RESUMO
Introduction Systemic lupus erythematosus (SLE) is a complex disease that is associated with significant mortality and an increased risk of hospitalization. Several validated instruments are available to measure disease activity in SLE patients. However, these instruments were not designed to screen for SLE patients at an increased risk of hospitalization. These instruments also fail to incorporate some data that are easily obtainable from electronic health records, such as the frequency of missed outpatient appointments. Methods All patients at a single academic medical center with an International Classification of Disease (ICD-10) code for SLE (M32) that were seen at least once between 2010 and 2017 were identified. Of these 3552 patients, 813 were randomly selected for chart review using a random number generator, and 226 were verified to have seen an outpatient rheumatologist and met the American College of Rheumatology Classification Criteria for SLE. Physician notes, laboratory values, and appointment information were reviewed, and relevant data were extracted. Weighted Cox regression models were used to estimate the risk of hospitalization and develop a screening algorithm, and receiver operating characteristic (ROC) curve analysis was performed to evaluate the algorithm. Results There were 160 patients with no lupus-related hospitalizations and 66 patients with such a hospitalization. In a multivariate analysis accounting for age, gender, and race, serum creatinine >1.20 mg/dL, white blood cell count > 10 (thousand)/µL, hemoglobin <11 g/dL, platelets < 180 (thousand)/µL, high risk immunosuppression use, missing between 0 and 20% of appointments, and missing ≥ 20% of appointments were associated with an increased risk of hospitalizations. Our proposed screening algorithm does well identifying SLE patients at risk of hospitalization (area under the curve (AUC): 0.90, 95% CI: 0.86-0.94). We recommend flagging patients with a score of ≥ 3 (sensitivity: 0.95; specificity: 0.54). Conclusions A new screening algorithm accounting for serum creatinine, white blood cell count, hemoglobin, platelets, high-risk immunosuppression, and the proportion of missed appointments may be useful in identifying SLE patients at an increased risk of hospitalization. Missing appointments may be a proxy for an underlying variable (such as access to health care) that is directly related to an increased risk of hospitalization.