Health systems around the world are aiming to improve the integration of health and social care services to deliver better care for patients. Existing evaluations have focused exclusively on the impact of care integration on health outcomes and found little effect. That suggests the need to take a step back and ask whether integrated care programmes actually lead to greater clinical integration of care and indeed whether greater integration is associated with improved health outcomes. We propose a mediation analysis approach to address these two fundamental questions when evaluating integrated care programmes. We illustrate our approach by re-examining the impact of an English integrated care program on clinical integration and assessing whether greater integration is causally associated with fewer admissions for ambulatory care sensitive conditions. We measure clinical integration using a concentration index of outpatient referrals at the general practice level. While we find that the scheme increased integration of primary and secondary care, clinical integration did not mediate a decrease in unplanned hospital admissions. Our analysis emphasizes the need to better understand the hypothesized causal impact of integration on health outcomes and demonstrates how mediation analysis can inform future evaluations and program design.