Background Treatment of latent tuberculosis infection (LTBI) in high-risk groups is an effective strategy for TB control and elimination in low incidence settings. A nine-month course of daily isoniazid (INH) has been the longest prescribed therapy; however, completion rates are suboptimal. We need data to guide TB program outreach efforts to optimize LTBI treatment completion rates. Methods We pooled seven (2009–2015) years of LTBI treatment outcome data. We computed the probability of INH treatment disruption over time by patient demographic and clinical risk factors. We used log-rank tests and Cox proportional hazards models to assess the risk factors for treatment disruption. Results We analyzed data from 12,495 persons with complete data on INH treatment initiation. Pediatric cases (0–17 years), recent contacts of active TB patients, and non-U.S.-born adults living in the United States ≤5 years represented 25.2, 13.0, and 59.2% of the study population, respectively. Overall, 48.4% failed to complete therapy. The median treatment duration was 306 days (95% CI: 297, 315). A significant drop in adherence could be observed around day 30 of treatment initiation. Indeed, by day 30 of treatment, 17.0% (95% CI: 16.4, 17.7) of patients had defaulted on therapy. Pediatric patients (HR = 0.83, 95% CI: 0.78, 0.89), recent contacts (HR = 0.74, 95% CI: 0.68, 0.81), patients with diabetes (HR = 0.77, 95% CI: 0.60, 0.98), and patients with HIV (HR = 0.39, 95% CI: 0.30, 0.51) had a lower risk of treatment default. However, black patients (HR = 1.57, 95% CI: 1.44, 1.70), Hispanic patients (HR = 1.54, 95% CI: 1.43, 1.66), and non-U.S.-born persons living in the United States ≤5 years (HR = 1.25, 95% CI: 1.18, 1.32) were significantly more likely to default on therapy. Conclusions In this analysis of INH treatment outcome, we see high levels of treatment discontinuation. On average, patients defaulted on their prescribed nine-month daily INH therapy within 30 days of initiating treatment, and those at increased risk of progression to active disease were most likely to do so. We highlight the need to introduce patient-centered programs to increase treatment adherence in this population. Electronic supplementary material The online version of this article (10.1186/s12889-019-7524-4) contains supplementary material, which is available to authorized users.