https://orcid.org/0000-0002-6416-5395
Figures
Abstract
Depression among adolescents is a global concern, including in South Africa, emphasizing the need for reliable screening tools. While the the Patient Health Questionnaire Adolescent (PHQ-A) and Center for Epidemiologic Studies - Depression Scale for Children (CES-DC) are commonly used mental health screening tools among adolescents, their psychometric properties in South Africa are not well-studied. This study aims to fill this gap by examining the psychometric properties of the PHQ-A and CES-DC among adolescents aged 10–14-year-olds in South Africa. First, a sample of 42 adolescents aged 10–14 years were recruited to participate in cognitive testing focus groups to determine their comprehension of the depression screening tool items. Secondly, a sample of 500 adolescents from 10 schools completed a tablet-based survey including socio-demographic information and the adapted depression screening tools. A subset of 123 adolescents underwent clinical diagnostic interview with the MINI-KID depression modules only. After two weeks, a subset of 145 adolescents had the initial tablet-based survey re-administered for test-retest reliability. Criterion validity was examined using ROC analysis, with the MINI-KID as the reference standard. Of the 500 adolescents surveyed, the majority were female (59.2%), with an average age of 12 years (SD = 1.42). The PHQ-A demonstrated good internal consistency (alpha = 0.82) and moderate test-retest reliability (ICC = 0.56). The CES-DC showed good internal consistency (alpha = 0.86) and moderate test-retest reliability (ICC = 0.68). The PHQ-A displayed good sensitivity (85.7%) and specificity (61.8%). Youden’s index indicated 7.5 as the ideal cutoff score for the PHQ-A. The CES-DC has a sensitivity of 95.2% and specificity of 56.9% and Youden’s index of 21. Factor structure for the PHQ-A included one factor loading for 9 items and 3 factor loading for 8 items; while for the CES-DC factor structure consisted of o three factor loadings. The findings show good psychometric properties of the PHQ-A and CES-DC and contributes to the limited research on mental health assessment tools for younger adolescents in Africa. Testing the psychometric properties of the PHQ-A and CES-DC for this population could help improve early detection and intervention for mental health conditions.
Citation: Mkhize M, van der Westhuizen C, Sorsdahl K (2025) Investigating the psychometric properties of the Patient Health Questionnaire for Adolescents (PHQ-A) and Center for Epidemiologic Studies - Depression Scale for Children (CES-DC) among young adolescents in South Africa. PLoS One 20(11): e0334658. https://doi.org/10.1371/journal.pone.0334658
Editor: Dickens Akena, Makerere University CHS: Makerere University College of Health Sciences, UGANDA
Received: December 23, 2024; Accepted: September 30, 2025; Published: November 17, 2025
Copyright: © 2025 Mkhize et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Data is managed in accordance with the University of Cape Town’s Intellectual Property and Research Data Management Policies. All databases are stored on the University of Cape Town’s firewall-protected server. The dataset has been de-identified to protect participant privacy; however, due to data sharing restrictions imposed by the University of Cape Town’s Human Research Ethics Committee (HREC), all data sharing is subject to the institution’s approved open access platform for research data. The dataset supporting the findings of this study is stored in the Alan J. Flisher Center for Public Mental Health (CPMH registry) on ZivaHub and may be accessed in accordance with Human Research Ethics Committee guidelines. The dataset supporting the findings of this study is available at ZivaHub (University of Cape Town) (email: Zivahub.uct.ac.za) under the DOI: https://doi.org/10.25375/uct.30529310.
Funding: The Sue Struengmann Initiative (SSI) provided funding for this work.
Competing interests: The authors have declared that no competing interests exist.
1. Introduction
The prevalence of mental health conditions among adolescents is high worldwide, particularly in Africa. Between 10% to 20% of adolescents experience a mental health condition [1,2], contributing 5% of disability-adjusted life years (DALYs) and represent a quarter of all years lived with disability (YLDs) among adolescents aged 10–19 [3]. Despite the limited prevalence research in Africa, reported prevalence of depression and anxiety ranges from 5% to 50% [4–9], but most studies focus on high-risk or older adolescents aged 15 and higher [6,7,10], with studies focusing on younger adolescents underrepresented Data from phase 1 of this study, a recent cross-sectional study of 621 adolescents aged 10–14 years in school settings in South Africa found that 33.5% reported experiencing symptoms of depression [11]. While these studies offer valuable insight into symptoms of depression and anxiety among this population, a major limitation is the use of tools that have not been assessed for reliability and validity within the context.
Although the gold standard for identifying adolescents with a mental health condition are structured clinical interviews administered by a trained mental health specialist, resource constraints and a shortage of mental health professionals pose a challenge in low and middle income countries (LMICs) [2,12,13]. Therefore, there is a need for validated, reliable, and culturally-appropriate assessment tools to screen mental health symptoms for children and adolescents [14]. According to a systematic review, the psychometric properties of only 11% of the existing screening tools in LMICs have been tested for use among children and adolescents [15]. Most of the available studies validating screening tools for depression were conducted in high-income countries (HICs) [16,17], with the most utilized screening tool being the Patient Health Questionnaire Adolescent (PHQ-A) [18] and the Center for Epidemiological Studies Depression Scale for children (CES-DC) [19,20]. The PHQ-A is a self-report tool adapted from the PHQ-9 and used to identify depression in adolescents aged 10–18 years [18,21]. However, only three studies all targeting older adolescents have been conducted in Africa [10,22–26]. A recent study examined the isiXhosa version of the PHQ-9 tool among 302 adolescents aged 10–19 in schools [24]. The study revealed an acceptable criterion validity with 91% sensitivity and 76% specificity in identifying adolescents with depression, based on a cut-off score of 10 or higher indicating depression [24]. Apart from this study, there are no others that have examined the psychometric properties of the PHQ-A among adolescents aged 10–14 years.
The Center for Epidemiologic Studies-Depression Scale (CES-D) has been modified for children and adolescents as the CES-DC [27]. It is a 20-item self-report tool designed for adolescents aged 6–17 [19,20]. While the CES-DC has been primarily tested for its psychometric properties in HICs [20,28], there is one study that includes data from Africa [29], with an internal consistency of α = .89 and sensitivity of 80%. However, there are limited data on the CES-DC’s psychometric properties in LMICs particularly in Africa. To test the criterion validity, a study in Rwanda found that the CES-DC is a good tool for depression screening among adolescents aged 10–17, with good internal consistency (α = .86) and test-retest reliability (r = .85) [30]. In South Africa, limited studies have used the CES-DC measure and reported on its criterion validity and reliability [31,32]. This study conducted in the Free State Province sampled 435 South African adolescents aged 13–17 using the CES-DC tool and found good internal consistency (α = .89) [32]. Although some studies have utilized the CES-DC tool to assess depression in older adolescents, there is limited research on its criterion validity and psychometric properties in the local context for South African adolescents aged 10–14 years.
In summary, despite the growing use of the PHQ-A and the CES-DC, very few studies in Africa have examined the psychometric properties of these measures [33,34] in the younger adolescent population group [35]. Existing self-report screening tools often target older adolescents, while for those under 14 years, researchers rely on parent and teacher reports [36,37]. To fill this gap, this study aimed to evaluate the psychometric properties of the self-report PHQ-A and CES-DC for application among South African adolescents aged 10–14 years.
2. Materials and methods
To examine the psychometric properties of the PHQ-A and the CES-DC, this study was conducted in two phases. During the first phase, we conducted cognitive interviewing focus groups to assess the participants comprehension of the measures. In the second phase, when the measures had been adapted and finalized, a larger survey was conducted to investigate the psychometric properties of the PHQ-A and CES-DC.
2.1. Setting
We collaborated with three non-profit organizations currently working in the Western Cape province, South Africa. For phase one, one school was approached; for phase 2, 10 schools were approached and assisted with recruitment of participants [11]. These schools are based in under resourced areas of the Cape Town metropole and surrounding areas.
2.2. Phase 1 recruitment of participants for cognitive testing
We received approval from the University of Cape Town Human Research Ethics Committee (HREC: 565/2020) and the Western Cape Education Department (20211018–6776). In collaboration with one of our partner schools, we recruited a sample of 42 learners for the cognitive testing phase. Participants were selected based on the following criteria: a) aged 10–14; b) enrolled in grades 4–7; and c) provided parental consent and adolescent assent.
2.3. Study procedure
To conduct the cognitive testing phase using a focus group discussion (FGD) format, learners in grades 4–7 were identified and approached by school staff. Once all learners were identified the researcher explained the study and the cognitive testing process to all the learners. The consent and assent forms were sent to all their guardians. Learners that returned their signed guardian consent and assent forms were recruited into the study at a convenient time at the school. The PHQ-A and CES-DC were administered to the 42 learners with four focus group discussions (FGD) held with learners. Each focus group session was led by two facilitators, including a social worker and a psychologist. Learners were grouped by grade levels [4–7], with 10–12 learners in each group. During the FGD, the participants’ comprehension of and response to questions were elicited guided by the structured discussion guide for cognitive testing [38]. Learners shared their perspective on the PHQ-A and CES-DC using a structured interview guide. After the session, learners received a gift and a snack box as a thank you for their participation.
2.4. Measures
Depression. The Patient Health Questionnaire for Adolescents (PHQ-A) is an adolescent version of the Patient Health Questionnaire – 9 item and is used as a screening measure to detect depression among adolescents aged between 10 and 18 years [18,21]. It is self-administered and consists of the nine criteria used in the DSM-IV diagnosis of depressive disorders [21]. The total score can range from 0 to 27, with a suggested cut-off score of 10 and higher indicating greater severity of depression.
The Center for Epidemiologic Studies – Depression Scale for Children (CES-DC) for children and adolescents is a 20 item self-report questionnaire for children and adolescents between the ages of 6 and 17 [19,20]. The total score can range from 0–60, with a suggested CES-DC cutoff score is 15 as being suggestive of depressive symptoms in children and adolescents [28]. That is, scores over 15 can be indicative of significant levels of depressive symptoms [28].
2.5. Results of cognitive testing
The researcher analyzed and organized the responses from the cognitive testing FDGs. Minor changes were made to the PHQ-A and CES-DC after the FGD cognitive testing based on the participant’s feedback to improve understanding, language, and wording. Three expert researchers were involved in reviewing and providing feedback on the adjustments before the next phase. The adapted measures were then included into the study survey conducted across the schools in collaboration with our partner schools [11]. Table 1 below illustrate the adjustments made to the PHQ-A and CES-DC (see S1 and S2 Tables).
2.5. Phase 2: Recruitment of participants for the study survey
Recruitment for phase 2 followed the same process for phase 1. Ten schools were approached and 621 participants were recruited between the ages of 10 and 14 years and data collected between February 2022 and July 2022 [11]. This study was part of a larger study investigating the prevalence and factors associated with depression and anxiety among young school-going adolescents in the Western Cape Province of South Africa [11]. Each participant was provided with a tablet to complete the survey using the user-friendly REDCap platform. The participants choose their preferred language to complete the survey, which was available in English [11]. The survey itself took approximately 45–60 minutes to administer, and an energizer activity was facilitated by the research team halfway through the allotted time [11]. The survey obtained a high response rate from participants (n = 504) who completed it in English out of 621 participants. For this paper, only 500 of the 504 participants are included in this analysis. Four participants were removed owing to missing data.
A team of trained clinical research assistants (including social worker, psychologist, and registered counsellor) administered the depression modules of the Mini Neuropsychiatric Interview version 7.0 (MINI) for Children and Adolescents (MINI-KID) in English to 123 leaners (approximately 5 per class). These 123 participants were randomly selected based on their consent, not their PHQ-A and CES-DC scores. The researchers offered the participants the option to participate in the clinical interview before, during, or after completing the survey.
To assess the test-retest reliability of the PHQ-A and CES-DC, a sample of 145 learners who had provided consent were selected two weeks after the initial administration of the survey. The test–retest interval of two weeks was selected, consistent with standard psychometric practice to provide substantial time to minimize recall bias and limit the likelihood of clinically significant changes in depressive symptoms among adolescents [25,39]. Participants were asked to complete sociodemographic questions, the PHQ-A and CES-DC measures only. All visits included refreshments, and all participants received a snack box for their participation.
2.6. Phase 2 measures
Alongside the PHQ-A and CES-DC mentioned previously, the survey included sociodemographic questions relating to participant sex, age, home language and level of education.
Clinical Diagnosis. The Mini Neuropsychiatric Interview version 7.0 (MINI) for Children and Adolescents (MINI-KID) [40], was designed as a brief structured diagnostic interview for the major psychiatric disorders in Diagnostic Statistical Manual of mental disorders (DSM-III-R, DSM-IV, and DSM-5) and International Classifications of Disease (ICD-10) psychiatric disorders to assess the 30 most common and clinically relevant disorders [40,41]. The MINI-KID depression module was utilized in this study. Although developed for use in HICs, the MINI-KID has been widely utilized in LMICs as a gold standard for clinical diagnosis of anxiety and depression among children and adolescents [34,42–44] including in Africa [5,30] and South Africa [45–47].
3. Data analysis
The survey received a high response rate from participants who completed it in English. As a result, this analysis concentrates solely on the English responses. The Statistical Package for Social Science (SPSS) (29.0) was used to analyse the data. Frequency distributions and descriptive statistics were calculated for the sociodemographic variables of the sample. Receiver operating characteristic (ROC) analysis was performed to evaluate the criterion validity of the CES-DC and the PHQ-A, using the MINI-KID as the gold standard. The coordinates of this curve were tabulated in SPSS, indicating the sensitivities and specificities at various cut-off points. Thereafter the Youden’s index was calculated and included on the table with the sensitives and specificities.
To evaluate the construct validity of the PHQ-A, the principal component analysis (PCA) with varimax rotation was conducted. The researcher first ensured that the underlying assumptions of PCA were met which included: Evaluating the data for normality, linearity, outliers, and factorability [48]. Normality was assessed by visually inspecting histograms and analyzing skewness and kurtosis. Linearity was assessed through scatterplot examinations of variable pairs to ensure a linear relationship. To assess normality, we conducted the Shapiro-Wilk test on each variable and used Q-Q plots to visualize distributions, confirming approximate normality. No outliers were detected for continuous variables, and box plots were produced. Second, we evaluated the sampling adequacy and factorability of the dataset using the Kaiser-Meyer-Olkin (KMO) measure. For the PHQ-A, the KMO value was.896, exceeding the recommended threshold of.6 [49]. Bartlett’s Test of Sphericity yielded a significant result (χ2 (36) = 1295.47, p < .001), indicating the data’s suitability for factor analysis and the sample size adequacy (n = 504) for PCA. Furthermore, we examined multicollinearity by calculating the correlation matrix and ensuring significant inter-correlations among variables without redundancy. After meeting these assumptions, we proceeded with principal component analysis (PCA) and opted for varimax rotation with Kaiser Normalization to extract factors based on eigenvalues >1 [48,50]. The factors explained over 63% of the variances, and we used a scree plot to retain factors before eigenvalues leveled off. Items with communalities between 0.0–0.3 were considered, and only those with factor loadings less than 0.30 were suppressed. Only items with factor loadings of an absolute value of 0.30 or greater were considered, and those instances where there were cross-loadings with a difference between factor loadings of less than 0.2, were considered [49].
The same steps were taken to evaluate the CES-DC PCA assumptions [48], and all criteria were met, including an adequate sample size (n = 504), absence of outliers (confirmed through a box plot), no missing data, and no multicollinearity. Sampling adequacy and factorability were assessed using the Kaiser-Meyer-Olkin (KMO) measure for the CES-DC, resulting in a value of.929. Bartlett’s Test of Sphericity was significant (χ2 (190) = 3510.70, p < .001), indicating the data’s suitability for factor analysis. Factor extraction parameters included eigenvalues >1, factors explaining more than 57% of variances and using a scree plot analysis to retain factors before eigenvalues leveled off. Items with communalities ranging from 0.0–0.3 were considered. Only items with factor loadings above 0.3 were considered and cross-loadings, where the difference between factor loadings was less than 0.2 were also considered [49].
Cronbach’s coefficients were used to evaluate the internal consistency of the PHQ-A, and CES-DC tools. To assess test-retest reliability, we calculated the intraclass correlation coefficient (ICC) to estimate the stability of PHQ-A and CES-DC scores in the first and second administrations using one-way random effects ICC.
4. Results
4.1. Sociodemographic characteristics
The socio-demographic characteristics of the sample are shown in Table 2. Out of all the participants, 504 (81.2%) decided to complete the survey in English, while 105 (16.9%) completed it in isiXhosa and only 12 (1.9%) in Afrikaans. Of the total of 500 adolescents surveyed, the majority were female (59.2%) with an average age of 12 years (SD = 1.42). More than two-thirds of the participants reported having a dual home language, with 395 (71.8%) speaking English, 214 (42.8%) speaking isiXhosa, 103 (20.6%) speaking Afrikaans, and 58 (11.6%) speaking another language. The participant’s family structure was complex, with participants reporting multiple living arrangements which indicated that more than half 364 (58.6%) live with a single parent, while more than a quarter 162 (26.1%) live with both parents and 95 (15.3%) do not live with their parents.
4.2. Internal consistency and test retest reliability of the PHQ-A and CES-DC
Only the English responses have been included in the analysis for the purposes of this paper. The results show good internal consistency for both measures, with the Cronbach alpha result of the PHQ-A being 0.83, 95% CI [.81,.85], and for the CES-DC 0.88, 95% CI [.87,.90]. as listed in Table 3. The test-retest reliability results for the PHQ-A indicate moderate reliability (ICC = .57, 95% CI.40−.69) [51]. Similarly, the CES-DC findings demonstrate moderate to good reliability (ICC = .65, 95% CI.51 −.74) in line with the PHQ-A results.
4.2.1. ROC curve analysis for PHQ-A.
The ROC curve analysis evaluated the sensitivity and specificity of PHQ-A against the MINI-KID depression module English only. Table 4 displays the properties of the performance of the PHQ-A ROC analysis. Results correspond to the ROC curve shown in Fig 1.
The Table 4 shows that a Youden’s index of 0.475 is highest at a cut-off score of 7.50 (rounded off to 8), suggesting that this score optimally balances sensitivity and specificity for the PHQ-A. Therefore, the recommended cut-off score, based on Youden’s index is 7.5 (rounded off to 8) for this context based on the results on ROC curve analysis as depicted in Fig 1, with a sensitivity of 85.7% and specificity of 61.8%. This suggests that the PHQ-A is fairly sensitive meaning it correctly identifies 86% of true positive case while it has a moderate specificity, correctly identifying 62% of true negative cases. Furthermore, the area under the curve (AUC) value of 0.764 suggests that the ROC curve, based on the PHQ-A, shows good discriminatory ability between the negative and positive cases.
4.2.2. ROC curve analysis for CES-DC.
The ROC curve analysis evaluated the sensitivity and specificity of CES-DC against the MINI-KID depression module. Table 5 displays the properties of the performance of the CES-DC ROC curve analysis. Results correspond to the ROC curve shown in Fig 2.
Table 5 shows that a Youden’s index of 0.521 is highest at a cut-off score of 21.50 (rounded off to 22), suggesting that this score optimally balances sensitivity and specificity for the CES-DC. Therefore, the recommended cut-off score, based on Youden’s index is 21.50 for this context based on the results of the ROC curve analysis as shown in Fig 2, with a sensitivity of 95.2% and specificity of 56.9%. Furthermore, the area under the curve (AUC) value of 0.797 suggests that the ROC curve, based on the CES-DC, shows good discriminatory ability to differentiate between the positive and negative cases.
4.4. Factor analysis: Principal component analysis for the PHQ-A
The principal component analysis (PCA) with varimax rotation was conducted to assess the underlying structure of the PHQ-A examining depressive symptoms among the 500 adolescents.
For this analysis, we conducted two analyses based on findings from previous research: (i) a PCA analysis including all the adapted PHQ-A 9 items and (ii) a second analysis with the adapted PHQ-A excluding the ninth question, which relates to suicide [52]. The findings from Table 6 and Fig 3 showed the data met the assumptions of sample adequacy (KMO = .89), the absence of multicollinearity, and significance of Bartlett’s test of sphericity. Two factors were extracted from the analysis – ‘depression’ and ‘somatic’. The first component showed high loadings on items relating to low mood, anhedonia, and negative thinking, while the second component showed higher loadings on items reflecting somatic symptoms such as sleep, appetite, and fatigue. The results from Table 7 and Fig 3 PHQ-A 8 items, data met the assumptions of sample adequacy (KMO = .87), the absence of multicollinearity, and significance of Bartlett’s test of sphericity. Only one principal component factor was extracted from the analysis with the removal of item 9. The scree plot confirms the extraction of the one component. A unidimensional principal component factor was indicated with all the principal component factors loaded into one factor.
4.6. Factor analysis: Principal component analysis for the CES-DC
The principal component analysis (PCA) with varimax rotation was conducted to assess the underlying structure of the adapted 20 items of the CES-DC examining depressive symptoms among the 500 adolescents.
Table 8 and Fig 4 of the analysis showed the data met the assumptions of sample adequacy (OKM = .92), the absence of multicollinearity, and significance of Bartlett’s test of sphericity. Factor extraction established three components with eigenvalues greater than one accounted for 28.0% of the variation in depressive symptoms among adolescents. Three principal component factors were extracted from which relate to somatic symptoms, depressed mood and positive affect.
5. Discussion
This study is one of the first to investigate the psychometric properties of the PHQ-A and CES-DC, which have been culturally adapted for adolescents aged 10–14 in school settings in the Western Cape Province of South Africa. Three key findings emerge from this research.
First, the study findings indicate that both the PHQ-A and CES-DC demonstrate good internal consistency and moderate test-retest reliability, suggesting that the items are highly correlated and measure the intended construct reliably among young adolescents in South Africa. Consistent with previous studies conducted in Africa, the South African adapted PHQ-A exhibited acceptable internal consistency [24,26,30,53,54] with a Cronbach’s alpha = 0.83. Notably, compared to a recent study conducted by Lovero and colleagues in 2022 involving 485 adolescents aged 12–19 years in Mozambique, our research identified slightly higher internal reliability for the English PHQ-A (Cronbach’s alpha = 0.83), in contrast to the Mozambican Portuguese language (Cronbach’s alpha = 0.80) [26].
While limited research has explored the internal consistency of the CES-DC in Africa [30,55]. By comparison, the study conducted by Betancourt and colleagues’ in Rwanda using Kinyarwanda language yielded results similar to ours [30]. Their study revealed an internal consistency coefficient of 0.86 for the CES-DC scale [30]. In contrast, although our study demonstrated higher internal consistency (Cronbach’s alpha = 0.88). The findings emphasize good internal consistency across the studies conducted among young African adolescents, with our study reporting slightly higher internal consistency.
Test-retest reliability is another aspect of reliability, as it assesses the consistency of scores over time, reflecting the stability of measurements. The moderate test-retest reliability coefficients obtained for the PHQ-A (ICC = 0.57) and the CES-DC (ICC = 0.65) suggest that these screening tools yield relatively consistent results when administered at different time points. Previous studies have also shown stability in scores for the PHQ-A and CES-DC over time [21,27,56]. By contrast, our study findings differ from those of Lovero’s study in Mozambique where they reported a good test-retest reliability coefficient of 0.70 for the PHQ-A [26], our research identified poor to moderate test-retest reliability for the PHQ-A (ICC = 0.57). Similarly, for the CES-DC, our study revealed moderate test-retest reliability (ICC = 0.65) in comparison to the good Pearson coefficient of 0.85 for test-retest reliability among Rwandan adolescents [30].
These findings underscore the impact of cultural influences as a potential source of variation in our findings. This is reflected in the high internal consistency of the depression screening tools and differences in test-retest reliability, emphasizing the need to consider both cultural and linguistic factors in research and mental health screening and assessment [26,57,58].
Although our study reported on the English PHQ-A and CES-DC responses, it is crucial to address the translation process of depression screening tools from English to African languages. After conducting cognitive interviews, several items were reworded to better reflect the psychological construct and improve linguistic and cultural clarity. This careful adaptation is a crucial step in cross-cultural research, as making a screening tool more comprehensible increases its validity when used in diverse contexts. To address this, it is important to consider a rigorous translation methodology and conduct cognitive testing interviews to check consistency of interpretation of the translation with the original language. The translation process may influence internal consistency, potentially leading to variations in tool interpretation and responses. This aligns with validation studies conducted in Kenya and Nigeria, emphasizing the significance of multiple language considerations [59,60].
The second finding of our study is the ROC curve analysis, which assessed the sensitivity and specificity of the PHQ-A and CES-DC against the MINI-KID depression module among adolescents in South Africa. Regarding criterion validity, our findings for the PHQ-A revealed a recommended cut-off score of 7.50 (rounded off to 8), based on the highest Youden’s index of 0.475, which balances sensitivity and specificity for this context. The sensitivity of 85.7% and specificity of 61.8% indicate that the PHQ-A is fairly sensitive in correctly identifying true positive cases, while also demonstrating moderate specificity in correctly identifying true negative cases. The Area Under the Curve (AUC) value of 0.764 suggests that the ROC curve for the PHQ-A shows good discriminatory ability in distinguishing between negative and positive cases of depressive symptoms among adolescents in South Africa.
Our findings are consistent with prior research. For example, Lovero and colleagues’ study also reported good sensitivity and specificity values (>0.70), and the Youden’s index identified an optimal cutoff score of 8 for the PHQ-A [26]. In contrast, our results differ from other studies conducted in South Africa [24], Kenya [54] and Nigeria [60], where good to high sensitivity and specificity values (>0.70–0.90) were observed. All three studies identified an optimal cutoff score of >10 or higher [24,54] compared to our findings.
The ROC curve analysis of the CES-DC identified a recommended cut-off score of 21.50 (rounded off to 22), based on the highest Youden’s index of 0.521, which optimally balances sensitivity and specificity for this screening tool. The sensitivity of 95.2% and specificity of 56.9% indicate that the CES-DC is highly sensitive in correctly identifying true positive cases, while showing moderate specificity in correctly identifying true negative cases. The AUC value of 0.797 suggests that the ROC curve for the CES-DC demonstrates good discriminatory ability in distinguishing between positive and negative cases of depressive symptoms among adolescents in South Africa. In contrast to our findings, a previous study assessed the CES-DC’s psychometric properties in Rwandan adolescents aged 10–19 using the MINI-KID clinical interview reported indicated that a CES-DC score of ≥ 30 yielded a sensitivity of 0.82 and a specificity of 0.72 in detecting depression, with an AUC of 0.83 [30]. The optimal score was found to be > 30, which is higher than the > 21 reported in our study, along with different sensitivity and specificity values compared to our findings.
The sensitivity, specificity, and discriminatory ability of these screening tools ensure their applicability for early detection and intervention of mental health conditions in this population. Early detection enables timely management of depressive symptoms among adolescents in clinical and school settings. The high sensitivity rate observed in our study plays a crucial role in achieving this objective by accurately identifying true positives and reducing the risk of false negatives [58]. It is important to note the potential for over identification of false positives due to moderate specificity in our results. Nonetheless, the specificity values were moderate (i.e., 56–62%) with context-dependent considerations. The clinical context may lead to lower specificity generating more false positives and adding burden to systems with limited resources; this might be navigated by means of a two-stage pathway (brief screening followed by a clinician interview or a second validated measure administered by trained clinicians or non-specialists) and context-adjusted cut-offs [58,61]. In the context of screening in schools and communities, where the objective is to avoid missing at-risk adolescents, lower specificity is a reasonable trade-off as long as positive screens are followed by organized secondary assessment and clear referral options [58,61]. However, in our context where early identification is paramount, this can be managed through strategies such as additional follow-up assessments to confirm diagnoses or considering the context of use, whether for initial mental health screening or comprehensive diagnostic evaluations [58,61].
The third finding of this study from the PCA analysis provide valuable insights into the factor structure of the PHQ-A and CES-DC screening tools. The PCA analysis of the PHQ-A revealed two distinct factors: depression and somatic symptoms. This finding is consistent with previous research that has identified similar factor structures in adolescent populations [25,62]. The extraction of these two factors suggests that the PHQ-A captures both emotional and physical symptoms of depression among adolescents in South Africa. The removal of the ninth item related to suicide from the analysis resulted in a unidimensional factor, indicating that this item may not align with the underlying structure of depressive symptoms in this population. However, these findings are consistent with previous studies indicating that item 9, assessing self-harm and suicidal thoughts, does not effectively differentiate between individuals likely to be depressed and those who are not, and also exhibits a lower factor loading [21,63]. Contrary to previous studies using all 9 items, suggesting a single factor loading for the PHQ-A, our study revealed two factor loadings for the PHQ-A 9 items [21,64]. Nevertheless, item 9 of the PHQ-A which measures self-reported self-harm or suicidal ideation, may be influenced by cultural norms and stigma, leading adolescents to underreport due to fear, shame or stigma [65]. Meanwhile, it is also observed that by contrast some adolescents tend to misunderstand what constitutes suicidal ideation, leading to a potentiality of false positive and burdening scarce clinical resources [66–68]. Therefore, these findings underscore the need for culturally sensitive administration by trained clinicians or non-specialists, in-depth understanding of each item meaning and structured clinical interviews to validate positive responses.
In contrast, the PCA analysis of the CES-DC identified three components: somatic symptoms, depressed mood, and positive affect, showcasing the multidimensional nature of depressive symptoms captured by the screening tool. This contrasts with prior research that reported a four-factor loading for the CES-DC to include: somatic complaints, depressed affect, interpersonal problems and positive affect [27,55,69]. By contrast, our findings reported three factor loadings which include somatic symptoms, depressed mood, and positive affect indicate that the CES-DC offers a comprehensive assessment of different symptoms of depression among adolescents. These differences underscore the importance of conducting PCA analyses in specific cultural contexts to understand the unique factor structure of screening tools for mental health assessment.
The findings of this study underscore important implications for practice, particularly suggesting that the PHQ-A could serve as a screening tool in schools and possibly in primary care settings. Its brevity and alignment with DSM criteria for depression symptoms make it suitable for use by trained clinicians,(e.g., lay counsellors or social workers). In school or community settings, a trained lay counsellor can employ this tool, provided there are adequate referral pathways for adolescents identified as at-risk. However, these implications should be interpreted in light of several limitations. First, only the English version of the instrument underwent cognitive testing and validation, introducing possible language bias of participants who preferred their home language (e.g., isiXhosa or Afrikaans). Second, the sample was drawn from the Western Cape, the findings may not be generalisable to school-going adolescents in other South African provinces other provinces in South Africa and other LMICs contexts. Third, while interviewers were blinded to PHQ-A scores, potential biases inherent to interviewer-administered diagnostics (e.g., interviewer effects) cannot be fully excluded, which may have introduced bias in the diagnostic interview process.
6. Conclusion
Despite the limitations, our findings provide insight on the reliability, validity, and psychometric properties of the PHQ-A and CES-DC screening tools for detecting depressive symptoms among adolescents aged 10–14 years in South Africa. These findings add to the sparse literature on mental health assessment among adolescents, emphasizing the significance of culturally relevant and reliable screening tools for this population. Furthermore, they assist in promoting early detection and intervention for adolescent mental health conditions. Further research is needed to validate these findings in larger and more diverse populations. By improving our understanding of the psychometric properties of these screening tools, we can better support the mental health of adolescents in South Africa and other LMICs contexts. Finally, these tools also hold potential for integration into school health programs and broader mental health systems, where early identification and intervention can be embedded within existing structures to promote scalable and sustainable impact.
Supporting information
S2 Table. Original and modified CES-DC items.
https://doi.org/10.1371/journal.pone.0334658.s002
(DOCX)
Acknowledgments
We would like to acknowledge and thank our community partner organizations, Community Keepers, Student for a Better Future, and Empilweni, as well as the respective schools and learners that participated in this study, for their contribution. We would also like to thank the Center for Public Mental Health (CPMH) staff for their assistance with collecting data for this study. The opinions expressed in this article are solely those of the authors, and do not represent the views, decisions, or policies of the institutions with which they are affiliated.
References
- 1. Silva SA, Silva SU, Ronca DB, Gonçalves VSS, Dutra ES, Carvalho KMB. Common mental disorders prevalence in adolescents: A systematic review and meta-analyses. PLoS One. 2020;15(4):e0232007. pmid:32324835
- 2. Patel V, Flisher AJ, Hetrick S, McGorry P. Mental health of young people: a global public-health challenge. Lancet. 2007;369(9569):1302–13. pmid:17434406
- 3. Erskine HE, Moffitt TE, Copeland WE, Costello EJ, Ferrari AJ, Patton G, et al. A heavy burden on young minds: the global burden of mental and substance use disorders in children and youth. Psychol Med. 2015;45(7):1551–63. pmid:25534496
- 4. Ndetei DM, Khasakhala L, Nyabola L, Ongecha-Owuor F, Seedat S, Mutiso V, et al. The prevalence of anxiety and depression symptoms and syndromes in Kenyan children and adolescents. J Child Adolesc Ment Health. 2008;20(1):33–51. pmid:25861005
- 5. Kinyanda E, Kizza R, Abbo C, Ndyanabangi S, Levin J. Prevalence and risk factors of depression in childhood and adolescence as seen in four districts of North-Eastern Uganda. BMC Int Health Hum Rights. 2013;13:19. pmid:23561039
- 6. Nkuba M, Hermenau K, Goessmann K, Hecker T. Mental health problems and their association to violence and maltreatment in a nationally representative sample of Tanzanian secondary school students. Soc Psychiatry Psychiatr Epidemiol. 2018;53(7):699–707. pmid:29651620
- 7. Ward-Smith C, Sorsdahl K, van der Westhuizen C. An investigation into symptoms of depression and anxiety and emotion regulation among older adolescents from low-income settings in South Africa. Compr Psychiatry. 2024;132:152476.
- 8. Osborn TL, Venturo-Conerly KE, Wasil AR, Schleider JL, Weisz JR. Depression and Anxiety Symptoms, Social Support, and Demographic Factors Among Kenyan High School Students. J Child Fam Stud. 2019;29(5):1432–43.
- 9. Jörns-Presentati A, Napp A-K, Dessauvagie AS, Stein DJ, Jonker D, Breet E, et al. The prevalence of mental health problems in sub-Saharan adolescents: A systematic review. PLoS One. 2021;16(5):e0251689. pmid:33989357
- 10.
Osborn TL, Venturo-Conerly KE, Wasil A, Schleider JL, Weisz J. 2019.
- 11. Mkhize M, van der Westhuizen C, Sorsdahl K. Prevalence and factors associated with depression and anxiety among young school-going adolescents in the Western Cape Province of South Africa. Compr Psychiatry. 2024;131:152469. pmid:38461564
- 12.
WHO. Mental Health Atlas. 2014.
- 13. Kieling C, Baker-Henningham H, Belfer M, Conti G, Ertem I, Omigbodun O, et al. Child and adolescent mental health worldwide: evidence for action. Lancet. 2011;378(9801):1515–25. pmid:22008427
- 14. Kwan B, Rickwood DJ. A systematic review of mental health outcome measures for young people aged 12 to 25 years. BMC Psychiatry. 2015;15:279. pmid:26573269
- 15. Ali G-C, Ryan G, De Silva MJ. Validated Screening Tools for Common Mental Disorders in Low and Middle Income Countries: A Systematic Review. PLoS One. 2016;11(6):e0156939. pmid:27310297
- 16. Beck A, Dryburgh N, Bennett A, Shaver N, Esmaeilisaraji L, Skidmore B, et al. Screening for depression in children and adolescents in primary care or non-mental health settings: a systematic review update. Syst Rev. 2024;13(1):48. pmid:38291528
- 17. Habtamu K, Birhane R, Demissie M, Fekadu A. Interventions to improve the detection of depression in primary healthcare: systematic review. Syst Rev. 2023;12(1):25. pmid:36829262
- 18. Johnson JG, Harris ES, Spitzer RL, Williams JBW. The patient health questionnaire for adolescents: validation of an instrument for the assessment of mental disorders among adolescent primary care patients. J Adolesc Health. 2002;30(3):196–204. pmid:11869927
- 19. Fendrich M, Weissman MM, Warner V. Screening for depressive disorder in children and adolescents: validating the Center for Epidemiologic Studies Depression Scale for Children. Am J Epidemiol. 1990;131(3):538–51. pmid:2301363
- 20. Roberts RE, Andrews JA, Lewinsohn PM, Hops H. Assessment of depression in adolescents using the Center for Epidemiologic Studies Depression Scale. Psychol Assess. 1990;2(2):122–8.
- 21. Kroenke K, Spitzer RL, Williams JB. The PHQ-9: validity of a brief depression severity measure. J Gen Intern Med. 2001;16(9):606–13. pmid:11556941
- 22. Sebera F, Vissoci JRN, Umwiringirwa J, Teuwen DE, Boon PE, Dedeken P. Validity, reliability and cut-offs of the Patient Health Questionnaire-9 as a screening tool for depression among patients living with epilepsy in Rwanda. PLoS One. 2020;15(6):e0234095. pmid:32530968
- 23. Materu J, Kuringe E, Nyato D, Galishi A, Mwanamsangu A, Katebalila M, et al. The psychometric properties of PHQ-4 anxiety and depression screening scale among out of school adolescent girls and young women in Tanzania: a cross-sectional study. BMC Psychiatry. 2020;20(1):321. pmid:32560705
- 24. Marlow M, Skeen S, Grieve CM, Carvajal-Velez L, Åhs JW, Kohrt BA, et al. Detecting Depression and Anxiety Among Adolescents in South Africa: Validity of the isiXhosa Patient Health Questionnaire-9 and Generalized Anxiety Disorder-7. J Adolesc Health. 2023;72(1S):S52–60. pmid:36274021
- 25. Makhubela M, Khumalo IP. Psychometric evaluation of the PHQ-9 in university students: Factorial validity and measurement equivalence across three African countries. Curr Psychol. 2022;42(21):18061–9.
- 26. Lovero KL, Adam SE, Bila CE, Canda ED, Fernandes ME, Rodrigues TIB, et al. Validation of brief screening instruments for internalizing and externalizing disorders in Mozambican adolescents. BMC Psychiatry. 2022;22(1):549. pmid:35962378
- 27. Radloff LS. The CES-D scale: A self-report depression scale for research in the general population. Appl Psychol Measur. 1977;1(3):385–401.
- 28. Weissman MM, Orvaschel H, Padian N. Children’s symptom and social functioning self-report scales. Comparison of mothers’ and children’s reports. J Nerv Ment Dis. 1980;168(12):736–40. pmid:7452212
- 29. Rochat TJ, Houle B, Stein A, Pearson RM, Bland RM. Prevalence and risk factors for child mental disorders in a population-based cohort of HIV-exposed and unexposed African children aged 7–11 years. Europ Child Adolesc Psychiatry. 2018;27(12):1607–20.
- 30. Betancourt T, Scorza P, Meyers-Ohki S, Mushashi C, Kayiteshonga Y, Binagwaho A, et al. Validating the Center for Epidemiological Studies Depression Scale for Children in Rwanda. J Am Acad Child Adolesc Psychiatry. 2012;51(12):1284–92. pmid:23200285
- 31. Martin L, Viljoen M, Kidd M, Seedat S. Are childhood trauma exposures predictive of anxiety sensitivity in school attending youth? J Affect Disord. 2014;168:5–12. pmid:25033472
- 32. Thurman TR, Luckett BG, Nice J, Spyrelis A, Taylor TM. Effect of a bereavement support group on female adolescents’ psychological health: a randomised controlled trial in South Africa. Lancet Glob Health. 2017;5(6):e604–14. pmid:28462880
- 33. Hoosen N, Davids EL, de Vries PJ, Shung-King M. The Strengths and Difficulties Questionnaire (SDQ) in Africa: a scoping review of its application and validation. Child Adolesc Psychiatry Ment Health. 2018;12:6. pmid:29344084
- 34. Nezafat Maldonado B, Chandna J, Gladstone M. A systematic review of tools used to screen and assess for externalising behaviour symptoms in low and middle income settings. Glob Ment Health. 2019;6:e13. pmid:31391945
- 35. Kuhn C, Aebi M, Jakobsen H, Banaschewski T, Poustka L, Grimmer Y, et al. Effective Mental Health Screening in Adolescents: Should We Collect Data from Youth, Parents or Both?. Child Psychiatry Hum Dev. 2017;48(3):385–92. pmid:27363421
- 36. Mughal AY, Devadas J, Ardman E, Levis B, Go VF, Gaynes BN. A systematic review of validated screening tools for anxiety disorders and PTSD in low to middle income countries. BMC Psychiatry. 2020;20(1):338. pmid:32605551
- 37. Sharp C, Venta A, Marais L, Skinner D, Lenka M, Serekoane J. First evaluation of a population-based screen to detect emotional-behavior disorders in orphaned children in Sub-Saharan Africa. AIDS Behav. 2014;18(6):1174–85. pmid:24623068
- 38. Collins D. Pretesting survey instruments: an overview of cognitive methods. Qual Life Res. 2003;12(3):229–38. pmid:12769135
- 39. Li X, Jin W, Han L, Chen X, Li L. Comparison and application of depression screening tools for adolescents: scale selection and clinical practice. Child Adolesc Psychiatry Ment Health. 2025;19(1):53. pmid:40346636
- 40. Sheehan DV, Sheehan KH, Shytle RD, Janavs J, Bannon Y, Rogers JE, et al. Reliability and validity of the Mini International Neuropsychiatric Interview for Children and Adolescents (MINI-KID). J Clin Psychiatry. 2010;71(3):313–26. pmid:20331933
- 41. Sheehan DV, Lecrubier Y, Sheehan KH, Amorim P, Janavs J, Weiller E, et al. The Mini-International Neuropsychiatric Interview (M.I.N.I.): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. J Clin Psychiatry. 1998;59:(Suppl 20):22–33;quiz 4-57.
- 42. Paul MA, Khan W. Prevalence of Childhood Mental Disorders Among School Children of Kashmir Valley. Commun Ment Health J. 2019;55(6):1031–7. pmid:29508178
- 43. Yatham S, Sivathasan S, Yoon R, da Silva TL, Ravindran AV. Depression, anxiety, and post-traumatic stress disorder among youth in low and middle income countries: A review of prevalence and treatment interventions. Asian J Psychiatr. 2018;38:78–91. pmid:29117922
- 44. Leffler JM, Riebel J, Hughes HM. A Review of Child and Adolescent Diagnostic Interviews for Clinical Practitioners. Assessment. 2015;22(6):690–703. pmid:25520212
- 45. Stockings E, Degenhardt L, Lee YY, Mihalopoulos C, Liu A, Hobbs M, et al. Symptom screening scales for detecting major depressive disorder in children and adolescents: a systematic review and meta-analysis of reliability, validity and diagnostic utility. J Affect Disord. 2015;174:447–63. pmid:25553406
- 46. Cluver L, Orkin M, Boyes ME, Sherr L. Child and Adolescent Suicide Attempts, Suicidal Behavior, and Adverse Childhood Experiences in South Africa: A Prospective Study. J Adolesc Health. 2015;57(1):52–9.
- 47. Baron EC, Davies T, Lund C. Validation of the 10-item Centre for Epidemiological Studies Depression Scale (CES-D-10) in Zulu, Xhosa and Afrikaans populations in South Africa. BMC Psychiatry. 2017;17(1):6. pmid:28068955
- 48. Hair JF Jr, L.D.S. Gabriel M, da Silva D, Braga Junior S. Development and validation of attitudes measurement scales: fundamental and practical aspects. RAUSP Manag J. 2019;54(4):490–507.
- 49.
Tabachnick BG, Fidell LS, Ullman JB. Using multivariate statistics: pearson. Boston (MA); 2013.
- 50. Kaiser HF. The Varimax Criterion for Analytic Rotation in Factor Analysis. Psychometrika. 1958;23(3):187–200.
- 51. Koo TK, Li MY. A Guideline of Selecting and Reporting Intraclass Correlation Coefficients for Reliability Research. J Chiropr Med. 2016;15(2):155–63. pmid:27330520
- 52. López-Torres S, Pérez-Pedrogo C, Sánchez-Cardona I, Sánchez-Cesáreo M. Psychometric Properties of the PHQ-A among a Sample of Children and Adolescents in Puerto Rico. Curr Psychol. 2019;41(1):90–8.
- 53. Anum A, Adjorlolo S, Kugbey N. Depressive symptomatology in adolescents in Ghana: Examination of psychometric properties of the Patient Health Questionnaire-9. J Affect Disord. 2019;256:213–8. pmid:31181377
- 54. Tele AK, Carvajal-Velez L, Nyongesa V, Ahs JW, Mwaniga S, Kathono J, et al. Validation of the English and Swahili Adaptation of the Patient Health Questionnaire-9 for Use Among Adolescents in Kenya. J Adolesc Health. 2023;72(1S):S61–70. pmid:36376148
- 55. Ayyash-Abdo H, Nohra J, Okawa S, Sasagawa S. Depressive symptoms among adolescents in Lebanon: a confirmatory factor analytic study of the Center for Epidemiological Studies Depression for Children. Acta Psychopathol. 2016;2(6):1–8.
- 56. Betancourt T, Scorza P, Meyers-Ohki S, Mushashi C, Kayiteshonga Y, Binagwaho A, et al. Validating the Center for Epidemiological Studies Depression Scale for Children in Rwanda. J Am Acad Child Adolesc Psychiatry. 2012;51(12):1284–92. pmid:23200285
- 57.
Lund C. Mental health in Africa: findings from the mental health and poverty project. Taylor & Francis; 2010. p. 547–9.
- 58. Kohrt BA, Kaiser BN. Measuring mental health in humanitarian crises: A practitioner’s guide to validity. Conflict Health. 2021;15:1–10.
- 59. Kaiser BN, Ticao C, Anoje C, Boglosa J, Gafaar T, Minto J, et al. Challenges in simultaneous validation of mental health screening tools in multiple languages: Adolescent assessments in Hausa and Pidgin in Nigeria. SSM Ment Health. 2022;2:100168. pmid:36712479
- 60. Nyongesa V, Kathono J, Mwaniga S, Yator O, Madeghe B, Kanana S, et al. Cultural and contextual adaptation of mental health measures in Kenya: An adolescent-centered transcultural adaptation of measures study. PLoS One. 2022;17(12):e0277619. pmid:36520943
- 61. van der Westhuizen C, Carvajal-Velez L, de Carvalho Eriksson C, Gatto J, Kadirova A, Samuels R, et al. Implications for Mental Health Promotion and Prevention Interventions: Findings From Adolescent Focus Group Discussions in Belize, Kazakhstan, and South Africa. J Adolesc Health. 2023;72(1S):S105–11. pmid:36229392
- 62. Granillo MT. Structure and Function of the Patient Health Questionnaire-9 Among Latina and Non-Latina White Female College Students. J Soc Soc Work Res. 2012;3(2):80–93.
- 63. Merz EL, Malcarne VL, Roesch SC, Riley N, Sadler GR. A multigroup confirmatory factor analysis of the Patient Health Questionnaire-9 among English- and Spanish-speaking Latinas. Cultur Divers Ethnic Minor Psychol. 2011;17(3):309–16. pmid:21787063
- 64. Long A, DeFreese JD, Bickett A, Price D. Factorial Validity and Invariance of an Adolescent Depression Symptom Screening Tool. J Athl Train. 2022;57(6):592–8. pmid:34902860
- 65. Hawton K, Saunders KEA, O’Connor RC. Self-harm and suicide in adolescents. Lancet. 2012;379(9834):2373–82. pmid:22726518
- 66. Plöderl M, Kralovec K, Yazdi K, Fartacek R. A closer look at self-reported suicide attempts: false positives and false negatives. Suicide Life Threat Behav. 2011;41(1):1–5. pmid:21309818
- 67. Haghish EF. Differentiating adolescent suicidal and nonsuicidal self-harm with artificial intelligence: Beyond suicidal intent and capability for suicide. J Affect Disord. 2025;378:381–91. pmid:39955075
- 68. Simon GE, Rutter CM, Peterson D, Oliver M, Whiteside U, Operskalski B, et al. Does response on the PHQ-9 Depression Questionnaire predict subsequent suicide attempt or suicide death? Psychiatr Serv. 2013;64(12):1195–202. pmid:24036589
- 69. Essau CA, Olaya B, Pasha G, Gilvarry C, Bray D. Depressive symptoms among children and adolescents in Iran: a confirmatory factor analytic study of the centre for epidemiological studies depression scale for children. Child Psychiatry Hum Dev. 2013;44(1):123–36. pmid:22714666