Conflict in the Indian Kashmir Valley II: psychosocial impact

The overall survey questionnaire assessed baseline demographics, confrontation with violence (results presented elsewhere [2]), mental health, health service usage, socio-economic functioning and sources of support. Mental health was assessed using a Self-Reporting Questionnaire (SRQ), with a reference period of 30 days preceding the survey. The SRQ is an instrument developed by the World Health Organization (WHO) to measure general psychological distress, especially in developing countries. It has good validity and reliability for adults (≥ 15 years) [5], and can be used both as a self- or interviewer-administrated questionnaire. It consists of 20 closed questions covering expression of distress, the total score corresponding to the sum of positive responses. Various studies have validated the use of the SRQ in India [6‐9]. Currently a cut off score of 11 or 12 is accepted [10] although this has been critiqued as being too high [11]. In our study we used a conservative cut-off score of 12, meaning those respondents scoring ≥ 12 are considered to be suffering from psychological distress.

Four categories of closed questions were applied to establish use of health services (categories: never; once; 2–3 times; 4+) and medications (Categories: never; 1–3 times; 4–6 times; 7+). Closed questions were also used to assess coping mechanisms for dealing with stress. The composition of categories for 'consequences of violence' and 'sources of support' was done with input from national staff.

To establish individual socio-economic functioning in relation to health during the past thirty days the H-section of the WHO-Disability Assessment Schedule-II (WHO-DAS-II) was used. This tool has good internal, convergent validity and good sensitivity for change [12].

The survey was forwarded and back translated from English to Urdu and phonetic Kashmiri and piloted prior to full implementation.

Data entry was standardised and checked by supervisors, entered into EXCEL and analysed in EPIINFO-2002. Because males and females differed significantly in the number of confrontations with violence [2], we used univariate analysis to stratify for gender to determine relationships between psychological distress (SRQ ≥ 12) and demographic details, living circumstances, confrontations with violence (witnessing, self-experiencing), health outcomes (physical symptoms, health service use), and socio-economic functioning. We excluded variables such as 'torture while being detained/held hostage' as these responses relate to a sub-sample of those surveyed. We also excluded exposure to violence from this analysis because the proximity to the violence was not defined in detail.

A multivariate statistical model was constructed to investigate relationships between mental health (SRQ ≥ 12) and the above-mentioned variables. We used a logistic regression model including variables that were significant in the univariate analysis (p < 0.05) with backward elimination. In our model we expected each type of event conferring an additional risk over and above any other event experienced. This is in accordance with studies reporting exposure to cumulative traumatic events as a risk factor for the development of PTSD [13, 14].

Psychological distress was mostly expressed through symptoms such as nervousness, tiredness, being easily frightened and headache (Table 1). The prevalence of suicidal ideation is striking: one-third of those surveyed had had thoughts of ending their life in the past 30 days. Over a third of respondents were categorized as suffering from psychological distress (SRQ ≥ 12) using the Indian validated SRQ (33.3%, 170; CI: 28.3–38.4). The design effect for the SRQ was 1.4. Females scored significantly higher (43.8% vs. 24.1%, OR 2.5; CI: 1.7–3.6; p < 0.001).

General methodological limitations, including sampling methodology, retrospective study design, and terminology, have been discussed previously [2]. There are, in addition, a number of potential limitations related to this specific analysis. First, as this is a cross-sectional survey, no causal inferences between violence and mental health can be conclusively made. Second, individual respondents may have implicitly used the presence of mental health symptoms as a deciding factor for whether they have experienced a traumatic event in case of doubt (i.e. recall bias [37]). We consider this as unlikely as we asked respondents to recall violent events but did not ask them to identify which events were traumatic. Finally, we used the SRQ to avoid labelling populations with a psychiatric diagnosis, but using a self-reporting questionnaire has obvious limitations. A comparative study in India of five questionnaires showed good internal consistency and a high discriminating ability with the SRQ having the best results [9], but in comparison to clinical interview, questionnaires only showed strong positive predictive value when a considerable compromise on sensitivity was made. It was concluded that the choice of an optimum cut-off score (to balance sensitivity and positive predictive value) should be adapted to individual settings, and recommend a higher cut-off score for resource-limited primary-care settings [9]. We used a high cut off score of 12, in line with this recommendation. But in the absence of clinical interview no detailed analysis of the mental health status is possible.

In the context of predominantly Urdu speaking population we considered, but did not use, cut off scores from other Urdu speaking cultures such as in Pakistan. A meta-analysis of psychiatric rating scales in Urdu [38] concluded that only a small number of instruments (including SRQ) were sufficiently evaluated. The same review concluded that for the SRQ no cross-culturally validated gold standard was used, cut-offs varied considerably, as did sensitivity (78–93%) and specificity (77–85%). We consider the Indian validation studies [9] as more appropriate because they used clinical interview as gold standard.