Suicidal ideation, distress, and related factors in a population of cancer patients treated in a general acute hospital

Participants were patients treated for cancer in a general acute hospital in Frankfurt, Germany (Markus-Krankenhaus) during a 6-month period in 2007 [20]. Included were all patients fulfilling the following criteria: minimum age of 18, sufficient command of German, no cognitive impairments, no pain or nausea, no medication influencing mental or cognitive abilities at the time of the interview, written agreement to participate in the study after obtaining informed consent. All patients had been informed by their physicians about their cancer diagnosis prior to obtaining informed consent.

All eligible cancer patients were approached by members of the department of psycho-oncology within the first 3 days of their stay and received extensive information about the aims, procedure, data collection, and possible risks and side effects of the study. Written informed consent was only obtained after sufficient time for deliberation (minimum 24 h) and answering possible questions to the patient’s satisfaction. Trained interviewers administered the psycho-oncological basic documentation short form (PO-Bado-SF) semi-structured interview and collected demographic data. Finally, participants were provided with the self-assessment instruments and an anonymous envelope in which they could be returned.

Statistical analyses were conducted using IBM SPSS 24 software package (IBM SPSS Inc., Chicago, Illinois, USA). All statistical tests were 2-sided and p-values < 0.05 were considered statistically significant. Data were descriptively analyzed calculating means, standard deviations, and absolute and relative frequencies. Bivariate associations were explored by either using cross tables or calculating Chi² tests as well as Phi- and Cramér’s V-coefficients, point-biserial-correlation-coefficients, or Kendall-Tau-b correlations. Effect sizes for the Cramér’s V-coefficient follow the convention by Cohen [26, 27] small effect: V = 0.1, medium effect: V = 0.3, strong effect: V = 0.5. Kendall-Tau-b (r_τ) effect sizes follow the convention of Kühnel and Krebs [28] with no association: 0.00 < r_τ < 0.05, small association: 0.05 < r_τ < 0.20, medium association: 0.20 < r_τ < 0.50, strong association: 0.50 < r_τ < 0.70, very strong association: r_τ > 0.70. Differences between groups were investigated with the t-test for metric and the Mann–Whitney-U-test categorical variables.

Overall, N = 478 patients were recruited, of those n = 226 (47%) had complete information in the instruments considered in this study. The average age was 59.8 years (SD = 12.4) and n = 142 (63%) were female. Nearly half (44.7%) of patients suffered from breast cancer, 15.5% had prostate or testicular cancer, 11.9% colon or rectum cancer, 8.8% stomach, esophageal or pancreatic cancer, other urological cancers had 6.2%, 5.8% had lung cancer, 4.0% other gynecological tumors, 0.4% hematological malignancies, and 2.7% other forms of cancer. For more information, see Table 1. The n = 266 patients who were included in the study were on average younger (t₍₄₇₆₎ = 4.49, p < 0.001).

The overall prevalence of SI was 15.0%. 11.9% of patients reported SI on several days over the last 2 weeks, 1.3% on more than half of the days, and 1.8% on almost every day.

Descriptive statistics for the PO-Bado-SF, QSC-R23, and PHQ-9 are shown in Table 2. Descriptive statistics for single items of all instruments can be found in table S1. Correlations between PHQ-9 item 9 and distress measures are shown in Table 3.

We found a medium to large correlation between the PO-Bado-SF and the QSC-R23 sum scores (r_τ = 0.46, p < 0.001).

When externally assessed with the PO-Bado-SF, the mean distress score was 7.9 (SD = 4.3). SI correlated weakly with externally assessed distress (r_τ = 0.19, p < 0.001). An indication for further psycho-oncological treatment (cutoff > 9) had 36.7% of patients. Subjectively, i.e., independently of the cut-off-value, the raters saw indications for further psycho-oncological treatment in 55.0% of patients. We found weak associations of SI with subjectively assessed indication for further psycho-oncological treatment (χ² = 15.33, V = 0.26, p = 0.002) as well as with having an indication by reaching the cutoff (χ² = 11.97, V = 0.23, p = 0.007). Furthermore, patients scoring above the PO-Bado-SF cutoff scored significantly higher in SI than patients below the cutoff (U = 6568.5, p = 0.031). See also tables S2 and S3 in the supplement.

The mean value of self-assessed distress with the QSC-R23 was 22.6 (SD = 18.6). There was a correlation of medium size between SI and self-assessed distress (r_τ = 0.31, p < 0.001). The cut-off value (> 34) was exceeded by 24.8% of patients. Thereby, those patients had an indication for further psycho-oncological treatment. An association of medium strength was found between SI and having an indication for further psycho-oncological treatment (χ² = 25.70, V = 0.337, p < 0.001). Here, also patients scoring above the cutoff differed significantly in SI from those scoring below the cutoff (U = 6083.5, p < 0.001).

The mean value of the PHQ-9 was 6.1 (SD = 5.1). We also found a correlation of SI with the remaining symptoms of a major depressive episode (MDE; PHQ-9 without item 9: r_τ = 0.32, p < 0.001). About a quarter (23.5%) of patients scored ten or more points in the PHQ-9 and thereby had symptoms sufficiently severe for at least a medium major depressive episode. Between SI and having symptoms indicative of an MDE, we found an association of medium strength (χ² = 42.26, V = 0.31, p < 0.001). Feeling bad about oneself was the item correlating most strongly with SI (r_τ = 0.35, p < 0.001), followed by feeling down, depressed, and hopeless (r_τ = 0.31, p < 0.001). Additionally, we found medium to strong correlations between feeling down, depressed, and hopeless and measures of distress (PO-Bado-SF: r_τ = 0.44, p < 0.001; QSC-R23: r_τ = 0.50, p < 0.001).

There were also associations between the individual distress factors and SI. The item of the PO-Bado-SF most strongly associated with SI was sadness (r_τ = 0.23, p < 0.001) followed by deficits in activities of daily life (r_τ = 0.18, p = 0.002). The QSC-R23 scale psycho-somatic afflictions correlated most strongly with SI (r_τ = 0.30, p < 0.001), followed by social restrictions (r_τ = 0.28, p < 0.001), and restrictions in daily life (r_τ = 0.26, p < 0.001).

Finally, we found associations between SI and demographic variables (table S3). Being in a steady relationship (χ² = 8.12, V = 0.190, p = 0.044), having children (χ² = 10.10, V = 0.211, p = 0.018), and the current functional status (χ² = 19.05, V = 0.230, p = 0.025) had medium effect strengths.

Our findings show that a sizable portion of cancer patients (15%) has had suicidal thoughts at least on several days over the last 2 weeks. This is comparable to the results of other studies that employed similar assessment methods in comparable populations. For example, Zhong et al. [29] who also looked at hospitalized patients and used a single-item assessment tool found that 15.3% of patients had suicidal thoughts in the past month. Likewise, Moreno-Montoya et al. [30] reported a prevalence of SI of 13.6% in hospitalized cancer patients. Similar percentages among different populations of cancer patients are reported in various studies [31, 32]. In comparison to the general population, not only are suicide rates higher in cancer patients, but also the prevalence of SI is elevated. For example, in Germany, the prevalence of SI in the general population is 8.0% [33] and 2.0% worldwide [34]. This underscores the fact that a cancer diagnosis constitutes a major risk factor for suicidality. Yet, a recent meta-analysis shows that the prevalence of SI in cancer patients is subject to wide variation [35]. The authors report prevalence rates ranging from 0.7to 46.7%. An older review reports an even bigger variety of SI ranging from 0.8 to 71.4% [36]. This wide span is likely due to population characteristics and variation in assessment methods [35, 36]. Another factor might be the transient nature of suicidality and SI making it more comparable to a state rather than a trait [5]. This is also exemplified by the fact that suicidal thoughts and behaviors can be observed during all stages of the disease trajectory in different populations [3, 4, 9, 37]. In our opinion, this highlights two crucial points. First, a better understanding of the interactions between SI and completed suicides in cancer patients is necessary. Up to now, there is little understanding of the timing of suicide risk and its associated factors and it remains unclear whether findings are consistent across different populations. A meta-analysis planned by Calati et al. [38] will address these issues. This also suggests that more research especially focusing on longitudinal designs is needed. Second, it highlights the need for health professionals to be aware of the fact that cancer patients are a risk population and that therefore the necessary precautions should be taken. A screening for suicidality would be one such measure. Yet, especially in acute settings, resources are scarce and the implementation of screening procedures is often met with resistance. For example, in Germany, certification as an oncological center or organ center by the German Cancer Society requires the establishment of a distress screening, which is often met with numerous barriers [39]. This makes the implementation of an additional screening seem unfeasible. Therefore, it should be further investigated whether established distress screenings are able to also identify patients at risk for suicidality. Lastly, education of health care workers concerning suicidality and associated risk factors in cancer patients is the essential basis for any prevention strategies [40] and is a desire expressed by health care workers themselves [41].

In our study, depending on the instrument, a quarter to a third of patients were clinically distressed (i.e., scoring above the respective cutoff). Clinically distressed patients agreed significantly more often with item 9 than non-distressed patients. This holds for all distress screening tools. Patients who scored above the cutoff (≥ 10) in the PHQ-9 and thus can be classified as showing symptoms of at least moderate major depressive episode [16, 24] also differed significantly from patients below the cutoff regarding their endorsement of item 9. This demonstrates the close association of clinically relevant psychic distress and SI also reported in other studies [10, 17, 42]. The more distress a patient experiences, the more frequent are suicidal thoughts. For example, in our study, all patients that reported having suicidal thoughts almost every day were classified as being clinically distressed in all distress screening instruments, except for one case in the FBK-R23. Among the different factors constituting psychosocial distress, hopelessness and loss of meaning are recurrent themes in many studies that investigated the connection between psychosocial distress and SI [30, 31]. We found not only a correlation between SI and hopelessness but also associations between being clinically relevant distressed and hopelessness, independently of the screening instrument. Interestingly, hopelessness as well as loss of meaning are also two constitutive factors of the demoralization syndrome [12, 43, 44], which according to recent research seems better able to capture reasons for distress in cancer patients than psychiatric disorders [10, 30, 45]. For example, mood and anxiety disorders seem to show similar prevalence rates in cancer patients (2.6–12.5%) as in the general population [46], whereas about 20% of cancer patients show clinically relevant symptoms of demoralization [13]. Demoralization is associated not only with psychosocial distress but also with SI and seems better able to explain distress in cancer patients than psychiatric disorders [10, 31, 45]. Furthermore, as Robinson et al. [44] have shown demoralization is negatively associated with patients will to live, making the demoralization construct uniquely suited to identify cancer patients at risk for suicide. This is corroborated by the fact that concerning SI, demoralization has explanatory power beyond that of mental disorders [45]. Fang et al. [10] investigated the role of demoralization in the interplay of distress and SI. They tested mediation models with either demoralization or depression as mediators and found a stronger effect size for demoralization as mediator. In clinical practice, it is often recognized that the description of distress in cancer patients in terms of mental disorders is highly problematic. Cancer patients often do not fulfill the required criteria for a diagnosis, because the diagnostic criteria were not developed to accommodate the specific situations of somatically ill patients. For example, in the case of anxiety disorders, many cancer patients report severe anxiety, and regarding the perceived existential danger in which patients are put by their diagnosis, these fears are of course not irrational. Therefore, a screening tool that is not only able to capture distress on a general level but would also be able to identify patients at risk for suicidality could be the first step in prevention. This is especially pertinent because demoralization predicts a breakdown of coping and thereby a loss of perceived self-efficacy, which, for example, is also an important assumption in the cry-of-pain-model [47].

Overall, in our study, different dimensions of distress either self- or externally assessed were associated with SI in cancer patients. Externally rated sadness and deficits in activities of daily life were most strongly associated with self-rated SI. This would suggest that those factors were most obvious to the raters and therefore warrant further investigation. Similarly, self-assessed psycho-somatic afflictions, social restrictions, and restrictions in daily life were most strongly associated with SI. This suggest that similar dimensions of distress, regardless which method of assessment was used, are indicative of SI. In order to educate health care workers to be more attentive not only to verbal but also non-verbal cues of distress, it seems important to identify especially pertinent cues and elaborate reliably methods of detection [1]. Finally, we also found associations between SI and being in a relationship, having children, and current functional status. It can be assumed that those indicate a protective factor, which is in line with what other studies have found [10, 48]. Being in a stable relationship is a source of social support and affords cancer patients with additional resources on which to rely in times of crisis. It is also possible that it helps in strengthening a sense of belonging and purpose, factors negatively affected by demoralization [43], and thereby has a preventative effect. Overall, research on preventative factors is still in its infancy and it would be recommendable to expand it.