A STARD-compliant prediction model for diagnosing thrombotic microangiopathies

Aim of the study was the definition of a predictive model for the initial diagnosis of thrombotic microangiopathies (TMA). We retrospectively collected data on all adult patients admitted to the Gemelli Hospital from 2010 to 2014. ICD-9 codes from primary diagnoses were used for TMA diagnosis. Demographic and laboratory characteristics on admission of patients with TMA were then compared with a random sample of 500 patients with other diagnoses. The prediction model was externally validated in a cohort from another hospital. Overall, 23 of 187,183 patients admitted during the study period received a primary diagnosis of TMA. LDH (OR 1.26, 95% CI 1.05, 1.63) and platelets (OR 0.96, 95% CI 0.94, 0.98) were the only independent predictors of TMA. The AUROC of the final model including only LDH and platelets was 0.96 (95% CI 0.91, 1.00). The Hosmer-Lemeshow (HL) test (p = 0.54) suggested good calibration. Our model also confirmed good discriminatory power (AUROC 0.72 95% CI 0.60, 0.84) and calibration (HL test p = 0.52) in the validation sample. We present a simple prediction model for use in diagnosing TMA in hospitalized patients. The model performs well and can help clinicians to identify patients at high risk of TMA.