Taiwan National Suicide Prevention Hotline callers’ suicide risk level and emotional disturbance difference during and before COVID-19

Although the COVID-19 pandemic is having mental health consequences for many people, the relationships between the pandemic and suicide risk is still unclear. Some studies revealed an increase in suicide mortality rate during the COVID-19 pandemic (e.g., Devitt, 2020), while others found no change or even a decrease (e.g., Pirkis et al., 2021). In this study, we addressed this concern via directly assessing the levels of suicide risk and emotional disturbance of the callers of Taiwan National Suicide Prevention Hotline (NSPH). To understand the influence of the pandemic on the callers, we compared the data of two equivalent groups: the pre-COVID-19 group (N = 108, 2013–2018) and the COVID-19 group (N = 108, February-May, 2020).

The suicide risk of each caller was assessed with a 9-option scale (5 of them reflecting the levels of suicide risk) by two well-trained raters according to what s/he said on the phone. The inter-rater reliability was pretty high (kappa $=.89$，$p<.01$). In order to assess the emotional disturbance of each caller, we conducted text analysis with LIWC (Linguistic Inquiry of Word Count; Pennebaker et al., 2015) for the transcript transferred from her/his narrative on the phone. LIWC comprises a dictionary which defines common words as about 71 psychological-linguistic categories and an application which is used to recode the words in a text to the percentages of LIWC word categories. In this study, we focused on 7 LIWC word categories which are evident to be the markers for depression or anxiety (see Shen and Rudzicz, 2017; Sonnenschein et al., 2018), namely First-person-singular pronoun (I), Negative emotions, Positive emotions, Sadness, Discrepancy, Health, and Anxiety. As people in Taiwan normally speak Chinese, the transcripts were processed by C-LIWC (Lin et al., 2020), the Chinese version of LIWC. This research has received approval from the Human Research Institutional Review Board (IRB) from National Chengchi University Research Ethics Center, Taipei Lifeline Association (TLA) and Taiwan Ministry of Health and Welfare to obtain the data.

1. Results and discussions

The frequency distribution of 5 suicide risk levels was different between these two groups [χ² ₍₄₎＝30.04, p < .01]. Specifically, comparing to the pre-COVID-19 group, the COVID-19 group had 38% more low suicide-risk callers who have no, occasional, or frequent suicide ideation ($89/108$ vs. $48/108$) [$Z=5.79$, $p<.01$], but 19% fewer high risk callers who have ongoing suicide plans or attempts ($7/108$ vs. $27/108$) [$Z=-3.74$, $p<.01$], suggesting a decrease in suicide risk during the pandemic.

Table 1 shows that 4 out of 7 LIWC word categories could distinguish these two groups. As sadness and anxiety are the exclusive markers for depression and anxiety respectively, the COVID-19 group showed lesser depression and more anxiety than the pre-COVID-19 group. As depression and suicide are strongly associated (Hawton et al., 2013), the text analysis results also suggested a decrease in the suicide risk of the COVID-19. While for the pre-COVID-19group, suicide risk was positively correlated with the percentage of the depression words (r = 0.52) and negatively correlated with that of the anxiety words (r = −0.32), no correlation was found for the COVID-19 group.

Table 1.

Mean (percentage), standard deviation (percentage) and t-value of CLIWC word categories in COVID-19 group and pre-COVID-19 group.

	Depression/Anxiety	COVID-19 caller (n = 108)	Pre-COVID-19 caller (n = 108)	t (214)
		M (SD)	M (SD)
I	D, A	7.46(2.86)	8.61(2.91)	-2.94 **
Negative Emotion	D, A	3.79(1.14)	4.22(1.39)	-2.49 *
Positive Emotion	D, A	2.80(1.70)	2.44(1.36)	1.68
Sadness	D	0.31(0.25)	0.44(0.41)	-2.79 **
Discrepancy	D, A	3.98(1.19)	3.83(1.33)	0.85
Health	D, A	1.58(0.99)	1.47(1.35)	0.73
Anxiety	A	0.63(0.50)	0.42(0.30)	3.42 **

*: p < .05. **: p < .01.

To sum up, the COVID-19 group showed lower suicide risk and lower depression but higher anxiety than the pre-COVID-19 group did. However, readers should not interpret these results without caution. As reminded in Tondon (2021) that the findings about suicide rates during the COVID-19 pandemic depend on research scale quite a lot, our data may not be suitable to represent the countries other than Taiwan. In fact, Kahil et al. (2021) systematically reviewed the papers related to this issue among 2000 more papers and concluded that the association between pandemics and suicide remains poorly supported. Also, the literature review in Suchandra et al. (2021) indicated the inconsistent findings about the impact of the COVID-19 pandemic on suicide. Thus, future research with a cautious design and sufficient representativeness is needed to clarify this issue.

2. Conclusion

With the NSPH callers in Taiwan as our target, this study demonstrates how the callers’ transcripts can be analyzed to understand the change on suicide risk associated with the COVID-19 pandemic. Although a decline in suicide risk is found during the pandemic in this study, the constraints from the sample representativeness and the study design should not be ignored before a conclusion can be drawn. Despite the fact that the association between pandemics and suicide still needs future research to clarify, one worth noting contribution of this study is that we have demonstrated how LIWC can be applied to extracting the psychological features behind the words used by clients, that would effectively expand the scope of suicide research.

Funding

This study was funded by Ministry of Science and Technology, Taiwan (109-2420-H-004-022-).

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.