Table 1. Possible study designs for estimating the tuberculosis burden in India.

Study design	Possible objectives	Existing data used	New data collection required	Current feasibility	Application
					Advantages	Disadvantages
Inventory study
Retrospective analysisa	Quantification of underreporting of diagnosed cases. Estimation of tuberculosis incidence. Demonstration of negligible underreporting	National tuberculosis surveillance database plus one or two national case-based databases – the exact number depending on objectives	None	Not feasible because multiple national case-based databases not available in India
Survey of sample of all providers, selected using lot-quality assurance samplinga	Demonstration of negligible underreporting	National tuberculosis surveillance database	Provider survey of random sample of all tuberculosis providers, selected using lot-quality assurance sampling	Not appropriate because underreporting known to be substantial in India
Survey of all providers in large areas suitable for capture–recapture analysisa	Quantification of underreporting of diagnosed cases. Estimation of tuberculosis incidence	National tuberculosis surveillance database plus two other case-based databases for each geographical area selected	Provider survey of all tuberculosis providers in random sample of large, self-contained geographical areas	Needs to be assessed	Generates comprehensive, direct estimate of underreporting at all levels	Assumptions regarding migration and probability of inclusion in each database. Error-prone because of reliance on probabilistic matching across multiple databases
Survey of all providers in sampled areasa	Quantification of underreporting of diagnosed cases. Demonstration of negligible underreporting	National tuberculosis surveillance database	Provider survey of all tuberculosis providers in random sample of geographical areas	Feasible for quantifying underreporting of diagnosed cases	Of the feasible studies, relatively inexpensive because fewer data need to be collected	Proportion of cases with no health system utilization estimated from self-reported household survey data. Level of underdiagnosis estimated from other new data collection or existing data with limitations
Survey of all providers in sampled areas with assessment of underdiagnosis	Quantification of underreporting and underdiagnosis by RNTCP and non-RNTCP providers	National tuberculosis surveillance database	Provider survey of all tuberculosis providers in random sample of geographical areas, including assessment of underdiagnosis	Feasible	Generates direct estimates of the greatest number of the parameters contributing to underreporting	Proportion of cases with no health system utilization estimated from self-reported household survey data. More expensive than assessment of only underreporting of diagnosed cases in sampled areas because of additional data collection
National prevalence survey	Estimation of national prevalence of active tuberculosis in adults. Assessment of the proportion of tuberculosis cases which are drug-resistant	None	For nationally representative sample of adults aged ≥  15 years: either Xpert MTB/RIF assay or X-ray screening plus two sputum samples if symptomatic or X-ray abnormal	Feasible	Generates direct estimate of national tuberculosis prevalence, with potential to assess extent of drug resistance	In comparison with other study designs, longer period of data collection and more expensive

RNTCP: Revised National Tuberculosis Control Programme.

^a Described in detail in the WHO guide for conducting inventory studies.²⁷