Psychiatric Disability Among Tortured Bhutanese Refugees in Nepal

Details of our methods, including the historical background of the Bhutanese refugees, have been described previously (14). We approached a representative sample of 526 tortured adult refugees matched with 526 nontortured refugees between March 20 and July 31, 1997; all of them had been interviewed in 1995 (13). The Center for Victims of Torture Nepal, a nongovernmental organization that was responsible for providing medical and psychosocial care to more than 1,200 tortured Bhutanese refugees, carried out this survey in collaboration with the Transcultural Psychosocial Organization as part of the Multisite Impact of Human-Made Disaster Study. The staff of the Center for Victims of Torture Nepal conducted a hut-to-hut survey in 1994 and identified 2,331 Bhutanese adult refugees who reported a history of physical torture. The 526 tortured participants were randomly selected from that list of 2,331 registered tortured adult refugees, and a comparison group of 526 nontortured participants were selected from neighbors of the tortured participants. The tortured and nontortured refugees were matched for age and gender. A difference in age of 10 years or fewer was accepted as an age match.

We were able to approach 946 (89.9%) of the 1,052 refugees in our sampling frame during the period in which we were allowed to collect the data. Of the 946 approached refugees, 879 (92.9%) were interviewed, 32 refused, 20 were out of the camp, four were not found, five had died, and six were too disabled by mental or physical illness to attend the interview. Of the 879 interviews, 20 were not completed because of mental state problems or deafness. Moreover, the results of 49 interviews were discarded because the interviewees were not in our sampling frame and had been interviewed by mistake. The remaining 810 participants consisted of 418 tortured and 392 nontortured refugees.

We used the depression, specific phobias, dissociative, PTSD, persistent somatoform pain, and generalized anxiety disorder modules of the Composite International Diagnostic Interview, version 2.1 (15). We studied ICD-10 disorders because, in contrast to DSM-IV, ICD-10 has been specifically developed for international use. International field testing in a wide range of cultures and settings has shown that the modules of the Composite International Diagnostic Interview that were included in this study have excellent interrater reliability and acceptable-to-good test-retest reliability (16).

Disability was measured with the World Health Organization (WHO) Short Disability Assessment Schedule (17). This instrument is a short version of the WHO Psychiatric Disability Assessment Schedule, which has been used for cross-cultural comparison of psychiatric disability after extensive field trials and studies in more than 20 countries (18–20). The WHO Short Disability Assessment Schedule is a semistructured interview, and two local physicians, familiar with the context, rated disability in four domains: personal care (washing, shaving, dressing, maintaining hygiene), social function (social activities and relationships with neighbors), household activity (cooking, cleaning, having meals together, decision making), and occupational activity (work or study productivity, relations with others during work or study).

Disability status was given to all those who were found to have any impairment in their expected roles in any domain of assessed functions. A dichotomous variable “any disability” (here onward referred to as “disability”) was created and, consistent with the conceptualization of the WHO Psychiatric Disability Assessment Schedule, defined as any restriction or lack of ability (resulting from a loss in function) to perform an activity in the manner or within the range considered normal for an individual in his or her sociocultural setting (17, 18).

On the basis of a brief medical investigation, the physician assessed the presence of significant current and past physical diseases, including diabetes, hypertension, heart disease, head injury, meningitis, encephalitis, and thyroid dysfunction.

The Composite International Diagnostic Interview was systematically translated and implemented after pilot testing, as discussed in detail elsewhere (21). The interviewers were two male physicians and three male and two female undergraduate students. They all received three weeks of training in administering the instruments. We anticipated the challenge of female respondents’ reluctance to answer private questions related to sex, due to cultural barriers. Accordingly, we assigned female interviewers to all female respondents. The interviews consisted of a medical section and a nonmedical section. The medical section, administered to all participants by one of the two physicians, consisted of the somatoform and dissociative disorders section of the Composite International Diagnostic Interview, a medical history, the WHO Short Disability Assessment Schedule, and a brief physical examination. The five lay interviewers administered all other measures.

All interviews took place in a confidential environment. Because most of our respondents were illiterate, we asked for verbal informed consent after explaining briefly our research methods and aims. Obtained consent was recorded on paper. We followed the Declaration of Helsinki’s recommendations guiding ethical research (22).

We used chi-square analysis and Student’s t test to examine univariate differences in disability and sociodemographic variables between the tortured and nontortured groups. We are reporting two-tailed test values, and we present Yates-corrected chi-square values for all two-by-two tables. Values for Student’s t tests are reported according to the results of testing the equality-of-variance assumption.

We assessed the following possible factors associated with disability: gender, age, marital status, education, religion, political activity, economic status, present physical disease, time since displacement, duration of torture, time since torture, and 12-month ICD-10 psychiatric disorders. With the exception of the variable age, continuous variables were dichotomized with the median value as the cutoff point. Associations between disability and all other possible factors among the tortured refugees were examined through calculation of odds ratios by using univariate logistic regression analyses. Main effects were tested in multivariate logistic regression analysis by simultaneously entering all of the variables that had shown some association (p<0.10) with disability in the univariate analyses. A separate multivariate analysis was run by adding to the model the interaction terms created for the interactions among age, gender, and the three predictor variables that were significant in the multivariate testing of main effects. We used multicollinearity diagnostic tests to check for collinearity among the independent variables in the model. Multicollinearity diagnostics indicated a conditioning index smaller than 30 and at most one variable proportion larger than 0.5 for a given root number, thus suggesting the absence of multicollinearity in these regression analyses (23). For univariate analyses, significance was set at 0.005 to reduce chance significance caused by multiple tests. Significance was set at 0.05 for multivariate analyses. All analyses were repeated by centering the continuous data and entering standardized variables (z scores) and interactions computed from the z scores, and the pattern of results stayed the same (data not reported). Data were analyzed by commercially available software (24).

There was no significant difference between the tortured and nontortured groups in terms of disability detected in the four domains (Table 1). The physicians identified disability in 21.3% of the tortured refugees and 23.7% of the nontortured refugees. Furthermore, out of the total 182 disabled, 127 (69.8%) were rated as disabled in all four domains, 20 (11.0%) in three domains, 10 (5.5%) in two domains, and 25 (13.7%) in one domain; the mean number of disabilities was 3.4 (SD=1.1). Among the disabled, only 15.7% (N=14) in the tortured group and 10.8% (N=10) in the nontortured group had been disabled for less than 1 year. The intraclass correlation coefficient for the correlation among the four different domains of disability was 0.86, and the 95% confidence interval (CI) was 0.83–0.87 (p<0.001). On the other hand, out of the total 499 respondents who had at least one 12-month ICD-10 diagnosis, there were 238 (47.7%) with only one disorder, 160 (32.1%) with two disorders, 67 (13.4%) with three disorders, 30 (6.0%) with four disorders, and four (0.8%) with five disorders; the mean number of disorders was 1.8 (SD=0.9).

Table 2 displays odds ratios for disability among the tortured refugees, based on univariate logistic regression analyses. Assessed disability was not associated with gender, age, marital status, economic status, religion, or political activity (p>0.005 for all tests of association). Having formal education appeared to protect against disability. Those who had present physical disease were at much higher risk for disability. Torture-related variables—duration of torture, time since torture, and time since displacement—were not found to be associated with disability.

Among the tortured respondents, we also assessed bivariate associations of disability with 12-month ICD-10 diagnoses of generalized anxiety, PTSD, persistent somatoform pain, specific phobia, dissociative disorder, and depression. Results of univariate logistic regression analyses are shown in Table 3. Except for generalized anxiety disorder and depression, we found bivariate associations between disability and all assessed psychiatric disorders. The results also indicate a dose-response relationship between number of psychiatric disorders and associated disability, because the odds for disability increased fourfold for those having more than three psychiatric disorders in the past 12 months.

Multivariate logistic regression analyses were conducted to adjust the odds of disability for each other by simultaneously entering in the model the following variables, which had some association (p<0.10) in the univariate analyses: gender, age, education, religion, political activity, present physical disease, PTSD, persistent somatoform pain, specific phobia, dissociative disorder, generalized anxiety disorder, and depression. Table 4 shows that the odds of disability for tortured refugees were increased if they had present physical disease, PTSD, or specific phobia. The pattern of results stayed the same when we added the terms for interactions among age, gender, and the three predictor variables to the multivariate model. None of the added interaction terms was significant at the 0.05 level. Similar multivariate analyses for the nontortured refugees suggested that current physical disease (odds ratio=3.3, 95% CI=1.9–5.7, p<0.0001), greater age (odds ratio=2.3, 95% CI=1.1–5.1, p=0.04), and generalized anxiety disorder (odds ratio=3.2, 95% CI=1.2–8.9, p=0.02) were associated with disability.