Interpersonal Sensitivity as Mediator of the Relations Between War Experiences and Mental Illness in War-Affected Youth in Northern Uganda: Findings From the WAYS Study.

Data Collection

Research assistants collecting data for the WAYS study were all university graduates with extensive training in data collection and interviewing skills. The research assistants were provided with training on the background of the WAYS study and further trained on how to conduct interviews. The research assistants were fluent in speaking and writing the native language of the participants (Luo) and the English language. Data collection took place in participants’ homes, nearby trading centers, or community halls. The data collected were on demographic characteristics; participants’ experiences before, during, and after the war; individual factors (e.g., interpersonal sensitivity); and mental health outcomes (PTSD, symptoms of depression/anxiety, and psychotic symptoms). The questionnaire took 30–45 min to complete.

The research assistants were accompanied by a clinical psychiatric officer who could take care of any mental health emergencies where there was a possibility for harm and who could make referrals to the regional referral hospital. Informed consent was obtained from all participants in accordance with ethical guidelines and approval from Gulu University Institutional Review Board and Uganda National Council for Science and Technology.

Impact of Events Scale

The Revised Impact of Events Scale (IES-R) was used to assess symptoms of PTSD at follow-up ( Horowitz, Wilner, & Alvarez, 1979 ; Weiss & Marmar, 1997 ). These symptoms were anchored to war-related traumatic experiences. The IES-R is a 22-item scale that indicates severity of PTSD symptoms with a Likert response format ranging from 0 ( not at all ) to 4 ( extremely ). Total symptom scores for each of the symptom’s clusters (Reexperiencing, Avoidance, and Hyperarousal) were computed by adding up the relevant item scores. The authors reported high test–retest reliabilities and internal consistencies of the three subscales, with alpha coefficients ranging from 0.79 to 0.92 ( Horowitz et al., 1979 ; Weiss & Marmar, 1997 ). In the current study, the IES-R demonstrated high internal consistency values for the total scale as well as the three subscales, ranging from α = .81 to α = .89.

Mental health outcomes

Subscales from the Acholi Psychosocial Assessment Instrument (APAI), which is a modified version of the African Youth Psychosocial Assessment Instrument, were used in the current study. The subscales included Depression/Anxiety Symptoms (18 items), Somatic Complaints (three items), and Conduct Problems (10 items). APAI is a field-based measure previously developed for use in Northern Uganda ( Betancourt et al., 2009 ). In APAI, depression and anxiety were a mixed set of items appearing as one scale. The Depression/Anxiety scale was represented by questions that indicate behavior specific to depression/anxiety, such as I have lots of worries , I sit alone , I think about suicide , and so forth. In this study, the Cronbach’s α values were 0.89 for the combined depression/anxiety items. Depression and anxiety symptoms commonly co-occur. Consequently, the questionnaire items that assessed depression and anxiety psychopathology were mixed together in one scale for common mental health problems, thus preventing them from being considered separately as distinct outcomes. Previous studies also showed a strong overlap among items in the Depression and Anxiety subscales ( Brodbeck, Abbott, Goodyer, & Croudace, 2011 ). For each question, responses were scored from 0 to 3, where 0 = never , 1 = rarely , 2 = sometimes , and 3 = always . The mental health problems were assessed at follow-up.

Psychotic symptoms

Four items indicative of psychotic symptoms (i.e., hallucinations, delusions, and persecutory feelings) were used in the current study: (a) sometimes I hear voices or see things other people do not see, (b) sometimes I feel that I have special powers, (c) sometimes I think that people are listening to my thoughts or watching me when I am alone, and (d) sometimes I think that people are against me. Hallucinations, delusions, and persecutory feelings are all common characteristics of psychotic symptoms. The items were scored from 0 to 3, where 0 = never , 1 = rarely , 2 = sometimes , and 3 = always . The psychotic symptoms scale had good psychometric properties (Cronbach’s α = .71). Psychotic symptoms were assessed at follow-up.

Brief Symptom Inventory

The Brief Symptom Inventory measures different domains of psychological distress experienced in the previous week ( Derogatis & Melisaratos, 1983 ). The Brief Symptom Inventory scale consists of 53 items scored on a 5-point response format ranging from 0 ( not at all ) to 4 ( extremely ). The interpersonal sensitivity subscale is indicated by four items (e.g., “Feeling that people are unfriendly or dislike you”). In this study, only the Interpersonal Sensitivity subscale is used in the analysis and was assessed at follow-up. The internal consistency of the Interpersonal Sensitivity subscale in the current study was acceptable at α = 0.82.

Statistical Analyses

First, correlations among demographic, predictor, and outcome variables were computed. Second, structural equation modeling (SEM) was used to assess the associations between the total number of war events experienced, interpersonal sensitivity, and mental health outcomes (PTSD, depression/anxiety, and psychotic symptoms). Regression models were fitted in line with the strategies outlined by Baron and Kenny (1986) in SEM ( Muthén & Muthén, 1998 ). A mixture of confirmatory factor analyses (CFAs) and multiple regressions were used to examine the relations between constructs in SEM. Constructs that are unobserved or latent variables (e.g., PTSD) are usually estimated by a factor analysis of data from theoretically related measures, such as observed or indicator variables ( Muthén & Muthén, 1998 ). Accordingly, CFA was used to estimate latent variables for the mental health outcomes by loading the indicators from the PTSD, Depression/Anxiety, and Psychotic Symptoms scales. Each factor was identified by fixing the first item loading for each factor to 1, estimating the factor variance, and then fixing the factor mean to 0, while estimating all possible item thresholds (four for each item given five response options) and remaining item loadings. We used weighted least squares mean- and variance-corrected robust methods (all item residual variances were constrained to 1) and used a probit link and THETA parameterization to estimate all higher order models ( Muthén & Muthén, 1998 ). Thus, model fit statistics describe the fit of the item factor model to the polychoric correlation matrix among the items.

The mediation models were assessed to examine the direct relationship between the number of war experiences and mental health outcomes and the indirect relationship between total number of war experiences and mental health outcomes via interpersonal sensitivity. All analyses were adjusted for sex. Models were fitted using the M plus software Version 7 ( Muthén & Muthén, 1998 ). Mediation analyses is appropriate for this study because the WAYS study assessed previous war experiences retrospectively (more than 6 years ago), whereas interpersonal sensitivity and mental health outcomes were assessed at follow-up, slightly more than a year after baseline assessment.

Results

Descriptive statistics and correlations of measures in the study are presented in Table 1 . Generally, participants reported an average of 24.7 ( SD = 6.1) traumatic war-related events, and 169 (37.3%) met diagnostic criteria for PTSD (≥33 on total PTSD score). The results of pairwise correlation analysis among study variables are presented in Table 1 . All variables measured in the mediation model (war experiences, interpersonal sensitivity, PTSD, depression/anxiety, and psychotic symptoms) were significantly correlated with each other ( Table 1 ).

The CFAs indicated moderate-to-high loadings on their respective factors: Interpersonal Sensitivity (β = 0.60, p < .05 to β = 0.83, p < .0001), PTSD (β = 0.61, p < .05 to β = 0.85, p < .0001), Depression/Anxiety (β = 0.57, p < .05 to β = 0.87, p < .0001), and Psychotic Symptoms (β = 0.52, p < .001 to β = 0.89, p < .0001). The comparative fit indices ranged from 0.96 to 0.97 and root mean square error of approximation from 0.04 to 0.06. Comparative fit index values larger than 0.95 and root mean square error of approximation values below 0.06 have been suggested to indicate excellent model fit ( Hu & Bentler, 1995 ; Kline, 2011 ).

Measurement Model

The SEM model, which assesses the strength of the direct relationships, indicated that there were significant direct associations between war experiences and PTSD ( Figure 1 ), symptoms of depression/anxiety ( Figure 2 ), and psychotic symptoms ( Figure 3 ). Interpersonal sensitivity was statistically and significantly related to war experiences and to all indicators of mental health problems (symptoms of PTSD, depression/anxiety, and psychotic symptoms). Interpersonal sensitivity accounted for the relationship between war experiences and all indicators of mental ill-health by statistically significant indirect paths. For symptoms of PTSD, about 55% of the effect of war experiences is partially accounted for by interpersonal sensitivity. Interpersonal sensitivity fully accounted for the total effects of war experiences on symptoms of depression/anxiety. Similarly, 53% of the effect of war experiences on psychotic symptoms was partially mediated through interpersonal sensitivity. Whereas the effects of war experiences on symptoms of PTSD ( Figure 1 ) and psychotic symptoms ( Figure 3 ) markedly attenuated but remained statistically significant, the effects of war experiences on symptoms of depression/anxiety ceased to be significant after including interpersonal sensitivity in the mediation model ( Figure 2 ).

Mediation effects of interpersonal sensitivity on the relations between past war experiences and posttraumatic stress disorder (PTSD). Total effect: β = 0.42, 95% confidence interval (CI) [0.32, 0.52], and total indirect effect: β = 0.19, 95% CI [0.11, 0.27]. The β below the continuous line from war experiences to PTSD represents the total effect of war experiences on PTSD, whereas the β above the dotted line represents the effect of war experiences after interpersonal sensitivity was added to the model as a mediator. Approximately 55% of the effect of total number of war experiences on PTSD is mediated through interpersonal sensitivity. The direct effect of the total number of war experiences on PTSD was attenuated markedly but remained statistically significant, β = 0.19, 95% CI [0.11, 0.27]. All analyses were adjusted for sex.

Mediation effects of interpersonal sensitivity on the relations between past war experiences and depression/anxiety. Total effect: β = 0.28, 95% confidence interval (CI) [0.18, 0.38], and total indirect effect: β = 0.03, 95% CI [−0.03, 0.04]. The β below the continuous line from war experiences to depression/anxiety represents the total effect of war experiences on depression/anxiety, whereas the β above the dotted line represents the effect of war experiences after interpersonal sensitivity was added to the model as a mediator. Apparently, the effect of total number of war experiences on depression/anxiety is fully mediated by interpersonal sensitivity. The direct effect of the total number of war experiences on depression/anxiety ceased to be significant, β = 0.03, 95% CI [−0.03, 0.04]. All analyses were adjusted for sex.

Mediation effects of interpersonal sensitivity on the relations between past war experiences and psychotic symptoms. Total effect: β = 0.25, 95% confidence interval (CI) [0.15, 0.35], and total indirect effect: β = 0.19, 95% CI [0.03, 0.24]. The β below the continuous line from war experiences to psychotic symptoms represents the total effect of war experiences on psychotic symptoms, whereas the β above the dotted line represents the effect of war experiences after interpersonal sensitivity was added to the model as a mediator. Approximately 53% of the effect of total number of war experiences on psychotic symptoms is mediated through interpersonal sensitivity. The direct effect of the total number of war experiences on psychotic symptoms attenuated markedly but remained statistically significant, β = 0.13, 95% CI [0.03, 0.24]. All analyses were adjusted for sex.

Each regression coefficient represents the number of standard deviation ( SD ) change in the outcome variable per SD change of the independent variable. For example, the regression of interpersonal sensitivity on war experiences indicates that a change of 1 SD in the number of war experiences is associated with a 0.40 SD change in interpersonal sensitivity. When regression analysis was carried out between mental health outcomes and interpersonal sensitivity, after adjusting for war experiences, the proportion of explained variance increased from R 2 = 0.19, F (4, 447) = 23.97, p < .001, to R 2 = 0.47, F (4, 447) = 73.63, p < .001, for PTSD; from R 2 = 0.08, F (4, 447) = 9.78, p < .001, to R 2 = 0.78, F (4, 447) = 292.86, p < .001, for symptoms of depression/anxiety; and from R 2 = 0.08, F (4, 447) = 9.06, p < .001, to R 2 = 0.15, F (4, 447) = 14.91, p < .001, for psychotic symptoms.