Study design, research site and ethics clearance
Gorongosa is a rural district, and the people rely mostly on agriculture for self-sustenance. The 2017 general population census determined that 176,845 people dwelled in the area. Since the end of the civil war (October 1992) and following the government’s postwar reconstruction efforts, districts such as Gorongosa were removed from the isolation of the initial peace years, which facilitated the reestablishment of manual agricultural and artisanal mining activities and consumption of mass media technologies. The government also rebuilt new markets, schools, and health posts. Despite these changes, large segments of the local population are still confronted with daily stressors such as unbridled poverty, unemployment, and domestic and community violence [30,34,35]. The Ethics Board of the University of Queensland approved the research project (Clearance No: 2009001021). Informed consent was obtained through the local authorities in Gorongosa district (center of Mozambique) and the legal guardians of young people which facilitated access to their research participation in their households and schools. Furthermore, all methods, including the presentation of supplementary information, were performed in accordance with the Declaration of Helsinki.
Participants
The total research participants were composed of n = 794, whereby n =430 were males (54.2%) and n =364 (45.8%) were females (Table 1). Their ages ranged between 9 and 21 years. The initial pilot interviews with key informants helped to establish that culturally defined conceptions of being young in the area ranged from 6 years old to puberty but also if the person does not have children yet. Following the local cultural procedures, the initial contacts with young people (6-8 years old) revealed immense difficulties in answering the questionnaires in the self-reported manner adopted in this study. In this regard, we increased the age threshold to the 9-21 range and established three age groups that were consistent with the age groups within the school system namely pre-secondary (G1, 9-15 yrs), secondary (G2, 16-19 yrs) and pre-university (G3, 20-21 yrs). The mean age of all participants was 17.24 years (SD 2.9).
We recruited the respondents as part of a community-based household survey (n = 595, 74.9%) and in two schools, a secular and a Catholic school (grades 6-10) (n = 199, 25.1%) in the Gorongosa Municipality. Most of the respondents were students n = 634 (79.8%), while the remaining approximately 20% stated they were involved in some form of employment. Of the students, there were pre-university students n = 429 (54%), secondary school n = 167 (21%), and primary school n = 45 (5.7%). Most of the participants were single n = 770 (97%) and only n = 23 (2.9%) declared to be married. The residence status of the participants was as follows: n = 272 (34.3%) lived with relatives other than the biological parents; n =217 (27.3%) lived with both parents (father and mother); n = 164 (20.7%) lived with a partner; n = 95 (12%) lived with the mother; n = 37 (4.7%) lived alone; and n = 9 (1.1%) lived with the father.
Respondents’ parental relations were as follows (Table 1): n = 530 (66.8%) stated that they had living parents, while n = 127 (16%) stated that they had no living parents. Yet, among those with parents, n = 107 (13.5%) stated that only the mother was alive while n = 27 (3.4%) only had a father. For those with living parents, n =433 (54.5%) of the participants stated their parents lived together, while n = 354 (44.6%) reported that the parents did not live together. In relation to siblings, n = 467 (58.8%) stated that they had five or more siblings, while n = 327 (41.2%) had four or less.
Qualitative and quantitative research approach
This study used mixed methods combining qualitative and quantitative analyses. The qualitative phase consisted of two pilot studies in Gorongosa (2012 and 2013). It involved individual and group interviews with young people (n =25) and five focus-group discussions with primary and secondary school teachers (n =30) to grasp insights about the mental health problems considered as common to the local young people, and the factors they regarded as either promoting or hindering psychological functioning. We conducted content and thematic analysis to draw out emergent themes that were consistent across respondents. Such themes were used to design a semi-structured questionnaire comprising the following: Demographic characteristics - gender, age, marital status, profession, residential status, parental relationship, and parental status. After capturing demographic information, participants were asked to rate “Yes” or “No” to 16 questions organized in three categories of risk factors at home, school, and community. Risks in the family (7 items) - “Beaten at home”, “fight at home”, “crying at home”, “insulted at home”, “going to school without eating [food insufficiency]”, “alcohol consumption at home”, and “perceptions of quality of family relations.” Risks in the community (4 items) - “Attacked while going to school”, “Sexual assault while going to school”, “Fear of attack” and “fear of something.” Risks at school (5 items) - “Insulted in the school”, “notebook stollen”, “text stollen”, “beaten at school” and “forced to have sex at school.”
The protective factors were “parents live together/separate”, “number of siblings”, “possession of a radio at home”, and “possession of television at home”, “working besides study” and “participation in church.” For both risk and protective factors, participants were given a score of 1 for “Yes” responses and 0 for “No” response. Participants who did not respond or who said the question was not applicable were treated as missing data. Positive responses of 50% and above of the items in each of the three categories of risk factors and protective factors were aggregated, resulting in an index of the extent to which participants were experiencing stressors in the family, school, and community. The qualitative interviews also identified a range of symptoms that young people associated with mental health problems, which overlapped with items of the psychological scales of depression, anxiety-related disorders, sleeping problems, and anxiety dreams used in this study.
Quantitative instruments
For the incidence and severity of depression we used the Beck Depression Inventory-II (BDI-II, 23 items) [44], which was designed to capture psychological and somatic symptoms manifested within a 2-week period. The instrument has been previously used in numerous empirical studies of depression in Western and non-Western societies as well as reviews and meta-analysis to determine its psychometric properties. Various studies have shown high levels of internal consistency of the BDI-II. For example, in a study of a multiethnic population of young people in the United States, the coefficient alpha was .89 [45]; in a study of Portuguese young students, the Cronbach’s coefficient α was .90 [46]; in a study of Hong-Kong adolescents, the Cronbach’s coefficient α was .94 [47]. Furthermore, in a study of HIV-positive adolescents in Malawi, the Cronbach’s coefficient α was 0.80 [23], while a study of young South Africans showed a DBI-II internal consistency of .84 [48]. In this study, the BDI-II was administered as a self-report screening measure. We conducted two pilot studies to adapt the BDI-II to the local context. Linguistic equivalents for each of the items on the questionnaire were established in Portuguese and Chi-Gorongose. The BDI-II was first translated from English into Portuguese and then into the local language by bilingual members from Gorongosa. The translations were then translated back into Portuguese by different interpreters. This validation process determined that the response format in the original BDI-II was not easily understood by participants. Consequently, the response scale of the BDI-II was truncated into a 3-point scale: “no” (scored 0), “sometimes” (scored 1) and “often” (scored 2). Based on pilot testing, two questions in the original instrument were re-worded and each divided into two questions. For instance, the question regarding “changes in sleeping pattern” was divided into two questions: “Do you have difficulty falling asleep?” and “Do you have difficulty staying asleep?” Furthermore, the question regarding “changes in appetite” was divided into two questions: “Do you lack appetite?” and “Do you crave food all the time?” As a result, the locally adapted version of the BDI-II comprised 23 questions. Scores were aggregated such that the maximum total score was 46. Because of the changes to the items and endpoints, no clinical cut-offs were applied, and the scale was interpreted as a continuous measure of depression symptoms. Cronbach’s alpha for this 23-item measure was strong (α = .917).
The Self-Report Questionnaire (SRQ-20) [49] is a tool designed by the World Health Organization to screen for common anxiety-related disorders in developing countries. It contains 20 items that measure general symptoms of distress associated with anxiety and depression. The SRQ-20 has been widely used in studies of mental health problems in both conflict and non-conflict societies [49,50,51,52]. The SRQ items are scored 0 ('no', symptoms absent) or 1 ('yes', symptoms present) and the item scores are summed to obtain a total score. Previous studies conducted in the study region in Mozambique using the SRQ-20 validated this scale [35,36,37,53]. During the pilot for this research, one question (“do you have poor digestion?”) was found to be confusing due to experience with food poisoning in the two schools, and so was dropped from analysis. Thus, the scores for this measure ranged from a minimum of 0 to a maximum of 20. Cronbach’s alpha for the scale indicated satisfactory reliability (α = .929).
Sleep disturbances and nightmare experiences were assessed using the Nocturnal Intrusions after Traumatic Experiences Questionnaire (NITE) [36,54]. It was designed to assess anxiety-related dreams and post-traumatic nightmares involving physical reactions. NITE scores can be analyzed in either a qualitative or quantitative fashion; it is a two-part (partially) structured questionnaire that contains 14 quantitative items and 3 qualitative ones. Yet only 14 quantitative items were used, which focused on general sleeping problems and the presence, content, and impact of anxiety dreams on respondents. NITE-14, items are scored 0 ('no', symptoms absent) or 1 ('yes', symptoms present). In this case, the scores were averaged such that the overall score ranged from 0-1. The NITE was established for content and linguistic equivalents and used in previous studies in the region [35,36,37,53], and the Cronbach’s alpha for this 14-item measure was strong (α = .943).
For the community-based survey (n = 595, 74.9%), research assistants administered the questionnaire in the form of a semi-structured interview, whereas in the schools (n = 199, 25.1%). The participants answered the questionnaires individually as self-reports. Since the class sizes were large, we divided the students in groups so that each student had a single chair and desk, and the research assistants oversaw the process in the classroom without input or interference from classroom teachers, which ensured that participants had confidentiality.
Statistical analysis
All results were examined using IBM SPSS Statistics (Version 27). Preliminary univariate analysis was completed on all variables to ensure consistency and to address the first aim of this report. As stated above, the internal consistency was excellent. To address the second aim, the chi-square test of independence was run on all the baseline variables (demographics, risk, and protective measures) to assess for group differences in the symptoms of BDI-II, SRQ, and NITE. This allowed us to ascertain the associations between the baseline variables and the incidence and severity of mental health issues in our sample. Exploratory binary logistic regression analyses were performed to ascertain which variables contribute most significantly to the prediction of mental health problems (BDI-II, SRQ and NITE) in the participants. All three questionnaires were converted to a binary form for this section of the analysis, to allow for ease of interpretation (mental health problems present or not present), with these being the outcome variable. Demographic variables, risk, and protective factors were added into the model as predictors. Odds ratios were examined for those predictors that were significant in predicting the presence of mental health problems among the research participants and to allow ease of interpretation.