ARTICLE IN PRESS Journal of Adolescence xxx (2010) 1–8 Contents lists available at ScienceDirect Journal of Adolescence journal homepage: www. elsevier. com/locate/jado Factors accounting for youth suicide attempt in Hong Kong: A model building Gloria W. Y. Wan a, Patrick W. L. Leung b, * a b Clinical Psychology Service, Hong Kong Sheng Kung Hui Welfare Council, 5/F, Holy Trinity Bradury Center, 139 Ma Tau Chung Road, Kowloon, Hong Kong, China 3/F, Sino Building, Clinical and Health Psychology Centre, Department of Psychology, The Chinese University of Hong Kong, Shatin, NT, Hong Kong, China b s t r a c t Keywords: Suicide ideation/attempt Family Psychopathology Life events/stressors Chinese youths This study aimed at proposing and testing a conceptual model of youth suicide attempt. We proposed a model that began with family factors such as a history of physical abuse and parental divorce/separation. Family relationship, presence of psychopathology, life stressors, and suicide ideation were postulated as mediators, leading to youth suicide attempt. The stepwise entry of the risk factors to a logistic regression model de? ned their proximity as related to suicide attempt.
Path analysis further re? ned our proposed model of youth suicide attempt. Our originally proposed model was largely con? rmed. The main revision was dropping parental divorce/separation as a risk factor in the model due to lack of signi? cant contribution when examined alongside with other risk factors. This model was cross-validated by gender. This study moved research on youth suicide from identi? cation of individual risk factors to model building, integrating separate ? ndings of the past studies. O 2009 The Association for Professionals in Services for Adolescents.
Published by Elsevier Ltd. All rights reserved. Introduction Youth suicide, being one of the three leading causes of death in young people, has been a focus of research. Various individual risk factors have been identi? ed (Gould, Greenberg, Velting, & Shaffer, 2003). Despite this success, not until recently are there attempts to develop complex theory-based models that draw together all those identi? ed risk factors and depict their interplay (Bridge, Goldstein, & Brent, 2006; Mann, Waternaux, Haas, & Malone, 1999). Correspondingly, empirical studies in this area are few (e. g. Foley, Goldston, Costello, & Angold, 2006; Fortune, Stewart, Yadav, & Hawton, 2007; Prinstein et al. , 2008; Reinherz, Tanner, Berger, Beardslee, & Fitzmaurice, 2006). Hence, we propose here a model of youth suicide attempt and test it in a sample of Chinese high school students. We aim at articulating and testing hypothetical pathways between family factors, psychopathology, life stressors, and suicidal behavior. Our model begins with consideration of family risk factors, including a history of physical abuse within the family, poor family relationship, and parental divorce/separation (Johnson et al. 2002; Gould, Fisher, Parides, Flory, & Shaffer, 1996; Gould, Shaffer, Prudence, & Robin, 1998; Liu, Sun, & Yang, 2008; Salzinger, Rosario, Feldman, & Ng-Mak, 2007). However, the latter’s association with youth suicidal behavior is no longer signi? cant or attenuated after controlled for parent-child or family relationship (Groholt, Ekeberg, Wichstrom, & Haldorsen, 2000). Family adversities are also known precursors of youth psychopathology (Fergusson, Woodward, & Horwood, 2000). The latter in turn is found to be a risk factor of suicidal behavior * Corresponding author. Tel. : ? 852 2609 6502; fax: ? 852 2603 5019.
E-mail address: [email protected] edu. hk (P. W. L. Leung). 0140-1971/$ – see front matter O 2009 The Association for Professionals in Services for Adolescents. Published by Elsevier Ltd. All rights reserved. doi:10. 1016/j. adolescence. 2009. 12. 007 Please cite this article in press as: Wan, G. W. Y. , Leung, P. W. L. , Factors accounting for youth suicide attempt in Hong Kong: A model building, Journal of Adolescence (2010), doi:10. 1016/j. adolescence. 2009. 12. 007 ARTICLE IN PRESS 2 G. W. Y. Wan, P. W. L. Leung / Journal of Adolescence xxx (2010) 1–8 (Brent, Baugher, Bridge, Chen, & Chiappetta, 1999; Osvath, Voros, & Fekete, 2004).
A wide range of psychopathology has been implicated, including internalizing/externalizing disorders, and substance use disorders (Brent et al. , 2004; Foley et al. , 2006; Gould et al. , 2003; Lee et al. , 2009; Shaffer et al. , 1996; Stewart et al. , 2006). Furthermore, the risk of suicidal behavior increases with the number of comorbid disorders and with the combination of mood, disruptive and substance abuse disorders (Brent et al. , 1999; Shaffer et al. , 1996). Suicide ideation is among the best predictors of suicide attempt (Prinstein et al. , 2008; Wong et al. , 2008).
Studies also suggest that the occurrence of life stressors may prompt suicide ideators into acting on their ideation, ending up in attempted suicide (Borges et al. , 2008; Liu & Tein, 2005). Based upon the above review, our model on youth suicide attempt starts with family risk factors such as a history of physical abuse and parental divorce/separation. We postulate that poor family relationship, psychopathology, life stressors, and suicide ideation act as mediators, leading to suicide attempt. Speci? cally, we hypothesize that a history of physical abuse and parental divorce/separation are associated with poor family relationship.
The latter is in turn related to the occurrence of psychopathology in youths. Comorbid internalizing and externalizing disorders then play a crucial role as risk factors to recent life stressors and suicide ideation. Finally, the latter two are risk factors directly linked to suicide attempt. Method Participants and procedure A total of 2754 grade 7–11 Chinese high school students were recruited to participate voluntarily in the study. They were randomly sampled from 15 mainstream high schools of diversi? ed academic rankings from different regions of Hong Kong.
The participating schools represented a convenience sample, since they were schools served by the School Counselling Service of Hong Kong Sheng Kung Hui Welfare Council, at which the ? rst author of this study worked. However, the participating schools covered a full range of academic rankings and a wide geographic spread across Hong Kong, and no speci? c bias in their sampling was noted. Thus, they were considered to be representative of local mainstream schools. Special schools of various kinds were excluded. Ethics approval was ? rst obtained from the relevant institutional authority.
School visits were then made to explain the objectives of the study. A total of 15 schools were contacted and all of them (100%) agreed to participate in the study. Informed written consents were obtained from parents of the randomly sampled students. The response rate was 94. 4%. The participants completed the self-report questionnaires during school hours. The returned questionnaires were screened for severe psychopathology and suicidal behavior. For ethical reasons, the corresponding school counselors would be alerted for such cases in order to take appropriate actions.
Measures Psychopathology The 1991 version of Youth Self-Report (YSR) was re-validated with satisfactory test–retest reliability and criterion validity for use with Hong Kong Chinese youths (Leung et al. , 2006). It evaluated the occurrence of psychopathology in the past 6 months. Since two items in YSR referred to suicidal/self harm behavior and they thus contaminated the relationship under investigation between psychopathology and suicidal behavior, the two items were removed in this study from the construct of internalizing problems as assessed by YSR.
Youth suicide ideation/attempt Self-report measures of suicide ideation/attempt had been found to be reliable primary data sources (Joiner, Rudd, & Rajab, 1999; Miranda et al. , 2008). Two measures were used to assess suicide ideation/attempt in this study. First, YSR had two items that referred to suicide ideation and attempt, respectively. However, a single-item measure for a variable was considered undesirable.
Hence, two short self-report questionnaires for suicide ideation and attempt were extracted respectively from an existing, longer questionnaire used in a previous local study which asked the occurrences and details of suicide ideation, communication, plan, and attempt (Ho, Leung, Hung, Lee, & Tang, 2000). This questionnaire was found useful in assessing the suicidal behavior of peers of suicide completers and attempters. Depending on whether a youth had consistently indicated suicide ideation and/or attempt both in the adapted questionnaires and with the corresponding item in YSR, he/she would be considered ategorically in this study as a suicide ideator and/or attempter, or not. Recent life stressors The Social Readjustment Rating Scale (SRRS; Holmes & Rahe, 1967) asked 30 stressful life events which might happen to the young people and their families in the preceding 12 months. In this study, items in the SRRS related to parental con? ict and divorce/separation were excluded, since these family events were separately examined elsewhere as family risk factors in our model. Items irrelevant to local young adolescents were also excluded (e. g. accepted at a college of your choice). Family relationship In this study, the Family Relationship Index (FRI), a composite of three subscales (i. e. , cohesion, expressiveness, and con? ict) of the Family Environment Scale (FES; Moos & Moos, 1986), was used as a measure of family relationship. The FRI had Please cite this article in press as: Wan, G. W. Y. , Leung, P. W. L. , Factors accounting for youth suicide attempt in Hong Kong: A model building, Journal of Adolescence (2010), doi:10. 1016/j. adolescence. 2009. 12. 007 ARTICLE IN PRESS G. W. Y. Wan, P. W. L.
Leung / Journal of Adolescence xxx (2010) 1–8 3 been found to correlate well with other measures of family dysfunction (Hoge, Andrews, Faulkner, & Robinson, 1989) and was widely used in research with Chinese youths (Locke & Prinz, 2002). Parents’ marital status The current marital status of the participants’ parents was enquired with reference to divorce/separation. History of physical abuse Participants were asked to report if they had experienced any physical abuse in the family since childhood. Three items were adapted from Childhood Trauma Questionnaire (CTQ) (Bernstein et al. 1994) and two additional ones that were relevant to the local Chinese context were speci? cally written for this study, e. g. , ‘‘People in my family had applied corporal punishment to me’’ and ‘‘After I was being physically punished, I had to go to see a doctor or could not go to school’’. Data analysis First, a series of logistic regression analysis would be conducted to explore the signi? cance of each individual risk factor separately in predicting suicide ideation and attempt. Except for parental divorce/separation, other risk factors were measured in this study in dimensional scales. They were turned into ategorical data in this logistic regression analysis using distributional cutoffs to de? ne relative deviance (see Table 1 for the exact de? nitions of cutoffs). Second, we tested our proposed model of youth suicide attempt, using hierarchical regression ? rst. The series of predictors would be entered in steps according to their positions in relation to suicide attempt in our proposed model. The risk factors of a history of physical abuse and parental divorce/separation would ? rst be forced into the regression equation, after controlling for effects of the background variables (including age, gender, and family income).
Family relationship would then be entered into the equation in the second step. The third batch of predictors would be internalizing and externalizing problems. They were followed in turn by suicide ideation and life stressors. We hypothesized a partial or complete mediational model in which the effects of factors entered ? rst would be attenuated or superseded by those of the subsequent factors in predicting youth suicide attempt. Lastly, path analysis would be conducted to directly test our proposed model of youth suicide attempt, using Lisrel 8. 71.
Path analysis has its strengths in examining the chains of in? uences between independent variables, and in postulating the possible cause-and-effect relations among variables for further investigation. An ordinary sample covariance matrix for path analysis would not be appropriate to deal with the dichotomous data of suicidal behavior and parental divorce/separation, as well as the kurtosis and skewness within the other dimensional data. Instead, an asymptotic covariance matrix should be used, analysis of which would require the use of an estimator that allowed for non-normality.
The weighted least squares (WLS) method, instead of the maximum likelihood estimator, had provision for such non-normality and was thus the appropriate estimator to be used in this study. Results The mean age of the 2754 participants was 13. 9 years (SD ? 1. 3, range ? 11–18). Among the participants, 55. 7% was male. There were missing data on suicidal behavior from 39 participants. Among the remaining 2715 participants, 252 participants (9. 2%) reported suicide ideation in the past 6 months. Among the 91 (3. 3%) reporting suicide attempt in the same period, only six (6. %) did not report suicide ideation. By gender, among the 1219 female participants, 167 (13. 7%) reported suicide ideation, while only two (3. 0%) of 66 (5. 4%) female suicide attempters did not report suicide ideation. The corresponding ? gures for male participants (1535) were 85 (5. 5%), four (16. 0%), and 25 (1. 6%). Table 1 Percentages of suicide ideators, attempters, and non-suicidal controls, association with various risk factors, and odds ratios (ORs) for predicting suicide ideation and attempt. Variables Cronbach’s Alpha Ideators (n ? 52) n Parents divorced/separated History of physical abusea Poor family relationshipb Internalizing problemsc Externalizing problemsc Frequent recent life stressorsd – 0. 83 0. 83 0. 89 0. 88 – 41 73 142 76 99 102 % 16. 3 29. 0 56. 3 30. 2 39. 3 40. 5 Attempters (n ? 91) n 19 29 51 32 45 44 % 20. 9 31. 9 56. 0 35. 2 49. 5 48. 4 Non-suicidal controls (n ? 2457) n 239 169 602 103 153 364 % 9. 7 6. 9 24. 5 4. 2 6. 2 14. 8 Ideators vs controls OR 1. 8** 5. 8*** 3. 8*** 13. 9*** 9. 8*** 3. 6*** (95% CI) (1. 2–2. 6) (4. 2–8. 1) (2. 9–5. 1) (9. 6–20. 1) (7. 1–13. 4) (2. 7–4. 8) Attempters vs controls OR 2. 3** 5. *** 3. 7*** 10. 7*** 11. 1*** 5. 2*** (95% CI) (1. 3–4. 0) (3. 3–8. 6) (2. 4–5. 7) (6. 6–17. 5) (7. 1–17. 5) (3. 4–8. 1) **p < 0. 01; ***p < 0. 001. a Cutoff at total score > ? 6 (80th percentile). b Cutoff at total score > ? 15 (80th percentile). c Cutoff at T-score > ? 64 (at clinical range, 92nd percentile). d Cutoff at number of recent life stressors > ? 4 (80th percentile). Please cite this article in press as: Wan, G. W. Y. , Leung, P. W. L. , Factors accounting for youth suicide attempt in Hong Kong: A model building, Journal of Adolescence (2010), doi:10. 1016/j. adolescence. 2009. 12. 07 ARTICLE IN PRESS 4 G. W. Y. Wan, P. W. L. Leung / Journal of Adolescence xxx (2010) 1–8 Table 1 presents the internal consistency coef? cients of the measures employed in this study. They were consistently satisfactory, ranging from 0. 83 to 0. 89. Table 1 also lists the results of separate logistic regression analysis of each risk factor, including the percentages of suicide ideators, attempters, and non-suicidal participants (i. e. , those reporting neither suicide ideation nor attempt) having various risks, as well as the odds ratios (ORs) of these risk factors in predicting suicide ideation and attempt.
Despite multiple testing of the group differences, such testing was all theory-driven (see the literature review above) and was not random so that statistical control of the effects of multiple testing was not required. All three family risk factors (i. e. , parental divorce/separation, history of physical abuse, and poor family relationship) were more prevalent among suicide ideators and attempters, and had signi? cant ORs. Among these risk factors, a history of physical abuse was the best predictor of suicide ideation (OR ? 5. 8, 95% CI ? 4. 2–8. 1) and attempt (OR ? 5. 3, 95% CI ? 3. 3–8. 6).
Nearly one third of ideators (29. 0%) and attempters (31. 9%), in contrast to 6. 9% of non-suicidal participants, had a history of being physically abused. With respect to the other two family risk factors, 16. 3% of ideators (OR ? 1. 8, 95% CI ? 1. 2–2. 6) and 20. 9% of attemptors (OR ? 2. 3, 95% CI ? 1. 3–4. 0), compared to 9. 7% of non-suicidal participants, reported parental divorce/separation, while 56. 3% of ideators (OR ? 3. 8, 95% CI ? 2. 9–5. 1) and 56. 0% of attemptors (OR ? 3. 7, 95% CI ? 2. 4–5. 7), compared to 24. 5% of nonsuicidal participants, reported poor family relationship.
Compared to family factors, internalizing and externalizing problems were even more associated with higher risks of suicidal behaviors. The ORs of internalizing problems for suicide ideation and attempt were respectively 13. 9 (95% CI ? 9. 6– 20. 1) and 10. 7 (95% CI ? 6. 6–17. 5), while the ORs of externalizing problems were 9. 8 (95% CI ? 7. 1–13. 4) and 11. 1 (95% CI ? 7. 1–17. 5). Nearly one third of ideators (30. 2%) and attempters (35. 2%), in contrast to 4. 2% of non-suicidal participants, had internalizing problems. The corresponding ? gures for externalizing problems were 39. 3% and 49. 5% vs 6. %. Life stressors in the past year also elevated the risk of suicide ideation (OR ? 3. 6, 95% CI ? 2. 7–4. 8) and attempt (OR ? 5. 2, 95% CI ? 3. 4–8. 1). About 40. 5% of suicide ideators and 48. 4% of attempters were reporting more frequent life stressors, compared to 14. 8% of non-suicidal participants. Table 2 shows the results of logistic regression analysis with forced entry of subsets of risk factors in steps, after controlling for background variables (i. e. , age, gender, and family income). With each successive entry of each subset of risk factors, the majority of the previous ones became insigni? ant so that in the ? nal regression model, only recent life stressors (OR ? 1. 01, p < 0. 01) and suicide ideation (OR ? 95. 7, p < 0. 001) signi? cantly accounted for youth suicide attempt. In other words, despite their initial statistical signi? cance when ? rst entered into the regression model, those risk factors such as a history of physical abuse, poor family relationship, and internalizing and externalizing disorders no longer signi? cantly accounted for youth suicide attempt, after life stressors and suicide ideation were later entered into the model.
This pattern of results indicated a mediational model largely compatible to our proposed model of youth suicide attempt. It should be noted that parental divorce/separation as a predictor was not signi? cant even when ? rst entered into the regression analysis alongside with a history of physical abuse. This risk factor was thus dropped in the later path analysis. Our proposed model of youth suicide attempt, in a form of a mediational model, was directly tested by path analysis. It achieved a very good ? t: c2 (6, N ? 2754) ? 39. 5, p < 0. 0001; GFI ? 0. 99; AGFI ? 0. 97; RMSEA ? 0. 045; NFI ? 0. 96; CFI ? . 97; RMR ? 0. 57 (Fig. 1). Weighted least squares standardized estimators of the model and their signi? cance according to the two-tailed z value are presented in Fig. 1. All paths shown were signi? cant at p < 0. 01. As shown in Fig. 1, a history of physical abuse, as a family risk factor, was linked directly to suicide ideation, as well as to the ? rst tier of mediators in the model, namely, poor family relationship, and externalizing and internalizing problems. They were in turn linked to suicide ideation. The externalizing and internalizing problems were additionally linked to recent life stressors.
Finally, suicidal ideation and life stressors were both associated with suicide attempt, with life stressors also linking to the suicide ideation as well. This mediational model with several tiers of mediators explained 48% and 87% of the variances in youth suicide ideation and attempt, respectively. Table 2 Logistic regression of risk factors in predicting youth suicide attempt, controlled for demographic variables. Blocks entered to the modela Deviance between blocks (c2)b 18. 37*** 1. 02*** n. s. 55. 59*** 72. 71*** 1. 02** n. s. 1. 17*** 1. 01* n. s. 1. 07** 1. 07*** 1. 09*** 168. 80*** 10. 5*** 1. 02* n. s. n. s. n. s. n. s. n. s. n. s. 95. 67*** 1. 01** ORs 1 2 3 4 5 1. Family factors History of physical abuse Parental divorce/separation 2. Poor family relationship 3. Psychopathology Internalizing problems Externalizing problems 4. Suicide ideation 5. Life stressors n. s. n. s. 103. 72*** n. s. : non-signi? cant. *p < 0. 05; **p < 0. 01; ***p < 0. 001. a The sequence of blocks entered into the logistic regression model; all factors entered were continuous variables except parental divorce/separation and suicide ideation. b Chi-square deviance of each block entered.
Please cite this article in press as: Wan, G. W. Y. , Leung, P. W. L. , Factors accounting for youth suicide attempt in Hong Kong: A model building, Journal of Adolescence (2010), doi:10. 1016/j. adolescence. 2009. 12. 007 ARTICLE IN PRESS G. W. Y. Wan, P. W. L. Leung / Journal of Adolescence xxx (2010) 1–8 5 0. 85*** 0. 52*** 0. 13*** 1. 00*** History of physical abuse 0. 39*** Poor family relationship 0. 19*** 0. 13*** Suicide ideation 0. 20*** 0. 15*** 0. 88*** Suicide attempt 0. 21*** 0. 17*** 0. 44*** 0. 79*** 0. 13*** 0. 33*** Externalizing problems 0. 42*** 0. 29*** Recent stressors 0. 82*** *p