Suicidal ideation among outpatients attending a public mental health clinic in Norway: rates, treatment effect, and associated characteristics

Abstract

Suicidal ideation is frequently reported by outpatients in mental health care. The aim of this study was to describe the rates of suicidal ideation (SI) in a heterogenous outpatient sample, the effect of treatment as usual (TAU) on SI, and characteristics of patients who report SI. Participants were outpatients receiving TAU at a public mental health clinic. They completed measures assessing levels of depression, anxiety, functioning, and health-related quality of life before (N = 2475) and after treatment (n = 559). More than half of the sample (57%) reported SI before treatment. For participants with SI, 55% experienced a reduction in SI, 38% reported no change, while 7% had increased SI. The overall reduction in SI reflected a moderate to large effect size. SI at pre-treatment was associated with more symptoms of depression and anxiety, impaired functioning, lower health-related quality of life, male sex, lower age, and not being in a relationship. SI is common among psychiatric outpatients and is more prevalent among patients with higher symptom severity, males, and people who are young and single. Treatment as usual is associated with moderate to large improvements for SI. However, a large proportion of patients report unchanged or increased SI, which indicates that additional interventions may be needed.

Keywords:

• suicidal ideation
• outpatients
• treatment
• anxiety
• depression

Suicidal ideation (SI) is associated with an elevated risk for subsequent suicide and constitutes a central component in numerous suicide theories (Jobes & Joiner, 2019; Klonsky et al., 2018). SI generally refers to a range of wishes, contemplations, and preoccupations with death and suicide. Even though there are known associations between SI and suicide, discussions about SI should not be limited to the context of suicide. SI should also be regarded as a serious mental health problem in its own right as it is associated with considerable psychological pain and distress (Goldney et al., 2001).

A Norwegian study found that clinicians reported SI for 17.3% of their psychiatric outpatients (Ose et al., 2021). Studies from other countries have found rates ranging from 18 to 40% (Naidoo & Collings, 2017; Ongeri et al., 2018). One study has investigated the rate of SI in a general outpatient sample using self-report questionnaires, in which SI was found to be present in 72.8% of the patients (Vera-Varela et al., 2022). Patients are generally much more reluctant to disclose SI in a face-to-face setting (Kaplan et al., 1994). In the study by Vera-Varela et al. (2022), SI was first assessed in an interview setting and later that day by using a self-report questionnaire. A total of 19.3% of patients disclosed SI to their clinicians compared to 72.8% in the self-report questionnaire. This could indicate that although clinical guidelines urge clinicians to ask about SI as part of routine assessment, SI may be underreported, and self-report questionnaires may be an important supplement when screening for SI.

Previous research has also investigated the effect of therapeutic interventions on SI. In a review study investigating whether psychotherapy for adult depression also affects SI, Cuijpers et al. (2013) were able to identify only three studies meeting the inclusion criteria. The treatment interventions included mindfulness-based cognitive therapy, cognitive restructuring, existential humanistic cognitive therapy, and cognitive behavioral therapy. A small (g = 0.12) and non-significant treatment effect was found. The study discussed whether the samples, which did not specifically target patients with SI, could be a possible reason for the relatively low effect size found. Based on the limited statistical power and methodological limitations, the review concluded that there was insufficient research to draw any conclusions at the time.

A more recent study found a large treatment effect (d = 1.5) on SI for adult patients with depression after short-term manual-based treatment (Schneider et al., 2020). The treatment interventions were empirically supported treatment (ESTs), and included approaches such as cognitive-behavioral therapy, dialectical behavioral therapy, emotion-focused therapy, and interpersonal therapy. In a related meta-analysis comprising 11 RCT studies, the effect of treatment interventions on SI was compared to active controls for outpatients aged 12 to 19 (Kothgassner et al., 2020). They found a significant difference between the groups, with specific treatment interventions being more effective (d = 0.31) in reducing SI than for active controls receiving routine clinical care. Notwithstanding, a significant reduction of SI (d = 0.87) was found from pre- to post-treatment in active controls receiving routine clinical care.

Many studies are specific regarding intervention, patient group, or both. Previous research has established that there is not always a match between the treatment given in randomized trials and the treatment offered in clinical practice (Johnson et al., 2016). Thus, while such studies may show that a given intervention has a desirable effect, it is not certain that it provides a credible measure of the effect of the treatment offered in ordinary clinical practice. Further, studies often exclude patients with high SI (Brooks et al., 2021). No studies have investigated the effect of treatment as usual (TAU) on SI in a general heterogeneous adult psychiatric outpatient population.

The question of what characterizes individuals who report SI has been addressed in previous research. Characteristics found to be associated with SI are psychiatric disorders, with depression appearing to be the most common (Naidoo & Collings, 2017; Ose et al., 2021). Anxiety is also a significant, yet weak, predictor of SI (Bentley et al., 2016). In addition, it has previously been assumed that more women than men report SI, while more men commit suicide, which has given rise to the so-called gender paradox in suicidal behavior (Canetto & Sakinofsky, 1998). However, recent research investigating SI in a Norwegian outpatient sample revealed only a small association between SI and gender, in addition, being single and of younger age was also associated with SI (Ose et al., 2021).

The first aim of this study was to explore self-reported rates of SI in a heterogenous psychiatric outpatient sample. The second aim was to investigate changes in SI following TAU. The third aim was to explore characteristics (symptoms, functioning, health-related quality of life, and sociodemographic factors) associated with SI when starting treatment.

Material and methods

Participants and procedure

Participants were outpatients receiving treatment at Nidaros District Psychiatric Center (DPC) in Norway. It provides mental health services to a population of approximately 115 000. The sample consisted of 2475 patients, of whom 63% were female, and the average age was 30.2 years (SD = 10.49, range = 18-75). Patients were diagnosed with a range of psychiatric disorders (ICD-10), with depression (25.8%) as the most prevalent. Comorbidity was present in 17.6% of the sample. For other demographic and diagnostic characteristics, see Table 1. There were no exclusion criteria, but it should be noted that some patients with specific diagnoses, such as OCD and schizophrenia, were treated at other, specialized facilities.

Table 1 Demographic characteristics of the patients attending outpatient treatment.

	n (%)
Age (n = 2475)
<20	112 (4.5)
20–29	1405 (56.8)
30–39	533 (21.5)
40–49	238 (9.6)
50–59	141 (5.7)
>60	46 (1.9)
Sex (n = 2475)
Male	910 (36.8)
Female	1565 (63.2)
Sick leave (n = 2246)
Yes	725 (32.3)
No	1521 (67.7)
In a relationship (n = 2255)
Yes	977 (43.3)
No	1278 (56.7)
Diagnoses (n = 1510)
Depression	389 (25.8)
ADHD	309 (21.0)
PTSD	167 (11.1)
Personality disorders	160 (10.6)
Bipolar disorder	94 (6.2)
GAD	84 (5.6)
Social anxiety disorder	71 (4.7)
Autism	54 (3.6)
Somatization disorder	45 (3.0)
Eating disorders	33 (2.2)
Other	65 (4.2)

Note. ADHD = Attention deficit hyperactivity disorder, PTSD = Post-traumatic stress disorder, GAD = Generalized anxiety disorder.

In accordance with the Norwegian healthcare system, patients are typically referred to treatment at DPC by their general practitioner, but some may also have been referred by other healthcare professionals. Patients under the age of 18 are not offered treatment at regular DPC but are referred to child- and adolescent mental health services. In a survey by Gråwe et al. (2008), it was found that outpatient treatment at Norwegian DPCs (as described by clinicians) typically included a combination of supportive psychotherapy (48.8%), psychopharmacological treatment (28.2%), cognitive psychotherapy (22.3%), counseling (13.9%), psychodynamic psychotherapy (12.9%), other individual treatment (12.7%), interpersonal psychotherapy (8.7%), and crisis interventions (7.5%). The percentages add up to more than 100% as therapists could report multiple interventions. In 2016, adult outpatients in Norwegian mental health care had a mean of 11.3 consultations per patient.

Upon admission to treatment, patients were asked to complete computerized self-administered questionnaires to assess their symptoms and level of functioning. The patients were asked to repeat the questionnaires after treatment had ended. Patients were able to complete the assessments electronically at home. Of the 2475 that completed the pre-treatment assessments, 559 (22.60%) also completed the post-treatment assessment. The difference between the number of pre- and post-treatment assessments is likely due to the nature of the study design, as the data collection is a continuously ongoing process. Therefore, a large proportion of the participants were still receiving treatment at the time of data extraction and lacked post-assessment. Participants were grouped into four different samples. The total number of patients who completed the initial assessment is referred to as total sample (N = 2475). Patients who completed both assessments are referred to as completers (n = 559). To monitor changes in SI during the treatment period, patients from the completers sample were grouped into two sub-samples referred to as SI-sample (n = 296) and no-SI-sample (n = 263). The mean number of psychotherapy sessions for treatment completers was 6.5 (SD = 8.0, range = 0-48), and days between pre- and post-treatment assessment was 200.4 (SD = 139.9, range 6-664).

Data were collected electronically between February 2020 and March 2022. Information about the study and consent were also given electronically, and only patients who provided consent for their responses to be used in the research were included in the sample. Data collected from pre- and post-treatment assessments were matched with the individual patients’ mental health records to provide information about diagnosis, marital status, and sick leave. Approval from the Regional Committees for Medical and Health Research Ethics (REK) was obtained (reference number: REK 2019/31836). The National Center for Research Data has also approved the project (reference number: NSD2020/605327).

Measures

Item nine of the Patient Health Questionnaire 9 (PHQ-9; Kroenke et al., 2001) was used to assess SI. This item asks whether “thoughts that you would be better off dead or hurting yourself in some way” have been present within the last two weeks. The response options are “not at all” (0 points), “several days” (1 point), “more than half of the days” (2 points) and “nearly every day” (3 points). This item has previously demonstrated a good ability to detect SI in individuals and have high convergent validity with other measures of the SI construct (Kim et al., 2021), and could be a predictor of subsequent suicide attempts (Simon et al., 2013).

The Patient Health Questionnaire Anxiety and Depression Scale (PHQ-ADS; Kroenke et al., 2016) was used as a composite measure for anxiety and depression by summing the scores from the Patient Health Questionnaire-9 (PHQ-9; Kroenke et al., 2001) and the General Anxiety Disorder-7 (GAD-7; Spitzer et al., 2006). Example items include: “Little interest or pleasure in doing things” and “Feeling nervous, anxious, or on edge”. Scores can range from 0 to 48. Cut-offs at 10, 20 and 30 have been suggested to represent thresholds for mild, moderate, and severe levels of depression-anxiety symptoms (Kroenke et al., 2019). Since suicidal ideation was measured using item 9 of PHQ-9, this item was removed from the PHQ-ADS measure, giving a possible range in scores of 0 to 44. Cronbach’s alpha was (.89) good.

The Work and Social Adjustment Scale (WSAS; Mundt et al., 2002) was employed as a measure for impairment in functioning. WSAS consists of five items, rated on a nine-point scale (0-8). The items address impairment in professional life, home management, social leisure activities, private leisure activities and close relationships. Example items include: “Because of my problem my ability to work is impaired” and “Because of my problem, my ability to form and maintain close relationships with others, including those I live with, is impaired”. A WSAS sum score below 10 indicates low impairment and is associated with subclinical populations. Sum scores between 10 and 20 indicate moderate functional impairment, while sum scores above 20 indicate severe impairment (Mundt et al., 2002). Its reliability and validity among psychiatric outpatients are well-supported (Pedersen et al., 2017). Cronbach’s alpha was (.81) good.

The EuroQol Visual Analog Scale (EQ-VAS; Herdman et al., 2011) was used to provide a self-reported measure on current global health status. The EQ-VAS is a vertical scale, ranging from 0 to 100, with 100 indicating the best imaginable health and 0 indicating the worst imaginable health. The item asks: “How good or bad is your health today?” EQ-VAS has proven to be a valid measure for health-related quality of life (Cheng et al., 2021).

Statistical analyses

Mixed measures ANOVA along with calculations of Cohen’s d were administered to examine the treatment effect on the study variables for patients with and without SI. One logistic regression analysis was conducted to investigate the characteristics of patients with SI. To investigate the probability of attrition bias, participants with complete data and patients with pre-treatment data only, were compared using t-tests. The variables included in the attrition analyses were sick leave, in a relationship, gender, age, and pre-treatment scores from item 9 from PHQ-9, PHQ-ADS, WSAS and EQ-VAS. No significant differences were found between the two samples (p-values ranging from .110 to .851), indicating no clear attrition bias.

Results

Figure 1 presents distribution of scores on SI for both samples before and after treatment. There was an overall reduction in SI after treatment, with 43-47% (43% in total sample, 47% in the completers sample) reporting no SI before treatment, which increased to 61.2% post-treatment. For the more severe levels of SI (indicated by answering having SI often and nearly every day) there was a reduction from 19.7% at pre-treatment to 13.8% after treatment. For the SI sample, 55.1% experienced a reduction in SI, 37.5% had no change, and 7.4% had increased SI.

Figure 1 Rates of SI at pre- and post-treatment.

Note. Pre-treatment (total sample) represents pre-treatment scores from the total sample (N = 2475). Pre-treatment (completers sample) and post-treatment bars represent pre- and post-treatment scores from the sample that completed the PHQ-9 at both times of assessment (n = 559).

Table 2 presents results from the mixed measures ANOVA exploring changes in SI, symptoms, quality of life, and functioning from pre- to post-treatment for patients with and without SI. In addition to reporting lower general quality of life, the SI-group reported generally more symptoms, compared to the no SI-sample. There was a significant improvement from pre- to post-treatment on SI (F = 53.66, p < .001), WSAS (F = 55.75, p < .001), PHQ-ADS (F = 217.27, p < .001), and EQ-VAS (F = 91.25, p < .001), with effect sizes ranging from small to moderate-large. The treatment effect sizes were largest for the SI-sample, and the interaction effects were significant (see Table 2).

Table 2 Changes in symptoms and functioning for the total sample, SI-sample, and no SI-sample.

Variable	Pre-treatment				Post-treatment				Time * group
	Non-SI		SI		Non-SI		SI
	M	SD	M	SD	M	SD	M	SD	F	p
PHQ-9 Q9	0.00	0.00	1.57	(0.80)	0.17	0.47	0.93	0.95	156.93	<.001
WSAS	18.17	8.14	23.89	7.30	16.00	9.16	20.12	9.57	4.08	.044
PHQ-ADS	22.59	8.75	29.15	8.08	17.24	9.56	21.90	10.12	4.90	.027
EQ-VAS	54.02	18.71	45.22	18.13	60.45	19.46	55.09	20.32	4.09	.044

Note. Sample sizes: N = 289-296 (SI-sample) and N = 254-263 (NO SI-sample). PHQ-9 Q9 = The Patient Health Questionnaire, item 9 (suicidal ideation). WSAS = Work and Social Adjustment Scale. PHQ-ADS = Patient Health Questionnaire Anxiety and Depression Scale (not including item 9). EQ-VAS = EuroQol Visual Analog Scale. SI = suicidal ideation.

There was a significant difference in SI (item 9 of PHQ-9) over time. The effect was moderate for the no-SI group (d = −0.51), and moderate-large for the SI group (d = 0.73). The effect size was obviously negative for the no SI-sample, indicating a slight increase in SI for patients with no SI at pre-treatment.

There was a significant difference in WSAS over time. Changes in WSAS were small for both groups, but slightly larger for the SI-group (d = 0.44 vs. 0.25). The mean post-treatment scores on WSAS in the no SI-sample were in the 10-20 range (18.2 and 16.0), suggested to indicate moderate impairment, while the scores in the SI-group were above 20 (20.1), indicating moderate-severe impairment.

There was a significant difference in PHQ-ADS over time. The effect was moderate for the no SI-sample (d = 0.58). For the SI-sample, the effect size was moderate-large (d = 0.79), approaching large. Post-treatment score for the SI-sample was 22, indicative of moderate levels of depression/anxiety symptoms, while the no-SI group had a score of 17 (mild-moderate).

There was also a significant difference in EQ-VAS over time. The effect was small for the no-SI group (d = −0.34), and moderate for the SI group (d = −0.51). Mean scores ranged from 45 to 60, which is more than a standard deviation lower than the general Norwegian population.

Table 3 presents results from the logistic regression analysis. The overall model was found to be statistically significant (χ²(7) = 526.69, p < .001), with a Nagelkerke R-squared of .30. Results indicated that symptoms of depression/anxiety were strongest associated with SI, followed by more severe impairment of functioning, and lower quality of life. It was also found that demographic factors such as being male, of lower age, and not in a relationship were characteristics associated with reporting SI. There was no significant association between being on sick leave and SI.

Table 3 Characteristics associated with SI at pre-treatment.

Variable	B	SE	Wald’s X^2	df	p	Exp(B)
Female	−0.45	0.11	17.03	1	<.001	0.64
Age	−0.02	0.01	7.25	1	.007	0.99
In a relationship	−0.72	0.11	47.19	1	<.001	0.49
Sick leave	−0.04	0.12	0.14	1	.712	1.05
PHQ-ADS pre	0.09	0.01	132.85	1	<.001	1.09
WSAS pre	0.03	0.01	15.00	1	<.001	1.03
EQ VAS pre	−0.46	0.46	11.98	1	.001	0.99

Note. Dependent variable: Q9 PHQ-9 pre, dichotomized (0 = no SI, 1 = SI), N = 2083.

Dichotomized variables: female (0 = male, 1 = female), in a relationship (0 = no, 1 = yes) and sick Leave (0 = no, 1 = yes). PHQ-9 Q9 = The Patient Health Questionnaire, item 9. WSAS = Work and Social Adjustment Scale. PHQ-ADS = Patient Health Questionnaire Anxiety and Depression Scale. EQ-VAS = EuroQol Visual Analog Scale. Pre = pre-treatment.

Discussion

It was found that SI is common among patients receiving psychiatric outpatient treatment, with 57% of the participants reporting current SI. A decrease in SI during treatment was observed, along with decreased symptoms of anxiety and depression, impairment of functioning, and an increase in health-related quality of life. Treatment was associated with large effect sizes on all measures for patients who reported SI upon admission. Demographic variables such as male gender and not being in a relationship were associated with reporting SI, along with symptoms of depression and anxiety, impairment of functioning, and lower health-related quality of life.

The first aim of the study was to explore SI rates among outpatients with mental health problems. More than half (57%) of patients reported SI, which is lower than the 72.8% reported by Vera-Varela et al. (2022). Reasons for this disparity could stem from the way SI was operationalized and characteristics of the samples. The rates found in the current study are different from interview-based assessment of SI, which may be vulnerable to underreporting (e.g. 19.3% in the study by Vera-Varela et al., 2022).

The second aim was to explore the changes in self-reported SI following TAU. The study demonstrated a significant reduction in SI from baseline to post-treatment. A total of 55% of patients with SI at baseline experienced a reduction during treatment, in which a moderate-large effect (d = 0.73) was found for the SI-sample. The treatment effect on SI was smaller (d = 1.50, Schneider et al., 2020), similar (d = 0.87, Kothgassner et al., 2020), and bigger (g = 0.12, Cuijpers et al., 2013) than the effect sizes found in previous studies. The differences between studies could also be reflected in the effect on depressive symptoms. Treatments showing larger effects on depression also show better results on SI (Keilp et al., 2018).

Of the previous studies, the study by Kothgassner et al. (2020) in which the control group received TAU is most like this study. It is promising that the effect sizes were similar (0.79 and 0.87) despite differences in age. The results of the Kothgassner study and our results thus indicate that patients with SI may benefit from routine treatment, although the interpretation is limited by lack of information regarding what interventions patients received. In the study conducted by Schneider et al. (2020), the age of participants in the sample was more like this study, but differed in that the sample was limited to patients with depressive symptoms. The observed effect sizes may also reflect pre-treatment differences and regression towards the mean effects, as the sample in Schneider et al. (2020) had higher initial levels of depressive symptoms and SI.

The effect size found in the meta-analysis by Cuijpers et al. (2013) was markedly lower than the other studies. This difference may be partially attributed to differences in samples (e.g. patients undergoing hemodialysis). Moreover, the treatment interventions given were less extensive than other studies including ours.

Noteworthy, patients from the SI-sample still reported having considerable amount of SI post-treatment. A total of 38.8% reported to still have SI after treatment. In addition, 7.4% of patients from the SI-sample experienced an increase in SI during treatment. These findings could be related to the fluctuating nature of mood disorders, unsuccessful treatment, and unresolved mental health issues. The rate is considerably higher than the rate of 3.2-4.1% in the general Norwegian population (Knudsen et al., 2021).

The third aim of the study was to explore characteristics associated with SI. It was found that a higher score on the composite measure for symptoms of anxiety and depression was associated with reporting SI at baseline. Furthermore, significant associations were also found between SI and impairment of function and low health-related quality of life. The relationship between psychological symptoms and SI is previously established in the field of suicide research (Goldney et al., 2001), and the present findings provide further support for this connection. Nevertheless, the findings emphasize the importance of being aware of SI in psychiatric outpatient populations. The high prevalence of SI in the current sample stresses the importance of routinely screening for SI in psychiatric outpatients. Previous research indicates that non-disclosure of SI is common in psychiatric settings (Høyen et al., 2022). Therefore, self-report measures should systematically be used as a supplement to clinical interviews.

There was also an association between SI and being male. This finding is notable as it breaks with the traditional notion that women report more SI than men (Canetto & Sakinofsky, 1998), and because no significant gender differences were found in a similar population (Ose et al., 2021). One possible reason for this association could have to do with more females reporting having a partner (48% vs. 36%), which aligns with previous research showing increased risk of SI among unmarried participants (Nystedt et al., 2019). A negative association was found between SI and being in a relationship. The direction of causality cannot, however, be determined with certainty. On the one hand, the social support associated with being in a relationship may serve as a protective factor against SI (Endo et al., 2014). On the other hand, SI can be thought to affect the likelihood of being in a relationship.

There are some limitations to consider in this study. Firstly, SI was operationalized and assessed using one item from the PHQ-9. While it has shown to be a reliable screening tool for SI in clinical practice, the small scale on which each item is rated may limit sensitivity to change during the treatment period. Another limitation is the lack of information available concerning the treatment given. As we have no information regarding what interventions patients received, we do not know if, how, and to what extent SI was targeted in treatment. Furthermore, SI was assessed at two points in time. More frequent assessments would give a more precise picture of how SI changes over time and be less prone to effects of chance. Also, the lack of follow-up data inhibits the possibility of capturing delayed treatment effects.

In summary, it was found that SI is common in this heterogeneous sample of psychiatric outpatients. Furthermore, patients who report SI tend to report more psychiatric symptoms and impairment of function compared to patients who do not report SI. SI was also found to be somewhat more common among male patients, younger patients, and patients who are not in a relationship. The present study indicates that treatment as usual is efficient in reducing SI, as more than half of patients with SI at baseline experienced a decrease in SI during treatment. However, more studies are needed to better explore interventions that are beneficial for patients with SI.

Acknowledgments

The authors wish to thank all patients and therapists at Nidaros DPS who contributed to this study.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Anonymized data is available upon request to the corresponding author.