Depressive symptoms after spinal cord injury (SCI) are linked with poorer participation and greater health service use, but long-term symptom heterogeneity remains incompletely defined. Objective: To identify model-based depressive symptom trajectories after SCI and examine their associations with 10-year rehospitalization, participation, complications, and related outcomes.

This secondary analysis used the US Spinal Cord Injury Model Systems (SCIMS) public-use dataset, release 2021ARPublic. Participants with a valid Patient Health Questionnaire-9 total score at year 1 and at least 1 additional valid score at years 5, 10, 15, or 20 formed the trajectory cohort ( N = 3116). Group-based trajectory models with 2–6 classes were compared. The primary endpoint was any rehospitalization during the prior 12 months at year 10. Secondary outcomes included participation, life satisfaction, pain, sleep problems, urinary tract infection, pressure ulcer, and falls. Associations were estimated with covariate adjustment and marginal standardization while accounting for posterior class-membership uncertainty.

A 5-class solution was retained: minimal symptoms (11.4%), stable-low symptoms (29.4%), early elevated-resolving symptoms (10.2%), late-elevating symptoms (20.2%), and persistent elevated symptoms (28.8%). Adjusted year-10 rehospitalization probability increased from 20.4% in the minimal-symptoms pattern to 23.5% in the stable-low pattern, 24.7% in the early elevated-resolving pattern, 28.1% in the late-elevating pattern, and 32.3% in the persistent elevated pattern. The persistent elevated and late-elevating patterns were also associated with lower adjusted participation and life satisfaction, greater pain and sleep burden, and higher probabilities of pressure ulcer and urinary tract infection.

Long-term depressive symptom patterns after SCI were associated with clinically meaningful differences in 10-year outcomes. However, because most participants contributed only 2 PHQ-9 observations and late-wave data were sparse, these trajectories, particularly the late-elevating pattern, should be interpreted as probabilistic vulnerability markers rather than stable individual trajectories or causal phenotypes.

This non-interventional secondary analysis of de-identified public-use observational data was not prospectively registered; registration was not mandatory because the study was neither an interventional study nor a systematic review.