To estimate the burden of disease in the Korean population, the KNBD study followed the approach of the GBD23 with several modifications: 1) disease classification and data sources were country-specific, 2) years lost due to disability (YLD) were determined using national disability weights, and 3) years of life lost (YLL) were calculated using the national standard life expectancy. Firstly, the study adjusted disease classification according to national representative data sources with expert consultation.4 Specifically, data on cancers were obtained from the Korea Central Cancer Registry, which is a national registry. Data from the Korean National Hospital Discharge Survey conducted by the Korea Centers for Disease Control and Prevention were used to estimate injuries. For the other diseases, we used the National Health Insurance Service (NHIS) claims data, which is a highly representative medical use database because NHIS covers approximately 97% of the population as a single insurer. Prevalent cases were defined by the presence of related International Classification of Diseases, tenth revision (ICD-10) codes for each disease, as principal diagnosis during inpatient or outpatient visits, in consultation with specialists. Incident cases were defined as newly diagnosed diseases during the washout period (0–5 years, defined by diseases) to exclude prevalent cases. Cause of death data were obtained from the Statistics Korea.

Secondly, national disability weights, a key variable for YLD estimation, were obtained from a self-administered web-based survey of a total of 496 physicians and medical college students,5 228 causes were evaluated by a paired comparison method. The study provided information on national disability weights with adoption of the GBD methodology, despite the need to improve its validity with revision by the consensus of experts using ranking scale on the different results of previously published disability weights, or by analysis with other methods.

Thirdly, YLL calculations were based on mortality data and national life expectancy estimates, which were determined using age- and gender-specific life tables published by the Statistics Korea. As the life table used in calculating the standard expected YLL is the key variable for estimating the burden of disease due to premature death, contextualization of YLL was performed. Furthermore, deaths assigned to so-called garbage codes have been redistributed to improve the validity of causes of death, as described in detail elsewhere.6

The application of the DALY estimate assuming a single-disease hypothesis fails to account for comorbid conditions.7 Even with the assumption of difficulty in identifying temporal order among the comorbid diseases for individuals, the burden of disease estimation without consideration of coincident illnesses is limited to being used as a summary measure of a multimorbid population.

Through the KNBD, we estimated comorbidity-adjusted DALYs by applying a multiplicative approach to some chronic diseases.8 The data were obtained from the Korea National Health & Nutrition Examination Survey, and disability weights of each disease assessed by using the EuroQol-5D (EQ-5D). As the number of comorbid diseases increased, disability weights increased, and the DALYs were the highest in the low-income and low-education level groups. A higher prevalence of comorbidity in the vulnerable group of socioeconomic levels was identified, suggesting the need to consider comorbidities when estimating population health using DALYs. The next KNBD project will aim to expand the diseases to be analyzed.

One of the factors promoting the validity of DALY is that it accounts for the severity level of diseases by using different weights for different stages of the disease. There has been no study on the burden of disease considering disease severity, and previous studies have each focused only on single diseases. The KNBD Study refined disability weights reflecting the severity of 60 causes, including cancers and stroke, by a survey of 605 physicians and medical college students.9

One of the biggest methodological challenges for the KNBD is deciding whether to convert to a prevalence-based approach or not. Following the publication of the GBD 2010 in 2013,1011 DALYs were calculated based on a prevalence-based approach to model the assumptions for estimation by Bayesian meta-regression tools such as DisMod-MR. In fact, issues with the incidence-based DALY metric, non-reflection of the current burden of sequelae and requiring estimates of both incidence and duration of disease, have led to the use of a prevalence-based approach in the GBD.

The KNBD mainly utilizes an incidence-based approach, because it allows the use of NHIS claims data that represent health status relatively broadly and accurately. From this perspective, although contextualization enhances the national acceptance of the results, it constrains direct international comparisons. The prevalence-based approach enables international comparison because the methodology is the same as that of the GBD. Accordingly, the KNBD 2012 firstly applied prevalence-based DALYs to the Korean population for the 30 leading causes, including cancers.12 However, the crucial problem was the lack of available national data on the distributions of sequelae and health states. Since the NHIS claims data are based on medical records and only the patients covered by national health insurance can be identified, it is difficult to estimate the probability of sequelae following all severity levels of cases. Although we tried to supplement the data by a systematic review of published literature, limited evidence exists for the distribution of health status. To estimate more reliable and valid prevalence-based DALYs of the Korean population, national epidemiological data resulting from population-based survey would be needed.

DALYs for different geographic regions or socioeconomic groups may allow for a more specific view of public health. For this reason, the GBD subdivided its estimations not only by disease classification, but also by subnational level. Previous studies emanating from the UK,13 Japan,1415 and China16 have shown that the subnational analysis of the burden of disease enables region-specific needs to be determined and provides evidence on national health policy. Subnational analysis provides the means to capture disease burden according to socioeconomic factors with regional characteristics and improves its implies to decision makers. Likewise, in Korea, the burden of disease and risk factors should be assessed by subnational and socioeconomic levels in the context of the nation to identify regional health variations. This will require both data availability and the development of methodology.

Besides the regions, variables that cause differences in health inequality are socio-economic factors (Gini coefficient, social trust status, income per capita), demographic factors (percentage of elderly population, fertility rate), health-related factors (prevalence of overweight, smoking, and drinking), and infrastructural factors (the number of physicians and hospital beds per population). Otherwise, a summary index can be developed to track gaps in the health status of the Korean population considering diverse variables, like the socio-demographic index developed by the Institute for Health Metrics and Evaluation.

Since the main data of the KNBD study were obtained from national published sources, including NHIS claim data, access to the database is possible one year after the actual event, and it takes even more time for a researcher to estimate and publish the results. This restricts the use of DALYs in the policy sector. Regularly updating estimates of the burden of disease enables us to monitor the trends of population health over time and assess the effectiveness of policies.17 Beyond estimations of the burden of disease by individual researchers, a platform for regular (annually, if possible) and timely production should be established as a national task.