Contents lists available at ScienceDirect Archives of Gerontology and Geriatrics journal homepage: www.elsevier.com/locate/archger Development and validation of the Korea Dementia Comorbidity Index (KDCI): A nationwide population-based cohort study from 2002 to 2013 Jae-Hyun Kim a,e , Ki-Bong Yoo b , Yunhwan Lee c,d, ⁎ a Department of Health Administration, College of Health Science, Dankook University, Cheonan, Republic of Korea b Department of Healthcare Management, Graduate School, Eulji University, Gyeonggi, Republic of Korea c Department of Preventive Medicine and Public Health, Ajou University School of Medicine, Suwon, Republic of Korea d Institute on Aging, Ajou University Medical Center, Suwon, Republic of Korea e Institute of Health Promotion and Policy, Dankook University, Cheonan, Republic of Korea ARTICLE INFO Keywords: Dementia Comorbidities Disease ABSTRACT This study develop and validate a simple and accessible measure of comorbidity, named the Korean Dementia Comorbidity index (KDCI), to assist in predicting the onset of dementia. This study used the National Health Insurance Service–Cohort Sample Database from 2002 to 2013 (n = 23,856). Cox proportional hazard model was used to estimate incident dementia (International Classification of Disease, 10th edition (ICD-10) codes: F00-F03, G30, G311), with a hazard ratio higher than 1.05 for each comorbid condition being assigned a score. Scores ranging from 1 to 4 were assigned based on the magnitude of the hazard ratio (HR): 1 (1.050 ≤ HR ≤ 1.099), 2 (1.100 ≤ HR ≤ 1.149), 3 (1.150 ≤ HR ≤ 1.199), and 4 (HR ≥ 1.200) Summated scores of comorbidities for each individual constituted the Korean Dementia Comorbidity Index (KDCI). Five patterns were extracted: (1) disease of the eye and adnexa; (2) endocrine and metabolic disease, and disease of circulatory system; (3) disease of the musculoskeletal system and connective tissue; (4) disease of the respiratory system; and (5) disease of the nervous system, and mental and behavioral disorders through factor analysis. Fitting performance by Akaike information criterion (AIC) of CCI by Charlson, CCI by Quan and KDCI adjusting for age and sex was 29,486, 29,488 and 29,444, respectively. Our analysis results on discriminatory abilities provided evidence that KDCI is superior to other comorbidity indices on incident dementia in terms of co- morbidity adjustment. Therefore, KDCI can be a useful tool to identify incident dementia. This has implications for clinical management of patients with multimorbidity as well as risk adjustment for database studies. 1. Introduction In South Korea, the prevention and management of dementia pa- tients have become an important national issue in that the prevalence of dementia among older adults over 65 years old has been gradually increasing. (Shin, Seo, Kim, Kim, & Lee, 2014) According to a Taiwa- nese study, (Lin, Tsai, Lai, & Wu, 2015) using a nationwide database, 55% of dementia patients had two or more comorbid conditions. Co- morbid conditions are more likely to affect the severity of dementia patients, socioeconomic burden and complexity of care (Prince et al., 2013). In addition, patients with a variety of comorbidity such as type 2 diabetes mellitus and hypertension might be at high risk for developing dementia (Mehta et al., 2015) and progression in dementia is more likely to be highly heterogeneous (Post, Merkus, de Haan, & Speelman, 2007). Thus, in the past decade, clinical and scientific interests in comorbidity have increased (Barnett et al., 2012; Diederichs, Berger, & Bartels, 2011). Although the definition of comorbidity is simple—the “means two or more medical conditions or disease pro- cesses that are additional to an initial diagnosis and directly or in- directly affects health outcomes (Feinstein, 1970)”—the underlying construct is considerably more complex. In epidemiologic and health services research, there may be at least four important reasons to measure comorbidity of patients: adjusting for confounding, identifying effect modification, using them as a pre- dictor of study outcomes and improving statistical efficiency (Groot, Becherman, & Lankhorst, 2003). In the past decade, applicable methods for classifying comorbidity conditions have been developed. One of the most common indices used is Charlson Comorbidity Index (Charlson, Pompei, Ales, & MacKenzie, 1987; Deyo, Cherkin, & Ciol, 1992), as valid and reliable methods available for clinical research to measure http://dx.doi.org/10.1016/j.archger.2017.06.001 Received 26 January 2017; Received in revised form 10 May 2017; Accepted 4 June 2017 ⁎ Corresponding author at: Department of Preventive Medicine and Public Health and Institute on Aging, Ajou University 164, World cup-ro, Yeongtong-gu, Suwon 16499, Republic of Korea. E-mail address: yhlee@ajou.ac.kr (Y. Lee). Archives of Gerontology and Geriatrics 72 (2017) 195–200 Available online 07 June 2017 0167-4943/ © 2017 Elsevier B.V. All rights reserved. MARK