Background Infection remains an important and leading cause of mortality in lupus. We aimed to establish a clinical prediction model for the 3 month all-cause mortality of patients with systemic lupus erythematosus (SLE) and invasive infection in the emergency department.

Methods SLE patients classified by 1997 ACR criteria complicating with invasive infection when admitted into the emergency department of our center between May 2015 and June 2018 were included. Invasive infection was defined as deep infection with definite evidence of pathogens or globally judged by the treating physician combining symptoms, lab and imaging tests. Patients clinical and laboratory characteristics at admission were retrospectively collected as baseline data. The outcome was all-cause death within 3 months since baseline.

Candidate predictors for multivariable logistic regression were selected by expert opinion based on clinical significance, previous studies and feasibility. A prediction model for all-cause mortality was established by combining independent predictors and evaluated by Receiver Operating Characteristic (ROC) curve analysis.

Results A total of 50/130 (38.5%) included patients died cumulatively within 3 month follow-up. Patients were predominantly female (91%) with a mean disease duration of 6.6 years. One hundred and eight (83.1%) patients had lung infection, while 23 (17.7%) patients had blood stream infection.

In the final multivariable logistic regression model, lymphocyte count <800/ul, urea >7.6 mmol/L, maximum prednisone dose in the past60 mg/d, qSOFA score and age at admission were independently predictive for all-cause mortality. However, the history of hydroxychloroquine use was protective.

In a combined prediction model, the six predictors were weighted by OR values, making the LUPHAS score ranging from 5 to 22 (table). All patients could be categorized to three group: low-risk (score 5–10), medium-risk (score 11–16) and high-risk (score 17–22). The mortalities were 11.1% (4/36), 32.8% (22/67) and 100% (22/22) in low-risk, medium-risk and high-risk patients, respectively. ROC curve analysis indicated that LUPHAS score could effectively predict all-cause mortality in this population (AUC=0.8595%CI=0.78–0.92). Furthermore, LUPHAS score performed better than the sole qSOFA score (AUC=0.6995%CI=0.59–0.78) in our cohort. The discriminatory performance of LUPHAS score was also superior than CURB-65 score (AUC=0.6995% CI=0.59–0.80) in the subgroup of patients with lung infection (n=108).

Conclusions In this large emergency cohort of lupus patients complicating with invasive infection, an impressive high mortality was recorded, underlining that special attention should be paid to these patients. Several independent predictors for all-cause mortality were successfully identified in our study. The real-world evidence-based LUPHAS score might be a promising tool for risk stratification in clinical practice.

Funding Source(s): None