Machine learning analysis of four cardiac surgery cohorts revealed that patients with multiple long-term conditions segregate into two distinct immune phenotypes with dramatically different surgical outcomes. Cluster 2 patients showed dysregulated tissue-resident macrophages, senescent T cells, and cardiomyocyte dedifferentiation, experiencing significantly higher rates of post-surgical organ injury compared to Cluster 1 patients who maintained immune system activation and healthier cardiac tissue profiles. Single-cell RNA sequencing of atrial biopsies confirmed these molecular signatures, with genetic variants affecting the differentially expressed genes altering 90-day mortality risk in UK Biobank data. This research transforms our understanding of surgical risk stratification by identifying immune homeostasis as the critical determinant of outcomes in multimorbid patients. The findings suggest that pre-surgical immune profiling could revolutionize patient selection and risk assessment, while targeting macrophage dysfunction and immunological aging may offer new therapeutic avenues for organ protection. However, as a preprint awaiting peer review, these promising results require validation through independent replication studies. The work represents a significant advance in precision medicine for cardiac surgery, potentially enabling personalized interventions to reduce mortality in our aging, increasingly multimorbid population.