The study followed a cross-sectional design. Data were gathered among TB patients across 11 PHC facilities in the Free State province from November 2019 to December of 2019. The Free State, with 10 994 new TB cases and a mortality rate of 10.6% in 2019 – the highest among the nine South African provinces²¹ – was purposefully selected for the study. Despite prevailing calls for integration, health services at the PHC level in this province are largely verticalised.^22,23 The National Mental Health Policy Framework has only recently endorsed including mental health data elements in the District Health Information System.²⁴ Prior to this, mental health statistics were not routinely captured and considered by the TB programme.

Given the constraint of limited resources, data collection was undertaken over one month, during which 208 patients were conveniently interviewed. Patients were eligible for the study if they were 18 years or older, on susceptible TB treatment, and could express themselves in English or Sesotho. Patients who were too ill to be interviewed and those who were already diagnosed with mental health conditions at the time of the research visit were excluded from the study. Patients undergoing TB re-treatment and those diagnosed with drug-resistant TB were excluded from the study as they tend to present with more severe and chronic TB disease involving complex treatment regimens and frequent hospitalisation.

To ensure that this study had sufficient statistical power (at least 80%), the G–Power calculator for sample size estimation was utilised.²⁵ In the analysis utilising sex as a predictor variable and given the preponderance of depression among females,⁸ it was assumed that 50% of women and 30% of men would exhibit symptoms of depression, with an odds ratio (OR) of 2.33 between the two groups. Considering a desired margin of error of 0.05, the calculation indicated that a minimum sample size of 190 patients would be necessary for a robust binomial logistic regression analysis. This sample would ensure that the study was adequately powered to provide reliable estimates.

A structured fieldworker-administered questionnaire was used to collect the data. Socio-demographic questions included age, sex, and social support. With three questions, the Oslo Social Support Scale (OSSS-3)²⁶ was used to measure patients’ perceptions of social support: Firstly, ‘How many people are so close to you that you can count on them if you have great personal problems?’ Response options included: 1 = ‘none’, 2 = ‘1 to 2 people’, 3 = ‘3 to 5 people’, and 4 = ‘more than 5 people’. Secondly, ‘How much interest and concern do people show in what you do?’ Response options included: 1 = ‘none’, 2 = ‘little’, 3 = ‘uncertain’, 4 = ‘some’, and 5 = ‘a lot’. Thirdly, ‘How easy is it to get practical help from neighbours?’ Response options included: 1 = ‘very difficult’, 2 = ‘difficult’, 3 = ‘possible’, 4 = ‘easy’, and 5 = ‘very easy’. The OSSS-3 has demonstrated a somewhat acceptable internal consistency of 0.6, primarily attributed to its brevity.²⁶ Despite this, the scale is widely used in diverse research settings because of its conciseness and ease of administration.²⁶

Possible depression was measured using the nine-item Patient Health Questionnaire (PHQ-9).^27,28 The scale assessed the frequency of depressed mood during the two weeks before the research visit. Responses were recorded on a 4-point Likert scale: 0 = ‘not at all’, 1 = ‘several days’, 2 = ‘more than half the days’, and 3 = ‘nearly every day’. Adjustments were made to the response set to enhance respondent comprehension such that ‘several days’ was defined as ‘1 to 7 days’, ‘half the days’ was defined as ‘8 to 11 days’, and ‘nearly every day’ was defined as ‘12 to 14 days’. The PHQ-9 has been found to be valid and reliable for use among TB patients in South Africa.²⁹ Clinical information extracted from patient records at the PHC facilities included TB disease classification, treatment duration, alcohol use, tobacco use, human immunodeficiency virus (HIV) status, and antiretroviral treatment (ART) status.

The questionnaire and consent form were available in English and Sesotho. A bilingual translator performed a forward translation of the research instruments from English into Sesotho. Then, another bilingual translator translated the Sesotho versions back into English. Thereafter, both translators compared the back-translated versions to the original versions of the research instruments, engaging in discussions to resolve any emerging discrepancies and reach a consensus.

A team of experienced bilingual fieldworkers, proficient in Sesotho and English, conducted face-to-face interviews in the patients’ preferred language. Participants were recruited with the assistance of healthcare providers. After their consultation session, the attending nurse referred eligible patients to trained fieldworkers waiting in private rooms within the PHC facility. The fieldworkers were trained to look for signs of distress and refer patients to a social worker or psychologist attached to the PHC facility for further assistance. The fieldworkers scored the patients on the PHQ-9 immediately after the interviews. Patients whose PHQ-9 scores suggested possible depression (≥ 5)^27,28 were referred to the social workers or psychologists for further evaluation and diagnosis. The questionnaire took approximately 10 minutes to complete.

Data were analysed using the IBM Statistical Package for the Social Sciences (SPSS) statistics (version 27).³⁰ Discrete variables were presented as frequency counts and percentages, and continuous variables were presented as means (standard deviations [s.d.]) and medians (inter-quartile range [IQR]). Chi-square tests of independence were used to assess whether there were significant differences between categorical variables. Sum scores were computed for the OSSS-3, ranging from 3 to 14. Higher scores represented more robust levels of social support.²⁶ Similarly, sum scores were computed for the PHQ-9, with scores above 4 suggesting possible depression.^27,28 Regarding the grading of symptoms based on the PHQ-9, total scores falling between 0 and 4 suggested minimal symptoms, scores ranging between 5 and 9 suggested mild symptoms, scores between 10 and 14 suggested moderate symptoms, and scores falling between 15 and 27 suggested severe symptoms.^27,28

Binomial logistic regression was performed to ascertain the factors associated with possible depression. The outcome variable was possible depression, defined as a score of ≥ 5 on the PHQ-9. The following key assumptions for the binomial logistic regression were inspected. Firstly, the independence of observations was examined using plots of standardised residuals and predicted values, and a random scatter of points around zero was demonstrated, suggesting independence. Secondly, the linearity of the logit of continuous variables was inspected using the Box Tidwell procedure. Based on this assessment, all continuous variables were linearly related to the logit transformation of the dependent variable. Thirdly, the occurrence of extreme outliers examined through inspection of studentised residuals identified two potential outliers with values greater than 2, while the inspection of the leverage values identified five high-leverage values. Since these unusual points were not because of data entry errors, the cases were maintained in the analysis. Fourthly, multicollinearity was assessed using a simple correlation analysis among continuous variables. The analysis established that the correlation coefficients were all less than 0.7, suggesting an absence of multicollinearity.

A test of the full model against a constant-only model was statistically significant (χ² = 26.410, p < 0.001, degrees of freedom [df] = 9), indicating that the predictors, as a set, reliably distinguished between possible depression and the absence of symptoms. The model explained 23.5% (Nagelkerke’s R²) of the variance of possible depression. The overall prediction success was 69.9%, with a specificity of 65.6% and a sensitivity of 73.6%. The positive predictive value was 68.9%, and the negative predictive value was 70.7%. Statistical significance was considered at a 95% confidence interval (CI) and p < 0.05.

Ethical approval for this study was received from the Health Sciences Research Ethics Committee, University of the Free State (UFS-HSD2019/1574/2611). The Free State Department of Health authorised access to PHC facilities to interview TB patients. Participation in the study was voluntary; prospective respondents were provided with information regarding the purpose of the research and potential risks to enable them to make an informed choice on whether to participate. Respondents signed informed consent forms consenting to the interviews and for their clinical information to be collected from the patient files. The paper-based data were captured in SPSS and secured in encrypted files on a password-protected computer. All data-gathering instruments, including informed consent forms and questionnaires, were kept in locked cabinets. If patients experienced emotional discomfort or distress during the interviews, or their PHQ-9 scores were higher than 4, suggesting possible depression, they were referred to a social worker or psychologist attached to the PHC facility for further assistance. Furthermore, all participants’ information was handled confidentially. The research team was required to sign statements agreeing to protect the security and confidentiality of participants’ information. Patients’ names and other identifying information were collected to verify their clinical information. However, the data were aggregated for dissemination. Efforts were made to ensure infection control during the data-gathering exercise. The patient interviews occurred in privately situated and well-ventilated rooms within the PHC facilities.

Table 1 depicts the socio-demographic and clinical characteristics of the 208 patients interviewed in this study. Most patients were male (n = 137, 65.9%). The mean age of the sample was 42.4 (s.d.: 15.2) years and the median age was 38.5 (IQR: 30.3–54.0) years. Most patients were diagnosed with pulmonary TB (PTB) (n = 174, 83.7%). The median TB treatment duration was 2.5 (IQR: 1–4) months. More than one-third of the patients were documented in the patient files as, respectively, alcohol (n = 81, 38.9%) and tobacco users (n = 69, 33.3%). About 6 in every 10 patients were co-infected with HIV (n = 118, 56.7%), with the majority (n = 84; 71.2%) being recorded in patient files as receiving ART. Just under half (n = 96; 46.2%) of the patients in this study screened positive for possible depression with 22.6% (n = 47), 18.8% (n = 39), and 4.8% (n = 10) presenting with mild, moderate, and severe symptoms, respectively. Half (n = 59; 50.0%) of the 118 patients co-infected with HIV screened positive for possible depression, with 26.3% (n = 31), 17.8% (n = 21) and 5.9% (n = 7) presenting with mild, moderate and severe symptoms, respectively. While possible depression was higher among TB-HIV coinfected patients than their HIV-negative counterparts, this difference was not statistically significant (χ² = 1.623, df = 11; p = 0.203).

Binomial logistic regression was performed to determine the effect of sex, age, TB disease classification, treatment duration, social support, tobacco and alcohol use, and HIV status on possible depression. The results in Table 2 depict that sex (OR: 3.2; CI: 1.76–5.83) and treatment duration (OR: 0.8; CI: 0.73–0.97) were independently significantly associated with possible depression. However, because of their theoretical association with the dependent variable, all the independent variables were considered in the adjusted model.³¹ After controlling for all variables in the model, possible depression was significantly associated with sex, TB disease classification, and treatment duration. Female patients were 3 times more likely to present with possible depression compared to their male counterparts (AOR: 3.0; CI: 1.25–7.32). Patients diagnosed with extra pulmonary TB (EPTB) were 2.7 times more likely to present with possible depression compared to those diagnosed with PTB (CI: 1.03–7.21). Alternatively, longer TB treatment was protective against possible depression (AOR: 0.8; CI: 0.73–0.95). Furthermore, Table 3 shows that after adjusting for confounding variables, the likelihood of possible depression was higher among HIV coinfected females (AOR: 2.8; CI: 1.22–6.56), EPTB patients (AOR: 3.5; CI: 1.11–11.50), and patients who were receiving ART (AOR: 2.5; CI: 1.13–5.46) relative to their respective counterparts, and less likely among those on TB treatment for longer (AOR: 0.8; CI: 0.67–0.97).