2021-10-18-世界卫生组织-Methods_used_by_WHO_to_estimate_the_global_burden_of_TB_disease_29页_733kb

Summary of WHO Methods to Estimate the Global Burden of TB Disease (2000-2020)

Core Content

The World Health Organization (WHO) employs a variety of methodological approaches to estimate the global burden of tuberculosis (TB) disease in terms of incidence and mortality, with a focus on disaggregation by HIV status, age, and sex. These methods are continuously refined to account for data gaps, biases, and changes in health systems, especially in the context of the COVID-19 pandemic.

Main Methods for Estimating TB Incidence

1. Case Notification Data with Expert Opinion

• Description: Combines case notifications with expert estimates of under-reporting and under-diagnosis.
• Usage: 39 countries.
• Formula: $ I = \frac{f(N)}{1 - g} $, where $ N $ is case notifications, $ f $ is a cubic spline or identity function, and $ g $ is the under-reporting proportion.
• Limitations: Relies on expert opinion, which may be biased or inconsistent.

2. TB Prevalence Surveys

• Description: Uses prevalence survey data and duration of disease estimates to calculate incidence.
• Usage: 29 countries.
• Two Approaches:
  • Approach 1: Divides prevalence by average disease duration.
  • Approach 2: Uses a compartmental model with susceptible, untreated, and treated cases.
• Assumptions: Epidemic equilibrium, uniform distribution of disease duration, and accurate data on treatment and mortality.
• Limitations: Insufficient precision in estimating prevalent cases on treatment.

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

• Description: Adjusts case notifications in high-income and some upper-middle-income countries for under-reporting and under-diagnosis.
• Usage: 139 countries.
• Adjustment Factors: Standard factor for most countries, country-specific for France, Republic of Korea, and Turkey.
• Consistency Checks: Monitoring of notification rate fluctuations and TB death to notification ratio.

4. Inventory Studies and Capture-Recapture Modelling

• Description: Uses multiple data sources to estimate the proportion of unreported TB cases.
• Usage: 8 countries.
• Assumptions: Closed population, homogeneity in detection probabilities, and consistent case definitions.
• Limitations: Requires at least 3 data sources and is not feasible in most high-burden countries.

Disaggregation by HIV Status

• HIV-Positive TB Incidence: Estimated using provider-initiated testing and HIV sentinel surveillance.
• Adjustment for Bias: Local polynomial regression is used to combine data sources, with different weights assigned based on data quality and representativeness.
• Indirect Estimation: In countries without specific HIV prevalence data among TB cases, it is derived from general population HIV prevalence.

Disaggregation by Age and Sex

• Data Collection: WHO has been collecting age-disaggregated TB notifications since 1995.
• Reporting Coverage: Improved over time, reaching 99% for smear-positive, 83% for smear-negative and extrapulmonary TB in 2012.
• Disaggregation Approach:
  • For countries with high case detection ratios (>85%), incidence is disaggregated proportionally to notifications.
  • For countries with low notifications, a Bayesian approach is used, drawing samples from a prior based on regional prevalence data and mathematical models.
  • For children, a model simulates the natural history of TB, considering BCG vaccination and HIV prevalence.

Incidence of TB Due to Mycobacterium bovis

• Estimation: Based on regional proportions of TB cases attributed to M. bovis.
• Uncertainty: Assumed a 50% standard deviation around the best estimate.
• Data Sources: Scientific studies, as routine reporting is lacking in many endemic countries.
• Mortality Estimation: Similar to incidence, but reduced by 20% to account for lower case fatality in extra-pulmonary TB.

Bacteriologically Confirmable TB Incidence

• Estimation: Calculated as the product of estimated TB incidence, proportion of notified pulmonary TB cases, and expected proportion of bacteriologically confirmed cases.
• Data Source: Based on high-income countries' data from 2019, with a weighted average of 83% and standard deviation of 5.9%.

Mortality Estimation (2000-2019)

• Data Sources: Vital registration systems and mortality surveys.
• Coding: Uses ICD-10 codes A15-A19 and B90, equivalent to ICD-9 codes 010-018 and 137.
• HIV-Related Deaths: HIV is recorded as the underlying cause, and TB as a contributory cause.
• Indirect Estimation: In countries without direct mortality data, mortality is derived from incidence and case fatality ratios.

Key Points

• Dynamic Models for 2020: New statistical and dynamic models were introduced for 16 and 111 countries for incidence and mortality, respectively, to account for pandemic disruptions.
• Challenges: Under-reporting, under-diagnosis, and lack of age- and sex-specific data remain significant challenges.
• Expert Input: Critical in countries with poor data quality or missing information.
• Regional Variability: Different methods are used based on the availability and quality of data, with a focus on regional and country-specific patterns.

Conclusion

WHO's methods for estimating TB burden are diverse, adaptive, and increasingly sophisticated. They aim to provide accurate and reliable data despite data limitations, incorporating both direct and indirect approaches, as well as expert opinion and statistical models. The integration of these methods ensures a comprehensive understanding of TB incidence and mortality, while also highlighting the need for improved surveillance and data collection, particularly for vulnerable populations like children and HIV-positive individuals.