Ethics statement: Ethical approval for the study was granted by the University of Ulster ethical committee. Written informed consent was obtained from the participants in this study for their involvement in the research.
The NISHS was the largest epidemiological study of mental health in NI. A multi-stage, clustered, area probability household sample was drawn based on the structure and information from the 2001 NI census. The sample size was 4,340 and the response rate was 68.4%. Data was cleaned and missing data collected or imputed prior to the analysis. See Bunting et al. for further details of the sampling procedures and quality assurance strategies [9] (link). The NISHS survey instrument was administered in two sections; all participants completed Section 1, section 2 was then administered to respondents who met the criteria for any core disorder, an additional 50% of individuals who were subthreshold core disorder cases, and a 25% sample of all other individuals (n = 1,986). This sampling strategy enabled the computation of weights to adjust for differential selection for Section 2. Weights to minimize the effects of bias included information relating to sample selection, nonresponse, and poststratification factors such as age, sex, and geographical region [14] . The NI population characteristics at the midpoint of the data collection period were used in these weight calculations.
The survey instrument was the WMH Composite International Diagnostic Interview (WMH-CIDI) [13] . This is a comprehensive, fully structured interview for the assessment of mental disorders according to the ICD-10 Classification of Mental and Behavioural Disorders: Clinical Descriptions and Diagnostic Guidelines (ICD-10) [15] and DSM-IV criteria [16] .
Lifetime suicidal behaviour was assessed using three questions from the suicidality section in part two of the WMH-CIDI: “Have you ever seriously thought about committing suicide?”, “Have you ever made a plan for committing suicide?”, and “Have you ever attempted suicide?”.
Traumatic events were assessed in the PTSD section of part two of the WMH-CIDI. Participants were presented with 28 types of traumatic events and asked whether they had experienced them during their lifetime and if they endorsed a particular event, they were asked the age at which they first experienced this event type. The research team identified events that were presumed to be conflict-related, drawing upon a previous study of conflict in Lebanon [17] (link). Individuals were assigned to a conflict-related category if they experienced any one of the following events from 1968 onwards: combat experience, peacekeeper in a place of war, unarmed civilian in a place of war, civilian in a place of ongoing terror, refugee, kidnapped, man-made disaster, beaten by someone other than parents or partner, mugged or threatened with a weapon, witnessed someone being killed or seriously injured, purposely caused injury or death, or saw atrocities. The event types classified as non-conflict related included rape and sexual violence, death or illness of a loved one or diagnosis with a life threatening condition. It is also likely that a proportion of unexpected deaths and traumatic events involving loved ones could be associated with the NI conflict, however we did not categorise these event types as conflict-related. This is therefore likely to be a conservative estimation of conflict-related trauma. Mental disorders were assessed on the basis of DSM criteria [16] again using the WMH-CIDI.
Chi squared tests were used to assess whether the difference in proportions between categories were statistically significant. The association between traumatic event types and suicidal ideation, plan and attempt, controlling for the effects of any lifetime mental disorder, was examined using logistic regression. The reference category for the logistic regression was not having endorsed suicidal ideation (“seriously considered suicide”). The analysis incorporated weights to adjust for the differential selection for Section 2, sample selection, nonresponse, and poststratification factors, age, sex, and geographical region [14] . All analyses were implemented using STATA version 10.0 [18] .
The NISHS was the largest epidemiological study of mental health in NI. A multi-stage, clustered, area probability household sample was drawn based on the structure and information from the 2001 NI census. The sample size was 4,340 and the response rate was 68.4%. Data was cleaned and missing data collected or imputed prior to the analysis. See Bunting et al. for further details of the sampling procedures and quality assurance strategies [9] (link). The NISHS survey instrument was administered in two sections; all participants completed Section 1, section 2 was then administered to respondents who met the criteria for any core disorder, an additional 50% of individuals who were subthreshold core disorder cases, and a 25% sample of all other individuals (n = 1,986). This sampling strategy enabled the computation of weights to adjust for differential selection for Section 2. Weights to minimize the effects of bias included information relating to sample selection, nonresponse, and poststratification factors such as age, sex, and geographical region [14] . The NI population characteristics at the midpoint of the data collection period were used in these weight calculations.
The survey instrument was the WMH Composite International Diagnostic Interview (WMH-CIDI) [13] . This is a comprehensive, fully structured interview for the assessment of mental disorders according to the ICD-10 Classification of Mental and Behavioural Disorders: Clinical Descriptions and Diagnostic Guidelines (ICD-10) [15] and DSM-IV criteria [16] .
Lifetime suicidal behaviour was assessed using three questions from the suicidality section in part two of the WMH-CIDI: “Have you ever seriously thought about committing suicide?”, “Have you ever made a plan for committing suicide?”, and “Have you ever attempted suicide?”.
Traumatic events were assessed in the PTSD section of part two of the WMH-CIDI. Participants were presented with 28 types of traumatic events and asked whether they had experienced them during their lifetime and if they endorsed a particular event, they were asked the age at which they first experienced this event type. The research team identified events that were presumed to be conflict-related, drawing upon a previous study of conflict in Lebanon [17] (link). Individuals were assigned to a conflict-related category if they experienced any one of the following events from 1968 onwards: combat experience, peacekeeper in a place of war, unarmed civilian in a place of war, civilian in a place of ongoing terror, refugee, kidnapped, man-made disaster, beaten by someone other than parents or partner, mugged or threatened with a weapon, witnessed someone being killed or seriously injured, purposely caused injury or death, or saw atrocities. The event types classified as non-conflict related included rape and sexual violence, death or illness of a loved one or diagnosis with a life threatening condition. It is also likely that a proportion of unexpected deaths and traumatic events involving loved ones could be associated with the NI conflict, however we did not categorise these event types as conflict-related. This is therefore likely to be a conservative estimation of conflict-related trauma. Mental disorders were assessed on the basis of DSM criteria [16] again using the WMH-CIDI.
Chi squared tests were used to assess whether the difference in proportions between categories were statistically significant. The association between traumatic event types and suicidal ideation, plan and attempt, controlling for the effects of any lifetime mental disorder, was examined using logistic regression. The reference category for the logistic regression was not having endorsed suicidal ideation (“seriously considered suicide”). The analysis incorporated weights to adjust for the differential selection for Section 2, sample selection, nonresponse, and poststratification factors, age, sex, and geographical region [14] . All analyses were implemented using STATA version 10.0 [18] .
Free full text: Click here
