Background Health insurance claims data are widely used for researchers to describe health care utilization, patterns of care, disease prevalence, drug and disease outcomes, and cost of care. In South Korea, National Health Insurance (NHI) claims database can provide enormous amount of detailed information because it is the system of universal mandatory health insurance. About 98.7% of general population of South Korea automatically registered as the insured and medicaid offers so-called the universal health care to the others. Moreover, automated billing system of hospital has been widely used in South Korea and over 99% of claims are using this system since 2005. It is useful to perform epidemiologic research and pharmacoeconomic study for Systemic Lupus Erythematosus (SLE) owing to its high representativeness of a large, defined population. However, measurement error such as inaccurate diagnostic code is a major problem to use claims data in epidemiologic studies. The accuracy of diagnostic code has not been high either in Korea or other country.
Objectives We aimed to develop an identification algorithm for validating the International Classification of Disease-tenth revision(ICD-10) SLE diagnostic codes in Korean national health insurance claims database.
Methods Individual copayment beneficiaries program for rare & intractable diseases including SLE started from July 2009 in South Korea. Patients registered in this system pays only 10% of total medical cost, though doctor’s official report for 1997 revised ACR criteria fulfillment is essential for registration. We analyzed all claims data of patients who had diagnostic code of SLE (M32) in 2010. We regarded patients who registered in this system as the gold standard SLE patient and examined the validity of several algorithms to define SLE diagnosis using clinical information including hospitalization, laboratory examination, and prescription data. We made more than 90 algorithms and diagnostic validity was calculated for each algorithm. Then we selected one algorithm which has high sensitivity and specificity and its comparability ratio is near to one.
Results Total of 32,058 patients who ever had diagnostic code of SLE was included in this validation study. Among them, 13,421(41.9%) patients were registered in individual copayment beneficiaries program and we considered them as true SLE patients. Among more than 90 algorithms, an algorithm consisting of hospitalization due to SLE(1≤), medication(concomitant prescription of immunosuppressant and hydroxychloroquine) or anti-dsDNA antibody test 2≤ or complement test(C3·C4) 2≤ for identifying SLE case from claims data. The algorithm had the high sensitivity, specificity, and positive predictive of 74%, 82%, and 75%. Its comparability ratio(number of SLE patients based on the algorithm to number of patients registered in the copayment beneficiaries program for rare & intractable diseases due to SLE) was 1.0.
Conclusions We developed an algorithm for identifying SLE patients with clinical information including hospitalization, laboratory examination, and prescription in claims data. This approach may facilitate estimation of prevalence and incidence, and pharmacoepidemiologic study using claims data.
Disclosure of Interest None Declared