Factors associated with tuberculosis lost to follow-up among individuals in Buffalo City Metro, Eastern Cape Province, South Africa, 2019 - 2022 (a retrospective cohort study)

¹Division of Public Health Surveillance and Response, National Institute for Communicable Diseases, Johannesburg, South Africa, ²School of Public Health, University of Pretoria, Pretoria, South Africa, ³Strategy and Organisational Performance, Eastern Cape Provincial Department of Health, Bhisho, South Africa, ⁴Africa Health Research Institute, KwaZulu-Natal, Mtubatuba, South Africa, ⁵Department of Family Medicine, School of Medicine, Faculty of Health Sciences, University of Pretoria, Pretoria, South Africa, ⁶Department of Public Health, Walter Sisulu University, South Africa

^&Corresponding author
Nothembelani Jongisa, Division of Public Health Surveillance and Response, National Institute for Communicable Diseases, Johannesburg, South Africa

Introduction    

Tuberculosis (TB) remains a significant public health threat globally, with an estimated 7.5 million newly diagnosed cases and of those, 1.3 million deaths occurred in 2022 [1]. Incomplete TB treatment is a major concern, as it can lead to the emergence of drug-resistant strains, which are significantly more difficult and expensive to treat [2]. Lost to follow-up (LTFU) during TB treatment contributes to this problem, as it allows the bacteria to continue multiplying and potentially develop resistance to available antibiotics. LTFU during TB treatment is one of the major obstacles in the fight against TB, which has serious implications for patients, families, and communities [3]. Studies have shown that LTFU can significantly increase the risk of onward transmission, such as drug resistance, which often hinders the progress towards TB elimination targets [4].

Globally, the United Nations adopted the Sustainable Development Goals (SDGs) in 2015 [5]. The SDGs are comprised of 17 goals, and SDG 3.3 aims to end the TB epidemic by 2030 [6,7]. Progress in lowering TB incidence was also hindered in 2020, with an annual drop of less than 2% [8]. This rate is significantly lower than the required 4% to 5% annual fall required to meet the End-TB Strategy's target of an 80% reduction [9,10].

According to the WHO Global Tuberculosis Report 2021, 9.9 million people had TB and resulted in 1.5 million deaths, with sub-Saharan Africa, especially South Africa (SA), bearing the largest burden [11]. SA continues to have the highest TB incidence globally, despite a decrease from 988 cases per 100,000 people in 2015 to 513 per 100,000 in 2020 [12,13]. In 2021, SA and sub-Saharan Africa reported TB patient LTFU rates of 25% and 23%, respectively, which were notably higher than the global average estimated at 16% [12]. TB treatment success rates in SA are greatly impacted by TB LTFU. Treatment success rates for TB patients in SA were 79.2% in 2019. Across provinces, the Eastern Cape is the lowest by 78% [14,15].

There have been significant efforts to improve TB outcomes in the Eastern Cape Province (ECP) in SA. However, the BCM is consistently experiencing a decline in drug-susceptible TB treatment throughout the years from 2020 to 2022, from 74.2% dropping to 67.4%. The downward trend in BCM as against other districts in the province necessitates investigations on factors contributing to TB LTFU [16]. Therefore, the study aimed to describe the clinical and demographic characteristics, quantify the burden, and study the predicting factors of LTFU in Buffalo City Metro (BCM) from 2019 to 2022.

Methods     

Study design and setting: we conducted a retrospective cohort study using secondary data from three interlinked electronic registers among TB laboratory-confirmed patients in BCM and ECP, SA, from 1^st January 2019 to 31^st December 2022. Buffalo City Metro (BCM) is a metropolitan municipality situated on the coast of ECP, in SA. It includes East London, Bhisho, and Qonce as well as two townships, Mdantsane and Zwelitsha. The metropolitan has the highest burden of TB (including drug-resistant TB) and HIV (AIDS) [17]. Buffalo City Metro (BCM), one of two EC metropolises and six districts, provides public health services via 74 fixed clinics, 11 mobile clinics, five community health centres, one regional, one tertiary, and two specialized hospitals.

Study population: the study population was laboratory-confirmed TB patients in BCM who were diagnosed between 2019 and 2022.

Sampling: we conducted convenient sampling by enrolling every laboratory-confirmed TB record from the Tier.Net database; therefore, we included the maximum possible sample for representation and target population inference.

Inclusion criteria: we included all those individuals with polymerase chain reaction (PCR), X-ray, clinical, culture, and GeneXpert positive for DS-TB in the BCM between 2019 and 2022, and who were initiated on TB treatment, registered on Tier.Net, and assigned a treatment outcome.

Exclusion criteria: we excluded TB case patients without an outcome assigned, children and teenagers were also excluded, such as patients aged 17 years or younger.

Data collection and management: data was obtained from Tier.Net. Tier.Net is a non-networked electronic program containing TB/HIV patient information that is used for surveillance and management to support patient management. Generated data is used to report routine program performance data into the web District Health Information System (DHIS) [18]. The patient data was manually transcribed from the Tier.Net patient view page onto a Microsoft Excel sheet. The data collection tool included the following variables: demographic data (age, sex, clinical data, HIV information, patient category, diagnostic test, TB type, registration type, and disease classification). The outcome variable was denoted as LTFU and not LTFU.

The following steps were taken to ensure data quality: data validation: we enforced mandatory fields for categorical variables and provided predefined options. To check the accuracy of dates and numerical values, validation rules were implemented. Data storage: the data extracted was password-protected and stored in the principal investigator's (PI) laptop, which is password-protected and only accessible to the PI and co-investigators. Data was stored on a secure server. A firewall and password protection were in place to safeguard the server. Only authorised persons had access to the data. Data cleaning: the data was anonymised, and unique identification numbers were assigned. Data was imported to STATA version 18 statistical software (Stata Corporation, College Station, Texas, United States) for coding and analysis.

Outcome: we determined predictors for TB LTFU among individuals with TB in the BCM in the past four years. TB LTFU was defined using the World Health Organization (WHO) definition, which defined it as an individual diagnosed with pulmonary tuberculosis (PTB) who begins TB treatment but discontinues it for two or more consecutive months without medical approval is considered a case of TB LTFU [19]. The outcome variable was coded as binary: 1 = LTFU, and 0 = not LTFU.

Definitions

Demographic variables: registration year - the year in which the patient was registered for TB treatment. Age - patient age at the time of TB registration, recorded in completed years; sex - male/female; settlement type - classification of patient residential type: rural, township, and urban; sub-areas - geographic sub-areas within BCM used for health reporting: Mdantsane, KWT/Bhisho, and East London.

Clinical variables: HIV information - HIV status (positive/negative/unknown) and ART; patient category - new TB case, retreatment def./fail, relapse, other; TB diagnostic test - PCR, X-ray, clinical, culture, smear, GeneXpert; TB type - pulmonary, extrapulmonary TB; registration type - new, transfer-in.

Data analysis: data was divided into two groups: Data of patients diagnosed with TB, registered for TB treatment on Tier.Net, and assigned a treatment outcome as LTFU were included in the study. Data of patients diagnosed with TB, registered for TB treatment on Tier.Net, and assigned a treatment outcome other than LTFU were also included as a comparison group. A data collection tool was developed, and data was recaptured into an Excel spreadsheet for data cleaning. The data was de-identified using the name, surname, and date of birth. Data were imported to STATA version 18 statistical software (Stata Corporation, College Station, Texas, United States) for coding and analysis. The primary outcome of the study was TB-infected individuals who were started on treatment and documented as lost to follow-up. Descriptive statistics were used to summarize the characteristics of the study participants. Association between categorical variables was analysed using the Chi-squared test and the t-test for continuous data. A multivariable logistic regression model was employed to identify the factors associated with TB LTFU. A P<0.20 as a cut-off value for inclusion in the multivariable logistic regression model was used. A multivariable model using a backward stepwise regression approach was used to select variables for inclusion in the final multivariable logistic regression model. Statistical significance was regarded as reached in the final model of the multivariable logistic regression if the P<0.05 level.

Ethical consideration: ethical approval was obtained from the University of Pretoria AAC and the Research Ethics Committee, with the following reference number: 174/2024. The Eastern Cape Provincial Health Research Committee Secretariat approved the study with the following reference: EC_202404_019. The researchers did not have direct contact with any of the TB cases included in the study. The researchers complied with the ethical principles of the Declaration of Helsinki and the South African National Health Research Ethics Council Guideline.

Results     

A total of 14,182 TB case patients were included and analysed between the period of 2019 and 2022. Of these 1,969/14,182, (14.8%) were classified as LTFU, while the other case patients were classified as the following outcomes (treatment complete, cured, treatment failure, and died) and were not analysed in detail in this study (Table 1).

Demographic and clinical characteristics of TB cases: LTFU: of the 1,969 TB LTFU, 1,801 (91.4%) had pulmonary tuberculosis. Males accounted for 1,432 (72.7%), and the median age was 40 years (IQR: 31-47 years). Those aged 30-44 and 45-59 years accounted for 939 (47.7%) and 471 (23.9%), respectively; 749 (38.0%) had human immunodeficiency virus, and 673 (34.2%) were on antiretroviral therapy. A significant portion of case patients had pulmonary TB, 1,801 (91.4%), and 90 (4.7%) had extrapulmonary TB. About 1,487 (75.5%) patients were registered as newly diagnosed, 119 (6.0%) were treated after relapse, 3 (0.2%) were re-treated after treatment failure and after default, and the remaining 18 (0.9%). About 1,035 (52.6%) of TB LTFU were HIV-negative, whereas 673 (34.2%) were on ART and 76 (4.2%) were HIV-positive but not on ART. Within this group, Gene Xpert was the diagnostic test that was mostly used in 727 (37.0%) cases, followed by smear microscopy at 655 (33.3%), and X-ray at 142 (7.2%). Of the total of 1,969 registrations, 1,759 (89.3%) were documented as new registrations, while 210 (10.6%) were transferred in (Table 2).

Temporal trends of LTFU rate by sub-areas and sex in BCM, 2019 - 2022: Figure 1 illustrates the decreasing trend of LTFU from 2019 to 2022. Over the past four years, the LTFU has declined by 9% from 18% to 9.1%, which suggests an improvement in the TB control measures. In Figure 2, East London observed the highest LTFU over the years, while Mdantsane observed the lowest trends over time. All sub-areas reported the decline of LTFU since 2021. Higher LTFU rates were observed among males compared to females. Decreasing LTFU cases across both sexes, with LTFU baseline of 158 cases in 2019 to 85 cases in 2022 among females and 363 cases in 2019 to 250 cases in 2022 among males. A significant decline in both males and females was seen in 2021 as compared to previous years, and in 2022, males showed a slight rebound to 2021 (Figure 3).

Factors associated with lost to follow-up among TB patients: to analyse factors associated with LTFU, the Poisson regression model was used, and significant differences were found across the various demographic and clinical characteristics. The univariable analysis has identified that males, older age population, some sub-areas, township dwellers, HIV status, TB registration, diagnostic test type, and pulmonary TB were found to be significantly associated with the TB LTFU. When taken to the multivariable analysis, the male patients in both models, the aIRR was 1.47, 95% CI: 1.33 - 1.63; P<0.001, and new TB cases (aIRR 2.96, 95% CI 2.28 - 3.83; P<0.001), the HIV-positive not on ART (aIRR: 1.9, 95% CI: 1.49-2.43; P<0.001), and pulmonary TB (aIRR 1.25; 95% CI:1.020-1.55; P=0.031), were associated with higher probability of TB LTFU. The 30-44 age group (aIRR 0.88, 95% CI 0.79-0.99; P=0.04), 45- 49 age group (aIRR 0.76, 95% CI: 0.67 - 0.87, P<0.001), patients residing in the township (aIRR 0.42, 95% CI: 0.37-0.50, P<0.001) and HIV positive on ART (aIRR of 0.80, 95% CI: 0.72-0.89; P<0.001) were associated with lower probability of TB LTFU (Table 3).

Discussion     

The objectives of this study were to determine and describe the demographic and clinical factors that are associated with TB LTFU among TB patients in the BCM. In this study, we determined predictors for TB lost to follow-up among individuals with TB in the Buffalo City Metro in the past four years. TB loss to follow-up was defined using the WHO definition, which defined it as an individual diagnosed with pulmonary tuberculosis (PTB) who begins TB treatment but discontinues it for two or more consecutive months without medical approval is considered a case of TB LTFU. In this study we found that males, the older age group, some sub-areas, township dwellers, HIV status, TB registration, diagnostic test type, and pulmonary TB were found to be significantly associated with the TB LTFU.

The study determined the following risk factors, which included males and the middle-aged group, who accounted for higher proportions. The above findings are consistent with studies conducted in Chennai, India, and in England and Wales, which found that sex and age contributed to the TB LTFU males are likely to be lost to treatment due to the fact that their health-seeking behaviour tends to delay may leave and migrate to other areas in search job opportunities as soon as they start treatment and start to feel better [20-22]. In our study, migration and health-seeking behaviour were not studied; however, these risk factors are documented widely, and this may assist in explaining the high proportions seen among men at middle ages.

The highest proportion of TB lost to follow-up was found in urban areas, 1,200 (62.8%), despite the metro having facilities in greater proximity; urban areas may pose unique challenges to TB treatment adherence, such as risk factors, like overcrowding in townships and constant mobility due to job seeking, which are more likely to contribute to negative outcomes in treatment adherence. These findings are consistent with studies conducted in developing countries where urban and township residents have a significantly higher risk of LTFU compared to rural populations. Geographic disparities in health outcomes are well-documented, with urban areas often exhibiting poorer health metrics due to socioeconomic factors. Similar findings were observed in a study conducted in England, where urban areas showed poorer TB treatment outcomes, which were influenced by socioeconomic factors, which subsequently led to increased risk of loss to follow-up [23]. However, our study findings are contrary to what is documented in other studies, where it is indicated that poor rural communities are the most affected by TB LTFU [24,25]. The results revealed by our study could be attributed to the lifestyle behaviours, inadequate living and working conditions of individuals residing in urban and township areas.

In our study, patients staying in rural areas contributed the lowest proportion of TB lost to follow-up patients. These findings are inconsistent with the two studies conducted in different settings in Ethiopia [25,26]. A study conducted in India also agrees with the findings, where they found the lowest proportions of LTFU in rural areas; however, age group is different, in the findings, 50 years and older reported high proportions [20].

This study revealed that male TB patients constituted the majority of LTFU patients, as compared with females. The results of our study align with prior studies, such as the study published in Open Forum Infectious Diseases, which suggests that males have a notably greater susceptibility to acquired tuberculosis, in comparison to males. Our data corresponds to the age-specific risk patterns that have been identified, especially within the 30-40-year age bracket. These correlations emphasize the need to take into account gender and age in strategies for preventing and treating tuberculosis [27].

The highest proportion of TB lost to follow-up was found in urban areas, 1,200 (62.8%). This situation can be attributed to the fact that health facilities may have more patients seeking health care compared to patients from remote areas, due to the accessibility of health facilities, since it is situated in an improved and semi-developed geographical area. This information is contrary to what is documented in other studies, where it is indicated that poor rural communities are the most affected by TB. This factor can be attributed to the lifestyle behaviours of individuals residing in urban and township areas [24,28].

Our study highlights the key important findings about predictors associated with LTFU among individuals with TB. In our study, the new TB cases and HIV-positive individuals not on antiretroviral therapy (ART) were statistically significant. However, the reasons why new TB cases were more likely to be LTFU could not be determined from the existing data due to the non-documentation of reasons for treatment discontinuation. Nevertheless, the findings from our study concur with the observational study conducted in China, where they observed that the majority of patients who were LTFU were lost during the initial phase of treatment in the first 3 months after starting their treatments [29].

HIV-positive individuals not on ART were statistically significant in the univariable and multivariable models. These findings are consistent with the study conducted in South Africa and in Ethiopia, where their findings demonstrated that HIV co-infection was associated with the rise of LTFU in the univariable analysis [30]; however, in the Ethiopian study, this relationship was not sustained in the multivariable analysis [31]. This suggests that HIV status may seem to influence LTFU when other factors are accounted for, even though these predictors may have a more significant impact when more factors are observed simultaneously.

Furthermore, the multivariate analysis revealed that individuals receiving ART had a lower likelihood of LTFU with aIRR of 0.80, 95% CI: 0.721-0.890, P<0.001. This finding is supported by the existing literature, which found that early initiation of ART was associated with a reduced rate of LTFU among HIV-positive TB patients in Ethiopia [32,33]. Likewise, the research study conducted in Nigeria revealed that the initiation of ART was mostly associated with improved TB treatment results, including decreased rates of LTFU [34]. The results are consistent across different settings with high burden, which underscores the importance of ART in improving TB treatment outcomes in HIV-positive individuals.

To get control of these obstacles, issues surrounding perceived barriers, such as improving access to essential services for the beneficiaries in urban and township areas to primary health settings by strengthening patient tracer teams in other parts of the metro, food security, and proper housing facilities, should be prioritised. Establishment and strengthening of the district/sub-district/sub-areas joint coordination by the departments of health and social development to align the community health workers and community care givers' work, such as sharing of geographic coverage, case management unification, standardisation of patient/client referrals and tracking, this will minimise duplication of services and promote optimal use of resource utilisation in the TB management. Case finding in males needs to be strengthened. Root-cause analysis should be done during program planning to unlock the bottlenecks in the population at risk.

Our study was a secondary data analysis; therefore, there were inconsistencies, missing data, and invalid data. To address these limitations, we took several steps to ensure the quality of the data. First, we conducted a data cleaning process to remove any incomplete or inaccurate records. We used the Poisson regression statistical method to adjust for any biases in the data. There may also be an under-representation of the TB-infected persons since the system does not capture all TB patients, such as those who do not come for medical assistance, patients treated by private sector healthcare providers, those who die at home, and those with no medical facility access or access to community-based or outreach services. The study focused on BCM, which limits the findings from being generalized to other districts. One more possible limitation of this study is that we excluded children under 18 years in the analysis of factors associated with TB LTFU. Another limitation was the reliance on programming to routinely collect data at the facility level, which lacked socioeconomic factors. The study provides evidence for targeted intervention.

Conclusion     

Our study identified the clinical and demographic factors. The middle-aged population emerged as the most susceptible age group, and males had a greater frequency of TB LTFU. Notably, patients on ART demonstrated a lower risk of developing TB LTFU. Geographically, individuals residing in urban areas were mostly affected by TB. The protective effect of ART highlights the importance of strengthening the integration of HIV and TB care. Furthermore, the geographical disparity in risk underscores the need for location-specific tuberculosis control strategies. The findings suggest the intricate relationships between demographic, geographic, and clinical risk factors that affect adherence to TB treatment. This emphasises the need to prioritise the targeted intervention towards male patients, individuals aged 30-59. A similar, larger, and more representative study is recommended on LTFU of TB patients; furthermore, a prospective study preferable a qualitative approach, is recommended to ascertain the risk factors associated with LTFU.

Competing interests     

The authors declare no competing interests.

Authors' contributions     

Conception and study design: Nothembelani Jongisa, Nqobile Ngoma, Ruvimbo Chingonzoh, Thomas Dlamini, Hetani Mdose, Alfred Musekiwa, Singilizwe Tinkili Moko, Zonwabele Merile, and Namhla Linda Mdingi-Buqa; data collection: Nothembelani Jongisa; data analysis and interpretation: Nothembelani Jongisa, Nqobile Ngoma, Ruvimbo Chingonzoh, Thomas Dlamini, Hetani Mdose, Alfred Musekiwa, and Singilizwe Tinkili Moko; manuscript drafting and guarantor of the study: Nothembelani Jongisa; manuscript revision: Nqobile Ngoma, Ruvimbo Chingonzoh, Singilizwe Tinkili Moko, Thomas Dlamini, and Namhla Linda Mdingi-Buqa. All the authors read and approved the final version of this manuscript.

Acknowledgments     

We would like to thank the Eastern Cape Department of Health and the Buffalo City Metro Health District team for allowing the researchers to access the database, and the University of Pretoria for authorization of the study. We would like to appreciate the scientific writing advice provided by the Provincial and National Institute for Communicable Diseases (NICD) epidemiologists, Provincial Department of Health - SOP, SAFETP teams, and Nqobile Ngoma.

Tables and figures     

Table 1: baseline characteristics of tuberculosis patients, Buffalo City Metro, Eastern Cape Province, 2019 - 2022

Table 2: socio-demographic and clinical characteristics of tuberculosis patients by lost to follow-up outcome, 2019 - 2022

Table 3: predictors of tuberculosis lost to follow-up among individuals, Buffalo City Metro, 2019 - 2022

Figure 1: Buffalo City Metro District tuberculosis lost to follow-up percentage trend analysis, 2019 - 2022

Figure 2: sub-areas trends for tuberculosis lost to follow-up cases, in Buffalo City Metro, 2019 - 2022

Figure 3: tuberculosis lost to follow-up trends by sex, Buffalo City Metro, 2019 - 2022

References