Identification of idiopathic inflammatory myopathy research cohorts using international classification of disease (ICD) codes: A systematic review

Objective: To summarize studies validating International Classification of Disease (ICD) coding for idiopathic inflammatory myopathies (IIM) case identification, focusing on dermatomyositis (DM) and polymyositis (PM). Methods: PubMed, Scopus, Embase, and CINAHL search was performed with a medical librarian (Prospero #CRD4202452377). Inclusion criteria required: English language, use of ICD-9 and/or -10 coding to identify cases, and clear method for validating cases. Results: 3684 citations were screened by title/abstract resulting in 69 full-text publications reviewed with 11 articles meeting inclusion criteria. Of the included studies, 5 evaluated only ICD-9, 5 evaluated only ICD-10, and one study evaluated both. All but one study in the US were single center, while those in Europe (n = 4) and Canada (n = 1) were population based. Reference standards varied, including Bohan and Peter (n = 5), 2017 EULAR/ACR classification (n = 2), expert opinion (n = 3), enrollment in Rheumatology Quality Register (n = 1) and muscle biopsy (n = 1). Four studies had a positive predictive value (PPV) of 0.9 or higher; two of these studies used coding associated with inpatient admission, and another required at least 2 codes (710.3) 3 months apart. Multiple coding instances decreased the sensitivity but increased PPV. Conclusions: ICD coding is a valuable tool for identifying cases in bioinformatic research. Only a few studies describe ICD code validation for IIM research cohort construction. The highest PPV’s are reported for DM and PM with the use of multiple coding instances, or those instances associated with inpatient admission.