Discrepancies in neglected tropical diseases burden estimates in China: comparative study of real-world data and Global Burden of Disease 2021 data (2004-2020)

Data accuracy and surveillance systems in assessing disease burden

The accuracy and reliability of data sources are fundamental to scientific research and resource allocation. WHO has advocated for countries to enhance national reporting systems to ensure that the quality and interpretability of the reported data better capture the real-world burden of diseases and formulate more effective response strategies.36 Our analysis of the disease burden for the six notifiable NTDs included in this study is based on data published by the China Public Health Science Data Centre, which serves as an open platform for disseminating data collected and reported by the China Information System for Disease Control and Prevention.3738 This system facilitates robust and timely data collection from the grass-root level in the public health system, supporting precise assessments of disease burden and the effectiveness of control measures.39 For instance, China's successful malaria elimination was achieved through the surveillance and response principle of “rapid reporting within one day, confirmed cases within three days, and effective intervention within seven days.”40 Similarly, for diseases such as echinococcosis and schistosomiasis, not only are systematic annual reporting mechanisms in place but also national funds and specialised control offices are dedicated to ensure meticulous data collection and monitoring at the community level.4142 Dengue in China is primarily a locally transmitted disease caused by imported cases, and the spread is closely monitored by a highly sensitive surveillance and early warning system.43 This system, coupled with efficient contact tracing, ensures that cases are rapidly identified, isolated, and eliminated. Given the near 100% fatality rate of rabies, China has stringent controls over the reporting and management of fatal cases.44 Leprosy, with only 200-300 cases annually, is nearing elimination in China.45 The transmission of visceral leishmaniasis in specific regions of China allows for more targeted disease surveillance and control efforts.46

The GBD database is a globally authoritative and comprehensive collection of disease burden data, sourced from censuses, household surveys, civil registrations, disease and health service records, air pollution monitoring, and satellite imaging, among others. Data are accessible via the Institute for Health Metrics and Evaluation at the University of Washington.47 The GBD project regularly updates its methods, disease categories, and country coverage to ensure data timeliness and accuracy. Each new GBD iteration, such as GBD 2021, GBD 2019, and GBD 2017, incorporates adjustments and optimisations to models, reflecting the latest scientific discoveries and technological advancements, ensuring the global applicability and impact of the findings.48 Additionally, the GBD team collaborates with national health departments to validate the accuracy and relevance of the models and estimates.

Underlying reasons for variations between data sources

Despite extensive international collaboration and methodological advancements, our study found notable discrepancies between real-world data and GBD estimates. The 17 year average DALY ratios between the GBD estimates and real-world data for the six NTDs ranged from 1.4 to 280 times. These differences may be attributed to the regional nature of diseases or variations in data collection and processing methods among countries.

Although GBD 2021 covers many regions and diseases globally, the data may not be complete for some rare or country specific diseases. For instance, China experiences two types of echinococcosis: cystic echinococcosis and alveolar echinococcosis.49 However, the GBD 2021 only includes data for cystic echinococcosis. Notably, alveolar echinococcosis, often referred to as “worm cancer,” imposes a much higher disease burden compared with cystic echinococcosis according to the real-world data. Even after accounting for the presence of both types of echinococcosis in China, the real-world DALYs remain substantially lower than the estimates provided by GBD 2021.50

Another contributing factor is the methodological differences in calculating life expectancy. GBD estimates standardised global life expectancy using the Sullivan's method. However, this method has limitations because potential changes in mortality rates over time are not accounted for, which may lead to differences in years of life lost calculations.3 In our study, we chose to use the national life expectancy for each corresponding year in China, which is updated based on the latest demographic data, health indicators, and mortality trends.3 This approach ensures that our calculations are more accurately aligned with China's conditions, capturing nuances such as population ageing and local health improvements that global estimates might overlook.

The GBD model does not fully account for the local epidemiological context and specific interventions implemented in China. Although the GBD team collaborates with national health departments to validate the accuracy and relevance of their models and estimates, gaps are present in capturing the unique public health strategies adopted by individual countries. In China, a variety of targeted strategies have been implemented to reduce the burden of NTDs (fig 1), such as schistosomiasis and echinococcosis.51 These strategies include improving public health infrastructure, which ensures better access to healthcare services and facilities; enhancing disease surveillance, which helps in early detection and prompt response to outbreaks; and conducting extensive health education campaigns, which raise awareness about prevention and control measures among the population.52

The GBD model is unable to accurately predict and reflect sudden outbreaks of diseases because the historical data and generalised assumptions may not fully account for rapid changes in disease dynamics or the impact of localised factors. One notable example is its failure to effectively capture the dengue outbreaks in China in 2014 and 2019.

Insights to advance the GBD model

Although GBD studies provide a vital scientific basis for global disease control and health resource allocation, these findings require continual updates and adjustments based on new data and scientific advancements.53 Reliance solely on GBD data may not fully capture the situation of certain diseases. Enhancement of the model's capacity to handle heterogeneity and disparities between global and national-level data is particularly crucial. Several key improvements are recommended to improve the predictive accuracy of the GBD model.

Model adaptability to local contexts could be improved. We recognise that there are apparent discrepancies between GBD 2019 and GBD 2021 (eg, echinococcosis, visceral leishmaniasis and dengue), which may raise questions about the model's stability and the consistency of the estimates over time. These differences can stem from various factors such as changes in data sources, updates in disease definitions, advancements in statistical modelling techniques, and the incorporation of new data. Given these substantial variations, assessment of the model's robustness is crucial, as is performance of sensitivity analyses to evaluate how changes in data and assumptions affect the estimates. The GBD model should be optimised to better adapt to the specific disease profiles and health data characteristics of different countries.54 This involves enhancing the analysis and integration of local data features, which would allow the model to more accurately reflect the unique epidemiological contexts and control measures implemented in various countries. For example, in China, robust national interventions and health strategies should be incorporated into the model adjustments to estimate local disease burdens more accurately.

Use of advanced modelling techniques, such as Bayesian hierarchical models, can better account for regional variations and improve the robustness of estimates.55 These techniques would allow the model to incorporate real-time data and adjust more dynamically to sudden changes in disease transmission patterns, including unexpected outbreaks like dengue.

A global data sharing platform should be established. Creation of a global, transparent, and interactive data sharing platform is vital. Such a platform should promote consistency in data at both global and national (or subnational) levels, support open access, and facilitate the sharing of data to enhance research quality and accuracy of policy making. Improvement of the model's ability to capture data for rare or country specific diseases, such as alveolar echinococcosis in China, which are currently under-represented in the GBD estimates is crucial. This development would also help to ensure that all relevant data are used, improving the model's ability to reflect disease conditions and trends.

Additionally, data validation and proper use should be emphasised. Researchers conducting local studies should prioritise validating their data before use, which aligns with WHO's recommendations to improve national reporting systems and ensure data quality and interpretability. As highlighted in the introduction, over 90% (284/317) of studies on disease burden in China in the past five years have used GBD data without adequate validation. To address this, both GBD and WHO should advocate the importance of data verification and encourage more cautious use of global data for local research. Enhancing these practices would lead to more accurate assessments of disease burdens and support the development of effective, context specific health strategies.

Limitations

This study also has some limitations. Firstly, the reported cases used in our analysis are based on the population of each region in China, and detailed individual case information is scarce. This aggregation can mask important nuances, such as the severity of the disease, individual treatment responses, and other case specific factors, which are crucial for a deeper understanding of disease dynamics and outcomes. Secondly, regional differences in resource allocation, variations in healthcare access, or inconsistencies in the implementation of reporting protocols etc, can lead to disparities in coverage and reporting practices. Thirdly, under-reporting of cases is a potential issue, which is a common in surveillance data, especially for diseases with non-specific symptoms or those that overlap clinically with other more prevalent diseases. For under-reporting, the 95% uncertainty intervals estimate provides an alternative reference range to compensate for the unstable point estimates caused by problems such as under-reporting; however, this study directly used annual report data from the China CDC, which did not include interval estimates. Fourthly, conducting active screening and enhancing the professional expertise of healthcare workers can improve disease detection rates, thereby more accurately reflecting the real-world disease burden of NTDs and reducing the likelihood of underestimation. Finally, we acknowledge that focusing solely on NTDs provides a narrower perspective. In future research, we plan to expand our analysis to include a broader range of diseases, such as cardiovascular diseases, diabetes, chronic kidney disease, cancer, and other non-communicable and infectious diseases.

Conclusions

The findings indicate that relying solely on global estimates, such as those from the GBD, may not adequately capture the true dynamics of DALYs caused by NTDs in China. We advocate for researchers to correctly approach and use data by integrating local epidemiological data into global health assessments, which is crucial for developing accurate, relevant, and effective public health policies and strategies. Enhancing the accuracy of disease burden estimates through localised data can better inform policy decisions and resource allocation, ultimately improving health outcomes and addressing the specific needs of affected populations. Such an approach ensures that public health interventions are tailored to the unique epidemiological context of each region, leading to more effective disease control and prevention efforts.