OBJECTIVE: To assess whether multi-year engagement by primary care practices in a pay-for-value program was associated with improved care for high-need patients. STUDY DESIGN: Longitudinal cohort study of 17,443 patients with 2 or more conditions who were assigned to primary care providers (PCPs) within 1582 practices that did and did not continuously participate in Blue Cross Blue Shield of Michigan's pay-for-value program (the Physician Group Incentive Program [PGIP]) between 2010 and 2013. METHODS: We used generalized linear mixed models, with patient-level random effects, to assess the relationship between whether practices continuously participated in PGIP and those practices' cost, use, and quality outcomes (derived from claims data) over a 4-year period. For most outcomes, models estimated the odds of any cost and utilization, as well as the amount of cost and utilization contingent on having any. RESULTS: High-need patients whose PCPs continuously participated in PGIP had lower odds of 30-and 90-day readmissions (odds ratio [OR], 0.65 and 0.63, respectively; P <.01 for both) over time compared with patients with PCPs who did not continuously participate. They also appeared to have lower odds of any emergency department visits (OR, 0.88; P <.01) and receive higher overall quality (1.6% higher; P<.01), as well as medication management-specific quality (3.0% higher; P <.01). We observed no differences in overall medical-surgical cost. CONCLUSIONS: Continuous PCP participation in a pay-forvalue program was associated with lower use and improved quality over time, but not lower costs, for high-need patients. National policy efforts to engage PCPs in pay-for-value reimbursement is therefore likely to achieve some intended outcomes but may not be sufficient to deliver care that is of substantially higher value.