Background

In 1993, the World Health Organization (WHO) declared global tuberculosis (TB) emergency to make TB a high priority [1]. Twenty-five years on, TB remains one of the leading infectious causes of illness and death worldwide [2]. Despite that TB is both preventable and curable, and efforts such as the implementation of directly observed treatment short course and coordinated national TB programs worldwide, approximately 10 million people fell ill with TB, of which 1.5 million died from the disease in 2018 [2]. The cumulative reduction in the TB incidence rate globally between 2015 and 2018 stood at 6% [2], imposing a significant delay in reaching the end TB milestone of 20% [3] reduction by 2020. TB control and elimination are critical challenges in many countries. However, the burden is disproportionately borne by 30 countries, mostly in Asia and Africa, accounting for 87% of the world’s TB (both pan-TB and drug-resistant TB) and TB/HIV cases [2].

In 2018, nearly one-third of the people with TB were estimated to be undiagnosed globally [2]. The delay in diagnosis and treatment is detrimental to the patients’ prognosis and perpetuates TB transmission in the community [4] and thus poses a great challenge to eliminating TB. Therefore, identifying the factors that lead to delayed TB diagnosis and treatment is imperative in developing interventions to reduce TB incidence substantially. Collectively, recent systematic reviews have provided empirical evidence associating sociodemographic, clinical, health system, and economic factors with delayed diagnosis and treatment of TB in different countries and regions [5,6,7,8,9,10,11]. However, delays in diagnosis and treatment vary across countries with a different burden of the disease. From what we know, no systematic reviews have addressed delayed diagnosis and treatment of TB among countries bearing most of the global TB burden. There is also a lack of reviews that triangulate qualitative and quantitative findings to provide a more complete and all-inclusive view of the matter. Therefore, a mixed-methods systematic review and meta-analysis were undertaken to derive the determinants and duration of diagnosis and treatment delays of pulmonary TB in the high TB-burden countries.

Methods

We structured this review following the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA)-statement [12]. The protocol of this systematic review has been published [13] and registered with the International Prospective Register of Systematic Reviews (PROSPERO) (Registration Number CRD42018107237).

Inclusion and exclusion criteria

In this review, we considered all studies conducted in the WHO high TB-burden countries—Angola, Bangladesh, Brazil, Cambodia, Central African Republic, China, Congo, Democratic People’s Republic of Korea, Democratic Republic of Congo, Ethiopia, India, Indonesia, Kenya, Lesotho, Liberia, Mozambique, Myanmar, Namibia, Nigeria, Pakistan, Papua New Guinea, the Philippines, Russian Federation, Sierra Leone, South Africa, Tanzania, Thailand, Vietnam, Zambia, and Zimbabwe. We included studies that reported on individual and interpersonal risk factors, social and physical environment, health systems, and policies associated with delayed TB diagnosis and treatment initiation published between 2008 and 2018. The factors could be self-reported, ascertained by health providers, or abstracted from medical charts or programs/administrative records.

We included study populations comprising presumptive TB (persons presenting with signs and/or symptoms suggestive of TB) and people with TB (new diagnosis, previously treated, and those without a known history of previous TB treatment) regardless of HIV and bacteriological status. We included observational (cross-sectional, case–control, retrospective, and prospective cohort design) and qualitative studies published in English or Chinese. Systematic reviews, meta-analyses, scoping reviews, intervention studies, publications in the form of letters and reviews, and studies lacking and/or unclear reporting of key outcomes were excluded.

Our primary outcomes were—(1) patient delay: the time interval between the onset of symptoms and the first encounter with healthcare professionals; (2) health system delay: the time interval between the first encounter with healthcare professionals and the diagnosis of pulmonary TB; (3) treatment delay: the time interval between TB diagnosis and TB treatment initiation; and (4) total delay: the time interval between onset of symptoms and TB treatment initiation. As there were no universal cut-offs [8] to a duration that constituted delay, we treated delay in this review as how they were defined in individual studies. We did not exclude studies based on the delay thresholds defined in individual studies.

Literature search strategy and study selection

First, we conducted a preliminary search of articles on PubMed and EMBASE to develop a set of appropriate Medical Subject Heading terms, index terms, and keywords [13], centered around three domains (population/problems: tuberculosis, outcomes: health-seeking behaviours; delays; barriers, countries: high burden countries). Using these identified search terms structured with Boolean logic operators (AND and OR), we contextualized the search strategies in PubMed, EMBASE, CINAHL, and PsycInfo (Additional file 1). The search fields included title, abstract, keywords, and text words. We also reviewed the reference list of key articles for additional studies. We managed all identified citations into EndNote X8 (Clarivate Analytics, Philadelphia, USA). Duplicates were removed, and the remainder was exported to Microsoft Excel (Microsoft Corporation, Washington, USA) for further assessment. AKJT and SRS independently screened the titles, abstracts, and full-text articles based on the inclusion and exclusion criteria. Interrater agreements for the titles and abstract screening between the reviewers were high (agreement = 98%, Cohen’s kappa = 0.95, and Krippendorf alpha = 0.95), and discrepancies were discussed. The two primary reviewers were able to resolve all the discrepancies without having to involve a third reviewer. The search and selection processes were conducted and presented in accordance with the PRISMA guidelines.

Data extraction

Study characteristics and data on risk factors were extracted independently by two authors (AKJT and SRS). We recorded study and participants’ characteristics, exposure variables (various factors associated with delays reported by individual study), primary outcome measures, and study quality assessment scores using a standard form. Data on variables to be included in the meta-analysis were extracted by one author (AKJT) and subsequently reviewed by a second author (KP). This included duration of delay (median and interquartile range/range and mean and standard deviation) and the effect sizes (crude and adjusted odds ratios) for exposures of interest.

Quality assessment

The quality of the selected non-randomized (quantitative) and qualitative studies was critically evaluated using the Newcastle–Ottawa Scale for cross-sectional studies, case–control studies, and cohort studies, and the Critical Appraisal Skills Program (CASP) tool, respectively [14,15,16]. For non-randomized quantitative studies, the assessment was made based on four main domains—(1) selection of samples (representativeness, sample size, definition and selection of cases and controls (for case–control studies), and non-response rate), (2) comparability of groups included in the analyses, (3) the ascertainment of exposures and outcomes, and (4) the statistical tests applied in the studies. A score of 1 was given to individual questions if the criterion was satisfied and 0 if the criterion was not satisfied or not justified. The highest possible score for cross-sectional studies was 10 (5 for selection, 2 for comparability, and 3 for outcomes). The highest possible score for case–control studies was 9 (4 for selection, 2 for comparability, and 3 for exposure). The highest possible score for cohort studies was 9 (4 for selection, 2 for comparability, and 3 for exposures). Studies that scored 0–3 were regarded as low quality (LQ), 4–6 were regarded as moderate quality (MQ), and ≥ 7 were regarded as high quality (HQ).

For qualitative studies, the assessment was made based on ten questions regarding the results, validity, and the value of the research. We gave a score of 1 if the paper fulfilled a criterion, 0.5 if we could not tell if the paper fulfilled a criterion, and 0 if the paper did not fulfill a criterion. A score of 0–5 equated to LQ study, a score of 6–7 equated to MQ study, and a score of ≥ 8 equated to HQ study. The final synthesized qualitative findings were graded based on the dependability and credibility of the findings using the ConQual approach [17].

Data synthesis and analyses

We described the studies by the populations, countries, study designs, and sample sizes. Countries were grouped by WHO region and categorized as low-income economies (LIC)—gross national income (GNI) per capita $1,025 or less in 2018; lower-middle-income economies (LMIC)—GNI per capita between $1,026 and $3,995; upper-middle-income economies (UMIC)—GNI per capita between $3,996 and $12,375 according to World Bank classification in 2019 [18]. We reported the independent variables significantly associated with the patient, health system, treatment, and total delays. Results from the multivariable analyses preceded bivariate for studies that reported both bivariate and multivariable analyses.

Median and interquartile range/range for the duration of delays in days were extracted and used to estimate a pooled median, i.e., median of study-specific medians [19]. We pooled weighted medians by incorporating study-specific sample sizes [19]. For patient delay, we excluded two studies [20, 21] from China with sample sizes > 10,000 because the pooled weighted medians were heavily skewed, including only estimates of the study with the largest sample size [21].

For independent variables (risk factors), effect sizes were extracted and used to calculate pooled odds ratios (OR) and their 95% confidence interval (CI). We pooled effect sizes of covariates from studies that utilized similar delay thresholds if data were available in more than two studies and duration of delays by meta-analysis using R (R Foundation for Statistical Computing, Vienna). We pooled effect sizes for studies that defined patient delay using threshold values of 14–15 days (n = 5), 20–21 days (n = 7), 28–30 days (n = 17); and health system delay using threshold values of 14–15 days (n = 5). We found five studies that reported treatment delay using a threshold value of 7 days. However, the studies did not report similar covariates with effect sizes that allowed pooling. Where adjustments for covariates had been performed, the data from the adjusted model were pooled.

We quantified between-study heterogeneity using Chi-square statistic Q, I2, and Tau [22]. We estimated the pooled OR and its 95% CI using a Bayesian random-effect model for each meta-analysis, which accounted for between-study heterogeneity [23]. The estimates for Tau and I2 statistics were presented together with the pooled estimates and the 95% CI. We used the inverse of the effect size variance to determine the pooling weights. We assessed the association of the primary outcomes and (1) sociodemographic and economic variables: sex, urbanicity; (2) behavioral variables: smoking, alcohol use, TB knowledge; and (3) clinical and health services-related variables: hemoptysis, weight loss, fever, chest pain, night sweats in the meta-analyses.

We extracted qualitative findings and sample quotes reported in qualitative and mixed-method studies verbatim. The extracted data were annotated and analyzed using NVIVO 12 (QSR International). We retrieved references deductively and applied thematic analyses to categorize the textual references. Two authors (AKJT and SRS) coded the data independently. Discrepancies, code definitions, and the emergence of sub-themes were discussed. The results were presented by income categories that the high-burden countries represent.

Results

Study selection

The systematic review process is presented in Fig. 1. A total of 4878 records were identified from electronic database searches. Following the removal of duplicates (n = 1189) and non-relevant records (n = 3383), 306 records were assessed for eligibility. Of these, 182 articles were further excluded. Finally, 124 articles were reviewed. A qualitative synthesis was performed for 36 studies. We conducted quantitative and narrative synthesis on 86 studies. Two mixed-method studies underwent both qualitative and quantitative/narrative synthesis. We found large heterogeneity among studies included in the meta-analyses—1 (7%) had I2 ≤ 50%, 14 (93%) had I2 > 50%, and 13 (87%) had I2 > 75%.

Fig. 1
figure 1

PRISMA flow diagram for identification of studies via databases

Full size image

Study characteristics and quality assessments

These studies described data from 18,759 presumptive TB and 131,142 people with TB [20, 21, 24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109], 1659 in-depth and structured interviews, and 87 focus groups [48, 96, 110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144,145] from 19 countries in three continents (Table 1). A total of 14 countries were classified as lower-income (LIC) and lower-middle-income economies (LMIC), and five were classified as upper-middle-income economies (UMIC) [18]. Patient delay was reported in 103 studies, health system delay in 29 studies, treatment delay in 18 studies, and total delay in 21 studies. Of the 30 high TB-burden countries, 11 countries were not included in this review, either due to data unavailability or lack of key outcome data (Fig. 2). After assessments of study quality, a total of 81 HQ studies, 40 MQ studies, and one LQ study were identified. Two mixed-methods studies scored MQ/HQ and HQ/MQ for the quantitative and qualitative components, respectively. The final synthesized qualitative findings were rated HQ (55%) and MQ (45%) using the ConQual method. Details of the assessments were illustrated in the Additional file 1. No studies were excluded based on the outcome of quality assessments; instead, the information was considered during data synthesis and interpretation.

Table 1 Characteristics of included observational and qualitative studies
Full size table
Fig. 2
figure 2

Geographical coverage of studies published between 2008 and 2018 included in this systematic review of delayed diagnosis and treatment of pulmonary tuberculosis. The 30-high tuberculosis (TB) burden countries which have been designated by the World Health Organization are outlined in black. Of them, countries with studies presenting various types of delay are categorized by the various colors. For example, countries shaded in green had studies presenting both patient and health system delay, and those with diagonal strips presented total delay too. Some of the high TB burden countries shaded in grey had no studies identified here or lacked key outcome data. The table on the left represents the TB incidence per 100,000 population of high TB burden countries in 2019. Rows shaded in grey represent countries that were not included in this review either due to data unavailability or lack of key outcome data

Full size image

Patient delay

The pooled median patient delay (Fig. 3) in LIC and LMIC was 28 days (95% CI 17–30). The pooled median patient delay in UMIC was 10 days (95% CI 10–20). The overall median patient delay in high TB burden countries was 16 days (95% CI 11–20). In the meta-analysis and narrative synthesis of quantitative data (Table 2), females (pooled OR 1.48. 95% CI 1.09–1.98, P = 0.01) were more likely to delay care-seeking for TB (Fig. 4). Qualitative studies highlighted limitations for women to seek healthcare [115, 117, 125, 126, 130]. Women reported economic constraints and power imbalances in the decision-making process as barriers to care-seeking [115, 117, 125, 126]. We further stratified the analysis by sex. We found that women were disproportionately affected by risk factors for patient delay (Fig. 5), such as unemployment, poor TB knowledge, and difficulties traveling a long distance to visit health facilities [43, 51]. Long-distance to health facilities was also reported by qualitative studies as a barrier to care-seeking [48, 111, 112, 114, 117, 126, 131, 136, 137, 140, 142, 145]. In addition to physical barriers, financial insecurities and economic challenges also compounded patient delay [20, 25, 28, 33, 38,39,40, 43, 50, 51, 66, 67, 69, 72, 76, 77, 83, 87, 88]. Among qualitative studies (Table 3), seven articles reflected on participant’s experiences where competing priorities of livelihoods and commitment to work and family led to individual care-seeking delay [48, 125,126,127, 134, 138, 145]. We also found that being rural residents in LIC and LMIC (Table 2 and Fig. 4) was associated with patient delay (pooled OR 1.75, 95% CI 1.01–2.94, P = 0.02). No studies from the UMIC were included in the meta-analysis for urbanicity. This review also reported other sociodemographic and economic risk factors for patient delay, such as lower education level and being older, unmarried, and unemployed. High indirect medical costs [48, 126], absence of health insurance, productivity, time, and income loss [48, 125, 127, 134, 138, 145] resulting from disease suffering further worsen household vulnerabilities and contribute to a delay in TB diagnosis and poor health outcomes (Table 2) [146].

Fig. 3
figure 3

Duration of patient delay by regions reported in high tuberculosis burden countries. Countries were grouped by WHO regions (AFR African region, AMR Region of the Americas, SEAR South-East Asia region, WPR Western Pacific region). Countries were also categorized as (i) LIC (low-income countries), (ii) LMIC (low-middle income countries), or (ii) UMIC (upper-middle income countries) as designated by the World Bank in 2019. Studies from LIC and LMIC were in bold. Patient delay (in blue) was pooled by the countries’ economic status using Weighted Medians of Medians methods by McGrath (2019). The estimates were weighted by sample sizes of the studies. Pooled results for LIC and LMIC were not presented separately due to insufficient studies from LIC. Duration of delay in days were presented in the log scale

Full size image
Table 2 Summary of risk factors for patient delay, health system delay, and treatment delay in high TB burden countries
Full size table
Fig. 4
figure 4

Association between sex of individuals, urbanicity and patient delay. Countries were grouped by WHO region (AFR African region, AMR Region of the Americas, SEAR South-East Asia region, WPR Western Pacific region) and categorized as (i) LIC/LMIC (low- or lower-middle-income countries), or (ii) UMIC (upper-middle-income countries) as designated by the World Bank in 2019. The reference group for sex (left panel) was male and urbanicity (right panel) was urban. The odds ratio (OR) were pooled (in blue) by countries’ economic status using Bayesian random-effects meta-analysis. Odds ratios are presented in the log scale

Full size image
Fig. 5
figure 5

Subgroup analysis of patient delay and selected covariates by sex of the individual. Tamhane et al. (2012), represented as square points, and Mfinanga et al. (2008), represented as round points, provided sex-specific association of patient delay and three covariates; i.e., being unemployed, having to travel long distances or long travelling time, and having poor TB knowledge. The sex-specific odds ratio, in the log scale, for males are presented in hues of blues and for females in hues of reds

Full size image
Table 3 Emerged themes and synthesized findings from qualitative studies
Full size table

Furthermore, poor TB knowledge (Table 2), unawareness of free TB treatment policies [30], low perceived susceptibility, and severity of TB was associated with patient delay [40, 48, 110, 113,114,115, 117, 119, 124, 126, 137]. However, the pooled estimates for TB knowledge (delay thresholds 21 days—(pooled OR 0.91, 95% CI 0.24–2.71, P = 0.62) and 28 days—(pooled OR 1.36, 95% CI 0.39–4.83, P = 0.25) were not significantly associated with patient delay in the meta-analysis (Fig. 6).

Fig. 6
figure 6

Association between TB knowledge, smoking, alcohol use and patient delay. Countries were grouped by WHO region (AFR African region, AMR Region of the Americas, SEAR South-East Asian region, WPR Western Pacific region) and categorized as (i) LIC/LMIC (low- or lower-middle-income countries), or (ii) UMIC (upper-middle-income countries) as designated by the World Bank in 2019. For TB knowledge, the top left panel pooled estimates from studies that defined patient delay threshold as 28 days. The bottom left panel pooled estimates from studies that defined patient delay threshold as 21 days. The reference group was TB knowledge (no or low). The top right panel represented pooled estimates for the association between alcohol use and patient delay. The bottom right panel represented pooled estimates for the association between smoking and patient delay. Both plots pooled estimates that defined patient delay threshold as 28 days. The reference groups were no smoking and no alcohol use, respectively. The odds ratio (OR) were pooled (in blue) by countries’ economic status using Bayesian random-effects meta-analysis. Odds ratios are presented in the log scale

Full size image

Perceived stigma and discrimination (Table 3) at the workplace, within the family, and community and associating TB with HIV deterred presumptive TB patients from care-seeking [110, 114,115,116,117, 121, 125,126,127,128,129,130, 135,136,137, 139,140,141,142,143,144]. From the qualitative data, we found studies that explained the use of alcohol and smoking to conceal health issues, especially among men, which resulted in delayed care-seeking [116, 128, 137, 138, 144, 145]. However, these lifestyle behaviors were not statistically significant in the meta-analysis, where the estimates from both sexes were pooled (Fig. 6). Several studies in Africa [110, 113, 114] highlighted superstitious beliefs that TB is caused by divine retributions of past misdeeds, sinful behaviors, and curses; thus, help is first sought from traditional or spiritual healers instead of a health provider. Besides, studies in Asia reported the misconception that TB is hereditary [117, 126].

Long chains of care-seeking through multiple non-formal or private health providers were also reported as a determinant of patient delay [110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128, 134, 135, 137, 138, 140, 142,143,144]. Qualitative data also suggested that the lack of trust in the public health care system perpetuated delays in care-seeking [122, 131, 136,137,138,139, 143, 144]. The inability of people with TB to recognize symptoms such as fever and cough that were not ascribed to TB intrinsically led to self-medication and treatment or waiting for symptoms to self-resolve due to low perceived disease severity [110, 119, 126, 128, 129, 134, 137, 140, 142]. Studies reported that the presence of cough [69, 70, 74, 80, 86, 88] was associated with patient delay compared to hemoptysis and weight loss that were perceived to be more severe (Table 2). However, the relationships between TB symptoms and patient delay were not statistically significant in the meta-analysis (Fig. 7).

Fig. 7
figure 7

Association between TB symptoms and patient delay. Countries were grouped by WHO region (AFR African region, AMR Region of the Americas, SEAR South-East Asian region, WPR Western Pacific region) and categorized as (i) LIC/LMIC (low- or lower-middle-income countries), or (ii) UMIC (upper-middle-income countries) as designated by the World Bank in 2019. The reference group was no symptom. The patient delay threshold was 28 days. The odds ratio (OR) were pooled (in blue) by countries’ economic status using Bayesian random-effects meta-analysis. Odds ratios are presented in the log scale

Full size image

Health system delay

The pooled median health system delay (Fig. 8) in LIC and LMIC was 14 days (95% CI 2–28). The pooled median health system delay in UMIC was 4 days (95% CI 2–4). The overall median health system delay in high TB-burden countries was 4 days (95% CI 2–4). We explored the association between sex and health system delay, and we did not find a significant relationship (Additional file 1). Twelve qualitative studies reported that poor practices and ignorance of TB among health providers at health facilities led to a delay in TB diagnosis [110, 111, 114, 116, 118, 122,123,124, 129, 132, 133, 135]. Seven qualitative studies explained that complicated administrative procedures at the health facilities [96, 119, 126, 130, 132, 133, 144], which could have resulted in longer waiting time [46], and complex referral system[144] that eventually prolonged health system delay. This review identified studies reporting that health system delays were associated with visiting lower-level facilities that did not provide TB services [34, 46, 72, 75, 90]. Six qualitative studies mentioned that inadequate resources and supplies in health facilities could have delayed TB diagnosis [96, 129, 130, 132, 138, 144]. Three quantitative studies reported that people with smear-negative TB were more likely to experience health system delay [81, 82, 91].

Fig. 8
figure 8

Duration of health system and treatment delay by regions reported in high tuberculosis burden countries. Countries were grouped by WHO regions (AFR African region, AMR Region of the Americas, SEAR South-East Asian region, WPR Western Pacific region). Countries were also categorized as (i) LIC (low-income countries), (ii) LMIC (low-middle income countries), or (ii) UMIC (upper-middle income countries) as designated by the World Bank in 2019. Studies from LIC and LMIC were in bold. Health system delay (in yellow) and treatment delay (in red) were pooled by the countries’ economic status using Weighted Medians of Medians methods by McGrath (2019). The estimates were weighted by sample sizes of the studies. Pooled results for LIC and LMIC were not presented separately due to insufficient studies from LIC. Duration of delay in days were presented in the log scale

Full size image

Treatment delay

The pooled median treatment delay (Fig. 8) in LIC and LMIC was 14 days (95% CI 3–84). The pooled median treatment delay in UMIC was 0 days (95% CI 0–1). The overall median treatment delay in high TB burden countries was 1 day (95% CI 0–14). One qualitative study noted that the geographical distance to health facilities, especially when treatment was initiated in separate institutions, delayed TB treatment initiation [141]. This could be exacerbated by residing in areas without health centers nearby [97]. Health system factors such as logistical issues in drug transportation [96, 141] and the absence of TB diagnostic services in local health facilities [96, 119] compounded delay in treatment initiation. Like patient delay, a qualitative study provided insights into TB stigma experienced by women resulting in the concealment of diagnosis, expulsion from their community, or isolation; thus, delaying access to TB care and treatment[127]. Four qualitative studies mentioned self-perception of health, unconvinced diagnosis and need for TB treatment, and the perceived low effectiveness of TB treatment led to a delay in TB treatment initiation [48, 96, 119, 141]. We also found that retreatment cases were more likely to delay TB treatment initiation [94,95,96,97,98].

Discussion

Our review is the first to focus on determinants of delayed TB diagnosis and treatment among high TB burden countries using evidence-based quantitative and qualitative information. Studies from high TB-burden LIC/LMIC reported longer median patient delay (28 days) than UMIC (10 days). Our findings were consistent with previous systematic reviews conducted in countries of different income levels [5, 8]. However, the median patient delay among UMIC in this review was shorter than the findings from observational studies conducted in other high-income countries [147, 148]. TB burden in high-income countries has been progressively reduced through improvements in socio-economic conditions, strong health systems components such as the delivery of TB services and universal health coverage, and social protection schemes [149]. Nevertheless, the high standards of living and wellbeing have shaped the notion that TB is not a significant concern, rendering a lower index of suspicion of TB and thus delaying TB care-seeking [150]. Notwithstanding, TB remains an issue, especially among hard-to-reach populations living in high-income settings like migrants [151, 152], who face challenges in accessing healthcare due to stigma, language barriers, and cost of testing and treatment [6].

The median health system delay in this review (LIC/LMIC: 14 days and UMIC: 4 days) was found to be shorter than previous systematic reviews conducted among countries of similar economies [5, 8]. As this review included studies conducted in the last decade, the improvement in health system delay may be attributed to the enhancements of healthcare systems [153] and the quality of TB laboratories [154]. The clinicians’ ability to consider TB as a differential diagnosis in high burden settings is also essential for early diagnosis and treatment [155]. However, there remains a paucity of data in several high TB burden countries, including seven in Africa (Central African Republic, Democratic Republic of Condo, Lesotho, Liberia, Namibia, Republic of the Congo, and Sierra Leone) and four in Asia (Democratic People’s Republic of Korea, Myanmar, Papua New Guinea, and Vietnam), potentially due to logistical challenges in conducting such studies.

Patient delay

While TB is a disease mainly affecting men[156], in our review, we found that women faced challenges in accessing TB care promptly in some settings due to resource constraints, power imbalances, and poor TB knowledge. However, there was a paucity of sex-specific data on the determinants of delay in TB diagnosis and treatment. It is imperative to recognize sex disparities in TB care-seeking. Women with TB in high burden countries experienced delays in diagnosis and treatment because of barriers to TB services. Therefore, future studies should report disaggregated data by sex to inform programs and interventions addressing sex-specific vulnerabilities in improving access to TB services among men and women.

Despite wide coverage of free TB diagnostic and treatment services in high burden countries [157], people with TB and their families, especially the poor, bear the impact of high economic costs [158]. Studies included in our review also reported that livelihood, work, and family were prioritized and led to a delay in care-seeking. These factors, coupled with the physical environments and impoverished living conditions [127, 138], plunged low-income households into a vicious cycle of impoverishments [159, 160], making TB elimination overtly challenging. Aside from the broad expansion of TB services that have been shown to reduce the financial burden on TB-affected households [161], it is also essential to ensure that financial and social protection policies are in place to protect those at risk of catastrophic TB costs and poverty.

Our pooled estimates showed that rural dwellers were significantly associated with patient delay. In the rural setting, access to healthcare facilities, particularly an institution that offers TB diagnostic services, might be lacking [5]. Rural populations were also more likely to have a lower health literacy [162], resulting in poorer health status and outcomes [163]. Nevertheless, the concept of urban–rural is dynamic and context-dependent, driven by migration, population, and economic growth over time [164]. The consistent findings of rural residence and patient delay in TB [5, 8, 165] suggest increased efforts tailored to the country’s specific circumstances in reaching the affected communities are required.

In countries where TB diagnostic and treatment services are provided for free, access to TB care is further challenged by poor knowledge and awareness regarding such policies, making presumptive TB seek treatment early [123, 135]. In addition to poor awareness about the free TB treatment policy, we also identified studies that reported poor knowledge regarding TB symptoms associated with a delay in TB care-seeking. Therefore, people with TB would delay care-seeking until only when their illness compromised their ability to work and earn livelihoods [115]. Conventionally, symptomatic individuals are linked to TB transmission, and they are regarded as the target group for TB case-finding activities using the TB symptoms screening approach [166]. However, TB transmission could also occur during the subclinical (asymptomatic) phase, particularly heightened during episodes of symptoms exhibition unrelated to TB pathologies, such as bouts of either acute or chronic cough [167]. As people with subclinical TB might not report any symptoms, they have lower awareness and motivation to seek care; thus, leading to a delay in TB diagnosis and treatment and potentially sustaining TB transmission [168] in the household and community. Therefore, a better understanding of subclinical TB, its transmission dynamics, and the implications for TB control efforts are needed. Nevertheless, individuals who exhibit TB symptoms, such as cough, are more likely to have a higher bacillary load and transmit infection [169, 170]. Therefore, it is crucial to ensure that ill and symptomatic persons with TB are reached, tested, and treated promptly.

Furthermore, misperception regarding the causes of TB was also found to delay TB care-seeking. When no one in the family is ever diagnosed with TB, presumptive TB did not self-initiate care-seeking or is discouraged explicitly by family members to seek TB diagnosis and treatment [117]. Therefore, it is imperative first to measure the level of knowledge, awareness, and practices regarding TB in settings where studies as such have yet to be conducted. The gaps identified could then be used to develop health education programs and interventions about TB. Studies have shown that health education programs and dissemination of TB information effectively improve TB knowledge and awareness [171], enabling care-seeking and increasing the identification of TB cases [172]. Furthermore, understanding the knowledge and practices of health professionals could be done in parallel to improve care and facilitate the early identification of TB [173].

TB stigma continues to be a major barrier for people to access TB diagnosis and complete treatment [10, 143]. Moreover, stigma could also reduce the use of face masks [6, 174], further contributing to infection transmission. Despite being an increasingly important agenda of TB programs worldwide, there is a paucity of data on stigma [175], particularly information from the perpetrators of stigma [176]. There is also limited evidence on effective interventions that can reduce TB-related stigma [177]. Considering the importance of stigma reduction in TB control and elimination efforts, stigma should be systematically measured. De-stigmatisation must include approaches in healthcare institutions and beyond for a more inclusive care plan.

In high TB-burden countries, we found that people who presented with cough, fever, and night sweats were more likely to delay TB care-seeking. This is consistent with another systematic review conducted among low and middle-income countries [5]. The attribution of these symptoms to other respiratory infections or smoking and the inability to link them to TB was claimed as one of the primary reasons causing a delay in seeking care [138]. Contrarily, more severe symptoms such as hemoptysis were more likely to reduce delays in care-seeking. Therefore, education and awareness-raising activities could be recalibrated to specifically highlight the possibility of TB besides other respiratory illnesses in the event of more general symptoms such as cough and fever. Simultaneously, health workers' awareness on this matter in high-burden settings should be raised to improve TB case findings.

Health system and treatment delay

Health system delay was more pronounced in LIC and LMIC than UMIC, likely due to the standard of health care, the strength of the national health systems, and the availability of resources. Among LIC and LMIC, a systematic review reported that the quality of health care in the public and private sectors was poor. The private sector relatively outperformed the public sector regarding the delivery of care and medicines availability [178]. The discrepancies in effectiveness and efficiency were highlighted as a facilitator to seek private healthcare, which eventually leads to a delay in TB diagnosis in a high TB burden setting like Cambodia[179]. Narrowing down to high TB-burden countries, the quality of public and private healthcare was also found to be below par, and systematic evaluations are needed to identify gaps in the TB care pathway [157].

Likewise, treatment delay was longer in LIC and LMIC than UMIC. The delay might be due to logistic factors such as long distance to treatment centers, availability of anti-TB drugs, and the absence of TB diagnostic services in local health facilities [96, 97]. Beyond systemic factors, individuals’ low perceived susceptibility and TB stigma could delay a person’s decision to initiate TB treatment [48, 127]. Interventions to decrease isolation post-diagnosis and social support should be provided to encourage prompt initiation of TB treatment [180]. Health providers also play a vital role in assisting people with TB to internalize the diagnosis and support them in decision-making [181].

Strengths and limitations

To our knowledge, this systematic review will be the first to focus on countries with a high TB burden, where most of the TB cases in the world [2] are found. As the list consisted of countries from LIC, LMIC, and UMIC, we attempted to discern the differences in the determinants of delayed TB diagnosis and treatment between these countries.

However, we found high levels of heterogeneity amongst the studies potentially due to clinical and methodological diversities. We included studies from different high TB-burden countries and economic statuses. While we have restricted the study populations to people with presumptive TB and people with TB, their sociodemographic profiles were diverse. We acknowledged the limitation in analyzing data comprising of all possible subgroups in this review. Furthermore, we included the different observational non-randomized studies in this review. The design differed by temporality and the potential biases, contributing to methodological diversity.

In our attempt for comprehensiveness, we retained the threshold of delays as to how they were defined in individual studies. While it might not pose severe concerns for the narrative synthesis and pooling of median delays, the utilization of the delay threshold defined by individual studies in the meta-analysis of risk factors could lead to misinterpretation. Therefore, we pooled effect sizes from eligible studies that utilized similar delay thresholds in the meta-analyses. We incorporated heterogeneity into random-effects models using the Bayesian approach [23], which could yield more accurate interval estimates than conventional methods, especially for studies with a small sample size and are heterogenous [23, 182, 183]. However, the incorporation of heterogeneity in the random-effects models would not fully account for the clinical and methodological diversity in the studies. Analyses of study-level covariates in a meta-regression may be relevant to further investigate heterogeneities by the differences in studies characteristics and populations. We did not perform a meta-regression in this review, and this method could be considered in future reviews of similar nature.

Nevertheless, caution in interpreting and extrapolating the findings from the meta-analyses is warranted. For the pooled median delays, the overall estimates were influenced by studies in UMIC with larger sample sizes. Therefore, we opined that the pooled estimates by economies would be more informative. In the meta-analyses of risk factors, the pooling of estimates from studies with similar delay thresholds limited the number of studies that could be included. Most of the independent variables were also grouped differently, and we could not standardize them all. Hence, the meta-analyses were only performed for selected variables. However, we strived to maintain this review's comprehensiveness by triangulating findings from narrative synthesis and thematic analyses of qualitative studies.

This review did not include data from all 30 high TB-burden countries due to the absence of key outcome data and research activities. Notwithstanding the potential lack of representativeness due to the scarcity of data from several countries, this review highlights the gaps in knowledge and provides insights into the determinants of TB diagnosis and treatment delay in high-burden countries. However, the heterogeneity of the data limited the generalizability of our findings to settings underrepresented in this review.

Conclusions

Our analyses revealed a substantial delay between the onset of TB symptoms and TB care-seeking among high burden countries, highlighting the need to continue to shape knowledge, change attitude, and raise awareness of the community, people at risk of TB, and the health providers. Specific vulnerabilities such as sex disparities in care-seeking, being older, and geographic isolation should be recognized and addressed through tailored approaches to improve access to TB services and early diagnosis [184]. It is also crucial to improve the consciousness of the society regarding TB to battle stigma, and networks [185] of support from within the families, the grassroots, and institutions could create an enabling environment for early care-seeking and treatment adherence and success. In contrast to patient delay, the shorter health system and treatment delay were encouraging. Nonetheless, TB programs should strive to test and treat TB by adopting WHO recommendations for same-day TB diagnosis [186] to further reduce TB transmission and mortality [187]. Higher-level policies and interventions such as health system strengthening, universal health coverage, and the provision of sustainable social welfare schemes are important to reduce delays, improve access to TB care, and ultimately achieve the global TB targets [188].