Validation of screening tools for common mental health disorders in the methadone maintenance population in Hanoi, Vietnam

Common mental health disorders (CMDs) – including depression, anxiety and post-traumatic stress disorder (PTSD) – are leading contributors to disability adjusted life years (DALYs) globally [1, 2] and are highlight prevalent among patients with opioid use disorder and those on methadone maintenance therapy (MMT) [3, 4]. CMDs in low and middle income countries (LMICs) such as Vietnam are less frequently identified and treated than in high-income countries (HICs) despite an increased share in burden of disease [2]. Further, providers often only identify severe mental disorders when confronted with visible psychotic or manic symptoms; less severe disorders are often overlooked, as CMDs such as depression may not be considered a disorder but rather a state of simply “thinking too much” [5]. Lack of funding, stigma surrounding mental health, mental healthcare accessibility, limited psychiatric human resources and infrastructure, and few validated screening tools have hindered the study of CMDs in LMICs such as Vietnam [6‐8].

Only a few CMD screening tools for depression and anxiety have been validated in Vietnam. These tools include the DASS-21 (Depression, Anxiety and Stress Scale), Zung SAS (Zung Self-Rating Anxiety Scale), GHQ-12 (General Health Questionnaire), SRQ-20 (Self-Reporting Questionnaire) and EPDS (Edinburgh Postnatal Depression Scale), which were validated against the Structured Clinical Interview for DSM-5 (SCID) or the Composite International Diagnostic Interview (CIDI) [9‐11]. Additionally, another study examined the psychometric properties of the DASS-21 among adolescents [12]. These studies did not always include a reference gold standard and were restricted to specific sub-populations, such as women, perinatal women and their male partners, and adolescents in the northern regions [9‐12]. While a wide range of CMD screening tools have been designed, developed and tested in HICs, many, particularly for PTSD, have not been locally validated in LMICs such as Vietnam [5]. Local validation ensures the accuracy of the screening tool and can facilitate improved detection and diagnosis of mental health disorders. This confirmation is especially important for populations at high risk of developing CMDs, such as patients with opioid use disorder or on MMT.

In Vietnam, over 220,000 people are estimated to inject drugs, and an estimated 51,000 have enrolled in MMT, often at clinics in their communities [13, 14]. These patients with opioid use disorder and those on MMT have a high risk of having an undetected CMD [15]. CMDs within the MMT population are up to 10 times more prevalent than in the general population [15], and still two to three times higher than in substance users not on methadone [16]. At MMT enrollment, patients are not routinely screened for CMDs. As such, an adequate record detailing CMD prevalence while in treatment thereafter are virtually nonexistent. Further, very few studies have estimated the prevalence of CMDs among populations with opioid use disorder or on MMT in Vietnam, and none have estimated the prevalence of PTSD [8, 17‐19]. CMDs can hamper MMT compliance [20] which can exacerbate challenges to continued engagement in MMT. This challenge suggests a need for screening and treatment of CMDs within MMT populations. However, to our knowledge, no studies have focused on validating screening tools for CMDs in an urban environment among the MMT population, almost 25% of whom are living with HIV (PLWH) and at an additional high risk for CMDs.

In order to address this research gap, we aimed to validate screening tools for depression, anxiety and PTSD in a MMT population at an urban clinic in Hanoi to ensure the tools’ accuracy when compared to a reference gold standard.

Our study aimed to validate tools to detect CMDs using the PHQ-9, GAD-7 and PC-PTSD-5 among the MMT population at an urban clinic in Hanoi. The screening tools validated in this study were developed in HICs and therefore required validation locally to determine appropriate cut-off scores, sensitivity, and specificity for the Vietnamese MMT population. The PHQ-9, GAD-7, and PC-PTSD-5 were selected to allow for individual screening of CMDs with separate tools and to our knowledge had not yet been validated in this population. We previously highlighted the importance of implementing well-designed studies that validate screening tools needed to detect CMDs [33]. Our validation of the PHQ-9, GAD-7 and PC-PTSD-5 are the first of their kind in the MMT population in Vietnam to our knowledge [34].

Using the MINI, we found that 11% of the participants had one or more CMDs while the prevalence of depression, anxiety and PTSD were 10.5, 4 and 2%, respectively. The prevalence of depression has been documented as high as 50% among methadone patients in HICs [15] with rates remaining consistently high within other countries in proximity to Vietnam such as China [35] (38%) and Malaysia [36] (44%) with the use of various screening tools. The literature suggests that overall CMD prevalence within Vietnam is lower than surrounding South East Asian countries, both within the general and MMT population. A study of MMT patients in northern Vietnam estimated a prevalence rate of 26.8% for any mental health pathology using the Kessler psychological distress scale [18]. Another study conducted among MMT patients in Vietnam using the DASS-21 estimated a 3.9% prevalence of mild to extremely severe depression and a 18% prevalence of mild to extremely severe anxiety [8]. Regional differences between CMD prevalence within Vietnam further complicate true rates. The majority of CMD studies in the MMT population have largely been limited to the northern rural provinces, documenting increased prevalence of anxiety or depression in rural (43.1%) as compared to urban areas (18.1%) [18, 19]. Prevalence of specific disorders remain lower in urban areas as seen for depression (12.2%) as identified using the MINI in Hai Phong for people who inject drugs [17]. Our findings remain higher than Vietnamese population estimates by the Ministry of Health for depression (2.8%) and anxiety (2.6%) [8].

A number of possible explanations for lower prevalence of CMDs in the MMT population exist. First, this rate may be due to a gender-biased sample, as 98% of participants were male. Globally, CMDs are more prevalent in females than males, especially anxiety disorders [2]. It is difficult to discern whether there are significantly more women who inject drugs who do not present for care as our male-predominant sample is consistent with similar studies of Vietnamese MMT patients [18]. Second, at MMT enrollment, patients are asked two questions to screen for severe mental health disorders with the potential for referral for psychiatric treatment. Among the study sample, up to 25 patients were referred for psychiatric services upon MMT initiation, which could have artificially decreased the prevalence of CMDs in our study. Third, the screening tools were developed using Western manifestations of CMDs according to the DSM-5 criteria; these criteria may not be as sensitive in detecting CMDs in the MMT population in Hanoi due to low prevalence rates and low cut-off scores for all three disorders and screening tools, respectively. CMDs may manifest with more somatic symptoms (such as headaches and neck pain, back pain, fatigue, and palpitations) among the Vietnamese population, as has been seen previously and in other LMICs [37]. Screening tools targeting somatic symptoms should be used in future screening tools to assess whether the prevalence of CMDs increases. Finally, cultural, household and stigma factors remain a leading explanation as to variable rates of CMD between MMT patients living in Vietnam [38].

The cut-off scores for two of the three screening tools tested were lower compared to respective cut-off scores in HICs. The PHQ-9 had an optimal cut-off score of ≥5 with a high sensitivity (95%) and high specificity (92%) compared to its validated HIC cut-off of 8–11, tiered according to severity [37]. The optimal cut-off score for the GAD-7 was ≥3 with a sensitivity of 94% and a specificity of 88%, one point below the HIC cut-off score for mild anxiety [25]. The optimal cut-off score for the PC-PTSD-5 Checklist was ≥4 with a sensitivity of 62.5% and specificity of 95.2%, which is the recommended cut-off score for further evaluation in HICs [23]. These findings have implications for clinical practice as lower cut-offs would prompt consideration of a confirming clinical assessment with fewer endorsed symptoms, meaning that clinicians needed to be more sensitive to the presence of a CMD than standard screen thresholds would suggest. Also in clinical practice, the PPV and NPV of these tools will depend on the prevalence of the CMDs. This population had a low prevalence of depression (10.5%), anxiety (4.0%) and PTSD (2%), resulting in NPVs near 100%, but lower PPVs. (Figs. 2, 4 and 6). In populations with higher CMD prevalence [35, 36], as previously documented in other studies discussed above, the NPVs will still be very high and the PPV will increase markedly. To our knowledge, there have been no validation studies of PTSD in the Vietnamese population.

Validation studies are critical for understanding how to apply a screening tool among different populations. However, the majority of CMD studies within Vietnam have focused on prevalence instead of validation [34]. Of the six studies validating mental health screening tools within Vietnam through 2014, none validated the PHQ-9, GAD-7 or any form of PTSD screening [34]. The SRQ-20 was validated in a district and community sample in rural Vietnam and reported lower cut-off scores than the WHO (World Health Organization) has previously recommended [39, 40]. Validations of the Zung SAS, EPDS and GHQ-12 among perinatal women in Vietnam similarly required lower cut-off scores [9]. Such findings alongside our own, highlight the utility of validation studies and suggest that screening tools for CMDs may not be accurately interpreted without prior validation.

Our study was limited to a single urban methadone clinic in Hanoi. We used translations of the PHQ-9 and GAD-7 from a previous study in a similar Vietnamese healthcare population [27]. The multi-step approach including independent comparison to a previously validated version; cultural applicability and comprehensibility; translation by an experienced, bilingual clinical staff person; and practice interviews is similar to methods from existing publications and our findings are consistent with those in other LMICs, minimizing lack of validation as a weakness [41‐45]. While the cut-off score supported by our data is lower than standard cut-off scores in HICs, it is consistent with approaches used in other LMIC countries in non-psychiatric settings to indicate mild depression and other work indicating a lower threshold for major depressive disorder in Southeast Asian populations [41‐46]. Our study included a large sample size with a high enrollment percentage (95% of clinic sampled) and blinding of the interviewers. We had a multidisciplinary team including both Vietnamese and American physicians, public health epidemiologists, lay health care workers and bilingual staff members to maximize adequate design of the study.

Our findings have several implications for future research. First, these findings need to be replicated to confirm their accuracy. Future validation studies could consider a mixed-methods approach to additionally investigate content validity of the measures. Further local adaptation and validation research could pilot screening tools that include more somatic symptoms or create novel screening tools that more purposefully capture local idioms and concepts of psychiatric distress. Samples should include more gender diversity when possible and include patients from multiple urban methadone clinics. Finally, in addition to establishing the accuracy of screening tools compared to a reference gold standard, implementation science research is needed to investigate the feasibility of integrating routine screening for CMDs, particularly in LMICs such as Vietnam, in order to close the research-to-practice gap in mental health care.

The prevalence of one or more CMD among the MMT patient population was 11%. Prevalence of depression, anxiety and PTSD were 10.5, 4 and 2%, respectively. Prevalence of CMDs remained higher than the general Vietnamese population but may be lower than other reported prevalence of CMDs within the MMT population due to patients with severe mental illness having been referred for treatment prior to study engagement and other cultural factors. Optimal cut-off scores for the PHQ-9 and GAD-7 were lower than respective cut-off scores in HICs while the optimal cut-off score for PTSD remained the same.