Structure of Mental Health System in Mozambique
The National Mental Health Program in Mozambique is managed by the Department of Mental Health at the National Directorate of Public Health at the Ministry of Health. At the Provincial Health Directorate, the Provincial Mental Health Program is part of the Department of Public Health and is responsible for coordinating all mental health activities implemented in the districts and reporting to the National Mental Health Program in the capital of Maputo, Mozambique. At the district level, the District Mental Health Supervisor coordinates and supports activities implemented at health facilities and reports these activities to Provincial Mental Health Program. The country has 25 psychiatrists (18 of which are Mozambican), 305 psychiatric technicians, 130 clinical psychologists, and 14 occupational therapists who provide services to an estimated 7% of public clinics. Since 1996, Mozambique has been a leader in Sub-Saharan Africa in training a task-shared cadre of mental health professionals (Psychiatric Technicians) who can diagnose and treat all major categories of mental illness, with a focus on psychopharmacology. In 2014, the Mozambican Ministry of Health accomplished their goal of placing at least one psychiatric technician at a primary care health facility within each of the 135 districts nationally (22,26); however, the vast majority of psychiatrists are located in the capital of Maputo, Mozambique.
Study Setting and Participants
This study was conducted in Sofala Province, (see Fig. 1), located in the central region of Mozambique with a population of approximately 2.2 million. The official language is Portuguese, with Cisena and Cindau common languages spoken in rural areas. Sofala has a literacy rate of 56.4%, infant mortality of 83.3 per 1000 live births, life expectancy of 50 years, and an HIV prevalence of 14% (27). As a whole, Sofala province has 166 health facilities, of which 25 (15%) have trained mental health staff. These staff include 3 Psychiatrists, 29 Clinical Psychologists, 28 Psychiatric Technicians and 1 Social Worker (28). The present study was conducted in 3 health facilities: 2 in Beira City (Macurungo and Chingussura), and 1 in Dondo (Dondo health facility). Beira is the capital of Sofala Province and the second largest city in Mozambique after the national capital of Maputo. Beira City has a population of approximately 500,000 individuals. Regarding health infrastructure, Beira City has 13 primary care health facilities, 1 quaternary-level central hospital, and several private health facilities. Dondo is the closest city to Beira (35 km), with 8 primary care health facilities serving a population of 91,000 (27). We selected the above-mentioned facilities because they: (1) had at least 1 psychiatric technician and clinical psychologist; (2) were high-flow facilities providing general primary healthcare; (3) provided comprehensive maternal and child healthcare; and (4) were generally representative of other urban and peri-urban primary care health facilities in Mozambique.
Adaptation of PHQ-9 to the Mozambican Context (PHQ-9-MZ)
The PHQ-9 is a self-administered nine-item screening tool for depression that refers to the past two weeks with likert scale responses of how often a person has been bothered by symptoms, including “0 = not at all”, “1 = several days”, “2 = more than half of the days”, and “3 = nearly every day”. This tool can be used to screen for depression in at-risk populations and to monitor the severity of depression and treatment response (18,29). The adaptation and creation of the PHQ-9-MZ occurred from February to April, 2016 with a structured process to ensure content, semantic, and technical equivalence. This focused initially on a series of translations of the English PHQ-9, with a focus on comprehensibility (does an item retain its original semantic equivalence), appropriateness (fit, relevance, compatibility with new cultural context), and a specific focus on ease-of-understanding given the low literacy levels of primary care patients in Mozambique. We followed a modified version of the WHO’s seven steps for the translation and adaptation process (30), with the addition of cognitive interviewing of primary care patients after the first PHQ-9 translation. First, we established a bilingual group of experts, including a local Mozambican psychiatrist (VFJC), an American psychiatric epidemiologist (BHW), and an experienced local Mozambican psychiatric technician (HF). Second, this group examined and discussed the structure of the English PHQ-9. Third, this group collaboratively translated each PHQ-9 item. Fourth, this group examined the translation and refined initial elements. Fifth, the group (led by HF and a second psychiatric technician blinded to the original instrument (PC)) administered the PHQ-9 to 12 primary care patients attending outpatient consultations at Beira Central Hospital, Ponta-Gêa, and Munhava health facilities in Beira City, Mozambique. After administering the PHQ-9, HF and PC guided patients in a cognitive interview process whereby they asked patients what they felt was the underlying significance of each question, whether the question was unclear or inappropriate, and if so, how they might suggest improving each question. Sixth, following cognitive interviews, VFJC, BHW, HF, and PC reviewed the cognitive interviewing data and engaged in a collaborative process of improving the instrument based on this feedback. Last, HF and PC again pilot-tested the final instrument amongst 4 local Mozambican health staff of various literacy working at Health Alliance International, all of whom found the adapted PHQ-9-MZ instrument comprehensible, appropriate, and easy to understand.
Adaptation of the MINI International Neuropsychiatric Interview to the Mozambican Context (MINI 5.0-MZ)
The MINI International Neuropsychiatric Interview 5.0 (MINI 5.0) is a short structured diagnostic interview developed in the United States for Diagnostic and Statistical Manual for Mental Disorders 4 (DSM-IV) and International Classification of Diseases 10 (ICD-10): Psychiatric Disorders. It is administered approximately in 15 minutes (31). The MINI includes a structured psychiatric interview for all common mental disorders and was used in this study as a gold standard diagnostic tool to validate the PHQ-9. In this study, the administration of the MINI 5.0-MZ took an average of 30–45 minutes.
For this study, we adapted the existing Brazilian Portuguese version of the MINI 5.0 to the Mozambican context (MINI 5.0-MZ). Following a similar method as the adaptation of the PHQ-9, we first recruited a group of local Mozambican mental health professionals (2 clinical psychologists and 3 psychiatric technicians) to collaboratively adapt the Brazilian MINI 5.0 to the Mozambican context and linguistic idioms. Second, the instrument was coded in RedCap for use on tablets by a local Mozambican study staff member (AM). Following coding, the same group of mental health staff re-reviewed the MINI 5.0-MZ in RedCap and focused on understanding, ease of use, and logical sequence of questions. Fourth, mental health professionals from each target health facility had a 2-day training for 2 to 3 hours a day on the correct use of the MINI 5.0-MZ. Fifth, mental health staff conducted role plays where one professional was a patient and the other administered the MINI 5.0-MZ, with notes taken and reviewed for instrument improvement. Sixth, the MINI 5.0-MZ was pilot-tested over 4 days among 14 primary care patients attending outpatient consultations at Macurungo heath facility in Beira City. Seventh, patients administered the pilot MINI 5.0-MZ were guided in a cognitive interview process whereby they were asked what they felt was the underlying significance of each question, whether the question was unclear or inappropriate, and if so, how they might suggest improving each question. Following cognitive interviews and the pilot implementation, the mental health professionals, along with VFJC, BHW, and AM reviewed the cognitive interview data and engaged in a collaborative process of improving the MINI 5.0-MZ based on this feedback.
Data Collection Procedures
From October to February 2019, two trained data collectors, supervised by AM, administered a survey using tablet-based RedCap data collection that included sociodemographic variables and the PHQ-9-MZ, to 503 randomly selected patients from the waiting room of antenatal, postpartum, and general outpatient consultations. While in the waiting room, a data collector randomly selected individuals and asked them if they would be willing to complete a survey on depression. The data collector then directed interested individuals to a private room to administer the survey if they were 18 years old or over and agreed to participate in the study by signing an informed consent form. Patients were excluded if they had an acute health condition or disability impeding their ability to complete the survey. This initial survey took approximately 30–40 minutes. Following this survey, patients were referred to the trained mental health professional (psychologist or psychiatric technician) who administered the MINI 5.0-MZ, as the gold standard diagnostic validation tool, blinded to the responses of the patient on the PHQ-9-MZ.
Data Analysis Procedures
Using Stata 15 we calculated the sensitivity, specificity, positive predictive values, negative predictive values, and diagnostic odds ratios across screening cut-points for the PHQ-9-MZ and PHQ-2-MZ, using the MINI 5.0-MZ as gold standard. The receiver operating characteristic curves were graphically examined and the area under the ROC (AUROC) was calculated for each instrument. To examine initial item response theory properties of the PHQ-9-MZ, the item discrimination (α; describing how well a given item can differentiate between patients with different levels of depressive symptoms), item location (b1; b2; b3; the level of the latent trait of depression where the probability of endorsing a given item is 50%), item factor loadings, and item uniqueness were calculated. The item information functions, the full PHQ-9-MZ test information function and standard error, and test characteristic curve and expected scores for different values of the latent trait of depression were also graphically examined. Last, the Cronbach’s alpha (32), including item-test correlations, item-rest correlations, average inter-item covariances, and the Cronbach’s alpha value if each item were to be removed individually were calculated.