Global Burden of Diseases Dataset, Methodology and Its Use in Rheumatic and Musculoskeletal Diseases

Abstract

The Global Burden of Disease (GBD) study has strictly adhered to the Guidelines for Accurate and Transparent Health Estimates Reporting [1, 2], which aims to assess global mortality and disability rates caused by major diseases, injuries, and risk factors [3]. Established in 1991, the study published its first results in 1993 [4]. The most recent update, released in May 2024, provided data spanning 1990 to 2021. With nearly 12 000 researchers from 163 countries and territories involved, [5] it has become the largest and most detailed scientific project to quantify health levels and trends worldwide [3].

The GBD study comprehensively evaluated various causes of death, injuries, risk factors, and health conditions, providing a critical basis for analyzing global population health and trends [2, 6-9]. It also helped the health sector identify overlooked health issues, reveal inequalities, predict dynamic trends, and guide policy formulation.

The GBD study covered 204 countries and territories, divided into 7 super-regions, 21 regions, and multiple custom regions, with one widely used being a five-level classification based on the Sociodemographic Index (SDI). The SDI, [6, 7] a geometric mean of the total fertility rate under age 25, mean years of education for those aged 15 and older, and lag-distributed income per capita, ranged from 0 to 1. It is divided into five categories: 0–0.47 for low SDI, 0.47–0.62 for low-middle SDI, 0.62–072 for middle SDI, 0.72–0.81 for high-middle SDI, and 0.81–1.00 for high SDI [10]. In addition, subnational analyses were conducted in 21 countries [6, 7].

A key feature of the GBD study was detailed age stratification, covering 51 age groups from birth to 95 years and older [8]. Age-standardized rates (ASRs) [11] were used to remove the influence of differing age structures between populations, ensuring more accurate comparisons of disease or mortality rates across populations.

The GBD study's data was sourced from hospitals, government agencies, surveys, and other databases worldwide. The Global Health Data Exchange (GHDx) platform [12] provided a directory of the GBD study input data sources. A detailed methodological information for various diseases, injuries, and risk factors can be found in GBD 2021 Methods Appendices (Table 1) [13]. A principle of GBD study was the comparability of research results. To achieve this, the GBD study employed a series of methods during data processing, including quality rating of studies, independent disease coding, and standardized data modeling analysis, to enhance consistency and comparability across different countries and data sources.

Data input methods for diseases and injuries were categorized into causes of deaths [7] and nonfatal health outcomes [6]. For the causes of death category, data primarily came from vital registration (VR) sources, accessed mainly through the World Health Organization mortality database. In countries with high-quality VR systems (characterized by high completeness and low rates of garbage coding), VR data typically served as the primary source of cause-of-death information. For specific diseases and in countries lacking complete VR systems, statistics were supplemented by additional data types. The data underwent processing, cleaning, and auditing. It was mapped to a unified GBD cause list, while disease-specific misclassification issues were addressed, miscellaneous codes reassigned, and noise reduced before being stored in the cause-of-death (CoD) database. The CODEm modeling method, specifically designed for GBD, was primarily adapted to estimate uncorrected mortality rates. This approach was supported by additional modeling techniques such as negative binomial models, natural history models, sub-cause proportion models, and prevalence-based models. These models accounted for the influence of covariates to obtain initial mortality estimates. The CoDCorrect process was then applied to align cause-of-death estimates with all-cause mortality estimates, producing corrected mortality rates. Years of Life Lost (YLL) were calculated by multiplying the number of deaths for each cause-age-sex-location-year combination by the standard life expectancy at each age.

Rheumatoid arthritis (RA) data sources included vital registration, verbal autopsy, and surveillance data, which, after standardization, ICD mapping, age-sex splitting, garbage code removal, and noise reduction, were stored in the CoD database. Covariates for modeling rheumatoid arthritis mortality included factors such as smoking, milk intake, healthcare quality index, alcohol consumption, mean BMI, mean cholesterol, education level, income, and the SDI. Uncorrected deaths were estimated by using these covariates in the CODEm modeling approach, and adjusted mortality was then calculated using the CoDCorrect process [13].

For nonfatal outcomes, the GBD 2021 nonfatal estimation process outlined eight essential steps: [6] (1) data identification and extraction to compile sources (e.g., systematic reviews, survey data preparation, disease registries, and case notifications); (2) data adjustments (e.g., crosswalking, bias adjustment, network analysis, and age-sex splitting); (3) prevalence and incidence estimation using DisMod-MR 2.1 or alternative modeling strategies (e.g., ST-GPR modeling and MR-BRT meta-regression modeling); (4) injury estimation; (5) severity distribution; (6) incorporation of disability weights; (7) comorbidity adjustments; and (8) calculation of Years Lived with Disability (YLD) by sequelae and cause. This series of analyses provided incidence, prevalence, and YLD for disease and injury sequelae. Adding YLL and YLD produced the DALYs [6]. Table 1 described the data input methodology for three common rheumatic diseases.

Data visualization enhanced the effectiveness of information delivery and promoted action in the policy making process [14]. The GBD study offered 52 powerful visualization tools that helped users gain deep insights into global health data. The GBD Results Tool allowed users to view and download global health estimates as CSV files. GBD Compare provided comprehensive analyses of how disease patterns had changed over time. GBD Foresight evaluated and predicted future scenarios from 2022 to 2050. More detailed information regarding GBD data visuals can be found on the Health Data Research and Analysis website (https://www.healthdata.org/interactive-data-visuals).

The GBD study not only provides reliable data and a transparent methodology, but also offers a variety of visualization tools. Despite some limitations in data sources and knowledge translation, the GBD study provides detailed estimates of the global disease burden, helping policymakers identify major health challenges, compare health disparities across demographics levels, evaluate the cost-effectiveness of interventions and address global health challenges collectively. In the coming years, GBD research has the potential to assess the global burden of disease more accurately, comprehensively, and realistically.

Shi-Hang Chen: conceptualization, data curation, visualization, validation, and writing – original draft. Yuan Tang: formal analysis, validation, resources, and writing – review and editing. Harry Asena Musonye: software development, investigation, visualization, validation, and writing – review and editing. Hai-Feng Pan: methodology, project administration, supervision, resources, funding acquisition, and writing – review and editing.

The authors declare that there are no conflicts of interest regarding the publication of this article. None of the authors of this article directly participated in the GBD study.