Comprehensive psychometric examination of the attitudes towards suicide (ATTS) in South Korea

The aim of this study is to examine the validity and reliability of the ATTS on a group of university students, which would contribute to the evaluation of ATTS as a useful tool of measuring attitudes toward suicide in South Korea with very high suicide rates.

The target subjects of the study were 241 undergraduates, including 105 pre-medicine undergraduates who took the cultural subject course “Modern Society and Health,” at Chung-Ang University, Seoul, Republic of Korea in 2013. 195 students (80.9 %) of the total subject population agreed to fill out a self-reported questionnaire, whereas 46 undergraduate students were unable to participate as they were in absence from class. The questionnaire included 34 items from the ATTS and were added 4 demographic items such as sex, age, major and the university year. The Institutional Review Board of Chung-Ang University approved the protocol, survey instruments, and consent document for this study (IRB no. 1041078-201311-HR-0093-02).

The ATTS was developed through two survey studies conducted in 1986 and 1996 [16]. The first version of ATTS was developed in 1986 based on the SOQ [5] and was expanded in the second 1996 study to reflect related atti-tudinal studies. ATTS was predominantly used in Europe to measure the attitudes toward suicide [16, 17], but in recent years has been used globally including within Asian countries such as South Korea and Japan [8, 10, 11, 13, 18], as well as Africa [9, 19] and others. In this study, we adopted the second version of ATTS which has better validity and reliability than its first version [16].

Exploratory factor analysis (EFA) of ATTS version 2 yielded 10 interpret-able factors, which accounts for 60 % of the total variance and Cronbach’s α coefficients for those factors ranged from 0.38 to 0.86 [2, 8, 10, 11, 14‐16].

The ver-sion 2 in 1996 contained a total of 34 items [16]. These were scored on a five-point Likert scale (agree completely, agree to a large extent, doubtful, do not agree and agree not at all). For easier comprehension, this study applied a reversed scoring system. For example, the participant’s answer of “agree completely” to the item “Suicide understandable if severe, incurable disease-people” would correspond to the minimum score of one according to the original scoring system, but is scored the maximum of five in this study. Therefore the stronger the participant felt about a particular subject, the higher their score would become.

For this study, the original second version of ATTS with 34 items was translated into Korean using a forward and backward translation. The process for a forward–backward translation of ATTS and its psychometric testing, which occurred between November 16 and December 2, 2013.

Firstly, the forward translation of the English version of the ATTS into Korean was undertaken by a psychologist and a psychiatrist. Then, the researchers reviewed these two versions and reached a consensus on the Korean draft version. Secondly, a Korean–Australian bilingual expert, translated the draft back into English. The translator and the researchers compared the backward-translated English version with the original one in terms of conceptual equivalence, and then the Korean version was revised slightly. Through this process, the Korean version was completed and made available for the validity and reliability assessment.

To identify any factors unique to Koreans’ attitudes, EFA was performed with the Korean version of ATTS and then, confirmatory factor analysis (CFA) applied with the Korean version using the number of factors resulted from the EFA earlier, to examine its structural validity. We performed EFA to extract factors using the principal component analysis (PCA) with promax rotation and factor loading >0.4 on each item was considered to belong to the corresponding factors [20]. The internal consistency of the preliminary version of ATTS was assessed using Cronbach’s α, and intraclass correlation (ICC) coefficients was used to assess test–retest reliability. ICC coefficients were interpreted with the following criteria: ICC coefficients <0.4, poor; 0.4< ICC coefficients <0.75, fair or good, ICC coefficients >0.75, excellent [21]. The assessment of the test–retest reliability was performed by administrating the questionnaire after a month to 71 consenting pre-medical undergraduate students. A 1 month interval was deemed sufficient since it was greater than the minimum of 2 weeks required for a wash-out effect of the previous questionnaire [22].

For CFA analysis, indices that were used to assess the fit of the model included: (1) X²-value/degree of freedom (df); (2) normed fit index (NFI); (3) Tucker-Lewis index (TLI); (4) Parsimony Normed Fit index (PNFI); (5) Root mean square error of approximation (RMSEA). With large samples, X²/df ratio of two or less is recommended for interpreting good fit. Cut-off value for interpreting good fit for the NFI and the TLI are 0.90. The recommended cut-off value for the PNFI is 0.80. Recommendations for interpreting fit for the RMSEA are <0.05 for good fit and between 0.05 and 0.08 as adequate fit [23, 24].

The general demographical characteristics of the study subjects are shown in (Table 1). In a total of 195 participants, approximately 17 % more men (58.5 %) participated than women (41.5 %). Most participants (58.5 %) were between 20 and 24 years of age, and the proportions of first, second, third, and fourth-year students were 45.1, 35.9, 5.6, and 13.3 %, respectively. Pre-medical students accounted for 46.2 %, social science for 20.0 %, natural science for 20.5 %, and art and human science for 13.3 % (Table 1).

As shown in Table 2, after the exclusion of two items with factor loadings <0.4, EFA revealed the following 11 interpretable factors: (1) Acceptability (2) Acceptability related incurable disease (3) Preventability (4) Tabooing (5) Unpredictability (6) Normal–common (7) Aging (8) Incomprehensibility (9) Suicidal process (10) Non-communication (11) Relation-caused (Table 2).

The 11 factors explained 62.9 % of the total variance and contained 32 of 34 items. Factor 1 and Factor 2 consist of items on the acceptability of suicide. Factor 2 focuses on the questions of the acceptability of suicide in those with incurable disease on two separate levels, one as the subject, that is myself or as the third person who is assumed to suffer from an incurable disease with the question raised being whether suicide is acceptable. Factor 3 is a section on the preventability of suicide and Factor 4 is composed of items related to the taboo of suicide: ‘Should or would rather not talk about suicide’, ‘Risk to evoke suicidal thoughts if asked about’. Factor 5 describes the unpredictability of suicide and Factor 6 consist of questions about whether suicide is Normal–common. The results of EFA analysis with Factor and their respective items are detailed in (Table 2).

The Cronbach’s α for the Korean version of the ATTS is 0.657 which being greater than 0.6, can be interpreted as having acceptable internal consistency as according to the general consensus on the cutoff value. Cronbach’s α values for the total 32 items ranged between 0.309 and 0.781, with all factors’ Cronbach’s α value greater than 0.3. (Table 2).

Factor 2 which comprises of items on acceptability related to incurable diseases had the highest value of 0.781, and Factor 1 which comprises of items on whether suicide is acceptable followed with 0.762. Factor 4 is composed of items related to the taboo of suicide followed with 0.560, then Factor 3 is a section on the preventability of suicide followed with 0.530. Factor 6 had Cronbach’s α of 0.516 and Factor 5 had Cronbach’s α of 0.497. Factor 7 contained 1 item that related to Aging followed with 1.0, Factor 9 with Cronbach’s α of 0.380, Factor 11 with Cronbach’s α of 0.364 and Factor 8 with Cronbach’s α of 0.339. Factor 10 had the lowest Cronbach’s α of 0.309.

The Korean version of the ATTS showed an excellent overall test–retest reliability, with an ICC coefficients of 0.782 (p < 0.001) and ICC coefficients values greater than 0.7 for 5 out of 11 factors (Factor 1–3, 6 and 9). The highest test–retest reliability was shown in Factor 1 with ICC coefficients value of 0.832, then Factor 3 with ICC coefficients value of 0.771. Factor 6 with ICC coefficients value of 0.758, Factor 2 with ICC coefficients value of 0.750 and Factor 9 with ICC coefficients value of 0.748. The lowest of the 11 factors was Factor 11 with a value of 0.354. The internal consistency and Cronbach’s α values for each factor, as well as the ICC coefficient values for the entire 11 factors of the ATTS are detailed in (Table 2).