2022-10-18-世界卫生组织-Methods_used_by_WHO_to_estimate_the_global_burden_of_TB_disease_59页_637kb

Summary of WHO Methods to Estimate the Global Burden of TB Disease

Core Content

The World Health Organization (WHO) employs a variety of methods to estimate the global burden of tuberculosis (TB) disease, including incidence, mortality, and specific subtypes like rifampicin-resistant (RR) TB and multidrug-resistant (MDR) TB. These methods are updated annually and adapted to address changes in data availability and quality, particularly in the context of the global pandemic and its impact on TB surveillance.

Main Methods for Estimating TB Incidence (2000–2019)

WHO uses four main methods to estimate TB incidence:

1. TB Prevalence Surveys

   • Incidence is estimated using prevalence survey results and duration of disease data, accounting for HIV coinfection and antiretroviral therapy (ART).
   • This method is used for 29 countries, representing 66% of global TB incidence in 2019.
   • It assumes disease duration follows a uniform distribution for different case categories (Table 1).
   • Limitations include the inability to precisely estimate prevalent TB cases on treatment and recall biases.

2. Notifications in High-Income Countries Adjusted by a Standard Factor

   • Case notifications from high-income and some upper-middle-income countries are adjusted for under-reporting and under-diagnosis.
   • This method is used for 139 countries, representing 6% of global TB incidence in 2019.
   • Adjustments are made using a standard factor, with some countries having country-specific adjustments (e.g., France, Republic of Korea, Türkiye).
   • Surveillance data in these countries are generally internally consistent, with checks for rapid fluctuations in notification rates and the M/N ratio (TB deaths / TB notifications).

3. Inventory/Capture-Recapture Studies

   • Used to quantify under-reporting of detected TB cases.
   • This method is applied to 8 countries, accounting for 17% of global TB incidence in 2019.
   • It involves capture-recapture modeling, which requires at least three independent data sources.
   • Assumptions include a closed population, homogeneity of detection probabilities, and consistent case definitions.
   • It is considered a robust method but is only feasible in a few high-burden countries.

4. Case Notification Data with Expert Opinion on Case Detection Gaps

   • Combines case notification data with expert assessments of under-reporting and under-diagnosis.
   • This method is used for 39 countries, representing 11% of global TB incidence in 2019.
   • Trends are estimated using mortality data, surveys of infection risk, or exponential interpolation.
   • It is considered less reliable due to the small number of experts involved and potential biases in data interpretation.

Methods for Estimating TB Mortality (2000–2019)

• Direct Measurement from Vital Registration (VR) Systems and Mortality Surveys

  • VR systems are the primary source for TB mortality data, using ICD-10 codes (A15-A19 and B90) and ICD-9 codes (010-018, 137).
  • In countries where VR systems only report underlying causes of death, TB is often not recorded as a contributory cause.
  • Adjustments are made for incomplete coverage and ill-defined causes of death.
  • In 2019, 58% of TB mortality (excluding HIV) was directly measured or imputed from VR and survey data.

• Indirect Estimation Using Case Fatality Rate (CFR)

  • In countries without direct mortality data, TB mortality is estimated by multiplying incidence estimates by CFR.
  • This method is used for the remaining 42% of TB mortality in 2019.

Special Methods for 2020 and 2021

• For 2020 and 2021, WHO used dynamic models for 28 countries where there were large reductions in case notifications, likely due to improved detection.
• For 22 countries that showed cumulative reductions of 10% or more in TB case notifications from 2020 to 2021, region-specific models were used.
• For high-income countries, the same methods as before 2020 were applied, with adjustments for under-reporting and under-diagnosis.
• For low- and middle-income countries (LMIC) without significant reductions in case notifications, pre-2020 trends were assumed to continue.

MDR/RR-TB Estimation

• In 2022, WHO introduced new methods to estimate time series data for MDR/RR-TB incidence and mortality from 2015 to 2021.
• These methods include:
  • Estimating the proportion of TB cases with MDR/RR-TB using data from surveys and expert opinion.
  • Calculating the incidence of MDR/RR-TB based on the proportion of MDR/RR-TB cases among all TB cases.
  • Estimating MDR/RR-TB mortality using mortality data and CFR.
• The proportion of MDR/RR-TB cases with fluoroquinolone resistance (pre-XDR-TB) is estimated using expert opinion and available data.

HIV-Positive TB Incidence

• Provider-initiated testing and counseling with ≥50% HIV testing coverage is the main source for estimating HIV prevalence among TB patients.
• Other sources include:
  • Sero-surveys of newly diagnosed TB cases.
  • HIV sentinel surveillance systems that include TB as a sentinel group.
• Data are combined using local polynomial regression with weighted least squares, where weights vary based on data source reliability (1 for nationally representative surveys, 0.2 for sentinel systems, and coverage rate for provider-initiated testing with coverage >50%).

Limitations and Considerations

• Methods based on expert opinion are generally unreliable due to limited data and potential biases.
• Capture-recapture modeling is only feasible in a few countries due to the need for multiple data sources.
• The assumption of a stable HIV prevalence ratio between prevalent and notified TB cases is used in some African countries.
• There are uncertainties in indirect estimation methods, particularly regarding case fatality rates and the proportion of TB cases that are smear positive.

Code and Transparency

• The code for implementing the described methods is available in a public repository, enhancing transparency and reproducibility.

Conclusion

WHO's methods for estimating the global burden of TB disease are diverse and context-dependent, reflecting the varying data quality and availability across countries. The use of dynamic models for recent years (2020–2021) and new approaches for MDR/RR-TB data represent significant advancements in the accuracy and comprehensiveness of TB burden estimates. However, these methods still face challenges, particularly in areas with limited data or high under-reporting rates.