Background: Post-diagnosis progression in systemic autoimmune rheumatic diseases (SARDs) is not well characterized across domains. Prior studies often evaluate domains separately or use composite indices, which can miss co-occurring changes. We assessed whether jointly modeling early post-diagnosis trajectories of healthcare utilization, overlapping autoimmune conditions, and comorbidity burden identifies progression patterns associated with mortality.

Methods: In UK Biobank, SARD participants and 1:4 age-, sex-, and landmark date–matched controls (n=16,576; 72% females) were followed for up to six years. Multivariable latent class trajectory modeling jointly characterized longitudinal patterns of SARDs-related healthcare revisits, non-SARD comorbidity accumulation, and comorbidity burden, classifying individuals to trajectory classes. Associations between trajectory classes and all-cause mortality were evaluated using landmark Cox regression. To predict high risk class, we developed prediction models using one and two years post-diagnosis data (elastic net, random forest, XGBoost).

Results: Three trajectory classes were identified. The high-progression class showed steep increases in healthcare revisits and comorbidity burden. Compared with healthy controls, mortality increased stepwise across trajectory classes (HR 1.90 [95% CI 1.51–2.39] for low; 2.51 [1.11–5.67] for moderate; and 4.92 [3.07–7.88] for high). Prediction models yielded moderate discrimination for identifying high-risk trajectory, with XGBoost performing best (AUC up to 0.75 using two-year data). Top predictors included co-occurring SARDs, age at diagnosis, biological age, C-reactive protein, Charlson comorbidity index and socioeconomic deprivation.

Conclusion: Early post-diagnosis disease trajectories among patients with SARDs were strongly associated with mortality. Joint trajectory modeling combined with early prediction may support risk stratification and inform targeted monitoring.