East Asian Arch Psychiatry 2011;21:58-63


The Validity of Proxy-based NEO-Five Factor Inventory Data in Suicide Research: a Study of 18- to 64-year-old Hong Kong Chinese Who Attempted Suicide
以近亲为基础的大五人格量表对18至64岁香港华籍自杀未遂 者之效度研究
SSM Chan, CSM Wong, HFK Chiu

Prof Sandra Sau-man Chan, MRCPsych, FHKAM (Psychiatry), FHKCPsych, Department of Psychiatry, The Chinese University of Hong Kong, Hong Kong SAR, China.
Ms Corine Sau-man Wong, BCogSc, MSocSc, Department of Psychiatry, The Chinese University of Hong Kong, Hong Kong SAR, China.
Prof Helen FK Chiu, FRCPsych, FHKAM (Psychiatry), FHKCPsych, Department of Psychiatry, The Chinese University of Hong Kong, Hong Kong SAR, China.

Address for correspondence: Prof Sandra Sau-man Chan, Department of Psychiatry, The Chinese University of Hong Kong, G/F, Multicentre, Tai Po Hospital, 9 Chuen On Road, Tai Po, Hong Kong SAR, China.
Tel: (852) 2607 6025; Fax: (852) 2667 5464; Email: schan@cuhk.edu.hk

Submitted: 28 December 2010; Accepted: 11 January 2011

pdf Full Paper in PDF


Objective: To examine the validity of proxy-based NEO-Five Factor Inventory ratings for 18- to 64-year- old Chinese who attempted suicide.

Methods: In all, 71 suicide attempters and their proxy-informants were recruited. Data based on structured interviews with the proxy respondents were compared with data obtained from interviews of the subjects themselves.

Results: For the 71 subject-proxy pairs, the overall correlations were fair to moderate (r = 0.30-0.45; p < 0.05) in all domains, except openness-to-experience. Informants with lower education levels tended to yield data that correlated less strongly with subjects’ self-reports. Spousal ratings and self-reports correlated significantly in all domains, except extraversion (r = 0.42-0.63; p < 0.05).

Conclusion: The results supported proxy-based data on NEO-Five Factor Inventory in research of suicidal behaviour in this age-group.

Key words: Personality tests; Suicide




结果:除了经验的开放性範畴,这71对小组於所有研究範畴均呈中等度的相关性(r = 0.30-0.45; p < 0.05)。教育水平较低的自杀者近亲,其报告数据与自杀未遂者自我评估的数据呈较低相关性。除了外向性,近亲配偶对各项範畴的评分与自杀未遂者自我评分显著相关(r = 0.42-0.63; p < 0.05)。




Suicide is a common tragic endpoint to a number of biopsychosocial adversities. Cumulative suicide death tolls have amounted to a long-term public health problem worldwide. The design of effective suicide prevention strategies should be based on scientific research that defines and quantifies risk, and protective factors in the target population. Despite potential methodological limitations of psychological autopsy (PA), the PA method is generally regarded as the hallmark for risk factor research of suicide, when conducted in a case-control manner: it is able to elicit psychological profiles and psychosocial circumstances through systematic, in-depth, diagnostic psychosocial interviews with informants who are knowledgeable about suicide victims.1,2 The validity of proxy-based data in suicide research has been examined indirectly by applying PA protocols to suicide attempters whose risk factor profile is largely comparable to suicide victims.3,4 Results supported best-estimate methodology for assessing psychiatric diagnoses, psychosocial circumstances, and details of suicidal behaviour.3,4

Reviews of cohort and retrospective case-based studies have shown a high prevalence of personality disorders (especially borderline and antisocial personality) among suicide decedents and vice versa, but the magnitude of the relationship between personality disorders and completed suicide is constrained by several factors.5 They include: (1) reliability and validity measurement of the categorical diagnosis of personality disorder; (2) the complex interaction between personality and life- stressors; and (3) the co-morbidity of Axis I Psychiatric Disorders.6 Use of a dimensional approach to personality assessment yields combinations of personality traits that might not fit into categorical diagnostic constructs. Some of these traits have been associated with completed suicide: hostility, hopelessness, helplessness, dependency, social disengagements, and self-consciousness.7 One of the available dimensional measures of personality, namely the Neuroticism Extraversion Openness–Personality Inventory Revised (NEO-PIR), depicts the 5-factor model of personality, assessing neuroticism, extraversion, openness-to-experience (OTE), agreeableness, and conscientiousness.8 The 5 domains and 30 facet scales show substantial reliability, validity, and longitudinal stability in clinical and non-psychiatric populations even across different cultural settings.9-12 Using the NEO-PIR, some retrospective case-control studies reported the association of high neuroticism and low OTE with suicidal behaviour.13,14 To date, little is known about the validity of proxy-based data on the NEO–Five Factor Inventory (NEO-FFI), when applied to the study of suicidal behaviours.


Our sample population included all suicide attempters aged 18 to 64 years, presenting to the psychiatric consultation- liaison service of a regional government-funded hospital in the New Territories East District during a 1-year period starting 1 June 2004. Its catchment population within the Hong Kong Special Administrative Region amounted to 400,000. All patients attending the Accident and Emergency Department or any non-psychiatric wards for a suspected suicide attempt were routinely referred to the psychiatric consultation-liaison service. We adopted the definition of suicide attempt / non-fatal suicidal behaviour as per the World Health Organization’s Multi-site Intervention Study on Suicidal Behavior. The latter states that “Non- fatal suicidal behavior with or without injuries is a non- habitual act with a non-fatal outcome that the individual, expecting to, or taking the risk, to die or to inflict bodily harm, initiated and carried out with the purpose of bringing about wanted changes”.15 All potential subjects who met the inclusion criteria were approached. Each subject was asked to nominate a knowledgeable proxy-informant to participate in the other part of this study. Written informed consent was sought from all the subjects and their proxy- informants. The study was approved by the Joint Chinese University of Hong Kong and New Territories East Clinical Research Ethics Committee (CRE 2003.040). Following a protocol adopted in 2 local PA studies on elders and adults,16,17 suicide attempters and their proxy-informants were interviewed separately by the 2 independent raters to ascertain the subjects’ Diagnostic and Statistical Manual of Mental Disorders (4th edition) Axis I diagnosis,18 NEO Personality Profile using the 60-item NEO-FFI,10 psychosocial profile and life circumstances surrounding the index suicide attempt. Results on proxy- subject concordance in domains other than the NEO-FFI have been reported.18 This paper focuses on proxy-subject concordance on NEO-FFI scores.

The Statistical Package for the Social Sciences version 11.5 (SPSS Inc., Chicago [IL], US) was used for all data analyses. The Pearson correlation coefficient was used to estimate the level of agreement in the NEO-FFI domain norm scores. Statistical significance was set at a p value of 0.05 or less.


During the study period, 108 individuals were referred for psychiatric assessment following a suicide attempt, 71 of whom agreed to participate in the study, yielding a response rate of 66%. Among the 37 individuals (9 men and 28 women) who were not recruited, 27 were lost to follow-up after they were discharged from hospital for the index attempt, while 10 refused to participate stating that “they did not want us to contact their relatives”, or “they did not have time”. The mean (standard deviation [SD]) age of these eligible but unenrolled subjects was 29 (10) years with the following diagnoses: no psychiatric diagnosis (n = 22, 60%), adjustment disorder (n = 12, 32%), and major depressive disorder (n = 3, 8%). Among these non- recruited subjects, the most common suicide attempt was drug overdose (n = 26) followed by superficial wrist laceration (n = 11).

Characteristics of Enrolled Subjects and Their Proxy-informants

A total of 71 consecutive subjects were enrolled; 26 (37%) were male and 45 (63%) were female, and their mean (SD) age was 35 (12) years. Based on interviews with subjects in the past 4 weeks, their current psychiatric diagnoses were: major depressive disorder (n = 29, 48%), dysthymia (n = 5, 7%), adjustment disorder (n = 25, 35%), bipolar affective disorder (n = 1, 1%), non-affective psychosis (n = 8, 11%), substance abuse / dependence (n = 3, 4%), substance- induced psychosis (n = 3, 4%), alcohol abuse / dependence (n = 4, 6%), and none of the foregoing (n = 2, 3%).

In all, 71 proxy-informants were interviewed, 39 of whom were female. Their mean (SD) age was 41 (12) years. The most commonly used method of attempted suicide was drug overdose (n = 42, 59%). Other methods included wrist laceration (n = 14, 20%), jumping from a height (n = 14, 20%), charcoal burning (n = 8, 11%), and hanging (n = 1, 1%). The mean (SD) time lag between the index suicide attempt and the research interview was 4.3 (5.7) weeks for the subjects and 7.7 (6.7) weeks for the proxy-informants.

Table 1 shows the cross-tabulation of informants’ characteristics and their education levels. The education level of male informants was not significantly different from that of female informants (χ² = 0.01; p = 0.94), while the informants from subgroups consisting of spouses or friends / siblings tended to have higher education levels than those of the parent-child subgroup (χ² = 11.4; p = 0.01).

Proxy-subject Correlations in NEO–Five Factor Inventory Domain Normscores

All proxy-subject pairs completed the NEO-FFI independently. The Cronbach’s alpha of the 5 factors according to subjects’ self-reports were as follows: neuroticism (0.79), extraversion (0.61), OTE (0.57), agreeableness (0.53), and conscientiousness (0.72). Corresponding Cronbach’s alpha values for proxy- informant reports were: neuroticism (0.90), extraversion (0.57), OTE (0.49), agreeableness (0.54), and conscientiousness (0.91).

Table 2 presents the Pearson’s correlation coefficients for all proxy-subject pairs and subgroups of proxy-subject pairs classified according to informants’ demographic characteristics. When all proxy-subject pairs (n = 71) were included in the analysis, the correlations were fair to moderate (r = 0.30-0.45; p < 0.05), reaching statistical significance across all domains except OTE (p = 0.13).

Results of subgroup analysis are shown as follows. Male informants (n = 32) attained statistically significant correlations for neuroticism (r = 0.48; p = 0.01) and conscientiousness (r = 0.45; p = 0.02), while the correlations in other domains failed to reach statistical significance (r = 0.13-0.34). Female informants (n = 39) attained higher correlations for extraversion (r = 0.36; p = 0.03) and agreeableness (r = 0.58; p < 0.001), but failed to attain statistically significant correlations in other domains (r = 0.14-0.35; p = 0.11-0.41). Among informants with a primary education only (n = 20), the correlations were insignificant across all domains (r = 0.12-0.34). On the contrary, informants with secondary or higher education (n = 51) attained statistically significant correlations in all domains except OTE (r = 0.36-0.41; p = 0.004-0.01).

The correlations for spousal ratings (n = 27) were statistically significant (r = 0.42-0.63; p = 0.001-0.04) across all domains except extraversion. The correlations did not attain statistical significance in most of the domains among subgroups who were not spouses (i.e. parent-child or siblings / friends), with the exception of extraversion in the parent-child subgroup. Notably, informants who were siblings or friends had generally received more education than other groups such as spouses and yet the correlations were generally smaller and failed to reach statistical significance across all domains (r = 0.14-0.51; p > 0.05).


For personality assessment, there are pros and cons to the self-reporting and observer-rating approaches. Self-reports are convenient and it is assumed that individuals have better knowledge of their inner feelings and behaviour than do external observers.19 Self-reports may give rise to problems of measurement validity associated with random responding and social desirability, yielding measures of self-concepts of social perceptions that are influenced by many factors other than the real traits themselves.20 Researchers who prefer observer ratings see them as being more objective and less susceptible to distortions caused by defensiveness or self- presentational strategies.21 Despite the inherent differences between observer-reports and self-reports, there is reason to believe that there are important similarities between self-observations and external observation.22 By now there have been dozens of community-based studies that show parallel structures for self-reports and ratings on NEO-PIR in which spousal ratings displayed similar primary and secondary loadings to the large self-report sample (coefficients of factor congruence ranged from 0.91 to 0.97).23 Similar parallelism was replicated cross-culturally in German subjects24 and Chinese subjects.25 The correlation between single peer observations and self-reports tended to be in the range from 0.3 to 0.5; correlations between 0.5 and 0.7 are not uncommon when spousal ratings are used in place of peer ratings or when 3 or 4 ratings are aggregated.8,10 Observer ratings are thus particularly useful criteria for the validation of self-report inventories.

There is evidence supporting the applicability of this parallel structure of the 5-factor personality model in psychiatric populations. In a study that investigated the criterion and incremental validity of personality reports from psychiatric patients and knowledgeable informants in predicting patients’ risky behaviour, both informants’ and subjects’ reports contributed significantly to the prediction of several behaviours and most strongly to social behaviours, even though correlations between the 2 sources of data were fair to moderate (i.e. 0.3-0.6), suggesting different sources provide unique information.26

When taken together the discrepancies, similarities and complementariness of observer-ratings and self- ratings for personality assessments, it is ideal to adopt a combined approach to yield valid personality assessments. Such an ideal situation, however, can never be achieved in a PA setting. The overall subject-proxy concordance in NEO-FFI domain scores in our study (r = 0.30-0.45) was comparable to a Chinese psychiatric sample (n = 159) in an earlier study,27 in which the self-report and spousal ratings were correlated in the range from 0.20 to 0.56 among all patients (psychotic and non-psychotic). In our study this observation held true, particularly for spousal ratings or informants with higher education levels. Due to the small sample size in this study, subgroup analyses by informant demographic characteristics further compromise statistical power, making it difficult to interpret the effect of different demographic factors on proxy-informant ratings.

In other studies, it was shown that psychiatric diagnoses such as depression and substance abuse cause distortions in personality appraisal, and the subject-proxy concordance is therefore compromised.11,28 In our study, suicide attempters suffered from diverse psychiatric morbidities that could affect personality assessment. However, we cannot tell the nature and degree of such effects, as the small sample size did not permit further subgroup analyses at adequate levels of statistical power. Also, without a local prevalence rate and representative profile of suicide attempters, we were not able to comment on the representativeness of our sample in relation to suicide attempters in our community, bearing in mind that some of them may not seek medical attention. Despite these methodological setbacks, the clinical heterogeneity in our sample was similar to the clinical characteristics of subjects studied in other PAs. A previous study had shown that completed suicides and medically serious suicide attempts are 2 overlapping populations that share common psychiatric diagnostic and other socio-demographic features.29 Such remarkable similarity may allow us to make inferences based on our study results, about the validity of proxy-based data on personality traits in PA settings. Application of our study results to PA methodology warrants caution, as there are inherent differences between our study and PA studies. For instance, the aftermath of completed suicide is often more prolonged, socially complicated, and emotionally charged for informants. The time lag between a completed suicide and a subsequent research interview is often 6 to 12 months for PAs, whilst in our study most informants and subjects were interviewed within 3 months from the index attempt. Such discrepancies might contribute to different degrees of recall bias, by virtue of time-dependent natural memory decay and emotional attrition in the different stages of grief. Furthermore, all our proxy-informants were nominated by the suicide attempters as having the best knowledge about them; while those in PA studies are usually the legally defined next-of-kin or convenient subjects available in various unpredictable psychosocial contexts. Our study also excluded suicide attempters who had no available informant. Thus, we had a biased sample of self-selected knowledgeable proxy-informants.

Low concordance in the domain of OTE may have potential implications on risk factor research in suicidal behaviour. Low OTE has been associated with suicidal behaviour in older depressed adults,13 though high OTE may be associated with readiness to report suicidal ideation and on the contrary may be protective of completed suicide.14 We speculate the reasons for low concordance may be due to questions related to this personality domain, and demands appraisal of inner values and attitudes to new experiences, which may be less observable in one’s behavioural profile and affective expression.

In conclusion, the current study provides evidence that knowledgeable proxy-informants of suicide attempters (particularly spouses or those with higher education levels) are able to provide valid personality assessment at a moderate level of agreement with the subjects’ self-reports. Moreover, this ensued in a research setting approximating to a PA.


The authors gratefully acknowledge the assistance of Dr PF Pang for his assistance in recruiting subjects, data collection and data cleaning. The authors would also like to acknowledge the support from the NIH- funded ICOHRTA Program (D43 TW054814) under The University of Rochester’s Center for Suicide Research and Prevention.


The study was supported by the Research Grant Council of Hong Kong (CUHK 4373/03M; Project Code 2140401).


  1. Hawton K, Appleby L, Platt S, Foster T, Cooper J, Malmberg A, et al. The psychological autopsy approach to studying suicide: a review of methodological issues. J Affect Disord 1998;50:269-76.
  2. Cavanagh JT, Carson AJ, Sharpe M, Lawrie SM. Psychological autopsy studies of suicide: a systematic review. Psychol Med 2003;33:395-405.
  3. Conner KR, Conwell Y, Duberstein PR. The validity of proxy-based data in suicide research: a study of patients 50 years of age and older who attempted suicide. II. Life events, social support and suicidal behavior. Acta Psychiatr Scand 2001;104:452-7.
  4. Conner KR, Duberstein PR, Conwell Y. The validity of proxy-based data in suicide research: a study of patients 50 years of age and older who attempted suicide. I. Psychiatric diagnoses. Acta Psychiatr Scand 2001;104:204-9.
  5. Linehan MM, Rizvi SL, Welch SS, Page B. Psychiatric aspects of suicidal behavior: personality disorders. In: Hawton K, van Herringen K, editors. The international handbook of suicide and attempted suicide. West Sussex, England: John Wiley and Sons Ltd; 2000: 147-78.
  6. Duberstein PR, Conwell Y. Personality disorders and completed suicide: a methodological and conceptual review. Clinical Psychology: Science and Practice 1997;4:359-76.
  7. Duberstein PR, Seidlitz L, Conwell Y. Reconsidering the role of hostility in completed suicide: a life-course perspective. In: Masling J, Borstein RF, editors. Psychoanalytic perspectives of developmental psychology. Washington, DC: American Psychological Association; 1996: 257-323.
  8. McCrae RR, Costa PT Jr. Different points of view: self-reports and ratings in the assessment of personality. In: Forgas JP, Innes MJ, editors. Recent advances in social psychology: an international perspective. Amsterdam: Elsevier Science Publishers; 1989: 429-39.
  9. Cheung FM, Leung K, Zhang J, Sun H, Gan Y, Song WZ, et al. Indigenous Chinese personality constructs: Is the five factor model complete? J Cross Cult Psychol 2001;32:407-33.
  10. Costa PT Jr, McCrae RR. Revised NEO Personality Inventory (NEO- PIR) and NEO Five-factor Inventory (NEO-FFI) professional manual. Odessa, FL: Psychological Assessment Resources; 1992.
  11. Costa PT Jr, Bagby RM, Herbst JH, McCrae RR. Personality self- reports are concurrently reliable and valid during acute depressive episodes. J Affect Disord 2005;89:45-55.
  12. McCrae RR, Costa PT Jr, Ostendorf F, Angleitner A, Hrebícková M, Avia MD, et al. Nature over nurture: temperament, personality, and life span development. J Pers Soc Psychol 2000;78:173-86.
  13. Duberstein PR, Conwell Y, Seidlitz L, Denning DG, Cox C, Caine ED. Personality traits and suicidal behavior and ideation in depressed inpatients 50 years of age and older. J Gerontol B Psychol Sci Soc Sci 2000;55:18-26.
  14. Heisel MJ, Duberstein PR, Conner KR, Franus N, Beckman A, Conwell Y. Personality and reports of suicide ideation among depressed adults 50 years of age or older. J Affect Disord 2006;90:175-80.
  15. Bertolote JM, Fleischmann A, De Leo D, Bolhari J, Botega N, De Silva D, et al. Suicide attempts, plans, and ideation in culturally diverse sites: the WHO SUPRE-MISS community survey. Psychol Med 2005;35:1457-65.
  16. Chen EY, Chan WS, Wong PW, Chan SS, Chan CL, Law YW, et al. Suicide in Hong Kong: a case-control psychological autopsy study. Psychol Med 2006;36:815-25.
  17. Chiu HF, Yip PS, Chi I, Chan S, Tsoh J, Kwan CW, et al. Elderly suicide in Hong Kong — a case-controlled psychological autopsy study. Acta Psychiatr Scand 2004;109:299-305.
  18. Chan SS, Pang EP, Chiu HF. Validity of best-estimate methodology for assessing psychosocial risk factors and making psychiatric diagnosis among suicide attempters. Hong Kong J Psychiatry 2007;17:55-63.
  19. Osberg TM, Shrauger JS. The role of self-prediction in psychological assessment. In: Butcher JN, Spielberger CD, editors. Advances in personality assessment. Hillsdale, NJ: Lawrence Erlbaum; 1990: 97-120.
  20. John OP, Robins RW. Accuracy and bias in self-perception: individual differences in self-enhancement and the role of narcissism. J Pers Soc Psychol 1994;66:206-19.
  21. Funder DC. Accuracy in personality judgment and the dancing bear. In: Buss DM, Cantor N, editors. Personality psychology: recent trends and emerging directions. New York: Springer-Verlag; 1989: 210-23.
  22. Digman JM. Personality structure: emergence of the five-factor model. Annu Rev Psychol 1990;41:417-40.
  23. Costa PT Jr, McCrae RR, Dye DA. Facet scales for agreeableness and conscientiousness: a revision of the NEO Personality Inventory. Pers Individ Dif 1991;12:887-98.
  24. Ostendorf F. Language and personality structure: toward the validation of the five-factor model of personality. Regensburg, Germany: S. Roderer Verlag; 1990.
  25. Yang KS, Bond MH. Exploring implicit personality theories with indigenous or imported constructs: the Chinese case. J Pers Soc Psychol 1990;58:1087-95.
  26. Ready RE, Watson D, Clark LA. Psychiatric patient- and informant- reported personality: predicting concurrent and future behavior. Assessment 2002;9:361-72.
  27. Yang J, Bagby RM, Ryder AG. Response style and the revised NEO personality inventory: validity scales and spousal ratings in a Chinese psychiatric sample. Assessment 2000;7:389-402.
  28. Conner KR, Zhong Y, Duberstein PR. NEO-PIR neuroticism scores in substance-dependent outpatients: internal consistency and self-partner agreement. J Pers Assess 2004;83:75-7.
  29. Beautrais AL. Suicides and serious suicide attempts: two populations or one? Psychol Med 2001;31:837-45.
View My Stats