Introduction

Globally, an estimated 10.7 million people developed tuberculosis (TB) disease in 2024, of whom 3.7 million (35%) were women [1]. TB incidence peaks in women during the reproductive years [2–4] and is a leading cause of mortality among women of childbearing age [5] Globally, an estimated 239,300 pregnant and 97,600 postpartum women developed TB in 2023 [6]. Pregnant women with TB have poor TB treatment outcomes [7,8] and their babies are at high risk of premature birth, low birthweight and stillbirth [8–11]. Failure to initiate TB treatment is a predictor of adverse pregnancy outcomes [12]. Prompt initiation of TB treatment is therefore critical for all newly diagnosed individuals, especially pregnant women, to improve both TB treatment and maternal and neonatal outcomes.

South Africa is a high TB, TB-HIV and multidrug-resistant/rifampicin-resistant (MDR/RR)-TB burden country [1]. In 2024, an estimated 249,000 people developed TB disease in South Africa, 57% were women ≥15 years, 5% had MDR/RR-TB and 54% of people with TB were living with HIV [1]. Addressing TB requires an integrated approach that strengthens diagnosis, facilitates linkage to care with rapid initiation of appropriate treatment regimens, and supports treatment adherence [13]. Early initiation of TB treatment is critical, as delays increase the risk of TB-related morbidity and mortality¹⁰ and contribute to ongoing transmission of Mycobacterium tuberculosis (M.tb) [14,15]. An estimated 53,906/249000 (20%) people with estimated TB disease in 2024 in South Africa were not diagnosed [1]. Between 12% and 30% of individuals diagnosed with TB in South Africa do not initiate treatment (“initial loss to follow-up”) [16,17]. Finding the missing people diagnosed with TB and linking them to care to start treatment is the central focus of the South African TB Recovery Plan initiated in 2022 [18,19].

In South Africa, time from TB diagnosis to TB treatment initiation varies between 0 and 144 days, depending on the diagnostic method used, and drug susceptibility test pattern of the infecting organism [20–23]. Delays in treatment initiation among pregnant women are of particular concern, as they already face increased risks for both adverse TB and pregnancy outcomes. There is limited evidence on the time to TB treatment initiation in pregnant women with TB. To address this gap, we assessed the time from TB diagnosis to treatment initiation among pregnant women in a routine programmatic setting in South Africa.

Methods

Results

Overall

Between October 2018 and December 2020, a total of 5,459 women aged 15–45 years were diagnosed with TB; 292 (5%) were pregnant. Pregnant women had a median age of 29.0 years (interquartile range [IQR]: 24.0–34.2) and non-pregnant women were 31.3 years (IQR: 25.5–36.8). HIV co-infection was common, affecting 61.9% of all women, with similar prevalence across pregnancy status. The majority of women (70.8%) had bacteriologically confirmed disease. A higher proportion of pregnant women (39.7%) were diagnosed in hospital compared to those who were not pregnant (28.1%). The majority of the cohort (94%) had previously accessed care at a primary healthcare facility for any health services. (Table 1).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Demographic and clinical characteristics of women with tuberculosis, by pregnancy status.

https://doi.org/10.1371/journal.pone.0344059.t001

Initiation of TB treatment

The proportion of women who were initiated on treatment was similar between pregnant and non-pregnant women, with 249/292 (85.3%) of pregnant women and 4427/5167 (85.7%) of non-pregnant women initiating TB treatment. The median time to treatment initiation was similar; 2 days (IQR: 0–7 days) for pregnant women and 2 days (IQR: 0–6 days) for non-pregnant women. Most women initiated treatment within the first weeks after their TB diagnosis, with 78.9% of non-pregnant women and 73% of pregnant women initiating treatment within 7 days and 86.6% of non-pregnant women and 83% of pregnant women by day 14. The cumulative initiation of TB treatment plateaued in both groups by 1 month, with 92% of non-pregnant women and 89% of pregnant women initiating treatment by the first month.. The time to treatment initiation differed significantly by pregnancy status (Log-rank test, p = 0.0064) and was slower in pregnant women. See Fig 1 and Fig 2.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Time to treatment initiation stratified by pregnancy status over 6 months.

https://doi.org/10.1371/journal.pone.0344059.g001

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Time to treatment initiation stratified by pregnancy status in the first month following diagnosis.

https://doi.org/10.1371/journal.pone.0344059.g002

Determinants of TB treatment initiation

In multivariable analysis (Table 2), women who were HIV-negative had a slightly higher likelihood of treatment initiation compared to those living with HIV [adjusted hazard ratio (aHR) 1.08; 95% confidence interval (CI): 1.01–1.15]. Women who were diagnosed at a PHC facility compared to those diagnosed in hospital (aHR 3.86; 95% CI: 3.55–4.20) and women with a clinical TB diagnosis compared to having bacteriologically confirmed TB (aHR 4.98; 95% CI: 4.59–5.41) were significantly more likely to initiate treatment.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Predictors of treatment initiation among women 15-45 years (N = 5459).

https://doi.org/10.1371/journal.pone.0344059.t002

In univariable analysis, pregnant women were significantly less likely to initiate treatment compared to non-pregnant women (HR 0.84; 95% CI: 0.74–0.95). However, after adjusting for age, HIV status, site of diagnosis, and method of TB diagnosis, pregnancy was no longer significantly associated with treatment initiation status (aHR 0.90; 95% CI: 0.79–1.03) (Table 2).

Discussion

This study determined time to treatment initiation for pregnant and non-pregnant women diagnosed with TB within a routine programmatic setting in two sub-districts in South Africa, thereby addressing the evidence gap on time to treatment initiation among pregnant women with TB in this setting. We found that 85% of women started treatment, irrespective of pregnancy status, however pregnant women demonstrated slightly longer time to treatment initiation and ultimately the proportion starting treatment did not reach the same level as non-pregnant women within six months. This is an important finding, indicating that pregnant women with TB need additional support linking to care following diagnosis.

In South Africa, time to treatment initiation varies by setting, population group, TB type, and presence of comorbidities. In our cohort, the proportion of women starting treatment following TB diagnosis, was comparable to that reported among pregnant women in the same setting [12] and similar to hospital-diagnosed patients in rural South Africa [25]. However, it was higher than the treatment initiation rates observed among TB patients diagnosed in an urban South African hospital [26] MDR/RR-DR-TB patients [22], and individuals living with HIV who were not on antiretroviral therapy (ART) [27].

While median time to treatment initiation was the same in pregnant and non-pregnant women (2 days), time to treatment initiation over 6 months was statistically different with pregnant women lagging behind non-pregnant women. The South African National Department of Health (DoH) recommends that people diagnosed with DS-TB should initiate treatment within 2 days and those diagnosed with DR-TB, within 5 days [28]. In our urban setting in Cape Town, pregnant women were able to access treatment within the targets set by the DoH for all patients diagnosed with TB. However, it is concerning that almost 30% of pregnant women and 21% of non-pregnant women had not started TB treatment within 7 days. Any health system interventions to improve treatment initiation rates in general (and time to treatment after 7 days) should include and prioritise pregnant women who appear to take much longer to initiate treatment than non-pregnant women. A systematic review reported a median delay of 14 days (95% CI: 3–84) between diagnosis and treatment initiation in low- and middle-income countries [29]. Two South African studies reported the median time to treatment was 0 days in an urban setting, when Xpert MTB/RIF was used as the diagnostic test [20,23]. Another study reported 14 days in a rural setting [21]. While median time to treatment in our study was short for both pregnant and non-pregnant women, explaining the lag experienced by pregnant women is more complex and may be driven by patient delays in health seeking and/or health system challenges.

The proportion of women living with HIV in this cohort was high (62%), regardless of pregnancy status and exceeded the 2023 estimate where half of all individuals with TB (54%) in South Africa were co-infected with HIV [1]. The high HIV prevalence within our cohort indicates a substantial burden of disease, where most women are simultaneously navigating TB, HIV, and reproductive health services. In South Africa, public antenatal services are not routinely integrated with other primary healthcare services, while TB and HIV services are integrated to varying extents [30, 31]. It is probable that women may have attended antenatal services in the period between diagnosis and treatment initiation. Despite national policy promoting decentralised, integrated TB care and rapid treatment initiation, [28, 32, 33], implementation in many facilities continues to reflect legacy vertical TB service models [34, 35]. Patients diagnosed in antenatal or hospital outpatient services are often referred to dedicated TB clinics for programme registration and treatment initiation, resulting in additional visits and avoidable delays. This fragmentation in care delivery may contribute to delays in initiating TB treatment among women who must access multiple, disconnected healthcare services. In addition, pill burden could also explain delays in initiating TB treatment by overwhelming women who are living with HIV and now have the added burden of additional pills and regimens [36, 37].

Overall, nearly a third of women in our cohort were diagnosed with TB in hospital, with a notably higher proportion among pregnant women compared to non-pregnant women. The majority of this cohort had TB that was bacteriologically confirmed but this was similar by pregnancy status. A hospital-based diagnosis may indicate delayed recognition of TB symptoms. Having a negative Xpert MTB/RIF (M.tb nucleic acid amplification test) test result on initial testing can reflect more advanced disease. Pregnancy can complicate TB diagnosis, as physiological changes and overlapping symptoms may mask classic TB presentations [38]. Healthcare providers are also hesistant to perform chest X-rays on pregnant women [39]. This diagnostic complexity can contribute to delays, which in turn increase the risk of adverse maternal, pregnancy, and neonatal outcomes. It is acknowledged that time to treatment initiation for women diagnosed with TB in hospital may be overestimated, as some women may have initiated TB treatment during their hospital stay. However, any treatment initiated in hospital or at discharge that was not captured in the electronic data feeding into the PHDC would not have been included in our dataset. It is important to note that registration for treatment initiation at the primary healthcare (PHC) level typically includes essential counselling and adherence support, which are critical for successful treatment outcomes. Although we were unable to determine whether some individuals may have initiated TB treatment at a hospital, registration and initiation at a PHC facility represent the optimal pathway for ensuring continuity of care and adherence support.

Given that national guidelines recommend evaluation for TB in pregnant women at all antenatal visits [40], and targeted universal TB testing in pregnancy [41], it is expected that pregnant women would be more likely to be tested for TB. At the same time, pregnant women may face competing health priorities and, where TB screening and testing requires an additional contact point, it may not be prioritised by them. There may also be a lack of awareness among women regarding TB symptoms and when to seek help. Lack of knowledge has been shown to be a predictor of treatment delays [42,43]. This highlights the need for strengthened community-level education and awareness initiatives as a critical intervention to improve early recognition of TB symptoms and promote timely health-seeking behavior, particularly among pregnant women. Increasing the demand for TB screening and testing by women of reproductive age and especially pregnant women is an important consideration for public health services.

Being pregnant was associated with a lower likelihood of initiating treatment but was not significant in the multivariable analysis. This suggests that the initially observed association may be confounded by other clinical (e.g., disease severity or comorbidities), or healthcare system (e.g., access to care, provider decision-making) factors. As the effect of pregnancy reduced in the multivariable model, the adjusted hazard ratio for site or mechanism of diagnosis almost doubled, highlighting the need to prioritise diagnosis at PHC and where possible using TB-NAAT testing. This finding highlights the need to examine the broader healthcare context to reduce TB treatment delays as structural or clinical complexities may influence care pathways and linkage beyond individual characteristics such as pregnancy.

The major strength of this study is that it fills a critical gap in the scientific literature. While treatment delays among TB patients are well documented, there are limited data on time to treatment initiation among pregnant women with TB. As the study used routine data, it demonstrates a mechanism for national TB programmes to evaluate time between diagnosis and treatment registration which may be used as an estimate for actual treatment initiation. In addition, this study provides crucial linkage to TB care data for women in their reproductive years, an under-researched group identified by the WHO, as facing unique health risks and challenges. These risks have significant implications for maternal, child, and public health. Failing to disaggregate data by sex and age represents a missed opportunity to understand and address the specific needs of this population, which is essential for developing better, safer, and more equitable TB care.

Using linked electronic health data curated by the Provincial Health Data Centre (PHDC), enabled the use of routine programmatic data from multiple integrated sources, that allowed us to evaluate time to treatment initiation for pregnant and non-pregnant women with TB in a programmatic setting.

The study had several limitations. First, the study made use of routine electronic data which was limited by the variables available. While we attempted to consider those factors affecting treatment initiation, no information on the broader social determinants of health were available. Time to treatment initiation will need further exploration including understanding the individual context and perspective. Second, missing or incorrect data leading to misclassification must be considered. Where women attended TB services or where women received TB treatment in hospital but recording in the electronic systems was incomplete, we may have over estimated the treatment delays. Third, in our dataset, the first available evidence of a clinical TB diagnosis is often the recorded date of treatment initiation. Consequently, for women who were clinically diagnosed with TB, the date of diagnosis and treatment initiation are frequently the same, which may limit the ability to distinguish between the timing of diagnosis and initiation of treatment. Fourth, we did not invesitgate TB and treatment initiation of women during the postpartum period, where the risk of TB is substantial. Fifth, our data was restricted to two health sub-districts around Cape Town and therefore cannot be extrapolated to the entire South Africa. Lastly, our dataset only included women who had a TB diagnosis and therefore we cannot say anything about those who had TB disease but who were not diagnosed.

While this study demonstrates good overall linkage to TB care and TB treatment registration, the time taken to start treatment in some women is concerning. Since we used linked routine data, time-related elements are already available in the PHDC, and consideration should be given to introducing alerts and reporting flags for all individuals diagnosed with TB. Prioritizing pregnant women could help ensure that opportunities for linkage to care after screening and diagnosis are not missed during future health service visits. Further work is needed to investigate how time-to-treatment initiation indicators and integration with existing community outreach services can improve timely access to TB care.

Conclusion

While pregnant South African women with TB largely started treatment, almost 15% were never started on treatment and delays in initiating TB treatment persist. Ongoing transmission risks, and increased morbidity and mortality for mother and infant require more strategic interventions to prioritise early treatment initiation. There should be a specific focus on pregnant women who have not initiated treatment within one month after TB diagnosis. Women diagnosed in hospital or bacteriologically are at especially high risk, and use of available digital tools should be considered to limit missed opportunities for treatment initiation during other contacts with health services

Acknowledgments

We wish to acknowledge staff at the Centre for Integrated Data and Epidemiological Research (CIDER) at the University of Cape Town and staff at the Western Cape Provincial Health Data Centre (PHDC) for their invaluable assistance, especially Mariette Smith and Florence Phelanyane, who were instrumental in assisting with the TB, HIV and pregnancy cascades. We are thankful to Dr Rory Dunbar at Stellenbosch University, for his data management expertise.