The Tuberculosis Cascade of Care in India’s Public Sector: A Systematic Review and Meta-analysis

Background India has 23% of the global burden of active tuberculosis (TB) patients and 27% of the world’s “missing” patients, which includes those who may not have received effective TB care and could potentially spread TB to others. The “cascade of care” is a useful model for visualizing deficiencies in case detection and retention in care, in order to prioritize interventions. Methods and Findings The care cascade constructed in this paper focuses on the Revised National TB Control Programme (RNTCP), which treats about half of India’s TB patients. We define the TB cascade as including the following patient populations: total prevalent active TB patients in India, TB patients who reach and undergo evaluation at RNTCP diagnostic facilities, patients successfully diagnosed with TB, patients who start treatment, patients retained to treatment completion, and patients who achieve 1-y recurrence-free survival. We estimate each step of the cascade for 2013 using data from two World Health Organization (WHO) reports (2014–2015), one WHO dataset (2015), and three RNTCP reports (2014–2016). In addition, we conduct three targeted systematic reviews of the scientific literature to identify 39 unique articles published from 2000–2015 that provide additional data on five indicators that help estimate different steps of the TB cascade. We construct separate care cascades for the overall population of patients with active TB and for patients with specific forms of TB—including new smear-positive, new smear-negative, retreatment smear-positive, and multidrug-resistant (MDR) TB. The WHO estimated that there were 2,700,000 (95%CI: 1,800,000–3,800,000) prevalent TB patients in India in 2013. Of these patients, we estimate that 1,938,027 (72%) TB patients were evaluated at RNTCP facilities; 1,629,906 (60%) were successfully diagnosed; 1,417,838 (53%) got registered for treatment; 1,221,764 (45%) completed treatment; and 1,049,237 (95%CI: 1,008,775–1,083,243), or 39%, of 2,700,000 TB patients achieved the optimal outcome of 1-y recurrence-free survival. The separate cascades for different forms of TB highlight different patterns of patient attrition. Pretreatment loss to follow-up of diagnosed patients and post-treatment TB recurrence were major points of attrition in the new smear-positive TB cascade. In the new smear-negative and MDR TB cascades, a substantial proportion of patients who were evaluated at RNTCP diagnostic facilities were not successfully diagnosed. Retreatment smear-positive and MDR TB patients had poorer treatment outcomes than the general TB population. Limitations of our analysis include the lack of available data on the cascade of care in the private sector and substantial uncertainty regarding the 1-y period prevalence of TB in India. Conclusions Increasing case detection is critical to improving outcomes in India’s TB cascade of care, especially for smear-negative and MDR TB patients. For new smear-positive patients, pretreatment loss to follow-up and post-treatment TB recurrence are considerable points of attrition that may contribute to ongoing TB transmission. Future multisite studies providing more accurate information on key steps in the public sector TB cascade and extension of this analysis to private sector patients may help to better target interventions and resources for TB control in India.

The objective of this systematic review and meta-analysis is to gain insight into Gaps 2 and 3 in the TB cascade of care-the proportion of TB cases presenting to government designated microscopy centers (DMCs) who fail to get diagnosed and the proportion of cases who are diagnosed with TB at government facilities but who do not get registered in treatment. Specifically, we aim to extract data on three variables: (1) the proportion of patients evaluated at DMCs who submit one sputum sample but fail to submit a second sputum sample (sometimes referred to as "diagnostic default"). This value can be used to estimate the number of new smear-positive and retreatment smear-positive TB cases "missed" during the diagnostic workup (Gap 2).
(2) the proportion of TB patients with initial negative sputum smears who fail to complete the multi-step diagnostic workup for smear-negative TB (Fig B), especially chest X-ray evaluation (provides insights into Gap 2).
(3) the proportion of patients diagnosed with TB who fail to get registered and started on TB treatment, also known as "pretreatment loss to follow-up" (PTLFU) or "initial default" (Gap 3).
For the third aim, we should note that the Government of India's Revised National Tuberculosis Control Programme (RNTCP) has a relatively restrictive definition of PTLFU as consisting of smear-positive patients diagnosed in a given district, who are assumed to be living within the same district, but who are not initiated on treatment within the quarter (three-month calendar period) in which they were diagnosed. This definition has many shortcomings, especially the fact that many patients are diagnosed in one district but eventually migrate to another district to start treatment; these patients are excluded from the reporting of local PTLFU statistics.
Moreover, the RNTCP does not formally report PTLFU statistics in its annual report. Rather, we can only estimate this figure by calculating the difference between the number of smear-positive cases newly diagnosed at microscopy facilities and the number of smear-positive cases registered for treatment annually. This value may overestimate or underestimate the proportion of PTLFU cases. Therefore, we conduct our meta-analysis of studies that independently assess and more accurately report local PTLFU rates to help crosscheck this national estimate of PTLFU.

Search strategy
A medical librarian searched PubMed, Embase, Web of Science, and the Cochrane Register of Controlled Clinical Trials for studies published between January 1, 2000 and February 26, 2015, without language restrictions, using search terms for "tuberculosis", "India", and "loss to followup", including "pretreatment loss to follow-up" and "initial default" (Table D). In addition, we carried out electronic searches of key Indian journals that may not be indexed in the above databases: the Indian Journal of Tuberculosis, Lung India, the Indian Journal of Chest and Allied Sciences, the India Journal of Public Health, and the Indian Journal of Community Medicine. Additional studies were identified by searching the reference lists of the primary studies and relevant review articles.

Inclusion and exclusion criteria
We included cross-sectional or cohort studies that audited records at government DMCs, tracked patients to understand attrition during the diagnostic workup for TB, or evaluated linkage of diagnosed smear-positive TB patients to care in the government sector. The studies had to evaluate at least one of the three variables listed above. Studies of private sector TB care, population-based studies, studies with field research conducted prior to the year 2000, and studies with solely qualitative methods were excluded. In addition, studies in which data were collected prior to the year 2000 (before the Government of India's DOT programme had broad coverage) were excluded. As per our quality criteria below, studies using convenience sampling or that include fewer than 150 patients were considered to be of very low quality and were also excluded from the meta-analysis.

Study selection
Citations identified by the search were independently assessed by two reviewers (authors RS and RN) for their eligibility (Fig C). Disagreements between the two reviewers were resolved by discussion between RS and RN or, if necessary, by consulting a third reviewer (S Satyanarayana).

Quality assessment
There are no well-recognized tools for evaluating the quality of studies included in systematic reviews of operational indicators within health systems (such as diagnostic default or pretreatment loss to follow-up). We created the following quality criteria relevant to the specific indicators being assessed in this systematic review (Table E): (1) Studies in which a dedicated research team was used to track pretreatment loss to follow-up (PTLFU) patients were rated as being higher in quality than studies that rely on self-report by local TB programs to evaluate PTLFU.
(2) Studies that assess PTLFU and loss to follow-up during the diagnostic workup within a shorter time frame (e.g., within 1-2 months after diagnosis) were rated as being higher in quality than studies that assessed these indicators 3 or more months after diagnosis, since tracking patients to find out accurate outcomes becomes much harder with time due to patient mobility.
(3) Studies using convenience sampling or that include fewer than 150 patients were considered to be of very low quality and were excluded from the meta-analysis.
Notably, some of these quality criteria do not apply to evaluating of the proportion of people with suspected TB who fail to submit two sputum smears, since this indicator can be easily assessed through an audit of DMC records and does not require patient tracking.

Data extraction and analysis
Two reviewers (RS and RN) independently extracted the data from each included study into a structured data extraction form. Disagreements were resolved by consulting a third reviewer (S Satyanarayana). From each study, we extracted information on the study design, location, setting (i.e., urban versus rural), sample size, study quality, and variables of interest (Tables F-H). We also extracted information on 95% confidence intervals (95% CIs) where available; if 95% CIs were not reported, we calculated these from the data provided, assuming an infinite population size.
We generated Forest plots for each variable for which data were available from at least five studies using Stata version 14 (College Station, TX, USA). We assumed that each study finding represents the local prevalence of a given indicator in that facility, city, or district in India. India is a diverse country with substantial differences in the quality of public sector services in every state. In addition, there are substantial differences in the socioeconomic status and cultural practices of the patient population in every state. Therefore, we allow that the proportion is likely to vary from study to study, representing meaningful local differences.
Given these assumptions, we conducted the meta-analyses for variables with data from 5 or more studies using a random effects model, and we performed meta-analysis even if there was substantial heterogeneity in the values in different studies. We report the pooled prevalence and heterogeneity (I 2 ) for the values from the included studies. The Forest plots and a narrative discussion of the results are included in the main text of this manuscript. Fig B. Algorithm for diagnosing smear-negative tuberculosis in India's Revised National Tuberculosis Control Programme (RNTCP) [1]. This algorithm is currently undergoing revision by the RNTCP. • Studies with data on failure to submit two sputum samples (n=6) • Studies with data on LTFU during the smear-negative diagnostic workup (n=3) • Studies with data on PTLFU (n=16) Note: Some studies have data on more than one variable. Tables   Table D. Search strategy to identify manuscripts regarding pretreatment loss to follow-up, failure to complete the diagnostic workup, and loss to follow-up on treatment for TB patients in India  Table G. Characteristics of the included studies for the meta-analysis of the proportion of patients who fail to complete the multi-step smear-negative TB diagnostic workup