Table 2.

Findings and lessons derived from the implementation of the Lagos hybrid inventory methodology.

WHOa inventory methods	Lagos hybrid inventory methodology	Findings and lessons
Electronic case-based national TBb surveillance system	Retrospective digitalization of paper-based surveillance to create electronic case-based surveillance system	Quality-assured digitalization requires robust database design and data management. Relational databases are required where each element has a unique ID, including each facility.
Electronic case-based database with records for patients with TB	Digitalization of facility and laboratory records with no notification behavior	The inability to distinguish between sites that diagnosed no patients with TB and those who kept no records of patients with TB treated is a weakness of retrospective designs. The decay of records was an issue. Some patients’ paper records were physically damaged in 5 out of 701 sites through poor warehousing or force majeure, impacting the quality of both verification and case finding. As 4 of these sites were engaged, this may have led to the underverification of notified cases, leading to an inflated estimate of overreporting. This can be mitigated by triangulation and minimizing the period between reference year (2015) and data collection (2017).
Electronic case-based database with records for patients with TB	High-specificity on-site verification in engaged HFsc with the likelihood of notification behavior	On-site human verification of notified cases using a deterministic 7 variable matching algorithm in engaged DOTSd centers with a high likelihood of notification proved a viable alternative to 100% field-based data entry. It was acceptable to providers because TB registers did not need to be removed from the premises, on-site data entry was minimized, and the total respondent burden was reduced.
Census of all HF (retrospective) or random sampling of all HF (prospective)	Hybrid mix of census and stratified probability proportional to size sampling methods among HF strata	The census of laboratories paired with the sampling of private HFs was robust because it allowed for triangulation of self-report, as well as extrapolation. However, the frequency of TB service provision (32%-36%) in the unengaged private HFs was significantly overestimated, so the point estimate of misreporting of clinical TB in the unengaged private sector has large uncertainty bounds. This large sampling error would have compromised the estimate of misreporting of B+e TB also, were it not for the ability to rely upon the TB diagnoses from a census of all laboratories.
Standardized TB case definitions	Broad case definition for TB (all other forms); standardized case definition for B+ TB	The use of a broader definition of clinical TB permitted a more complete accounting of TB treatment coverage that includes overdiagnosis and overtreatment. Documentation of the frequency of diagnosis of TB without bacteriological testing in the unengaged sector is important information for public health stakeholders. The clinical diagnosis was a very small proportion of TB treatment found (11%).
Presence of unique identifiers for record linkage for deterministic record linkage	Use of multi-variable probabilistic record linkage algorithms with sensitivity analysis, combined with an independent review	Probabilistic record linkage with WHO-recommended software underestimated notification due to low sensitivity for name reversal. Use of Excel add-in (Fuzzy Lookup) allowed for matching across and between variables but is not syntax driven and provides no audit trail. The ability to account for name order reversal is important to avoid bias in misreporting estimates.
National in scope or sampled “self-contained” geographical areas	Subnational in scope, but using buffer zone sensitivity analysis to permit estimation and adjustment for cross-border health care seeking	Buffer zone sensitivity analysis is straightforward to conduct and permits robust subnational and urban inventory studies.
No recommendation to study misreport between levels of the TB surveillance system	Comparisons of aggregated data to identify misreporting between administrative levels of the notification system	The addition of within-surveillance system misreporting enhanced the value of the study for local stakeholders.
No recommendation to include the study of the underlying rationales and solutions to suboptimal notification behaviors in addition to its quantification	In-depth interviews with health care providers and focus group discussions with surveillance offers were undertaken	Focus group discussions and in-depth interviews provided context for quantitative findings of misreporting, critical to engage stakeholders in identifying roots of and solutions to the notification problem.

^aWHO: World Health Organization.

^bTB: tuberculosis.

^cHF: health facility.

^dDOTS: Directly Observed Treatment Short-Course.

^eB+: bacteriologically confirmed tuberculosis