BACKGROUND: Collaborative care models have been shown to improve mental and physical health, but their effectiveness varies. Implementation science frameworks identify measures at the structural (eg, sociocultural context, public policies), organizational, provider, innovation, and patient levels that may facilitate or impede collaborative care effectiveness. OBJECTIVE: To describe commonalities and variation in multilevel measures associated with the implementation of Care of Mental, Physical, and Substance-Use Syndromes (COMPASS), a large-scale collaborative care intervention for depression, diabetes, and cardiovascular disease. DESIGN: Qualitative study using semistructured descriptive data obtained from annual site visit reports and supplemental site surveys. PARTICIPANTS: COMPASS care teams from 8 health care systems serving 3854 patients with active depression and poorly controlled diabetes and/or cardiovascular disease. INTERVENTION: COMPASS included weekly case reviews with a consulting physician and psychiatrist, a patient-tracking registry, and monitoring of hospital and emergency department use. MAIN MEASURES: Site visit reports were analyzed with Atlas.ti software to qualitatively describe implementation measures and their variation across sites. KEY RESULTS: Nine measures were identified that impacted implementation efforts across health systems: (1) challenges in health systems' organizational environments, (2) prior care coordination experience, (3) physician engagement, (4) care team trust and cohesion, (5) care manager training and experience, (6) patient enrollment length, attainment of clinical targets, and frequency/content of care manager contacts, (7) patient-tracking registries, (8) quality improvement and outcomes monitoring reports, and (9) patients' social needs. CONCLUSIONS: Understanding multilevel measures impacting COMPASS implementation could increase the success of future collaborative care implementation efforts.