Objectives To estimate the proportion of adult primary care outpatients who are clinically detected and initiate treatment for depression and alcohol use disorder (AUD) in low-income and middle-income country (LMIC) settings.
Design Five cross-sectional studies.
Setting Adult outpatient services in 36 primary healthcare facilities in Sodo District, Ethiopia (9 facilities); Sehore District, India (3); Chitwan District, Nepal (8); Dr Kenneth Kaunda District, South Africa (3); and Kamuli District, Uganda (13).
Participants Between 760 and 1893 adults were screened in each district. Across five districts, between 4.2% and 20.1% screened positive for depression and between 1.2% and 16.4% screened positive for AUD. 96% of screen-positive participants provided details about their clinical consultations that day.
Primary outcomes Detection of depression, treatment initiation for depression, detection of AUD and treatment initiation for AUD.
Results Among depression screen-positive participants, clinical detection of depression ranged from 0% in India to 11.7% in Nepal. Small proportions of screen-positive participants received treatment (0% in Ethiopia, India and South Africa to 4.2% in Uganda). Among AUD screen-positive participants, clinical detection of AUD ranged from 0% in Ethiopia and India to 7.8% in Nepal. Treatment was 0% in all countries aside Nepal, where it was 2.2%.
Conclusions The findings of this study suggest large detection and treatment gaps for adult primary care patients, which are likely contributors to the population-level mental health treatment gap in LMIC. Primary care facilities remain unfulfilled intervention points for reducing the population-level burden of disease in LMIC.
- primary care
- alcohol use disorder
- clinical detection
- low- and middle-income countries
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Strengths and limitations of this study
This was a multicountry survey of diverse clinical settings, with large sample sizes, structured interviews and used of validated screening tools to identify cases.
The methods used here are demonstrably flexible and are replicable in other low-income settings, particularly for monitoring and evaluation purposes.
We used highly sensitive and non-specific coding criteria for our primary outcomes (ie, detection and treatment initiation), and so outcome misclassification is possible.
Mental, neurological and substance use disorders (MNS) contribute significantly to the Global Burden of Disease and account for 1 in every 10 lost years of health globally.1 In 2010, the absolute disability-adjusted life-years (DALYs) due to MNS disorders was 258 million DALYs, which was 10.4% of the total disease burden. MNS disorders were also the leading cause of years lived with disability globally.1 In addition, MNS disorders act as a significant risk factor for premature death2 and also account for substantial adverse social and economic consequences.1 3 Depression accounts for 40.5% of DALYs caused by mental and substance use disorders while alcohol use disorders (AUDs) account for 9.6%.4
In low-income and middle-income countries (LMICs) the population-level treatment gap is estimated to be between 76.3% and 91.9% for depression and between 94.9% and 97.2% for AUD.5 There is an emerging evidence base demonstrating that depression and AUD can be treated by primary care providers in LMIC6; WHO’s mhGAP guidelines support integration of mental health services into primary care as a means of narrowing the treatment gap.7 People affected by depression and AUD often present in primary health care facilities8 9 though not specifically for these disorders. These disorders co-occur with both acute and chronic medical problems; an untreated mental health disorder worsens the prognosis for the comorbid condition.10 11 The population-level treatment gap can be reduced by enhancing the capacity of primary care staff to detect, diagnose and treat these disorders.12 Clinical detection of depression is estimated to be 47% from a meta-analysis of 41 studies conducted in primary care settings13 and 42% for AUD from a meta-analysis of 12 studies.14 However, these meta-analyses were not able to identify studies conducted in LMIC settings, where the majority of people with depression and AUD in the world live, and where a paucity of mental health in the pre-service training of general health care providers as well as competing demands in under-resourced health care systems likely compromise the ability of clinicians to detect mental disorders. Little is therefore known regarding the detection levels of depression and AUD in primary care settings in LMICs.
The aims of this report were to estimate the proportion of adult primary care outpatients who are clinically detected and initiate treatment for depression and for AUD in LMIC settings.
Context, setting and participants
PRIME is a 6-year multicountry research programme consortium which, in collaboration with national and district Ministries of Health, has developed Mental Health Care Plans to support delivery of services for mental disorders in the public sector in Ethiopia (Sodo District), India (Sehore District, Madhya Pradesh State), Nepal (Chitwan District), South Africa (Dr Kenneth Kaunda District, North West Province) and Uganda (Kamuli District).15 16 Two key research questions of the PRIME evaluation, detailed in a separate report,17 were to assess the change in detection and change in initiation of treatment among adults presenting in primary health facilities, as a consequence of implementing the district mental health care plans. These questions were investigated by conducting cross-sectional facility-based patient surveys before and after the mental health care plan implementation. The two populations of interest for the study were the adult patients who screened positive (1) for depression and (2) for AUD.
Details about the study settings and clinics are in table 1.
The choice of included clinics was determined by the availability of staff who were planned to have authority to detect/diagnose, prescribe and/or refer for depression and AUD, which, per the respective country’s mental health care plan included clinics with health officers, medical officers, health assistants and auxiliary health workers, nurses and doctors. In South Africa, one of the PHC clinics was excluded from the study due to mental health care plan training occurring prior to the baseline survey round. In South Africa, nurses in the primary care clinics could refer suspected cases to a physician (for both disorders) and provide brief counselling (for non-dependent AUD) and so these clinics were included. Only patients attending for chronic care services (eg, HIV, tuberculosis, diabetes, hypertension, etc) were eligible for this study, reflecting eligibility for treatment in accordance with the South African mental health care plan. In Nepal, clinicians working in two clinics received mental health training prior to the baseline survey round; these clinics are included for clinical-level descriptive reporting but participants from these clinics are excluded for analysis here. All clinics were government run, aside one in Ethiopia.
For each country and disorder the sample size was set to have 80% power and a two-sided alpha of 0.05 for the two primary aims: (1) to detect a change in detection and (2) to detect a change in initiation of evidence-based treatment for depression and for AUD between the baseline and follow-up round. The baseline round was set prior to implementation of the mental health care plan and follow-up round was scheduled to start at least 18 months later, which assumed the plan had been fully embedded for several months. Country teams in Nepal and Uganda planned to conduct an interim survey round immediately after embedding, to assess the short-term effect of implementation on detection and treatment. Findings from interim and follow-up surveys will be detailed in future reports. Depending on country, the baseline level of detection was assumed to be 0%–5%, and at follow-up targeted to reach 20%–30%. The required sample size was adjusted to account for the possibility of false positives generated by screening tools, using site-specific figures for the positive predictive value. Within each country, whichever disorder required the higher sample size dictated the sample size. This sample size was increased by a factor of 2–3× to facilitate equity analyses, for example, comparisons of outcomes by sex or by socioeconomic status. The target sample sizes for each country in each study round were 1000 in Ethiopia, 760 in India, 1400 in Nepal, 1200 in South Africa and 1800 in Uganda. Within each country the total target sample size was then allocated by clinic in proportion to measures of clinic outpatient volume, such that busier clinics were allocated a larger recruitment target than less busy clinics.
Details of the sampling and data collection procedures are given in table 2.
Sampling and recruitment
Inclusion criteria for participation were as follows: above age of majority in the country (ie, 16 years or 18 years); fluency in a local study language; time and ability to complete the full interview; and willingness to provide informed consent. Exclusion criteria were as follows: incapacity to provide informed consent (eg, presence of severe intellectual disability, currently experiencing an acute medical issue). The research team for each country trained its interviewers to assess eligibility.
Logistic and cultural constraints dictated the sampling procedure within each country, meaning that random selection for a representative sample was not always possible. In Ethiopia and Uganda, to minimise disruption, consecutive sampling of all eligible patients on registration minimised the amount of time research staff spent in the clinics. In South Africa, research staff provided a group orientation to the study to all patients in the waiting room, and then asked interested patients to self-nominate for recruitment. In India, research assistants approached every fifth patient registering at reception and assessed them for eligibility. Among eligible patients approached, >94% consented to participate in all countries.
Data collection procedures
The facility interview comprised a screening questionnaire, an exit questionnaire and/or a clinical consultation form. Interviewers administered the two-part screening questionnaire, with part 1 used to identify probable cases. For probable cases and optionally for probable non-cases, interviewers completed part 2 of the screening questionnaire. For these same participants, the interviewer completed the exit questionnaire with the participant and/or requested a consultation form from that participant’s clinician. In Uganda, patients maintained their own medical files in a notebook which is handed over to clinicians during consultations, and so the clinical consultation form data were synonymous with exit questionnaire data. The data collection flow chart for each country is shown in online supplementary figures 1a-1e.
The sections within the screening interview questionnaire are described in table 3.
Data collection measures
Part 1 of the screening questionnaire consisted of sections on sociodemographic characteristics, screening for depression, depression symptoms in the past 12 months (aside in Ethiopia) and screening for AUD. A probable case of depression was a participant who was Patient’s Health Questionnaire (PHQ-9) positive or had recent depression symptoms, and a probable case of AUD was a participant who was Alcohol Use Disorder Identification Test (AUDIT) positive. Probable cases completed disorder-specific sections about recent (12 months) history of treatment seeking for their most recent episode of symptoms, and about internalised stigma. Part 2 consisted sections about sociodemographic characteristics and disability status. The exit questionnaire and clinical consultation form were thematically similar, and consisted of a mix of open-ended and closed-ended questions about that day’s clinical consultation, and specifically about diagnoses, advice, referrals and prescriptions. Each country developed their own exit questionnaire and clinical consultation form to enable data collection by interviewers with the participant and by the clinician directly, respectively, in recognition of the context-specific nature of patient–clinician interactions and the local idioms of distress. In addition to the questionnaire items described here, which were largely consistent across PRIME country sites, research teams included country-specific questions and sections, which will be described in future reports. See PRIME website (http://www.prime.uct.ac.za) for the purpose-built sections of the PRIME questionnaires.
Mental health measures
The denominators for the primary outcome measures consist of those participants who screen positive for depression or for AUD using the PHQ-9 and AUDIT, respectively. The PHQ-9 is a widely used screening tool that has been validated for use in all five countries.18–22 In this study, Cronbach’s alpha for the PHQ-9 ranged between 0.74 in India and 0.80 in Nepal. A score of 10 or more on the PHQ-9 screen was considered a positive screen. The AUDIT tool has been validated in India, Nepal and South Africa23–25 and in countries neighbouring Uganda.26 27 In this study, Cronbach’s alpha for AUDIT ranged from 0.66 in Uganda to 0.88 in Nepal. An AUDIT score of ≥20 was considered to indicate alcohol dependence, while scores of 16–19 and 8–15 were classified as harmful and hazardous drinking, respectively. A score of 8 or more in Ethiopia, India and Uganda, 9 or more in Nepal and 16 or more in South Africa was considered a positive screen. The higher cut-off score in South Africa was set to account for services being targeted to those with harmful and dependent alcohol use.
The numerators for the primary outcome measures were derived from participant exit questionnaire data when available (India, Nepal, South Africa, Uganda), and alternatively from the clinician consultation form data (Ethiopia). Given the sparse level of detail patients were expected to recall and/or clinicians were expected to record, cross-country variation in the terminology around detection and treatment, as well as expected low levels of detection and treatment at baseline, highly sensitive and non-specific coding criteria, detailed in table 4, were adopted and used for the outcomes’ numerators. These criteria were informed by WHO’s mhGAP guidelines.7
Outcome assessors (TR, SS and SDR) independently double-coded the outcomes for detection (yes/no) and treatment (Yes/No) with 99% initial scoring agreement. Disagreements were resolved through further discussion.
First, the sociodemographic characteristics and mental health screening scores of participants were summarised by presenting the median and IQR for continuous measures and counts and percentages for categorical measures. Second, for depression screen-positive participants, the numbers and proportions who had outcome data were detected for depression and who had initiation of minimally adequate evidence-based treatment were reported. The same figures were reported for AUD screen-positive participants and AUD. Finally, depression and AUD detection figures are reported for participants who were depression and AUD screen-negative and who did not have depression symptoms over the past 12 months. These latter figures are indicators—though not definitive evidence—for either misdiagnosis or overdiagnosis. All analyses were conducted in Stata 14.1 (StataCorp), and stratified by country. (see online supplemental file ‘stata do file code.docx’)
All participants gave written or verbal informed consent prior to being interviewed (table 2). The informed consent form made clear there would be no negative effects for non-participation. In South Africa, participants were provided a 30 rand (US$~2.80) supermarket voucher as a token of appreciation. In all countries, participants who endorsed questionnaire items about suicidality were referred to a provider in the clinic.
The institutional review boards of the WHO (Geneva, Switzerland), University of Cape Town (South Africa), College of Health Sciences of Addis Ababa University (Ethiopia), Indian Council of Medical Research (New Delhi, India), Sangath (Goa, India), Nepal Health Research Council (Kathmandu, Nepal), Makerere University (Kampala, Uganda) and the National Council of Science and Technology (Kampala, Uganda) reviewed and approved the protocols and informed consent procedures for this study.
The PRIME interventions and evaluations were informed through Theory of Change workshops held in each country.28 These workshops included national-level and district-level representatives, health service providers and, in some countries, mental health service users.
The demographic and mental health screening characteristics of participants are detailed in table 5.
Ages ranged from a median of 28 years in Uganda to 46 years in South Africa. The majority of participants in all countries were female, from 51% in India to 79% in Uganda. The proportion of participants who screened positive for depression ranged between 20% in India to 4.2% in Uganda, for depression symptoms in the past 12 months ranged between 8% in Uganda to 14% in India, and for AUDIT between 1% in Uganda and 16% in Ethiopia. These probable cases were asked to complete the Exit interview and/or their clinicians asked to complete a consultation form. Outcome data were available for 96% of these participants (from clinical consultation forms or exit questionnaires), though exit questionnaire completion was lower for depression screen-positive participants in Uganda (48/80, 60%), and AUD screen-positive participants in South Africa (38/43, 90%) and Uganda (18/23, 78%), respectively. Non-completion was primarily due to participants having to leave the clinic immediately or the interviewer being unable to locate the participant after their consultation.
The proportions of screen-positive participants who were detected and who started treatment are presented table 6.
Among depression screen-positive participants, detection of depression ranged from 0% in India to 11.7% in Nepal. Small proportions of screen-positive participants received treatment (0% in Ethiopia, India and South Africa to 4.2% in Uganda). Among AUD screen-positive participants, detection of AUD ranged from 0% in Ethiopia and India to 7.8% in Nepal. Treatment was 0% in all countries aside Nepal, where it was 2.2%.
For probable non-cases, detection of depression ranged from 0/557 (0.0%) in India to 3/74 (4.0%) in Nepal, and detection of AUD ranged from 0/557 (0.0%) in India to 2/74 (2.7%) in Nepal. Treatment was almost entirely absent (see table 7).
This study establishes the magnitude of the detection gap for adults attending primary care in diverse LMIC settings. There were low levels of detection of depression from screen-positive participants in Ethiopia, Nepal, South Africa and Uganda and no detection in India. There was no detection of AUD among screen-positive participants outside Nepal and Uganda. Conversely, there was almost no evidence of misdiagnosis or overdiagnosis of depression or AUD among participants who screened negative.
The detection figures observed here are substantially lower than the average figures found by Mitchell et al for detection of depression (47%) and for AUD (42%) by primary care providers in high-income countries.13 14 As studies of clinical detection in LMIC settings were not available for these meta-analyses, this study fills a key gap in our understanding of the detection gap globally. The consistency of findings across these five diverse settings likely provides insight across LMIC settings generally. The health service organisations in this study varied considerably in catchment size, services offered and provider types (table 1), and facility attendees varied considerably by age, sex, educational attainment and symptom severity (table 5). Yet detection was consistently poor. These findings provide insight into how the population-level treatment gap in LMIC is at least partially attributable to a facility-level detection gap.
While detection levels were low, the facility-level treatment gap approached 100% in most settings. The treatment outcome definitions used here were broad and included provision of advice or referrals to specialist care, both of which clinicians were able to dispense prior to implementation of the PRIME mental health care plans. The first barrier for providing treatment at the health facility is detection, and, as reported above, across all settings detection levels were extremely low. The findings of this study indicate consistent missed opportunities for providing evidence-based care across these diverse LMIC settings. Improving clinical detection and treatment of depression and AUD by primary care providers remains an area where intervention development is required; the PRIME consortium will evaluate its own interventions—the plans for which have described in detail29–33—in follow-ups to this report.
The PRIME Facility Detection study uses widely validated screening tools to identify probable cases. While diagnostic interviews are the gold standard for identifying cases, screening tools are usable by trained interviewers rather than clinicians, and so the methods used here are more easily replicable for monitoring and evaluation activities in other LMIC settings. Our outcome data were collected directly from the participants and clinicians rather than from a health management information system, which is also a replicable monitoring method for LMIC settings.
There are several limitations to our study. First, screening tools (PHQ-9 and AUDIT) were used to identify probable cases of depression and AUD; an unknown number of screen-positive cases are actually false positive cases. Further, a 100% detection figure is not a desirable goal as it indicates diagnosis among false positive cases. Screening misclassification is likely to be similar in the follow-up round, and so the denominators for detection and treatment are equally biased across rounds and allow a valid comparison across time to be made. Second, non-random sampling was used to select patients in some countries. While the samples may not be representative of the facility-attending population, the same sampling plan will be used in follow-up rounds, enabling valid comparisons for the study’s primary findings. And, given the non-random sampling and use of screening tools it is not appropriate to interpret the proportions of participants who screen positive as prevalence figures for cross-country comparisons. The loss to follow-up for screen-positive participants in Nepal, South Africa and Uganda could result in biased findings as the diagnostic characteristics of lost participants are unknown.
A third limitation concerns the outcome definitions for detection and treatment. Given the limitations of using patient-reported and clinician-reported data, with issues around recall and specificity, we opted to use extremely sensitive yet non-specific thresholds of evidence for coding detection or treatment as having occurred. The detection and treatment figures reported here should therefore be regarded as the upper bound of possibility: some of those coded as having been detected with depression may have other mental health disorders, and some of those coded as having treatment may not have what is considered to be minimally adequate evidence-based care. Again, the bias due to these misclassifications will be equal in the follow-up round. Use of cross-country coding criteria facilitates comparisons across diverse settings, and in future reports each country can use these criteria and/or develop their own more locally appropriate and specific criteria. This process has been completed in Sodo District, Ethiopia, where detection outcomes have been reported separately for using specific criteria for depression and non-specific criteria for common mental disorder.34
We plan to repeat this survey in each of the implementation sites. By comparing the baseline versus follow-up figures within each country, we will be able to determine whether the level of detection and level of initiation of evidence-based treatment for depression and for AUD has increased after implementing mental health care plans. Further to this we will compare the change in detection among probable non-cases (table 7), which is an indicator of inappropriate diagnosis; district health manager can use two detection figures to recalibrate their training and supervision systems. Also using follow-up data, we will also be able to assess whether the improved detection and improved treatment provision is equitable by age, sex and other socioeconomic factors. And, with the help of Theory of Change framework and process evaluation data collected over the implementation phase,17 we will try to explain the reasons for improvement/non-improvement of detection and initiation of treatment for depression and for AUD, along identifying with the factors relating to detection. As each country developed its own Theory of Change framework, it will be possible to contrast five frameworks with five sets of follow-up findings, and then to identify the essential characteristics of an effective strategy to improve detection.
Further research can identify the patient-level, clinician-level and system-level characteristics associated with detection as a means of further refining interventions. Some of these characteristics are already potential targets for intervention and have been identified in previous studies: on the patient level, those who have higher level of perceived need35 and lower levels of internalised stigma13 14 are more likely to receive a diagnosis. On the clinician level, detection improves with longer consultation time,36 37 adequate training, a stronger therapeutic alliance13 and with contractual incentives.38 And on the health system level, detection is likely to improve with the availability of medications and health providers at PHC level who have the authority to prescribe psychotropic medication, as well as referral pathways to counsellors. Also important are buy-in and support from leadership who give priority to mental health,1 governance and supervisory structures to develop and execute standardised protocols,39 and a functional health management information system40 to monitor and feed back on clinical activity. A combination of these patient-level, clinician-level and system-level characteristics may explain some of the substantially lower detection figures for depression and AUD found here from our LMIC settings compared with those found in meta-analyses by Mitchell et al in HIC settings.
The findings of this study suggest large detection and treatment gaps for adult primary care patients, which are likely contributors to the population-level mental health treatment gap in LMIC. Primary care centres remain unfulfilled intervention points for reducing the population-level burden of disease in LMIC.
Thanks to Katamba Mutyaba, Nabukko Sarah, Kirangi Juliet, Kasiiri Joweria, Namwase Suzan and Mwebesa Julius in Uganda; Tasneem Kathree, Palesa Mothibedi, Primrose Mathakga and Deanna Carter in South Africa; Anup Adhikari in Nepal; and the study participants.
AF and RS contributed equally.
Contributors Conceptualisation: SDR, MJ, IP, FK, JN, CL, AF and RS. Design: SDR, TR, GM, NPL, AB, IP, FK, JN, CL, AF and RS. Data acquisition: TR, GM, VM, SS, NPL, OS and JS. Data analysis: SDR. Data interpretation: SDR, CL and RS. Drafting SDR, TR, CL and RS. Critical revision: GM, VM, SS, NPL, OS, JS, MJ, AB, IP, FK, JN, CL, AF and RS. Final approval: SDR, TR, GM, VM, SS, NPL, OS, JS, MJ, AB, IP, FK, JN, CL, AF and RS. Accountability: SDR, TR, GM, VM, SS, NPL, OS, JS, MJ, AB, IP, FK, JN, CL, AF and RS.
Funding This study is an output of the PRogramme for Improving Mental health carE (PRIME). The material has been funded by UKaid from the UK Government (Department of International Development).
Disclaimer The views expressed do not necessarily reflect the UK Government’s official policies. The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.
Competing interests None declared.
Patient consent Not required.
Ethics approval WHO (Geneva, Switzerland), University of Cape Town (South Africa), College of Health Sciences of Addis Ababa University (Ethiopia), Indian Council of Medical Research (New Delhi, India), Sangath (Goa, India), Nepal Health Research Council (Kathmandu, Nepal), Makerere University (Kampala, Uganda), and the National Council of Science and Technology (Kampala, Uganda)
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Statistical code (Stata) is available as supplementary material. While we cannot make the dataset publicly available, we will consider all request to provide a minimal dataset to interested researchers via the PRIME consortium Expression of interest form here: http://www.prime.uct.ac.za/contact-us.