Pos1472 accuracy of Disease-Specific Icd-10 Code for Incident Systemic Lupus Erythematosus; Results From a Population-Based Cohort Study Set in Norway

Abstract

BackgroundAccurate estimations of incidence rates are important in all epidemiological studies. Register derived International Classification of Disease (ICD) codes appear near ideal case defining items being seemingly specific and easy to access in large populations, but their applicability in complex clinical syndromes like SLE is still unclear.ObjectivesWe aimed to clarify accuracy of ICD-10 coding in incident SLE by directly comparing incidence rates from code-based case-definitions and verified SLE diagnosis by expert clinical assessment in a defined population.MethodsFrom administrative data, we identified all individual cases who were registered at least once with the SLE-specific ICD-10 code M32 (i.e M32.1, M32.8 or M32.9) during 1999-2017 in three Southeast Norway counties (including Oslo), with total population 2.1 million. We manually reviewed the chart of every M32 coded case to confirm or reject SLE diagnosis. In cases found not to have SLE, most likely cause of inaccurate ICD-code was recorded. For the analyses of incidence rates by ICD-code, we applied a five-year washout period (1999-2003) and included only cases with first time M32 code from 2004 and onwards. Incidence rates were estimated from five case-definition; (a-c) first occurrence of one or more, two or more and three or more M32 codes from 2004-2017, respectively, (d) SLE diagnosis confirmed by chart review and (e) SLE classified by 1997 American College of Rheumatology classification criteria of SLE. To define accuracy, we applied incidence rate ratios obtained from dividing M32-derived incidence rate to those from SLE diagnosis.ResultsWe identified 1975 unique cases registered with a M32 code from 1999-2017. Chart review confirmed SLE diagnosis in 936 of the 1975 cases (45 %), while 1033 (52%) were found to have conditions other than SLE (most frequently another systemic connective tissue disease or cutaneous lupus).Of the 936 cases with confirmed SLE diagnosis, 323 (34%) were incident in the years 2004-2017 (Table 1). Figure 1a-c shows the incidence curves by different SLE case-definitions. Overall, the incidence rate ratio from two or more M32 code divided by SLE diagnosis was 2.1 (95 % confidence interval 1.8-2.4). Accuracy of ICD-coding was low for incident SLE across all ages, except in those under 25 years were the incidence rate ratio was 1.0 (95 % confidence interval 0.8-1.4) (Figure 1d). When reducing washout period to two instead of five years, the incidence rate ratio from two or more M32 codes divided by SLE diagnosis increased to 4.4 (95% confidence interval 3.8-4.7).ConclusionCase definitions based solely on ICD-10 code gave incidence rates of SLE twice as high as when cases were defined by expert clinical assessment, with a maximum discrepancy of seven times more in elderly (70-79 years of age) to a minimum of no discrepancy before 25 years of age.Table 1.Number of incident cases 2004-2017 in study area by different case definition of SLETotal, nWomen, nMen, nFirst occurrence of ICD code 2004-2017¹ICD code of SLE≥ 1 in time period1035867168ICD code of SLE≥ 2 in time period67058288ICD code of SLE≥ 3 in time period54347865Verified SLE diagnosis by chart review32327944SLE by 1997 ACR criteria²28924940¹No ICD code of SLE 1999-2003 ²American College of Rheumatology classification criteria SLEAcknowledgementsThis work was funded by DAM Foundation, The Norwegian Council for Musculoskeletal Health, Norwegian Women’s Public Health Association, Vivi Irene Hansens fund for SLE research and Ragna and Egil Eiken’s grant to promote rheumatism research.Disclosure of InterestsSigrid Reppe Moe: None declared, Hilde Haukeland Consultant of: UCB advisory board. Have not recieved any personal payment. Payment to employer/organisation., Cathrine Brunborg: None declared, Torhild Garen: None declared, Antonella Botea: None declared, Nenad Damjanic: None declared, Gro Wivestad: None declared, Heidi Kverneggen Øvreas: None declared, Thea Bøe: None declared, Anniken Orre: None declared, Sella Aarrestad Provan: None declared, Øyvind Molberg: None declared, Karoline Lerang: None declared.