The study was conducted in the oncology in-patient ward and out-patient clinic at Princess Marina Hospital, the main cancer referral centre for the southern part of the country, which serves a diverse range of cancer patients. The out-patient clinic and in-patient unit are housed within the same building and clinical staff move between both units.

This is a descriptive study that used a cross-sectional design. It was conducted from February 2023 to August 2024. Convenience sampling was employed for several reasons. The study was conducted as part of a training programme, which restricted what could be accomplished because of limited research time. In addition, in-patient turnover was limited because patients who lived far from the hospital needed to remain hospitalised for the duration of their treatment, including procedures typically conducted on an out-patient basis, such as chemotherapy infusions, and as a result, out-patient recruitment was higher than in-patient recruitment.

To be precise, 302 oncology patients agreed to participate in this study. Any client with a diagnosis of cancer who was 21 years old or older and who was an in-patient or out-patient seen in the oncology unit at Princess Marina Hospital was included in the study. Patients who were too sick to answer the questions (e.g. those with altered mental status or who were being treated in the high dependency ward or who were acutely ill needing emergent medical attention) were excluded from the study.

In-patients and out-patients were recruited on high-volume clinic days when oncology physicians were seeing out-patients. A general announcement was made in the ward and in the out-patient waiting area, and those who showed interest were approached personally and provided information about the study. Those who were willing to participate provided verbal consent.

Consented participants completed a questionnaire in either Setswana or English, and assistance was provided by the study team for participants unable to complete it by themselves because of their level of education or physical disability.

The structured questionnaire collected self-reported demographic and medical history data, including age, gender, average income, level of education, comorbidities, history of mental health illness and family history of mental illness. The history of mental illness included any psychiatric diagnosis that the patient reported, including depression. This was a cross-sectional study that looked at the current mental state of the participants; therefore, previous diagnosis of depression before cancer diagnosis was recorded but not considered during analysis. Participants were asked to subjectively rate their current pain level on a pain scale from 1 to 10, where 1 was no pain and 10 was extreme pain (the worst pain of their life). Participants then completed the Patient Health Questionnaire-9 (PHQ-9) and the World Health Organization QOL scale Brief Version (WHO-QOL-BREF) questionnaires.

Sensitive topics were handled with utmost care and addressed in a confidential manner and with respect for the participants. Participants who exhibited signs of depression or suicidality, based on their answers to the PHQ-9 (i.e. self-report), were referred for appropriate care.

All the instruments used in this study were translated and back-translated into Setswana by independent translators to ensure accuracy and quality.

The PHQ-9 is a freely available screening tool for depression that has been validated for use in Botswana.²⁰ It is a tool that consists of nine screening questions, and each question is scored from 0 to 3 points with a maximum score of 27. High scores on the PHQ-9 indicated that at the time the questionnaire was administered, the participant had symptoms suggestive of depression. The WHO-QOL-BREF tool measures various subjective aspects of QOL including physical health, psychological health, social relationships and environmental factors that lead to decreased QOL. Each of 20 questions is scored on a Likert scale from 1 to 5, and the overall score (out of 100) grades the QOL of the individual.²¹ The maximum score is 100, and the minimum score is 0, where the higher the score, the better the perceived QOL. In similar studies, the internal consistency was high for all domains of the WHO-QOL-BREF (Cronbach’s α ≥ 0.731).²² In this study, Cronbach’s α = 0.8, suggesting that the internal consistency of each part of the WHO-QOL-BREF was good.

The primary outcome was the prevalence of depression in the population of cancer patients at Princess Marina Hospital. The secondary outcomes were factors associated with depression and QOL. Data were collected on Microsoft Excel and exported for data analysis to Statistical Package for Social Sciences (SPSS for Windows 29.0.2.0 SPSS Inc., Chicago, IL, United States) and Stata Statistical Software (StataCorp. Stata Statistical Software: Release 15. College Station, TX: StataCorp LP; 2017). Descriptive statistics were presented in frequency tables and charts. Categorical variables were expressed in proportions or percentages. Continuous variables were summarised by means and standard deviation or medians and interquartile range. T-test was used for comparing continuous variables, while categorical variables were assessed by the chi-square test. Depression was diagnosed using a PHQ-9 score of 9 or higher, based on a meta-analysis indicating that this cut-off offers optimal sensitivity and specificity.²³ This cut-off was used in the dichotomous analysis, and in addition, because the severity of depression is directly proportional to the PHQ-9 score, the score was also analysed as a continuous variable. The WHO-QOL-BREF score was considered a continuous variable where scores closer to 0 were indicative of poor QOL and those closer to 100 indicated good QOL. Missing demographic data were noted during analysis and indicated in the tables where relevant.

A Poisson regression model with robust standard errors was used to estimate the association between demographic and clinical factors and the prevalence of depression. This model was chosen because the prevalence of depression in our study population was 35.8%, which exceeds 10%, making the outcome common; in such cases, prevalence ratios (PR) provided by Poisson regression are more accurate and interpretable than odds ratios (ORs). Univariable and multivariable analyses were performed, respectively. In the univariable analysis, each factor was tested individually to determine its association with depression. Following that, a multivariable model was constructed using a forced-entry (a priori) approach, including all collected demographic and clinical variables regardless of their univariable significance. This was done to rigorously control for potential confounding and to determine the independent effect of each predictor, which is represented as the adjusted prevalence ratio (aPR).

Based on prior literature, we predicted a prevalence of 25% for depression among cancer patients,²⁴ and thus we required a minimum sample size of 288 participants with a standard deviation of 1.96 for a 95% confidence interval and a degree of error estimated at 5%.

This study followed the principles outlined by the Declaration of Helsinki. Ethical clearance to conduct this study was obtained from the Institutional Review Board of University in Botswana, the Ministry of Health and Wellness and Princess Marina Hospital (Ref. No. UBR/RES/IRB/BIO/GRAD/170, HPDME: 13/8/1 and PMH 2/11AII [208]). Verbal consent was obtained from each of the participants. Data were collected, and a code was assigned to each participant. The link between the code and the participant’s details was kept in a separate place such that there was no bias during data entry and analysis. Verbal consent was chosen as a mode as the approval for this study was sought during the coronavirus disease 2019 (COVID-19) pandemic, necessitating a model that would keep the researchers and the participants safe.

To be precise, 302 oncology patients (28% in-patients and 72% out-patients) at Princess Marina Hospital were recruited and consented. The prevalence of depression among our study population was 35.8% (108) (95% confidence interval [CI]: 30.3%, 41.5%).

Socio-demographic characteristics (Table 1) were compared between participants with and without depression; however, no significant differences were found. The median age was 55 years (interquartile range [IQR] 43.5–66), and 71% were female. One-third of the participants were married, and the majority were Christians (92%). Most of the participants (210, 70%) were unemployed and therefore did not have a personal income; the remaining 92 participants reported an average income of $29 per month.

The subjective mean pain score reported by participants was 4 (± standard deviation [s.d.] 3), and participants with depression reported higher mean pain scores than those without depression (6 vs. 4, p < 0.001). Between those with and without depression, there were no differences in prevalence of chronic diseases, including HIV, personal history of mental illness and hypertension (p-value > 0.05). There was also no difference in family history of mental illness between those with and without depression (p = 0.378). While not statistically significant, a higher proportion of participants with type 2 diabetes were depressed (52.6%) compared to those without (34.6%) (Table 2).

There was no statistically significant difference between the type of cancer and participants with depression (Table 3).

In univariable analysis, none of the socio-demographic and clinical characteristics showed a statistically significant association with depression. Similarly, in the multivariable model adjusting for all covariates, most factors, including age, sex, ethnicity, marital status, occupation, education, religion, patient status, HIV status, hypertension and history of myocardial infarction, remained not significantly associated with depression (all p > 0.05).

In both univariable and multivariable models, participants with a new cancer diagnosis had a significantly higher prevalence of depression compared to those currently on treatment (aPR = 1.10; 95% CI: 1.01–1.19; p = 0.03). Participants in the palliative phase or post-treatment did not show statistically significant differences in depression prevalence compared to the reference group (palliative: aPR = 0.75; 95% CI: 0.51–1.10; p = 0.15; post-treatment: aPR = 0.99; 95% CI: 0.88–1.11; p = 0.89).

However, type 2 diabetes was a significant predictor of depression in the adjusted model. Participants with diabetes had a 16% lower prevalence of depression compared to non-diabetic participants (aPR = 0.84; 95% CI: 0.71–0.99; p = 0.033) (Table 4).

Depression was significantly associated with reduced QOL scores across all four domains, including physical health, psychological health, social relationships and environment (Table 5). There was a negative correlation between depression and the domains of physical health, psychological health, social relationships and environment (Table 6). In contrast, pain was positively correlated with depression scores, suggesting that higher pain levels are associated with increased depression scores, while the opposite is true for QOL scores.

Pain was identified as a major factor associated with depression, affecting multiple aspects of QOL

The significant prevalence of depression among cancer patients (35.8%) contrasted with the low rate of reported prior mental illness (1%), which highlights the need for routine screening for depressive symptoms. Increasing mental health awareness can facilitate early identification and management, ultimately enhancing patient care. Integrating mental health services, such as counselling and psychiatric support, into cancer care is essential for improving treatment outcomes.

The link between pain and depression highlights the importance of comprehensive pain management strategies, as prioritising effective pain relief interventions can improve QOL and possibly improve depressive symptoms.

This study paves the way for further research that needs to be done in depression among cancer patients. It is crucial to adopt a longitudinal design to examine the causal relationships between pain, depression and QOL over time, which will help in developing effective and efficient interventions. Furthermore, beyond establishing causality, it is essential to evaluate the effectiveness of both existing and future interventions to improve the management of patients facing both depression and cancer.

There was only one validated tool used in this study (PHQ-9).¹⁷ Although the WHO-QOL-BREF has been used in similar settings, little is known about the psychometric properties (validity and reliability) and the cultural sensitivity of the domains included in the instruments.²⁷ The stage of cancer was not assessed because majority of participants were unaware of their staging, and this was a self-administered questionnaire, which may impact physical abilities, mental health and QOL. The in-patients had a slower turnover as many of the stable patients travelled from distant areas around the country in order to receive treatment and thus had long stays in the hospital. As a result, it was difficult to get a good in-patient representation in this study because of low turnover. Causality between pain, depression and QOL could not be determined because of the cross-sectional design of the study.¹⁷

This study has notable strengths, including being the first to quantify depression and QOL specifically in cancer patients in Botswana, with the potential public health benefit of making the case for incorporating mental health screening, psychosocial support and pain control into Botswana’s oncology care. Another strength is its use of a locally validated measure of depression, the PHQ-9, thus strengthening the validity of the findings.