Background/Purpose: The electronic health record (EHR) serves as a powerful tool to enable researchers to collect a large cohort of patients across the healthcare system.  To assemble these patients, algorithms are needed to identify these patients accurately. We have previously validated and published algorithms to identify systemic lupus erythematosus (SLE) subjects in the EHR using ICD-9 billing codes, labs, and keywords. Currently, there are no published algorithms using ICD-10-CM codes. We aimed to develop algorithms using ICD-10-CM codes and clinical data to identify SLE patients accurately in the EHR. Methods: We analyzed data from a de-identified version of Vanderbilt’s EHR that contains over 2.8 million subjects with longitudinal data.  We identified 7399 potential SLE subjects with at least one count of the SLE ICD-9 (710.0) or ICD-10-CM (M32.1*, M32.8, or M32.9) codes.  Of these subjects, we randomly selected 200 as a training set for chart review to identify true case status.  A subject was defined as a case if diagnosed with SLE by a Vanderbilt or external rheumatologist, nephrologist, or dermatologist (specialist).  We selected the following algorithm components based on clinical knowledge and available data: SLE ICD-10-CM codes, positive anti-nuclear antibody (ANA) (titer ≥ 1:160), and ever use of antimalarials. Positive predictive values (PPVs) and sensitivities were calculated for ICD-10-CM codes and combinations of the above algorithm components. Of the 200 subjects, 88 had ICD-10-CM codes with 13 missing clinic notes and excluded from the analysis. Ten subjects had unsure or “probable” SLE diagnoses by a specialist and were not counted as cases.   Results:  Table 1 provides the PPVs and sensitivities of the algorithms. PPVs were higher for algorithms using ICD-10-CM codes compared to those using the ICD-9 code.  Algorithms that used only ICD-10-CM codes without clinical data had PPVs from 71 to 92%.  When adding a positive ANA, PPVs slightly increased and ranged from 69 to 100%. PPVs also slightly increased with adding ever antimalarial use with PPVs from 76 to 95%. The algorithms with the highest PPVs of 100% were 1) ≥ 4 counts of the ICD-10-CM codes and ANA positive and 2) ≥ 4 counts of the ICD-10-CM codes and ANA positive or ever antimalarial use.   Conclusion: Overall, the PPVs for algorithms using ICD-10-CM codes were higher compared to PPVs for ICD-9 codes.  Adding clinical data to ICD-10-CM codes slightly improved the algorithms’ PPVs but may not be required. Since “probable” cases were treated as not cases, PPVs could be underestimated. ICD-10-CM codes can identify SLE cases accurately in the EHR. Studies are underway to investigate the performance of algorithms that combine both ICD-9 and ICD-10-CM codes with clinical data.  

Table 1.