Using patient-reported measures to predict hospitalisation in a population-based lupus cohort.

Abstract

Objective: SLE is a multisystem autoimmune disease where periods of disease activity, often difficult to predict, can cause irreversible disease damage. This study aimed to develop a patient-centric predictive model using real-world data that can identify patients with SLE at a higher risk of hospitalisation compared with the general SLE population.

Methods: This observational, retrospective analysis used data from the Georgians Organized Against Lupus (GOAL) cohort from 2011 to 2013. The GOAL cohort is a population-based SLE cohort that collects yearly self-report surveys covering participants' sociodemographic characteristics, clinical characteristics and perceived SLE symptoms (using the Systemic Lupus Activity Questionnaire (SLAQ)). GOAL data were linked to the Georgia Hospital Discharge Database to collect participants' all-cause hospitalisation events in the 6 months following survey completion. A two-step approach was used to predict all-cause hospitalisations-logistic regressions selected a list of GOAL predictors that were subsequently included in the classification and regression tree (CART) models to generate patient subsets based on estimated hospitalisation rates.

Results: There were 846 participants who completed 1486 surveys. Participants who were hospitalised within 6 months after survey completion were more likely to be younger, living in poverty and have more reported SLE symptoms than participants without a hospitalisation. CART modelling identified participants who reported any weight loss without trying, severe fatigue and Raynaud's symptoms as most likely to have an all-cause hospitalisation: one in three (34%) patients in this subset were hospitalised in the 6 months following survey completion, 2.6-fold the hospitalisation rates of the overall GOAL cohort (13%) and 6.8-fold the rate in the subset with the lowest hospitalisation rate (5%).

Conclusions: This study suggests that patient-reported SLE symptoms and disease activity, specifically certain components of the SLAQ, may be of value in SLE risk management when considering hospitalisation reduction as a treatment goal.