Objective: To describe the course of suicide ideation (SI) in primary-care based late-life depression treatment, identify predictors of SI, characterize the dynamic relationship between depression and SI, and test the hypothesis that collaborative care decreases the likelihood of reporting SI by decreasing the severity of depressive symptoms. Methods: This was a secondary analysis of a randomized controlled trial comparing collaborative care to usual care for late-life depression. Participants were 1,801 adults age 60 and older from eight diverse primary-care systems. Depression was measured using the Hopkins Symptoms Checklist (HSCL-20). SI was operation- alized using one item from the HSCL-20. Predictors of incident SI were identified by a series of univariate analyses followed by multiple logistic regression. A mediator analysis was conducted to test the hypothesis that the effect of collaborative care on SI can be ascribed to the intervention’s effect on depressive symptoms. Results: The prevalence of SI was 14% (N􏰥253); the cumulative incidence over 24 months was 21% (385). The likelihood that SI emerged after baseline was highly dependent on change in depression (odds ratio: 5.38, 95% confidence interval: 3.93–7.36, df􏰥81, t􏰥10.66, p 􏰦0.0001). As hypothesized, the effect of collaborative care on SI was mediated by the treatment’s effect on depression. Conclusion: SI is not uncommon in depressed older adults being treated in primary care. The likelihood that depressed older adults will report SI is strongly determined by the course of their depression symptoms. Providers should monitor SI throughout the course of depression treat- ment. (Am J Geriatr Psychiatry 2007; 15:1024–1033)