Risk factors for psychological distress in Northern Ireland

Abstract

Background:

Stress-related mental ill health and its disorders are considered by the World Health Organization (WHO) to be the new world epidemic and their prevalence rates seem to be increasing worldwide.

Aims:

To examine and identify sub-populations at risk for psychological discomfort in Northern Ireland and map the relative impact of potential predictors.

Methods:

A sample of 4,638 respondents to the NIHSW-2001 survey was analysed with latent class analysis and latent class factorial analysis. Latent class multinomial logistic regression assessed the impact of a range of predictors on class membership.

Results:

Five sub-populations were differentiated. All subgroups at risk for anxiety and depression were characterized as being younger and female. Disability and adverse life events were strong predictors of risk. Long-standing illness and housing worries were predictors of medium and high risk membership. The effect of civil unrest was significant only for the medium-risk subgroup; marital status and income did not affect group membership.

Conclusions:

Because all five subgroups showed a different probability, but a similar profile of endorsing GHQ-12 items, it could be hypothesized that an underlying continuum dimension of anxiety and depression is present in the Northern Irish population.

Keywords

Psychiatric epidemiology community mental health latent class factorial analysis social psychiatry risk factor for ill mental health LC multinomial logistic regression

Introduction

Anxiety and depression are among the most prevalent disorders impairing mental health worldwide (WHO World Mental Health Survey Consortium, 2004). Their negative impacts, as debilitating factors on several domains of life, are well documented in the American (Kessler, McGonagle, & Zhao, 1994; Kessler, Tau Chiu, Demler, & Walters, 2005; Regier, Narrow, & Rae, 1990) and European literature (Jenkins et al., 2003; Lewis et al., 2003; McConnell, Bebbington, McClelland, Gillespie, & Houghton, 2002). Worldwide, the majority of cases of such psychological disorders are still left untreated or dealt with in an inadequate manner (Kessler et al., 2005; Kessler et al., 2009; McConnell et al., 2002).

In Northern Ireland, a long-standing political conflict (colloquially called ‘the Troubles’), resulted in the violent breakdown of community relations and has had a dramatic impact on the mental health and psychological well-being of this exposed population (Kelleher, 2003; O’Reilly & Stevenson, 2003). In a survey on the prevalence of psychiatric disorders and the need for psychiatric care in the District of Derry (Londonderry), McConnell et al. (2002) reported that 12.2% of the Derry population was estimated to have had a psychiatric disorder in the previous year and this was due to the combination of social deprivation and civil unrest.

Two national representative surveys were conducted to assess the well-being of the Northern Irish household population (NIHSW-1997 and NIHSW-2001). A comparative analysis of the findings of the two surveys (Miller, Devine, & Schubotz, 2003) revealed that mental health in Northern Ireland has improved with the changing of the sociopolitical climate. However, the reporting of stressful life events by younger respondents was higher in 2001 compared to 1997. Concerns regarding the mental health and psychological distress of young people in Northern Ireland have been raised by McWhirter (2004) and by Cairns and Lloyd (2005). In particular, McWhirter (2004) underlined the existence of a steadily increasing rate of suicide from 1983 among the 16–35-year-old population.

The Young Life and Times survey from 2004 to 2009 (Cairns & Lloyd, 2005; Schubotz & Devine, 2008; Schubotz & McMullan, 2010) investigating attitudes and the mental/emotional health of 16-year-olds in Northern Ireland, uncovered psychological distress in about 24% (2004) and 26% (2009) of respondents. A consistent gender difference (more females were often distressed than males) was established in all surveys. In 2004, the major cause of stress was academic performance, followed by family and financial problems. The 2009 survey confirmed this trend, showing that 43% of youths from a disadvantaged background reported greater emotional distress. However, in 2009 physical appearance was the most important cause of concern (64% of females and 28% of males), followed by academic performance, self-esteem and self-confidence. Finally, a gender difference was confirmed over time in all surveys.

O’Reilly and Stevenson (2003) reported that psychological distress in Northern Ireland was higher than the other regions of the UK. Furthermore, Miller et al. (2003), comparing the findings of the Scottish Health Survey 1998, the Health Survey for England 2000 and the Northern Ireland Health and Social Wellbeing Survey 2001, revealed that 21.1% of Northern Irish respondents, compared to 15% of Scottish and 14% of English respondents, had shown signs of psychological discomfort during the previous 12 months. In particular, Northern Irish women seemed more affected by anxiety than men and scored higher than men in the GHQ-12 scale (Goldberg & Williams, 1988) compared with the other two regions of the UK. Miller et al. (2003) emphasized that the marital status of respondents, adverse life events (in particular, the illness of a family member) and lack of social support were among the most important factors in predicting mental health problems.

Regarding the determinant of mental ill health and psychological distress, research carried out in six European countries (ESEMeD/MHEDEA 2000 Investigators, 2004) stressed the relationships between socio-demographic factors and the prevalence of anxiety and depressive disorders. In Great Britain, Paykel, Abbot, Jenkins, Bruga and Meltzer (2003) and in Northern Ireland, McConnell et al. (2002) reported that unemployment and social deprivation seemed associated with an increasing rate of anxiety and mood disorders. In a study among unemployed men in a western area of Northern Ireland, Bunting and NISHS Research Team (2006) demonstrated the negative impact of unemployment on the mental health of males.

The research findings concerning the Northern Irish population required further investigation in order to clarify the following questions: (1) are latent homogeneous subgroups with different degrees of psychological discomfort (anxiety/depression, social dysfunction and loss of confidence) present among the respondents of the NIHSW-2001 survey; (2) which covariates are associated with mental discomfort across the emergent latent classes; and (3) are all subgroups equally affected by the impact of these covariates?

The objective of this paper is to examine the latent homogeneous subgroups for different degrees of mental discomfort in relation to the impact of social and psychological factors. The rationale is that uncovering sub-populations at risk for socio-psychological discomfort and with a different susceptibility to social determinant and adverse life events, could have important implications for prevention and clinical intervention, among those exhibiting mental ill health symptomatology.

Methods

Sample

A random sample of 5,000 addresses were chosen from the Valuation and Land agency list (VLA), an updated list of private households made available to the Northern Ireland Statistics and Research Agency, Central Survey Unit (McWhirter, 2002; Miller et al., 2003). Face-to-face interviews with 5,205 individuals were carried out from February 2001 to July 2001 using a computer-assisted personal interview schedule (CAPI). To incorporate missing data, the sample was reduced to 4,638 and consisted of 58% females with an age range of 16–97 years (M = 47.12, SD = 18.58).

Measures

General Health Questionnaire-12 (GHQ-12)

The GHQ-12 (Goldberg & Williams, 1988) was administered by trained lay interviewers and used to assess recent changes in the perceived well-being of the respondents. The 12-item version of the GHQ (Goldberg & Williams, 1988) is a widely used measure for the assessment of mental health with high scores suggesting worsening of psychological condition (Miller et al., 2003). The symptoms of anxiety are mainly depicted by items 1, 2 and 5 of the scale; the symptoms of depression by items 6, 9, 10 and 11. These symptoms are considered among the inclusion criteria for anxiety and depression in accordance with the DSM-IV (APA, 2000). For our purposes, the 12 GHQ items were taken as dependent variables and, using a classic method of coding, were dichotomized in the input matrix of Mplus 6.11 (Muthén & Muthén, 1998–2010) as follows: scores 1 and 2 = 0; scores 3 and 4 = 1. Both in psychological practice and in analysis a uni-dimensional measure has frequently been shown to fit the data (Hankins, 2008). This will be our starting point for a model fit. While the underlying construct (GHQ) within a factor model implies a continuum, it is in practice used by taking a number of sub-populations (threshold) based on the level of severity. For this reason the factor model has been combined with a latent class model in order to establish the potential sub-populations and their associated risk factors.

Covariates

A number of variables were screened for their significance in relation to mental health as assessed by the GHQ-12: gender; age (continuous); income (coded as < £3,000, £3,000–£9,999 and £10,000–£24,999 compared to > £25,000); health-related variables, such as long-standing illness (yes/no) and disability (yes/no); marital status (single, separated/divorced, widowed compared to married/cohabiting/same-sex); effects of Northern Irish civil unrest ‘the Troubles’ on the environment and on the personal and family life of the respondent (not very much at all, just a little, quite a bit, a lot); stressful life events having occurred in the past 12 months (no worry or stress, just a little, quite a lot, a great deal of worry or stress); amount of perceived control over own life (undecided, disagree/strongly disagree compared to agree/strongly agree); worries about housing and safety concerns about the environment in which they live (yes/no); and perceived social support.

Perceived social support

Perceived social support was operationalized using a seven-item scale for family and friends taken from the 1987 Health and Lifestyle survey (Cox et al., 1987). Each of the items considers statements: ‘There are people I know among my family and friends’ (1) ‘Who do things to make me happy’; (2) ‘Who make me feel loved’; (3) ‘Who can be relied on no matter what happens’; (4) ‘Who would see that I am taken care of if I needed to be’; (5) ‘Who accept me just as I am’; (6) ‘Who make me feel an important part of their lives’; (7) ‘Who give me support and encouragement’. Responses range from ‘not true’, ‘partly true’ to ‘certainly true’.

Statistical analysis

A latent class factorial model (figure 1) with a number of explanatory measures was used to describe the relationship between the items of the GHQ-12. In order to establish sub-population profiles within this factor structure, a confirmatory one-factor mixture model was tested, thereby combining a factor analysis model (CFA) within a latent class structure (LCA). In addition, a CFA was conducted for perceived social support, incorporating factor scores. Both the GHQ-12 and perceived social support are usually used as equally weighted summed indexes, thus implying that a one-factor model (strictly speaking, parallel factor models) can be taken as a reasonable description of the data. The latent class factorial analysis (LCFA) was conducted using the Mplus software 6.11 (Muthén & Muthén, 1998–2010) in order to test for sub-populations based on patterns of endorsement of GHQ-12 items. To establish sub-population profiles within this factor structure, two to six class model solutions were explored. The following statistic criteria were used to find the best-fitting model: Akaike information criterion (Akaike, 1974); Bayesian information criterion (Schwartz, 1978); sample-size adjusted Bayesian information criteria (Sclove, 1987), which give the parsimony index, and the entropy value that indicates a good separation of classes. In addition, two likelihood ratio tests, Vuong-Lo-Mendell-Rubin (Vuong, 1989) and Lo-Mendell-Rubin (Lo, Mendell, & Rubin, 2001) were estimated. Typically, a range of fit indices are used to provide the best model fit with lower information criteria values indicating the most parsimonious and best-fitting models. A multinomial logistic regression was used in order to: (1) further characterize the subgroups observing the shift of membership as a function of the covariates; and (2) assess the differential impact on class membership of socio-demographic factors, adverse life events and perceived social support.

Figure 1.

Latent class factorial model.

The GHQ-12 items are shown as the expression of one factor and five sub-populations (latent classes). Differences across these sub-populations were modelled in terms of a number of predictors. The analysis was accomplished within the framework of an LCFA.

Results

GHQ-12

The results from a one-factor model (Table 1) indicate factor loadings ranging from 0.75 to 0.91. A one-factor model provided an adequate description of the data (CFI = 0.983; TLI = 0.979; RMSEA = 0.055, 90% CI = 0.052–0.059; WRMR = 2.539).

Table 1.

Estimates, standardized YX factor loadings for GHQ-12 and probability by class of endorsing each of the GHQ-12 items.

GHQ-12 item	Factor loading	Item endorsement (%)	Latent classes
			Probability of endorsing by class
			C1 (31.5%)	C2 (15.9%)	C3(7.2%)	C4 (2.5%)	C5* (42.8%)
Concentration	0.776	{19.1}	0.150	0.421	0.701	0.956	0.010
Sleep	0.772	{20.6}	0.165	0.464	0.748	0.968	0.010
Useful	0.745	{13.9}	0.106	0.301	0.557	0.905	0.009
Decision	0.821	{09.6}	0.037	0.176	0.472	0.931	0.001
Strain	0.823	{28.2}	0.257	0.666	0.895	0.993	0.010
Difficulties	0.879	{14.6}	0.047	0.332	0.772	0.992	0.000
Activity	0.778	{20.7}	0.174	0.476	0.753	0.968	0.011
Problems	0.879	{10.7}	0.020	0.194	0.649	0.989	0.000
Depressed	0.905	{21.5}	0.098	0.595	0.928	0.999	0.001
Confidence	0.880	{15.6}	0.063	0.355	0.760	0.989	0.001
Worthless	0.893	{08.8}	0.017	0.140	0.515	0.974	0.000
Happy	0.892	{12.8}	0.027	0.268	0.755	0.994	0.000

Reference group.

Note: Values shown in curly brackets indicate the percentage of respondents endorsing items with a higher level of psychological morbidity.

Perceived social support

In terms of perceived social support, factor loadings from a one-factor model ranged from 0.71 to 0.90 with individual loadings as follows: item 1 = 0.71; item 2 = 0.84; item 3 = 0.89; item 4 = 0.89; item 5 = 0.82; item 6 = 0.90; item 7 = 0.88. The inclusion of one correlated error between items 1 ‘made to feel loved’ and 2 ‘made to feel happy’ (0.48) had little impact on the overall factor structure. A one-factor model provided an adequate description of the data (CFI = 0.994; TLI = 0.990; RMSEA = 0.055, 90% CI = 0.048–0.062; WRMR = 1.551).

GHQ-12 endorsements

An LCA was conducted using the GHQ factor model in order to establish the sub-population structure of the construct (i.e. LCFA). In addition to the factor loadings shown in Table 1, the percentage of individuals who endorsed each of the items in terms of an increased level of strain during the past year is shown for the overall population within curly brackets, and then given for each of the sub-populations (within parentheses) in terms of the percentage falling within each group. The probability of endorsement within each of the five sub-populations is also shown in Table 1.

Confirmatory one-factor mixture model

A series of sub-populations were examined within the context of a one-factor model for the 12 observed measures relating to the GHQ-12. Two to six sub-populations (Table 2) were considered to explain the pattern of responding within the observed measures. A five-class model was chosen, as indicated by the fit statistic criteria reported in Table 2, to classify the sub-population profiles presented within the one factor for the GHQ-12. The reference class for the model (Class 5), labelled as having almost no risk for psychological discomfort (Figure 2), indicated a very low probability of positively endorsing any of the 12 items, representing 42.8% of the sample. A further 31.5% of the sample (Class 1), labelled as being at low risk, presented with a somewhat raised level of endorsement for the GHQ-12 items. The results from this sub-population also indicated a higher probability for the endorsement of strain and a slightly elevated level of: (a) lack of concentration; (b) lack of sleep; (c) less enjoyment; and (d) depression. This elevated level of responding was also present in all of the other classes, with the exception of those in the reference group.

Table 2.

Information criteria for the selection of the best model fit.

Model	AIC	BIC	SSABIC	Entropy	V-L-M-R LRT	L-M-R LRT
2-class	36397	36558	36479	0.926	.000	.000
3-class	34758	34932	34847	0.862	.000	.000
4-class	34505	34692	34600	0.811	.000	.000
5-class	34438	34638	34539	0.725	.005	.006
6-class	34434	34647	34542	0.715	.001	.002

Note: AIC =Akaike information criterion; BIC = Bayesian information criterion; SSABIC = sample-size adjusted Bayesian information criterion; VLMR-LRT = Vuong-Lo-Mendell-Rubin likelihood ratio test; LMR-LRT = Lo-Mendell-Rubin likelihood ratio test.

Best-fitting model is in bold type. The table shows that the 5-class model has the lowest BIC and SSABIC values. It retains a good separation of classes (entropy 0.725) and a significant LRT.

Figure 2.

Graphic representation of GHQ-12 subgroups with increasing levels of risk for psychological discomfort.

Those in Class 2, representing 15.9% of the population, and those in Class 3, representing 7.2% of the population, occupied the middle group (Figure 2). Class 2, labelled at medium risk, was very likely to endorse the items of concentration, sleep, strain, less enjoyment and depression, thereby showing a low pattern of responding to the items addressing issues relating to decision making, overcoming difficulties, problem solving and feelings of worthlessness. A similar pattern of responding was also present among 7.2% of the population who presented with an even higher endorsement rate on all 12 items (Class 3). The smallest sub-population (Class 4), labelled as neurotic-depressed, represented by 2.5% of the sample, was constituted by those individuals who had a high probability of having endorsed all items (Figure 2). The respondents of this group displayed symptoms of anxiety and depression.

Predictors of membership within the sub-populations

Where the reported odds ratios are less than 1, the inverse has been used for ease of interpretation. With the exception of individuals who tended to endorse positively all 12 questions (Class 4, the neurotic-depressed group), females had a greater probability of being in all of the other sub-populations when compared with those in the reference group – those with the lowest probability of positively endorsing any of the 12 statements (Class 5).

The odds indicate that as sub-populations showed a higher probability for positively endorsing the symptoms, then the group had a higher preponderance of females, with the exception of those who were most likely to endorse all of the questions positively, the neurotic-depressed group (Class 4). In comparison to the reference category, the odds for being female increased as the sub-populations moved further away (Class 1: OR = 1.39; Class 2: OR = 2.24; Class 3: OR = 2.49) from the reference class (Class 5), with the exception of those in the neurotic-depressed group (Class 4: OR = 1.95).

Age was an important predictor in all of the sub-populations, when compared to the reference group. The younger respondents were more likely to have a chance of belonging to one of the at-risk sub-populations that were being compared with the reference category. The odds shown in Table 3 are for a one-year shift in age. No differences were found in the relationship between marital status across groups.

Table 3.

Odds ratios and 95% confidence intervals for subgroups at risk of anxiety/depression compared with the reference group.

Predictors	Low risk	Medium risk	High risk	Neurotic-depressed
	31.5%^a	15.9%^a	7.2%^a	2.5%^a
Gender: male vs female^b	1.39 (1.04–1.79)*	2.24 (1.72–2.94)***	2.49 (1.72–3.57)***	1.95 (1.14–3.33)*
Age (16–96)^b	1.02 (1.01–1.03)**	1.03 (1.02–1.04)***	1.03 (1.02–1.04)***	1.04 (1.03–1.06)***
Marital status: Single	0.87 (0.60–1.27)	0.70 (0.43–1.05)	1.03 (0.61–1.77)	0.77 (0.28–2.15)
Sep/Div	0.89 (0.53–1.49)	0.91 (0.56–1.48)	1.05 (0.58–1.92)	1.25 (0.59–2.65)
Widowed	1.48 (0.96–2.30)	1.33 (0.82–2.16)	1.67 (0.92–3.02)	1.52 (0.56–4.14)
Long-standing illness	1.09 (0.80–1.48)	1.69 (1.24–2.29)**	2.31 (1.48–3.61)**	2.53 (1.29–4.93)*
Long-standing disability	2.86 (2.00–4.10)***	4.13 (2.88–5.92)***	3.92 (2.34–6.55)***	10.59 (5.31–21.14)***
Feel safe	0.78 (0.37–1.62)	0.66 (0.37–1.16)	0.61 (0.30–1.22)	0.61 (0.26–1.39)
Control own life: No control	1.13 (0.70–1.83)	1.59 (1.00–2.53)	1.77 (0.90–3.14)	5.11 (2.38–10.94)***
Unsure	3.12 (1.64–5.94)**	4.71 (2.53–8.79)***	8.75 (4.36–17.53)***	24.08 (10.92–53.12)***
Housing worries	3.24 (1.83–5.75)***	3.91 (2.14–7.15)***	3.57 (1.79–7.16)**	2.93 (1.33–6.48)*
Adverse events	3.53 (2.85–4.37)***	7.45 (5.90–9.41)***	15.82 (11.56–21.66)***	45.68 (26.94–77.46)***
Troubles: Environment	1.08 (0.95–1.22)	1.25 (1.09–1.43)**	1.03 (0.86–1.24)	1.31 (0.99–1.66)
Self/Family	0.98 (0.87–1.11)	1.05 (0.92–1.20)	1.38 (1.16–1.63)**	1.09 (0.87–1.38)
Income: < £3,000	0.96 (0.67–1.37)	0.83 (0.56–1.24)	1.43 (0.61–3.35)	1.06 (0.51–2.22)
£3,000–£9,999	1.15 (0.89–1.49)	1.36 (0.98–1.78)	2.23 (1.30–3.80)*	1.33 (0.84–2.09)
£10,000–£24,999	0.92 (0.74–1.14)	0.86 (0.68–1.09)	1.78 (0.90–3.13)	0.71 (0.42–1.22)
PSS^b	1.75 (1.39–2.33)***	2.86 (2.32–3.70)***	4.17 (2.86–5.88)***	5.26 (3.33–8.33)***

Percentage of respondents in each latent class.

Significant odds ratios were inverted to aid interpretation.

Gender: female = 0; male = 1; Age is centred; Marital status: single, separated/divorced, widowed vs married (reference group); Long-standing illness/disability: no = 0, yes = 1; Feel safe: no = 0; yes = 1; Control over own life: no control, unsure vs control (reference group); Housing worries: no = 0, yes = 1; Stressful life events in the previous 12 months: no worry or stress = 0; yes = 1; Effects of ‘the Troubles’ on the environment/on self/on family: no = 0; yes = 1; Income: < £3,000; £3,000–£9,999; £10,000–£24,999 vs > £25,000 (reference group).

PSS = perceived social support.

p < .05, ** p < .01,***p < .001.

In the medium-risk, high-risk and neurotic-depressed groups, those with a long-standing illness were much more likely to endorse GHQ-12 symptoms (OR = 1.69, 2.31 and 2.53, respectively) when compared with the reference category. On the other hand, having a disability was associated with membership in all four classes when compared with the reference sub-population (OR = 2.86, 4.13, 3.92 and 10.59, respectively). When individuals were asked whether they felt safe in the area where they lived, the responses did not seem to be specifically related to membership of any given class.

When compared to the reference sub-population, those with a somewhat greater probability for endorsing items (i.e. the neurotic-depressed group) were more likely to be undecided or feel that they had no control over their lives compared to individuals who felt they had control over their lives (OR = 5.11). No significant differences were found in the other groups. Furthermore, in all of the classes, respondents were more likely to endorse symptoms if they felt that they had little or no control over their lives compared to individuals who felt they had control over their lives (OR = 3.12, 4.71, 8.75 and 24.08, respectively).

When asked about the extent of recent adverse life events (ranging from 1 = none to 4 = a great deal), the odds of not being in the reference class dramatically increased (OR = 3.53, 7.45, 15.82 and 45.68, respectively). This pattern of sub-population separation was also present in response to the question relating to worries in relation to housing (OR = 3.24, 3.91, 3.57 and 2.93, respectively).

Two questions were asked about the conflict in Northern Ireland. The first was related to the effects of the conflict on the local environment and the other asked about the effects of the conflict on the respondents and their family. Again, responses were obtained on a four-point scale. The question related to the effects of the conflict on the local environment only showed a clear statistical demarcation in terms of those at medium risk of psychological discomfort (OR = 1.25). When asked about the effects of the conflict on themselves or on their family, this was only significantly related to individuals at high risk for psychological discomfort (OR = 1.38).

Previous studies have shown the strong effect of the conflict in Northern Ireland on mental health (Bunting, Murphy, O’Neill, & Ferry, 2011; Wang et al., 2011). In order to check the significant result further given the number of covariates, a separate analysis was conducted using only age, gender and both conflict-related measures. The results from this analysis indicated that there was not a statistically significant effect between those at low risk of psychological discomfort and the reference group in terms of the effects of the conflict on the local environment (OR = 1.13, 95% CI = 0.97–1.30, p = .105) and the effects of the conflict on self/family (OR = 1.07, 95% CI = 0.94–1.22, p = .285); a result in keeping with those in Table 3. However, in relation to those at medium risk of psychological discomfort, it was previously found that the effect of the conflict on the local environment was only significantly related to mental health. In the current analysis, both the effects of the conflict on the local environment (OR = 1.32, 95% CI = 1.18–1.49, p = .001) and on self/family (OR = 1.18, 95% CI = 1.05–1.33, p = .006) were statistically significant. Regarding individuals at high risk of psychological discomfort, the effects of the conflict on self/family were again found to be statistically significant (OR = 1.57, 95% CI = 1.35–1.84, p = .001), and the effect on the local environment remained not significant. The final comparison was between individuals classified as being neurotic-depressed and the reference group. On this occasion the results were statistically significant for the effect on the local environment (OR = 1.31, 95% CI = 1.04–1.64, p = .022) and self/family (OR = 1.56, 95% CI = 1.21–2.01, p = .001).

Individuals who were at low risk, medium risk and neurotic-depressed were no more likely to be earning < £3,000, £3,000–£9,999 or £10,000–24,999 compared to > £25,000 (reference group). On the other hand, respondents who were at high risk were more likely to endorse symptoms if they earned £3,000–£9,999 compared to the reference group (OR = 2.23). When individuals were asked whether they had social support, the responses were specifically related to the sub-populations (OR = 1.75, 2.86, 4.17 and 5.26, respectively). In comparison to the reference category, the odds of perceiving that they had lower social support increased as the sub-populations moved further away from the reference class.

Discussion

This study aimed to uncover, in a random sample of the household population in Northern Ireland, latent homogeneous subgroups with different degrees of risk for psychological discomfort and then examine the impact of a range of predictors for membership in these sub-populations. The results of LCFA (Muthén & Muthén, 1998–2010) suggested that the sub-population profile is better explained by five latent classes (Figure 2) within the one factor for the GHQ-12 (Goldberg & Williams, 1988). Results from the one-factor model of the classically scored form of the GHQ-12 indicated that a one-dimensional structure can satisfactorily describe the data. Other researchers (Hu, Stewart-Brown, Twigg, & Weich, 2007; Shevlin & Adamson, 2005) have reported a more complex factor structure, but this has invariably been in the context where the scoring system has been on a four-point scale.

The one-factor model assumes the presence of an underlying continuum, with the five subgroups representing different levels of severity in terms of psychological distress. Four sub-populations all exhibited increasing risk of psychological distress compared with the fifth sub-population that was considered as a reference group. The four sub-populations showed a similar endorsing pattern of GHQ-12 items at different levels of probability. In particular, they all endorsed difficulties in concentration, loss of sleep, constant strain, loss of enjoyment in normal activities and depression. These results indicated that in the general population in Northern Ireland, anxiety and depression were more likely supported by an underlying continuous dimension, in which clinical cases only express the severity of the symptomatology. Interestingly, all groups displayed a feeling of self-worth, except in the case of a younger population who were at a high risk for psychological distress.

From the range of predictors that were introduced to explain differences between membership in the other four groups and the reference group (individuals with a very low probability of endorsing any of the 12 items), it was apparent that as the symptom profile worsened, the more likely respondents were to report having: (1) a disability; (2) a long-standing illness; (3) having recently experienced adverse life events; and (4) lack of perceived social support. Indeed, the most important predictors associated with psychological distress seem to be the occurrence of several adverse life events during the past year. For the neurotic-depressed, the probability of having adverse life events triples compared to those in the high-risk group. Interestingly, only respondents classified as neurotic-depressed reported having a lack of control over their own lives and were also less worried about housing problems, which may be explained by their pathology. Overall, income level was not a predictor of mental ill health in Northern Ireland.

Our results confirm the importance of health and social predictors for the well-being of individuals in accordance with the findings of Gispert (2003) and Harrison, Barrow, Gask and Creed (1999). They also indicate the key roles of gender and age in determining class membership by revealing that all groups who were at risk were more likely to be female and younger. This finding confirmed the gender difference in respect of anxiety and depression reported in Europe (ESEMeD/MHEDEA 2000 Investigators, 2004) and in the UK (Cairns & Lloyd, 2005; Jenkins et al., 2003; Schubotz & McMullan, 2010; Singleton, Bumpstead, O’Brien, Lee, & Meltzer, 2001).

Contrary to the findings of Miller et al. (2003), feeling safe in one’s own environment was not found, in this research, to be related to class membership. The effects of the civil unrest on both the environment in which the respondents lived and on themselves and/or on their family, appears to have had no systematic effect on class membership, with the possible exception of those in the medium-risk group for the environment, and in the high-risk group in terms of self or family. The effect of the conflict seems to be modulated by experiencing adverse life events. In this regard, it is difficult to disaggregate the effects of the social conflict from socio-economic predictors. From the analysis in Table 3 and the revised multinomial logistic regression using only the predictors of age, gender and both conflict-related measures, it was evident that the conflict could be seen as having even greater importance in terms of mental health.

The lack of impact of ‘the Troubles’ on the well-being of respondents may be explained by the fact that all groups at risk of anxiety and depression were younger compared to the control group. The 16–24-year-old age group has been living in a more politically relaxed climate than the previous generation and is concerned with age-related stressors. This finding is in accordance with McWhirter (2004), Cairns and Lloyd (2005) and Schubotz and Devine (2008).

The British Psychiatric Morbidity Survey 2000 (Jenkins et al., 2003) pointed out that in England, Wales and Scotland the middle-aged group (45–55-year-olds) was more likely to display symptoms of anxiety and depression. Instead, this current study found that the younger segment of the Northern Irish population was significantly associated with stress and depression. These findings highlight potential regional differences throughout the UK.

Limitations

Some limitations of this study are evident. First, the cross-sectional nature of the study does not permit causal inferences between predictors and group membership. Second, the lack of providing a weighting procedure may produce biased results by not adjusting for differential probability of selection and non-response. Finally, the loss of 567 respondents who did not complete information on a range of predictors and the over-representation of women (58%) highlight possible limitations to the generalizability of the findings.

Conclusion

The findings of this research concerning the emergence of a young female population in Northern Ireland at high risk of psychological distress suggest the need of educational and clinical interventions aimed at lowering such a risk. The youngest age group constitutes a landmark in a person’s life due to the socio-psychological changes associated with this developmental stage. This is a time when redefinition of the self and discovery of self-worth occur, prompting the emergence of psychological distress. By becoming independent, the young person encounters for the first time a myriad of issues including housing problems, social pressure for employment and/or career, and sometimes solitude. In sum, all these factors can heavily impact on their mental and psychological well-being. Further research, using longitudinal studies, is needed to investigate the possible early predictors that could explain the difficulty of the younger population in coping with adverse life events.

References

Akaike

(1974). A new look at the statistical model identification. IEEE Transactions on Automatic Control, 19, 716–723.

APA (American Psychiatric Association). (2000). Diagnostic and statistical manual of mental disorders, 4th edn, text revision. Washington, DC: APA.

Bunting

B. P.

Murphy

S. D.

O’Neill

S. M.

Ferry

F. R.

(2011). Lifetime prevalence of mental health disorders and delay in treatment following initial onset: Evidence from the Northern Ireland Study of Health and Stress: An epidemiological study. Psychological Medicine. DOI: 10.1017/S0033291711002510.

Bunting

& NISHS Research Team. (2006). Mental health in unemployed men. Western IFH Action Research Project. Retrieved from http://ulster.academia.edu/FloraTodaroLuck/Papers/1239255/mental_health_in_unemployed_men.

Cairns

Lloyd

(2005). Stress at sixteen: The young life and time survey. Research update 33. Belfast: ARK.

Cox

B. D.

Blaxter

Buckle

A. L. J

Fenner

N. P.

Golding

J. F.

Gore

… Whichelow

(1987). The health and lifestyle survey: A preliminary report. London: Health Promotion Trust.

Gispert

(2003). Socio-demographic and health-related correlates of psychiatric distress in a general population. Social Psychiatry and Psychiatric Epidemiology, 38, 677–683.

Goldberg

Williams

(1988). A user’s guide to the General Health Questionnaire. Windsor: Nfer-Nelson.

Hankins

(2008). Factor structure of the twelve item General Health Questionnaire (GHQ-12): The result of negative phrasing? Clinical Practice and Epidemiology in Mental Health, 4, 10.

10.

Harrison

Barrow

Gask

Creed

(1999). Social determinants of GHQ score by postal survey. Journal of Public Health Medicine, 21, 283–288.

11.

Stewart-Brown

Twigg

Weich

(2007). Can the 12-item General Health Questionnaire be used to measure positive mental health? Psychological Medicine, 37, 1005–1013.

12.

Jenkins

Lewis

Bebbington

Bruga

Farrel

Gill

Meltzer

(2003). The National Psychiatric Morbidity Surveys of Great Britain, 2000: Initial findings from the household survey. International Review of Psychiatry, 15, 29–42.

13.

Kelleher

(2003). Mental health and the ‘troubles’ in Northern Ireland: Implications of civil unrest for health and wellbeing. Journal of Epidemiology & Community Health, 57, 474–475.

14.

Kessler

R. C.

Auguillar-Gaxiola

Alonso

Chatterji

Lee

Ormel

J. …

Wang

P. S.

(2009). The global burden of mental disorders: An update from the WHO World Mental Health (WMH) survey. Epidemiologia e Psychiatria Sociale, 18, 23–33.

15.

Kessler

R. C.

McGonagle

K. A.

Zhao

(1994). Lifetime and 12-month prevalence of DMS-III-R psychiatric disorders in the United States: Results from the National Comorbidity Study. Archives of General Psychiatry, 51, 8–19.

16.

Kessler

R. C.

Tau Chiu

Demler

Walters

(2005). Prevalence, severity and comorbidity of 12-month DMS-IV disorders in the National Comorbidity Survey Replication (NCS-R). Archives of General Psychiatry, 62, 617–627.

17.

Lewis

Bebbington

Bruga

Farrel

Gill

Jenkins

Meltzer

(2003). Socio-economic status, standard of living, and neurotic disorder. International Review of Psychiatry, 15, 91–96.

18.

Mendell

N. R.

Rubin

D. B.

(2001). Testing the number of components in a normal mixture. Biometrika, 88, 767–778.

19.

McConnell

Bebbington

McClelland

Gillespie

Houghton

(2002). Prevalence of psychiatric disorder and the need for psychiatric care in Northern Ireland. British Journal of Psychiatry, 181, 214–219.

20.

McWhirter

(2002). Health and social care in Northern Ireland: A statistical profile. Belfast: Information and Analysis Unity (HSSPS).

21.

McWhirter

(2004). Equality and inequalities in health and social care in Northern Ireland: A statistical overview. Belfast: Department of Health, Social Service and Public Safety.

22.

Miller

Devine

Schubotz

(2003). Secondary analysis of the 1997 and 2001 Northern Ireland Health and Social Wellbeing surveys. Belfast: ARK.

23.

Muthén

L. K.

Muthén

B. O.

(1998–2010). Mplus user’s guide, 6th edn. Los Angeles, CA: Muthén & Muthén.

24.

O’Reilly

Stevenson

(2003). Mental health in Northern Ireland: Have ‘the Troubles’ made it worse? Journal of Epidemiology and Community Health, 57, 488–492.

25.

Paykel

E. S.

Abbot

Jenkins

Bruga

Meltzer

(2003). Urban–rural mental health differences in Great Britain: Findings from the National Morbidity Survey. International Review of Psychiatry, 15, 91–96.

26.

Regier

D. A.

Narrow

W. E.

Rae

D. S.

(1990). The epidemiology of anxiety disorders: The Epidemiologic Catchment Area (ECA) experience. Journal of Psychiatric Research, 24, 3–14.

27.

Schubotz

Devine

(2008). Young people in post-conflict Northern Ireland. Dorset: Russell House Publishing.

28.

Schubotz

McMullan

(2010). Mental and emotional health of 16-year olds. ARK Research Update, 67. Belfast: ARK.

29.

Schwartz

(1978). Estimating the dimension of a model. Annals of Statistics, 6, 461–464.

30.

Sclove

S. L.

(1987). Application of model-selection criteria some problems in multivariate analysis. Psychometrika, 52, 333–343.

31.

Shevlin

Adamson

(2005). Alternative factor models and factorial invariance of the GHQ-12: A large sample analysis using confirmatory factor analysis. Psychological Assessment, 17, 231–236.

32.

Singleton

Bumpstead

O’Brien

Lee

Meltzer

(2001). Psychiatric morbidity among adults living in private households, 2000. London: The Stationery Office.

33.

ESEMeD/MHEDEA 2000 Investigators. (2004). Prevalence of mental disorders in Europe: Results from the European Study of the Epidemiology of Mental Disorders (ESEMeD) project. Acta Psychiatrica Scandinavica, 109 (suppl 420), 21–27.

34.

Vuong

G. H.

(1989). Likelihood ratio tests for model selection and non-nested hypotheses. Econometrica, 5, 307–333.

35.

Wang

P. S.

Aguilar-Gaxiola

Al Hamzawi

A. O.

Alonso

Andrade

L. H.

Angermeyer

… Kessler

R. C.

(2011). Treated and untreated prevalence of mental disorder worldwide: Results from the World Health Organization World Mental Health (WMH) surveys. In Thornicroft

Szmukler

Mueser

Drake

(Eds), Oxford textbook of community mental health (pp. 50–66). New York: Oxford University Press.

36.

WHO World Mental Health Survey Consortium. (2004). Prevalence, severity and unmet need for treatment of mental disorders in the World Health Organization World Mental Health Surveys. Journal of the American Medical Association, 291, 2581–2590.