From Assessment to Action: A Research Prototype for SIPAT-Based Multidomain Psychosocial Visualization and Prioritization in Lung Transplant Candidates

Background: Psychosocial assessment is a core component of lung transplant candidate evaluation, but total scores and broad candidate categories do not necessarily show how individual psychosocial concerns co-occur or which modifiable domains require further clinical attention. This study describes the development and internal evaluation of a clinical decision-support application that reorganizes Stanford Integrated Psychosocial Assessment for Transplantation (SIPAT) data into structured multidomain profiles. Methods: The application was developed using a retrospective single-center dataset of 496 adult lung transplant candidates. It integrates the SIPAT total score and candidate category with seven SIPAT-derived domain indicators, domain burden scores, cohort-referenced z-scores, graphical displays, and a ranked summary of domains for clinical review. All seven indicator targets were prespecified deterministic functions of SIPAT items or domain scores obtained during the same assessment. Random forest algorithms with sigmoid calibration were used to transform these targets into percentage-scaled display values; they were not trained using independently assessed clinical outcomes. Analyses included descriptive statistics, Spearman correlations, exploratory clustering, resampling-based cluster stability assessment, threshold sensitivity analyses, and subgroup analyses by age, sex, and primary pulmonary diagnosis. Results: Upper display-band classifications were identified for depression-related concerns in 32 candidates (6.5%), anxiety-related concerns in 11 (2.2%), nicotine-related concerns in 56 (11.3%), alcohol-related concerns in 22 (4.4%), illicit drug-use concerns in 11 (2.2%), social-support deficits in 40 (8.1%), and non-adherence-related concerns in 7 (1.4%). Exploratory clustering yielded a low-burden majority group (n = 432), a nicotine-dominant group (n = 53), and a small multidomain-elevation group (n = 11). The generated percentage-scaled indicators were positively associated with their conceptually corresponding SIPAT domains (Spearman’s ρ = 0.321–0.774) and with the total SIPAT score (ρ = 0.406–0.802; all p < 0.001). Sensitivity analyses showed that the smaller clusters were less stable under bootstrap resampling. These findings demonstrate internal alignment with the source instrument but do not constitute validation against independent clinical outcomes. Conclusions: The application provides an early-stage framework for organizing and visualizing SIPAT information and identifying domains that may warrant additional clinical assessment. Its outputs should be interpreted as SIPAT-derived decision-support indicators, not as independently validated probabilities of future clinical events. Prospective studies are required to evaluate usability, clinical impact, and associations with longitudinal outcomes.