Validation of self-reported medical condition in the Taiwan Biobank

• Published in: Journal of Epidemiology p. JE20240110
• Main Authors: Wu, Chi-Shin, Hsu, Le-Yin, Shen, Chen-Yang, Chen, Wei J., Wang, Shi-Heng
• Format: Journal Article
• Language: English
• Published: Japan Epidemiological Association 2024
• Subjects: biobank; claims data; concordance; self-reported medical condition
• ISSN: 0917-5040; 1349-9092
• DOI: 10.2188/jea.JE20240110

Background: This study aimed to validate self-reported medical conditions in the Taiwan Biobank (TWBB), in which participants were inquired about 30 disease conditions, by comparing them with claims records from Taiwan's National Health Insurance (NHI) claims database.Methods: We identified 30 clinical diagnoses using ICD-CM codes from ambulatory and hospital claims within the NHI claims database, matching diseases included in the TWBB. The concordance between self-reports and claims records was evaluated using tetrachoric correlation to assess the correlation between binary variables.Results: A total of 131,834 participants aged 30–70 years with data from the TWBB and NHI records were included. Concordance analysis revealed tetrachoric correlations ranged from 0.420 (chronic obstructive pulmonary disease) to 0.970 (multiple sclerosis). However, several disorders exhibited lower tetrachoric correlations. The concordance was higher among those with higher education attainment, and lower among married individuals.Conclusion: The concordance between self-reports in the TWBB and NHI claims records varied across clinical diagnoses, showing inconsistencies depending on participant characteristics. These findings underscore the need for further investigation, especially when these variables are crucial to research objectives. Integrating complementary databases such as clinical diagnoses, prescription records, and medical procedures can enhance accuracy through customized algorithms based on disease categories and participant characteristics and optimize sensitivity or positive predictive values to align with specific research objectives.