Network-guided symptom targets in maintenance hemodialysis using in silico interventions: a multicenter cross-sectional study.

backgroundPatients receiving maintenance hemodialysis (MHD) experience multiple concurrent symptoms with substantial heterogeneity in symptom burden. However, a reproducible evidence framework to translate symptom interdependencies into intervention priorities is lacking. To address this gap, we use model-based in silico perturbation simulations to quantitatively rank modifiable symptom targets, providing non-causal decision support for symptom-focused management in resource-constrained settings.

methodsWe used convenience sampling to recruit adults receiving MHD from 13 hemodialysis centers in Southwest China between February and June 2025. Symptoms were assessed using the modified Dialysis Symptom Index. We identified symptom clusters using exploratory factor analysis and estimated symptom networks to characterize interconnections and centrality. Latent profile analysis defined symptom-burden subgroups. To explore interventions, we performed in silico perturbation analyses with NodeIdentifyR, simulating both alleviating and aggravating interventions. Changes in the total number of active symptoms served as the primary outcome for comparing effects on symptom network structure.

resultsA total of 962 participants were included. Seven symptom clusters were identified: emotional, sexual and cardiopulmonary dysfunction; gastrointestinal; musculoskeletal; sleep disturbance; neurologic; and skin discomfort. Two symptom-burden profiles were identified by latent profile analysis, representing Profile 1 (lower-burden groups) and Profile 2 (high-burden group), with clear clinical interpretability. Feeling nervous showed the highest centrality, and fatigue demonstrated the strongest bridging role across clusters. Model-based perturbation analyses ranked candidate priorities. Simulated alleviation of feeling anxious, worrying, and feeling nervous was associated with estimated reductions in overall burden of 17.3%, 16.0%, and 14.4%, respectively. Simulated exacerbations of feeling irritable, nervous, and sad were associated with estimated increases in overall burden of 21.2%, 19.4%, and 18.5%, respectively.

conclusionsOur findings suggest emotional symptoms are priority candidates for intervention in patients receiving MHD. They may provide non-causal decision support for symptom management and resource allocation. Future longitudinal studies and intervention trials are needed to further evaluate both their effectiveness and generalizability.