https://orcid.org/0000-0003-2401-1065
Figures
Abstract
Tuberculosis (TB) continues to be the leading cause of death from infectious diseases, disproportionately affecting populations in low- and middle-income countries, where diagnostic and financial barriers impede timely care. While alternative treatments may offer symptomatic relief, the cornerstone of TB management remains anti-tubercular therapy (ATT) as recommended by the World Health Organization (WHO). Yet, many individuals with TB-like symptoms first consult traditional healers and AYUSH (Ayurveda, Yoga, Naturopathy, Unani, Siddha, and Homeopathy) practitioners, resulting in delays in diagnosis and treatment initiation. This quasi-experimental study evaluated the impact of strengthening linkages between non-allopathic and allopathic providers, through formal referral systems, alongside the implementation of a free chest X-ray (CXR) screening system for presumptive TB patients. A mixed methods design found the significant role of friends and family in influencing care-seeking patterns, while a regression analysis within a Difference-in-Differences approach found the intervention to reduce diagnostic delay by over eight days (-8.6, 95% CI: −16.6, −0.5). The accelerated pathway to diagnosis was observed irrespective of which provider was first visited by the patient, holding promise for future such interventions, especially in high-burden regions, with significant reliance on informal providers.
Citation: Sodhi R, Goel S, Baghel R, Rade K, Khaparde K, Dabas H, et al. (2026) Strengthening TB care pathways: Integrating informal and AYUSH practitioners for early diagnosis and referral. PLOS Glob Public Health 6(4): e0004416. https://doi.org/10.1371/journal.pgph.0004416
Editor: Julia Robinson, PLOS: Public Library of Science, UNITED STATES OF AMERICA
Received: February 23, 2025; Accepted: March 16, 2026; Published: April 8, 2026
Copyright: © 2026 Sodhi et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All of the data pertaining to this is available in a Github repository: https://github.com/ridhimasodhi/PatientPathway.
Funding: The baseline and endline assessments were funded by Wave 08 of the TB REACH program (Stop TB Partnership), awarded to the William J. Clinton Foundation. This grant supported the research and report preparation – but did not cover manuscript preparation. The Stop TB partnership had no involvement in study design, data collection, analysis, the decision to publish, or manuscript preparation. The authors received no additional funding for data compilation, statistical analysis, or manuscript preparation.
Competing interests: The authors have declared that no competing interests exist.
Abbreviations: ADITYA, Accelerating Diagnosis and Initiation of TB Treatment through your Action; CBNAAT, Cartridge-based Nucleic Acid Amplification Test; CI, Confidence interval; CME, Continuing Medical Education; CXR, Chest X-ray; DiD, Difference-in-differences; DxD, Diagnostic delay; FDC, Fixed Dose Combination; HCP, Health care provider; ICT, Information and Communication Technology; IDI, In-depth interview; IQR, Interquartile range; NAAT, Nucleic Acid Amplification Test; N-HCP, Number of health care providers consulted; OLS, Ordinary least squares; PD, Patient delay; TB, Tuberculosis; Xpert, Cartridge-based Nucleic Acid Amplification Test
Introduction
Tuberculosis (TB) remains a major public health challenge, particularly in low- and middle-income countries, where economic burdens exacerbate the disease’s impact [1–4]. Reliance on unqualified healthcare providers further delays diagnosis, increasing catastrophic expenditure and morbidity rates [2–4]. Dependence on informal providers is typically more common in rural areas [5], and studies indicate that more than 75% of primary care visits in India are made to them [5–7]. Provider shopping, and misdiagnosis, from unqualified as well as qualified providers, is common, and often results in prolonged delays, stretching for weeks or months before treatment begins [8–11]. Compounded further by non-specific antibiotic use and delay in diagnostic testing, these factors continue to be detrimental to TB elimination efforts within and outside India [8,12–18]. While significant provider-engagement efforts have been made [19–22], not many have integrated non-allopathic practitioners [2,4]. In Chhattisgarh, the region where this study is based, the breakdown of treated ailments by type of healthcare provider indicates that reliance on AYUSH (Ayurveda, Yoga & Naturopathy, Unani, Siddha, and Homeopathy) practitioners is 4.4% in rural areas and 1.4% in urban areas [23]. Systematic referral mechanisms, however, can be helpful in accelerating patient pathways, and help reduce catastrophic costs associated with such delays [2,24,25]. Set within Project ADITYA (Accelerating Diagnosis and Initiation of TB Treatment through your Action), the study evaluates the impact of formalizing referrals between non-allopathic and allopathic providers, whilst providing free diagnostic testing (Chest X-Ray or CXR; Nucleic Acid Amplification Test or NAAT) through an incentive and voucher-based model. Beyond analyzing TB patient pathways and their evolution during the intervention, this study provides a scalable roadmap for implementing similar strategies across other regions.
Methods
Drug sensitive TB Care services under Project Aditya
Project ADITYA was launched in February 2021 in Durg district, Chhattisgarh, as an inclusive private sector model for TB diagnosis and treatment, and was implemented through September 2022. Initiated with mapping all providers in the region, encompassing both qualified (AYUSH and MBBS/MD), and unqualified (those without a recognized medical degree) providers, alongside pharmacists, the project aimed to reduce diagnostic and treatment delays among DS-TB patients, through three components: a) voucher-based system providing free CXRs for presumptive TB patients through referrals to private laboratories, b) ensuring engaged providers can provide patients access to government funded/ free Xpert (Cartridge-cased NAAT) and fixed dose combination drugs (FDCs), and c) establishing an Information and Communication Technology (ICT) system for continuous patient management for TB patients diagnosed at facilities associated with Project Aditya. Engagement with both allopathic and non-allopathic providers was conducted through repeated interactions, including orientation meetings and continuing medical education (CME) sessions organized by the program team, and was supported by informal communication channels (e.g., WhatsApp groups) to facilitate information exchange. Of the more than 1,600 providers mapped during the intervention, 64% were engaged with the project, and 35% actively referred/notified patients. A total of 18 CMEs were organized during the intervention, which brought together all engaged providers, and were used to continually update providers on TB guidelines, project updates, and address any questions as well. The CXR referral incentive was set at a modest INR 100 (~ 1 USD), which was received by AYUSH and unqualified practitioners if a referred patient got a confirmed TB diagnosis. Upon diagnosis, these providers could also serve as treatment supporters, receiving a sum of INR 500 at each treatment initiation and completion, totaling INR 1,000 or ~12 USD for each successfully treated patient.
Study design
We used a mixed methods quasi-experimental intervention design, collecting quantitative and qualitative data concurrently [26,27]. The quantitative component employed a Difference-in-Differences (DiD) approach to assess the impact of the intervention on TB patient pathway, while thematic analysis of qualitative interviews examined how care seeking patterns evolved over time. The parallel trends assumption underlying the DiD framework is supported by the geographic and demographic uniformity of the study area across baseline and endline periods. Also, no other major TB interventions were introduced in the region during the study period, reducing the risk of time-varying confounding.
Questionnaire design
A semi-structured questionnaire was used for in-depth interviews (IDIs) at baseline and endline. Lasting 45–120 minutes, the interviews assessed patient pathways, covering the journey from symptom onset to treatment completion. The endline assessment included additional questions for those engaged with Project Aditya, which included aspects of treatment support received during the treatment, an additional feature of the intervention. These questions were not analyzed as part of the assessment, as the latter focuses on patient pathway until the diagnosis/ treatment initiation.
Data source
Three data sources were used. First, Ni-kshay, the Government of India’s national TB surveillance portal, provided programmatic data used to sample TB patients for both baseline and endline assessments. Second, the Project ADITYA Management Information System (MIS) was used to identify project beneficiaries. This included both presumptive patients (who received CXR) and confirmed TB patients (who were further assessed through clinical and diagnostic assessment). Demographic details such as age, gender, and diagnosing district were extracted from these programmatic sources and later verified through patient interviews. Third, interview responses collected during the baseline and endline surveys provided detailed information on care-seeking patterns, diagnostic and treatment delays, and patient experiences. These datasets together informed both the quantitative and qualitative components of the analysis.
Data selection (inclusion & exclusion criteria)
All adult (>=18 years) DS-TB patients, undergoing treatment in Durg, Chhattisgarh, were eligible. Individuals with drug-resistant or extrapulmonary forms of TB were excluded.
Sampling method
Proportionate random sampling was used for baseline, ensuring representativeness across age and gender. The endline employed stratified proportionate random sampling, with strata defined by whether the patients were ADITYA beneficiaries. Type of provider at first contact was also considered for proportional representation.
Sample size & sampling process
A total of 101 individuals were interviewed at baseline (34 women and 67 men) and 203 at endline. The larger endline sample reflected the inclusion of two strata - intervention and non-intervention groups. Sample size determination balanced pragmatic and scientific considerations: ensuring sufficient power to estimate intervention effects while capturing qualitative depth for thematic saturation [28–30].
Patients were pre-screened using a combination of Project MIS & Nikshay database, in order to ensure a) demographic representativeness, and b) confirm accurate tagging of patients across the intervention and control groups during the endline. Patients were telephonically contacted to obtain consent; approximately 60–70% declined, due to migration, work commitments, stigma, or the interview’s duration. The denial rate reported for similar qualitative health studies, especially those involving disadvantaged or stigmatized populations, ranges from 30% to 89% [31,32]. No financial or material incentives were offered, which likely contributed to the latter. The final sample size is described in Table 1.
Outcomes of interest
Two primary objectives were defined. First, to conduct a holistic assessment of the TB patient pathway, examining symptom onset, time to diagnosis, provider shopping, diagnostic testing, and associated expenditures, alongside understanding evolving patient and provider attitudes between the two time periods. The second objective was to quantitatively assess the intervention’s impact on patient delay, number of providers visited prior to diagnosis, and diagnostic delay, independent of temporal, clinical and demographic factors. The outcomes of interest were grouped under two domains for ease of assessment:
- TB Patient Pathway: First point of contact for the individual (First POC), Patient delay (PD), First healthcare provider (First HCP), Primary reason for choosing provider, Number of providers consulted prior to diagnosis (N-HCP), and Diagnostic delay (DxD)
- Impact of intervention: Patient delay (PD), Number of providers visited (N-HCP), and Diagnostic delay (DxD)
Patient delay (PD) is defined as the number of days between symptom onset, and the first contact with an HCP. Diagnostic delay (DxD) is defined as the number of days between the first consultation with an HCP and confirmed diagnosis/ treatment initiation. Total delay is the sum of PD and DxD, and reflects the time from symptoms to diagnosis/ treatment initiation.
Analysis
Qualitative interviews were coded thematically to capture patient experiences, barriers to care, and other pathway elements. Quantitative analysis was used to assess differences in key outcomes across intervention and non-intervention groups. Nonparametric tests (Kruskal Wallis rank-sum test) assessed statistical differences, while causal effects were estimated using DiD. Whenever presented, confidence intervals (CI) represent 95% interval values. For patient and diagnostic delay, Ordinary Least Squares (OLS) with robust standard errors were used. To explore distributional heterogeneity, quantile regressions were estimated at the median (τ = 0.5) and upper quartile (τ = 0.75). For the number of providers consulted, a Poisson model was used considering the non-negative and discrete nature of the variable. The standard modeling specification is provided in the supplementary text. Covariate selection was done basis a combination of theoretical relevance and prior literature on TB care-seeking behaviour. A series of models were fitted, sequentially adding demographic characteristics (age, gender, type of household, occupation), the first point of contact, first provider type, loss of income or/and job, symptom severity, comorbidities, TB history of individual/family, and primary drivers for choosing a provider. Best fit was measured by adjusted R Squared.
Robustness checks
To test the validity of our findings across the outcome distribution, particularly given the positively skewed nature of delay variables, quantile regressions at the 50th and 75th percentiles were conducted. Sparse categories (fewer than 10 observations) for first point of contact and occupation were combined to address convergence issues. Koenker and Machado’s pseudo R-squared values are reported for these models.
Quantile treatment effect (QTE) plots across the 10th to 90th percentiles (τ = 0.1 to 0.9) are provided in the supplementary information to illustrate intervention’s impact across the distribution of patient delay [Fig A in S1 Appendix]. Potential overfitting and multicollinearity were assessed using the Generalized Variance Inflation Factors (GVIF), all of which were well below the accepted threshold of two [Tables A-C in S1 Appendix]. Additional sensitivity checks included analyzing the impact of different provider types on diagnostic delay [Table D in S1 Appendix], as well as number of providers visited [Table E in S1 Appendix].
Software
Audio recordings were manually transcribed in Excel and organised by thematic categories. The cleaned data were stored as CSV files for quantitative and qualitative analysis. ATLAS.ti was used for thematic coding, and RStudio (R version 2022.07.01) for statistical analysis and visualisation, using the packages dplyr, tidyr, gtsummary, and ggplot2.
Data consolidation and confidentiality
Audio recordings were stored on a password-protected hard drive accessible only to the core research team. Transcriptions were anonymised by removing identifiable information, and each participant was assigned a unique identification code prior to analysis.
Ethical considerations
Institutional ethics clearance was obtained for both study phases. The baseline received clearance from the Sigma Institutional Review Board, New Delhi (Ref. No. 10052/IRB/20–21; 22 January 2021), and the endline from the TRIOs Institutional Review Board, New Delhi (Ref. No. 101/2022/013/I/CHAI; 15 December 2022).
Recruitment for the baseline occurred between 17 February and 10 March 2021, and for the endline between 1 January and 1 February 2023. Informed consent was sought for all interactions between the patient and researcher, and no minors were interviewed.
Results
Demographic summary
A total of 101 and 203 DS-TB patients were interviewed during the baseline and endline phases, wherein the latter were further segregated by intervention (123) and non-intervention groups (80) [Table 2]. Majority of patients (>61%) interviewed were male across the phases, with mean age among the baseline cohort to be significantly lower (37 years vs 41 years and 45 years for control and endline groups, respectively; p = 0.003). The share of unemployment across phases varied between 40% to 48%. Patients with reported TB history in self or family varied between 18% and 21%. The endline cohort had a significantly higher share of individuals with severe symptoms (43%-46% vs 26%, p = 0.007), but lower share with comorbidities (27%-34% vs 57%, p < 0.001). While distress sales due to TB-related expenditures were similar across phases (20%-21%), the share of those losing jobs were higher at baseline (25%) than at endline-control (5%) and endline-intervention (11%).
TB patient pathway
First point of contact (First POC).
Majority of TB patients (73% - 84%) reported sharing symptoms with friends and family [Table 2]. Other points of contact include community healthcare workers (CHWs), pharmacists, village doctors, and AYUSH practitioners. Endline assessment included an additional probe asking if individuals went directly to allopathic providers, to which 10% and 21% of individuals responded in admission from the intervention and control group, respectively. Such a probe was not present for the baseline assessment.
Patient delay (PD).
Individuals from the endline cohort reportedly took 10 median days to seek medical consultation from the onset of their symptoms, relative to 35 median days taken by the baseline cohort [Table 2, Fig 1].
Note: Dotted lines represent cohort medians, x-axis censored beyond 100th day for visual clarity. Negligible data lost (6 and 7 points lost for patient and diagnostic delay, respectively).
First healthcare provider (First HCP).
Individuals from the endline cohorts were significantly more likely to seek consultation from an allopathic provider (78% and 63% for the control and intervention groups, respectively), relative to those in the baseline cohort (47%). Subsequently, those from endline were less likely to seek consultation from pharmacy, as well as unqualified providers. The share of those consulting with AYUSH were highest from the endline-intervention group (13%) [Table 2].
Drivers behind choosing provider.
Primary factors influencing provider choice include affordability, family networks, proximity, and doctor referrals. Reliance on proximity decreased between phases – from 43% during baseline to 8.8% and 12% for endline control and intervention groups, respectively [Table 2]. Reliance on referrals, on the other hand, increased from 14% (baseline) to 24%-25% (endline). Affordability was a key concern for the baseline group (26%), relative to those from the endline group (7.5% and 4.1% for control and intervention groups, respectively).
Number of total providers visited (N-HCP).
The number of total providers visited before diagnosis decreased from 3 median providers at baseline to 1 median provider at endline [Table 2]. For both the control and intervention cohorts of endline, most patients (>=98%) were diagnosed after visiting one or two providers, relative to less than 20% at baseline [Table 2].
Diagnostic delay (DxD).
Days between the first provider visit and TB diagnosis reduced significantly between baseline (Median = 27, Mean = 32) and endline (Median = 3, Mean = 10 for control; Median = 4, Mean = 17 for intervention) phases [Table 2]. Mean diagnostic delay typically increased with the number of providers, but the variation was better explained by the assessment periods [Table 3, Fig 2]. Median values remained relatively stable, while mean values shifted more sharply, suggesting that the change in delay was driven by extreme values, rather than the middle percentiles.
Note: The dashed line connects group medians.
Impact of intervention (Difference in Differences - DiD)
DiD was used to estimate the impact of the intervention on three variables – patient delay, number of providers visited prior to diagnosis, and diagnostic delay. A series of regression models were fitted before deciding on the models presented, decided based on a combination of model parsimony and adjusted R-squared values.
Patient delay (PD).
The intervention is not observed to have any impact on patient delay, a finding consistent with OLS and quantile regressions [Table 4, Fig A in S1 Appendix]. The decline in patient delay between baseline and endline is estimated to be driven significantly by time factors (-19.4 days, C.I.: -26.8, -12.0), with consistent effects at the median (-23.9 days) and 75th percentile (-17.9 days). Comorbidities are observed to significantly reduce patient delay by about 5 days, an effect consistent at the median (3.6 days), but not at the upper (75th) quartile of the distribution. Symptom severity, on the other hand, is estimated to decrease patient delay by 5 days, but with a lower effect at the median (2.2 days), which goes insignificant at the upper quartile. Choosing providers based on proximity was linked to increase delay by more than a week (9 days, CI: -0.5, 18.3), but the association was not significant at the median and 75th percentile.
Number of providers visited (N-HCP).
While the model estimates the intervention to have negligible impact on the number of providers visited (IRR = 1.02, CI: 0.77, 1.27), temporal effects alone explain about 50% decline in providers consulted (IRR = 0.49, CI: 0.23, 0.75) [Table 5]. Compared with allopathic providers, patients who first sought care from AYUSH providers (IRR = 1.68; CI: 1.16–2.36), pharmacies/self-medication (IRR = 1.46; CI: 1.14–1.86), or unqualified providers (IRR = 1.48; CI: 1.18–1.85) are estimated to have significantly higher provider visits prior to diagnosis.
Diagnostic delay (DxD).
The intervention is expected to significantly reduce diagnostic delay by more than 8 days (-8.6, CI: -16.6, -0.5), but the effect is not significant at the median or the 75th percentile [Table 6]. Time effects account for a significant majority of the reduction (-11.2 days, CI: -21.7, -0.7), and the effects stay consistent at the median (−21.00 days) and upper quantile (-17.6 days) [Table 6]. History of TB in patient or family was associated with a slight reduction in diagnostic delay, but no other covariate showed consistent association. The OLS model predicted diagnostic delay to be consistently shorter in the intervention group, across all provider types [Fig 3], a finding consistent with another model specification with an interaction term on provider type [Table D in S1 Appendix].
Discussion
To our knowledge, this is the first quasi-experimental mixed-method study to assess the impact of an all-inclusive provider intervention on the TB patient pathway. By reflecting real-world care-seeking behaviour, where many patients do not initially consult MBBS/MD-qualified doctors, the analysis finds evidence underscoring the impact of incentivized referral mechanisms between providers on reducing diagnostic delay associated with preference for alternative and unqualified doctors, provider shopping, and lack of access to diagnostic services. The findings also highlight the significant influence of friends and family in care-seeking patterns, reinforcing the need for family centered and community-based approaches for TB elimination [Fig 4, Table 7].
Note: The figures given against certain factors (such as share of AYUSH doctors visited) are taken from Tables 2–5, and are approximated.
TB patient pathway
First point of contact.
Social networks and personal beliefs have historically been found to account for a majority (42%) of the care-seeking behaviour [33]. Advances in telemedicine may influence this in times to come, but our results as well as prior literature find that individuals typically discuss their symptoms first with a trusted friend or a family member, and not a doctor [34,35]. If at all a faster or initial contact is made with a qualified professional, the same is typically linked to proximity with the doctor, or linkage through employment-linked facilities [36–38].
Patient delay (onset of symptoms to seeking professional care).
In the National Tuberculosis Prevalence Survey 2019–2021, it was observed that up to 64% of those with presumptive TB symptoms or signs did not immediately seek care [39]. Prior research finds that individuals are likely to visit a medical professional (or not) based on advice given by close contacts, but prefer seeking a professional opinion in case of conflicting guidance by different friends and family members [33–35]. Community-based approaches encouraging a consultation with medical professional are hence, likely to have a positive effect on timely care-seeking [33,34,39]. Median patient delay in our study was just above a month (Median = 35, Mean = 42), aligning with existing estimates [8,40]. This is primarily linked to initial misattribution of symptoms to minor ailments such as seasonal flu or cough, and limited awareness of TB symptoms. Similar behavioral patterns, including perceiving symptoms as non-serious or believing that they will resolve on their own, have been documented in several Indian and global studies [31,35]. Subsequently, self-diagnosis or/and home remedies are typical treatment patterns followed, prior to consultations with qualified or/and unqualified doctors [37,38,41]. In addition, economic factors also influence care seeking. A 2017 study from South India found income to be associated with both patient delay and provider choice [9]. In our analysis, individuals with government employment sought professional care almost a week earlier than those in agricultural occupations, and more than two weeks earlier than those who are self-employed, likely reflecting differences in income and the opportunity cost of time [42,36,43]. In terms of prior health metrics, both symptom severity and comorbidities significantly affected care-seeking behaviour, though in opposite directions: severe symptoms prompted consultation approximately five days earlier, whereas comorbidities tended to delay it, as also reported in earlier studies [13,15,40].
“I just had a cough and did not think it would be something like TB. I thought it will go away in a few days, and used home remedies. When it did not go away for 8-9 days, I consulted with ASHA who referred me to visit a doctor in the city” – a patient when asked why they chose home remedies
“I followed their advice for a week but when there was no change in the symptoms, I consulted with a doctor” – a patient who bought medication from a pharmacy, following their parents’ advice
First Provider Type & Drivers behind choosing providers.
When (professional) care is eventually sought, informal providers are the first provider choice for 30% of patients at baseline. This is aligned with earlier literature indicating that stigma, limited resources, and familiarity with local providers often shapes early care-seeking, making it difficult for the individual to pursue formal medical consultation [14,35–37]. Such patterns are particularly pronounced in high-burden, low-resource settings, where traditional and informal healthcare options are more accessible and trusted than formal facilities [13,44–46]. Consistent with earlier research [37,47,48], our baseline data shows that only 47% consulted with an allopathic doctor (at first), while 23% directly went to a pharmacy. In addition to misdiagnosis, the ease of accessing pharmacies, especially during COVID 19, was also found to be a critical factor.
“I had cough for a couple of days and then it disappeared. I thought that it must been because of the weather change. But then a couple of months, it came back. I decided to see a doctor then” – a patient who first sought help from an unqualified doctor
“It is nearby, and they treat properly, everybody visits that doctor in our village, so I also went there” – a patient who sought help from an unqualified doctor
Diagnostic Delay & Number of Providers visited.
Reliance on unqualified practitioners, is a particular detriment to accurate diagnosis, owing to their limited diagnostic capabilities, as well as frequent misdiagnosis and delayed referral patterns, even when symptoms do not subside [4,13,37,47–50]. Many unqualified practitioners, if not all, dispense allopathic treatments such as antibiotics and injections, without formal training, or approvals [51,52]. When mired by multiple consultations and misdiagnosis with other qualified providers, it typically takes weeks or even months, for an accurate diagnosis to be reached [53,54]. Interventions such as the one assessed, have the potential to advance knowledge among unqualified providers as well, limiting unethical provider behaviour [7]. Our estimated baseline delay (27 median days) is consistent with both recent evidence (22–32 median days) [55,56], and earlier systematic reviews (31 median days) [57]. Effective referral mechanisms are therefore critical to link patients with qualified practitioners, who are better equipped to identify TB-like symptoms and initiate diagnostic testing [4,13,49,50,53,54]. Other significant determinants of diagnostic delay include comorbidities and symptom severity, which operate in opposite directions: comorbidities may mask TB symptoms and delay diagnosis [13,15,58,59], while more severe symptoms prompt earlier care-seeking.
Impact of intervention
While no significant impact on patient delay and number of providers (visited prior to diagnosis) is observed, the intervention is associated with a significant reduction in diagnostic delay (8 + days), even after adjusting for temporal effects and demographic, clinical, and care-seeking characteristics. The results, while significant, are not consistent across the patient pool, especially those who have moderate delays. This suggests that the intervention likely mitigated extreme diagnostic lag, plausibly benefiting the most at-risk, but stronger engagement efforts are needed to bring about a reduction across the presumptive patient pool. The encouraging insight is that the intervention reduced diagnostic delay irrespective of the provider type visited by the patient [Fig 3], which suggests that formal referral linkages were instrumental in reducing time between provider visits. The same can also help reduce associated catastrophic costs, often attributed to initial alternative treatments and misdiagnosis [60,61]. Previous research has found that strengthening provider-provider and CHW-provider linkages can stimulate better use of referral services and healthcare facilities [62]. In our study, while job losses were observed to decline between phases (25% at baseline to 5%-11% at endline), distress selling remained at 20%, across the three cohorts. While we cannot be certain if reductions in diagnostic delay affected these metrics, better policies, which combine benefits across the continuum of the patient pathway, are needed to reduce the financial burdens faced by the most vulnerable [25,60,61,63]. Studies in India, and outside, find that informal providers are willing and able to play more formal roles (referral, treatment support) if properly engaged and supported [7,64]. Combined with evidence from other Indian states, and high-burden TB regions, the intervention illustrates itself as a scalable solution to accelerate timely diagnosis, while asking for a more sustained effort if one is to bring about long term changes [45,46,53].
COVID 19 and time effects
Much of the baseline period (early 2021) coincided with the months following the first wave of COVID-19, whereas the endline (early 2023) reflected a period of post-pandemic normalization in healthcare services. This broader context may partly explain the reduction in both patient and diagnostic delays, which were captured as time effects in the regression model. It may also account for the greater reliance on allopathic providers observed during the endline period. Post-pandemic recovery, including the restoration of diagnostic services, heightened public awareness of respiratory illnesses, and renewed trust in formal healthcare, likely influenced these shifts in care-seeking behaviour and the reduction in the number of providers visited prior to diagnosis. These explanations are offered as post-hoc hypotheses to contextualize observed trends, and were not formally analyzed in this study. Nevertheless, they highlight how system recovery and behavioural shifts can reshape patient pathways.
Future research
Future studies could assess the implementation challenges of such inclusive provider engagement models, with particular focus on provider experience. Qualitative feedback from providers on their motivations and perceived barriers, especially when combined with quantitative metrics on referrals made, could inform the design of more effective interventions. In addition, estimating cost effectiveness of similar programs across diverse epidemiological and health-system contexts could guide targeted investments, not only for TB but also for other health conditions.
Conclusion
Our analysis demonstrates that strengthening provider linkages between formal and informal providers, supported by ease of diagnostics, can substantially reduce diagnostic delays among individuals with TB-like symptoms. Key elements, such as leveraging community and familial networks, using CXR as a cost-effective screening tool, and building trust-based referral systems, offer a scalable framework to streamline TB care pathways. The approach is replicable across other high-burden regions, both within and beyond India, and can be especially useful where informal providers are the first point of contact. Future provider engagement strategies should consciously integrate community healthcare workers and informal providers, broadening the system’s reach to the very start of the patient pathway and enabling greater reductions in diagnostic delay.
Limitations
While comprehensive, certain limitations from the study must be acknowledged. First, recruitment challenges, common in qualitative field research, may have introduced some sampling bias. While most interviews were conducted at participants’ homes, others were held in neutral locations such as workplaces or parks to reduce stigma and time constraints. Despite these efforts, some groups, particularly migrants and highly mobile workers, may have been underrepresented. Second, although interviews were designed to optimize recall of care-seeking timelines, minor inconsistencies may have affected the precision of delay estimates. However, as the DiD analysis focuses on relative changes, any recall bias is likely non-differential across periods. Third, the study was conducted in a predominantly urban district of India, but with significant rural and per-urban pockets. The findings, hence, may not be generalizable to regions with different healthcare infrastructure, socio-economic conditions, and cultural practices. Fourth, unmeasured variables may have influenced patient pathways beyond those captured by our covariates. Finally, the study period overlapped with the COVID-19 pandemic, making it difficult to fully separate intervention effects from broader post-pandemic changes in care-seeking behaviour.
Supporting information
S1 Appendix. Additional tables and figures consisting of robustness tests, diagnostic tests on multicollinearity.
The text also includes additional tables to support understanding relationship of comorbidities, symptom severity and occupation on patient pathway (choice of first point of care, first provider type, and mean values of patient delay & diagnostic delay).
https://doi.org/10.1371/journal.pgph.0004416.s001
(DOCX)
References
- 1.
Global Programme on Tuberculosis & Lung Health. WHO Global TB Report - 2025. 2025 Nov. Available: https://www.who.int/publications/i/item/9789240116924
- 2. Thapa P, Narasimhan P, Jayasuriya R, Hall JJ, Mukherjee PS, Das DK, et al. Barriers and facilitators to informal healthcare provider engagement in the national tuberculosis elimination program of India: An exploratory study from West Bengal. PLOS Glob Public Health. 2023;3(10):e0001390. pmid:37792715
- 3. Hossain S, Zaman K, Quaiyum A, Banu S, Husain A, Islam A, et al. Care seeking in tuberculosis: results from a countrywide cluster randomised survey in Bangladesh. BMJ Open. 2014;4(5):e004766. pmid:24871537
- 4. Thapa P, Hall JJ, Jayasuriya R, Mukherjee PS, Beek K, Das DK, et al. What are the tuberculosis care practices of informal healthcare providers? A cross-sectional study from Eastern India. Health Policy Plan. 2022;37(9):1158–66. pmid:35920775
- 5. Kumah E. The informal healthcare providers and universal health coverage in low and middle-income countries. Global Health. 2022;18(1):45. pmid:35477581
- 6. Bloom G, Standing H, Lucas H, Bhuiya A, Oladepo O, Peters DH. Making health markets work better for poor people: the case of informal providers. Health Policy Plan. 2011;26 Suppl 1:i45-52. pmid:21729917
- 7. Das J, Chowdhury A, Hussam R, Banerjee AV. The impact of training informal health care providers in India: A randomized controlled trial. Science. 2016;354(6308):aaf7384. pmid:27846471
- 8. Sreeramareddy CT, Panduru KV, Menten J, Van den Ende J. Time delays in diagnosis of pulmonary tuberculosis: a systematic review of literature. BMC Infect Dis. 2009;9:91. pmid:19519917
- 9. Van Ness SE, Chandra A, Sarkar S, Pleskunas J, Ellner JJ, Roy G, et al. Predictors of delayed care seeking for tuberculosis in southern India: an observational study. BMC Infect Dis. 2017;17(1):567. pmid:28806911
- 10. Deo S, Singh S, Jha N, Arinaminpathy N, Dewan P. Predicting the impact of patient and private provider behavior on diagnostic delay for pulmonary tuberculosis patients in India: A simulation modeling study. PLoS Med. 2020;17(5):e1003039. pmid:32407407
- 11. Satyanarayana S, Subbaraman R, Shete P, Gore G, Das J, Cattamanchi A, et al. Quality of tuberculosis care in India: a systematic review. Int J Tuberc Lung Dis. 2015;19(7):751–63. pmid:26056098
- 12. Sharma AK, Gupta N, Verma S, Chandran A, Dixit R. A study on procedural delay in diagnosis and start of treatment in drug resistant tuberculosis under RNTCP. Indian J Tuberc. 2019;66(3):394–401. pmid:31439186
- 13. Storla DG, Yimer S, Bjune GA. A systematic review of delay in the diagnosis and treatment of tuberculosis. BMC Public Health. 2008;8:15. pmid:18194573
- 14. Suganthi P, Chadha VK, Ahmed J, Umadevi G, Kumar P, Srivastava R, et al. Health seeking and knowledge about tuberculosis among persons with pulmonary symptoms and tuberculosis cases in Bangalore slums. Int J Tuberc Lung Dis. 2008;12(11):1268–73. pmid:18926036
- 15. Ananthakrishnan R, Jeyaraju AR, Palani G, Sathiyasekaran B. Care seeking behavior of the TB patients who were registered in an urban government tuberculosis control in Chennai, Tamilnadu, India. J Clin Diagn Res. 2012.
- 16. Said K, Hella J, Mhalu G, Chiryankubi M, Masika E, Maroa T, et al. Diagnostic delay and associated factors among patients with pulmonary tuberculosis in Dar es Salaam, Tanzania. Infect Dis Poverty. 2017;6(1):64. pmid:28335816
- 17. Titahong CN, Ayongwa GN, Waindim Y, Nguafack D, Kuate AK, Wandji IAG. Patient-pathway analysis of tuberculosis services in Cameroon. Trop Med Infect Dis. 2021;6:171.
- 18. Zhang L, Weng T-P, Wang H-Y, Sun F, Liu Y-Y, Lin K, et al. Patient pathway analysis of tuberculosis diagnostic delay: a multicentre retrospective cohort study in China. Clin Microbiol Infect. 2021;27(7):1000–6. pmid:33421578
- 19.
Joint Effort for Elimination of Tuberculosis. JEET Report (2018-2020). 2021. Available: https://www.finddx.org/wp-content/uploads/2023/02/20220105_rep_jeet_2018-2020_FV_EN.pdf
- 20. Sodhi R, Penkunas MJ, Pal A. Free drug provision for tuberculosis increases patient follow-ups and successful treatment outcomes in the Indian private sector: a quasi experimental study using propensity score matching. BMC Infect Dis. 2023;23(1):421. pmid:37344775
- 21. Sodhi R, Vatsyayan V, Panibatla V, Sayyad K, Williams J, Pattery T, et al. Impact of a pilot mHealth intervention on treatment outcomes of TB patients seeking care in the private sector using Propensity Scores Matching-Evidence collated from New Delhi, India. PLOS Digit Health. 2024;3(9):e0000421. pmid:39259731
- 22. Yassin MA, Kimbo L, Wandwalo E, Rashid A, Dzokoto A, Weber U, et al. Leveraging Global Fund’s investments to expand innovative public-private provider engagement in TB. IJTLD Open. 2024;1(6):250–7. pmid:39021451
- 23.
Key Indicators of Social Consumption in India: Health; NSS 75th Round (2017-2018). Ministry of Statistics and Programme Implementation, India; 2019 Nov p. Table A8. Available: https://www.mospi.gov.in/sites/default/files/publication_reports/KI_Health_75th_Final.pdf
- 24. Thapa P, Narasimhan P, Beek K, Hall JJ, Jayasuriya R, Mukherjee PS, et al. Unlocking the potential of informal healthcare providers in tuberculosis care: insights from India. BMJ Glob Health. 2024;9(2):e015212. pmid:38413099
- 25. Chatterjee S, Das P, Stallworthy G, Bhambure G, Munje R, Vassall A. Catastrophic costs for tuberculosis patients in India: Impact of methodological choices. PLOS Glob Public Health. 2024;4(4):e0003078. pmid:38669225
- 26.
Creswell JW, Plano Clark VL. Designing and conducting mixed methods research. Third Edition ed. Los Angeles: SAGE. 2018.
- 27.
Kennedy TD, Edmonds WA. An applied guide to research designs: quantitative, qualitative, and mixed methods. Second Edition ed. Los Angeles: SAGE. 2017.
- 28. Robinson OC. Sampling in interview-based qualitative research: A theoretical and practical guide. Qual Res Psychol. 2014;11:25–41.
- 29. Fusch PI, Ness LR. Are We There Yet? Data Saturation in Qualitative Research; The Qualitative Report (Volume 9). 2015. Available: https://scholarworks.waldenu.edu/facpubs/455/
- 30.
Ritchie J, Lewis J, editors. Qualitative research practice: a guide for social science students and researchers. London; Thousand Oaks, Calif: Sage Publications; 2003.
- 31. Robinson OC. Sampling in interview-based qualitative research: A theoretical and practical guide. Qualitative Research in Psychology. 2014;11:25–41.
- 32. Bonevski B, Randell M, Paul C, Chapman K, Twyman L, Bryant J, et al. Reaching the hard-to-reach: a systematic review of strategies for improving health and medical research with socially disadvantaged groups. BMC Med Res Methodol. 2014;14:42. pmid:24669751
- 33. Berkanovic E, Telesky C, Reeder S. Structural and social psychological factors in the decision to seek medical care for symptoms. Med Care. 1981;19(7):693–709. pmid:7266118
- 34. Onuegbu C, Harlock J, Griffiths F. Navigating the Lay Referral System for Treatment-Seeking Decisions During Illness in the Digital Age: A Qualitative Study of Adults Living in Slums in Nigeria. Sociol Health Illn. 2025;47(2):e13863. pmid:39614807
- 35. Lyons N, Bhagwandeen B. Examining healthcare needs and decisions to seek health services among Venezuelan migrants living in Trinidad and Tobago using Andersen’s Behavioral Model. Front Public Health. 2023;11:1212825. pmid:37900018
- 36. Andersen RM. Revisiting the behavioral model and access to medical care: does it matter?. J Health Soc Behav. 1995;36(1):1–10. pmid:7738325
- 37. Mistry N, Lobo E, Shah S, Rangan S, Dholakia Y. Pulmonary tuberculosis in Patna, India: Durations, delays, and health care seeking behaviour among patients identified through household surveys. J Epidemiol Glob Health. 2017;7(4):241–8. pmid:29110864
- 38. Samal J. Health-seeking Behavior of Tuberculosis Patients in an Eastern Indian State: A Cross-sectional Study and Review of Literature. Journal of Datta Meghe Institute of Medical Sciences University. 2024;19(4):722–8.
- 39.
Government of India, Ministry of Health and Family Welfare. National Tuberculosis Prevalence Survey 2019-2021. 2021.
- 40. Santos JA, Leite A, Soares P, Duarte R, Nunes C. Delayed diagnosis of active pulmonary tuberculosis - potential risk factors for patient and healthcare delays in Portugal. BMC Public Health. 2021;21(1):2178. pmid:34837969
- 41. Liefooghe R, Baliddawa JB, Kipruto EM, Vermeire C, De Munynck AO. From their own perspective. A Kenyan community’s perception of tuberculosis. Trop Med Int Health. 1997;2(8):809–21. pmid:9294551
- 42. Alkhawaldeh A, ALBashtawy M, Rayan A, Abdalrahim A, Musa A, Eshah N. Application and use of Andersen’s behavioral model as theoretical framework: a systematic literature review from 2012–2021. Iran J Public Health. 2023.
- 43. Akalu TY, Clements ACA, Gebreyohannes EA, Xu Z, Bai L, Alene KA. Risk factors for diagnosis and treatment delay among patients with multidrug-resistant tuberculosis in Hunan Province, China. BMC Infect Dis. 2024;24(1):159. pmid:38308252
- 44. Gugssa Boru C, Shimels T, Bilal AI. Factors contributing to non-adherence with treatment among TB patients in Sodo Woreda, Gurage Zone, Southern Ethiopia: A qualitative study. J Infect Public Health. 2017;10(5):527–33. pmid:28189508
- 45. Washington R, Potty RS, Rajesham A, Seenappa T, Singarajipura A, Swamickan R, et al. Is a differentiated care model needed for patients with TB? A cohort analysis of risk factors contributing to unfavourable outcomes among TB patients in two states in South India. BMC Public Health. 2020;20(1):1158. pmid:32709228
- 46. Pradipta IS, Idrus LR, Probandari A, Lestari BW, Diantini A, Alffenaar J-WC, et al. Barriers and strategies to successful tuberculosis treatment in a high-burden tuberculosis setting: a qualitative study from the patient’s perspective. BMC Public Health. 2021;21(1):1903. pmid:34670527
- 47. Kapoor SK, Raman AV, Sachdeva KS, Satyanarayana S. How did the TB patients reach DOTS services in Delhi? A study of patient treatment seeking behavior. PLoS One. 2012;7(8):e42458. pmid:22879990
- 48. Mistry N, Rangan S, Dholakia Y, Lobo E, Shah S, Patil A. Durations and Delays in Care Seeking, Diagnosis and Treatment Initiation in Uncomplicated Pulmonary Tuberculosis Patients in Mumbai, India. PLoS One. 2016;11(3):e0152287. pmid:27018589
- 49. Bam K, Bhatt LP, Thapa R, Dossajee HK, Angdembe MR. Illness perception of tuberculosis (TB) and health seeking practice among urban slum residents of Bangladesh: a qualitative study. BMC Res Notes. 2014;7:572. pmid:25163889
- 50. Suseela RP, Shannawaz M. Engaging the private health service delivery sector for TB care in India—miles to go!. Trop Med Infect Dis. 2023;8:265.
- 51. Sudhinaraset M, Ingram M, Lofthouse HK, Montagu D. What is the role of informal healthcare providers in developing countries? A systematic review. PLoS One. 2013;8(2):e54978. pmid:23405101
- 52. Khare S, Purohit M, Sharma M, Tamhankar AJ, Lundborg CS, Diwan V, et al. Antibiotic Prescribing by Informal Healthcare Providers for Common Illnesses: A Repeated Cross-Sectional Study in Rural India. Antibiotics (Basel). 2019;8(3):139. pmid:31491900
- 53. Yimer S, Bjune G, Alene G. Diagnostic and treatment delay among pulmonary tuberculosis patients in Ethiopia: a cross sectional study. BMC Infect Dis. 2005;5:112. pmid:16343350
- 54. Steen TW, Mazonde GN. Pulmonary tuberculosis in Kweneng District, Botswana: delays in diagnosis in 212 smear-positive patients. Int J Tuberc Lung Dis. 1998;2(8):627–34. pmid:9712276
- 55. Chandra A, Kumar R, Kant S, Krishnan A. Diagnostic Pathways and Delays in Initiation of Treatment among Newly Diagnosed Tuberculosis Patients in Ballabgarh, India. Am J Trop Med Hyg. 2021;104(4):1321–5. pmid:33617478
- 56. Balasubramnian A, Francis PT, Leelamoni K, Rakesh PS, Lalu JS. Diagnostic and treatment delay among new pulmonary tuberculosis patients in Southern India: A cross-sectional study. Indian J Public Health. 2022;66(Supplement):S60–5. pmid:36412476
- 57. Sreeramareddy CT, Qin ZZ, Satyanarayana S, Subbaraman R, Pai M. Delays in diagnosis and treatment of pulmonary tuberculosis in India: a systematic review. Int J Tuberc Lung Dis. 2014;18(3):255–66. pmid:24670558
- 58. Chen H-G, Liu M, Jiang S-W, Gu F-H, Huang S-P, Gao T-J, et al. Impact of diabetes on diagnostic delay for pulmonary tuberculosis in Beijing. Int J Tuberc Lung Dis. 2014;18(3):267–71. pmid:24670559
- 59. Vishnu Sharma M, Arora VK, Anupama N. Challenges in diagnosis and treatment of tuberculosis in elderly. Indian J Tuberc. 2022;69 Suppl 2:S205–8. pmid:36400510
- 60. Mesfin MM, Newell JN, Madeley RJ, Mirzoev TN, Tareke IG, Kifle YT, et al. Cost implications of delays to tuberculosis diagnosis among pulmonary tuberculosis patients in Ethiopia. BMC Public Health. 2010;10:173. pmid:20353567
- 61. Oxlade O, Piatek A, Vincent C, Menzies D. Modeling the impact of tuberculosis interventions on epidemiologic outcomes and health system costs. BMC Public Health. 2015;15:141. pmid:25884339
- 62. Give C, Ndima S, Steege R, Ormel H, McCollum R, Theobald S, et al. Strengthening referral systems in community health programs: a qualitative study in two rural districts of Maputo Province, Mozambique. BMC Health Serv Res. 2019;19(1):263. pmid:31035983
- 63. Amare D, Ambaw F, Alene KA. Effect of integrating traditional care with modern healthcare to improve tuberculosis control programs in Ethiopia: a protocol for a cluster-randomized controlled trial. Trials. 2023;24(1):582. pmid:37697429
- 64. Thapa P, Narasimhan P, Hall JJ, Jayasuriya R, Mukherjee PS, Das DK, et al. Exploring current and potential roles of informal healthcare providers in tuberculosis care in West Bengal, India: A qualitative content analysis. PLOS Glob Public Health. 2025;5(11):e0004085. pmid:41191607