caregiver communication training.