|Home | About | Journals | Submit | Contact Us | Français|
To assess whether social supports (proxies) can detect the presence of suicide ideation in a clinical sample of depressed adults 50 years of age or older, and to additionally assess the potential impact of depression symptom severity on patient–proxy concordance in reports of patient suicide ideation.
Cross-sectional data were collected regarding Axis I diagnoses, severity of depressive symptoms, and suicide ideation in a clinical sample of 109 patients 50 years of age and older. Patients were administered study measures by trained interviewers. Patients’ social supports completed proxy measures of these same variables. We assessed concordance in self- and proxy-reported suicide ideation, employing global suicide ideation items derived from depression scales and more fine-grained suicide ideation items drawn from multi-item suicide ideation measures. We investigated patient–proxy concordance regarding the presence of patient suicide ideation.
Patients who endorsed suicide ideation and were concordantly seen by their social supports to be suicidal reported significantly greater depressive symptom severity than patients concordantly reported to be nonsuicidal. Patients’ social supports reported significantly less depressive symptom severity in patients who endorsed suicide ideation yet who did not appear to be suicidal to them.
Our findings suggest that family and friends can broadly ascertain the presence of suicide ideation in depressed middle-aged and older adults, yet in doing so may largely be responding to their broad perceptions of depressive symptom severity in patients and not specifically to the presence of suicidal thoughts.
Older adults have high rates of suicide in Canada and worldwide,1,2 necessitating preventive interventions for those at risk. Late-life suicide often occurs in the context of untreated depression.3 Although few older adults who die by suicide have seen a mental health specialist in the days and weeks before their deaths,4 as many as 50% to 75% see a primary health care provider in that time frame,5 creating an opportunity for clinical detection of suicide risk.6 Effective detection of suicide risk is vital to clinical suicide prevention initiatives3,6 and to safe clinical treatment with older adults.
Expression of suicide ideation is associated with risk for death by suicide,7,8 and is thus an important clinical risk indicator.9 Adults 55 years of age or older who presented to hospital following self-harm had a high risk for suicide in the following year; risk for suicide remained elevated during the next 15 years.10 Swedish researchers, reporting on a retrospective study8 of adults 65 years of age or older who died by suicide, indicated that nearly 75% were reported to have communicated a wish to die or suicide ideation to a family member or acquaintance in the year prior to suicide.
Although expression of suicide ideation is associated with risk for suicide,1 the absence of expressed suicide ideation is not evidence of the absence of suicide risk.11,12 Some at-risk patients deny suicide ideation to health care providers,8 impeding risk detection.13,14 In one study,15 only 11% of patients seen in primary care in the month prior to dying by suicide communicated suicidal intent. In another study,16 78% of patients receiving or recently discharged from mental health care services denied suicide ideation minutes to hours prior to taking their lives.16
Given that older adults tend to deny or underreport depressive symptoms,13,14 clinicians are advised to interview their social supports. Canadian national guidelines for assessment and intervention with older adults at risk for suicide identify social supports as potential proxies for collection of collateral information when assessing patient suicide risk.17 However, evidence is limited regarding the concordance of patient and proxy reports of suicide ideation,18,19 and little is known about the potential impact of patient depression symptom severity on patient–proxy concordance.
A Swedish study18 of depression symptom reporting among adults 90 years of age or older indicated that social supports identified significantly less suicide ideation than was revealed during patient examination by a physician. A US study19 among home care patients 65 years of age or older similarly indicated that social supports identified less suicide ideation than did clinical interviewers. Previous studies assessed suicide ideation with a single global suicide ideation item derived from structured diagnostic interview measures, potentially limiting detection of patient suicide ideation.
Although available data suggest that clinicians do a better job of detecting suicide ideation than do collateral informants,18 researchers have yet to explore self-proxy concordance of suicide ideation among older recipients of mental health services. This information may be clinically useful, given that older adults are often accompanied to health care appointments by social supports, and may help identify at-risk older adults who underreport suicide ideation. The purpose of our study was to assess the concordance in suicide ideation as self-reported by mental health patients 50 years of age or older and proxy-reported by their social supports, and to investigate whether self- and proxy reports of patient depression symptom severity are associated with concordance in reports of patient suicide ideation. We assessed concordance of patient–proxy reporting using a variable assessing presence of suicide ideation based on self-report and interviewer rating scales assessing suicide ideation to serve as a higher bounds estimate of the presence of patient suicide ideation. We hypothesized that concordance would be positively associated with patient depressive symptom severity, such that patients with greater depression symptom severity would be more likely to report suicide ideation and to be reported by proxies as being suicidal relative to those with lower levels of depression symptom severity.
We recruited a sample of adults 50 years of age or older from inpatient and outpatient psychiatric services associated with 3 teaching hospitals in Rochester, New York, including a community hospital, a tertiary care facility, and an academic medical centre.20 Trained interviewers assessed for the presence of mental disorders in patients with the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Axis I disorders.21 Researchers administered a set of interviewer-rated and self-reported depression and suicide ideation assessment tools to study patients. Patients were asked to nominate a family member, friend, colleague, or other close social support person to be contacted and invited to participate in a research interview assessing concordance in patient and proxy perceptions of patient mental health symptoms and diagnoses. The patients’ social supports were also asked to complete study measures regarding the patients’ current depression and suicide ideation. Patients and their social supports separately provided written informed consent to participate in our study and were interviewed separately by different researchers. Our study received research ethics approval from the University of Rochester Medical Center Research Subjects Review Board.
The SSI is a 19-item interviewer-rated measure of suicide ideation during the past week.22 The initial 5 items screen for suicide ideation: assessing the presence and intensity of the wish to live, the wish to die, comparison of reasons for living with reasons for dying, active thoughts of suicide, and more passive thoughts of letting life slip away. The remaining items assess the presence and severity of suicide ideation, presence of a suicide plan, deterrents to suicidal behaviour, preparations for a suicide attempt, and anticipation of suicide. Only patients endorsing either SSI item 4 (active suicide ideation) or item 5 (passive suicide ideation) were administered the remaining items. The SSI 5 initial screening items demonstrated acceptable internal consistency (Cronbach’s alpha) for our study for self- (α = 0.89) and proxy-reports of patient suicide ideation (α = 0.86). The SSI has demonstrated acceptable internal consistency with older adults (α = 0.88) and significant association with the Geriatric Suicide Ideation Scale, an older adult-specific measure of suicide ideation.12
The BDI-II is a 21-item self-report measure of cognitive, affective, and vegetative symptoms of depression during the past 2 weeks.23 The BDI-II contains a single item assessing presence and severity of suicide ideation, scored from 0 (“I don’t have any thoughts of killing myself”) through 3 (“I would kill myself if I had the chance”). This item assesses both suicidal thoughts and wishes. The full BDI-II demonstrated acceptable internal consistency for our study for self- (α = 0.94) and proxy-reported ratings of patient depression symptom severity (α = 0.95). The BDI-II has shown acceptable internal consistency with older adults (α = 0.86) and significant positive associations with measures of depression and anxiety, negative associations with measures of psychological well-being among community residing older adults,24 and sensitivity to clinical change among older primary care patients receiving treatment for generalized anxiety disorder.25
The HDRS is a commonly used, clinician-administered measure of depressive symptom severity during the past week. For the purpose of our study, we used a 24-item version of the HDRS, incorporating prompts and probes to standardize its administration.26 The HDRS contains a single item assessing the spectrum of suicide-related thoughts and actions, including presence of death ideation, suicide ideation, and (or) suicidal behaviour, and is scored from 0 (absent) through 4 (attempts at suicide). The full HDRS demonstrated acceptable internal consistency for our study for patient self- (α = 0.80) and proxy-reported depressive symptom severity (α = 0.80).
We assessed the presence of patient suicide ideation categorically, based on endorsement of either SSI active (item 4) or passive (item 5) suicide ideation items (categorized as a suicide ideator), compared with neither (nonsuicide ideator) and categorically scored BDI-II and HDRS suicide ideation items (0 = nonsuicide ideator and >0 = suicide ideator). We then collapsed all of the suicide ideation measures into a composite any suicide ideation variable to assess a higher bounds estimate of the presence, compared with absence, of patient suicide ideation, given differential levels of sensitivity and specificity to suicide ideation among the different measures and to approximate a best estimate of the presence of patient suicide ideation. Frequencies and percentages were computed for the prevalence of patient self- and proxy-reported suicide ideation.
We assessed concordance in self- and proxy reports of patient suicide ideation by computing percentage agreement and Cohen’s27 kappa statistic for the presence, compared with absence, of suicide ideation. To assess potential associations between depression symptom severity and suicide ideation, we created a 4-point concordance variable based on similarities and differences in self- and proxy-reported suicide ideation for the any suicide ideation variable. These were categorized as agreement–suicide ideation, in which both the patient and the proxy reported a presence of patient suicide ideation (categorized as yes/yes or Y/Y), disagreement–proxy denies, in which the patient endorsed suicide ideation and the proxy denied patient suicide ideation (yes/no or Y/N), disagreement–patient denies, in which the patient denied suicide ideation and the proxy endorsed patient suicide ideation (N/Y), and agreement–no suicide ideation, in which both the patient and the proxy denied the presence of patient suicide ideation (N/N). We explored associations between concordance and depression, as measured by total scores on the BDI-II and HDRS, excluding their respective suicide ideation items. Omnibus ANOVAs were computed to assess potential between-group differences; planned comparisons between the 4 groups were computed for significant omnibus F tests using Tukey’s Honestly Significant Difference Test. All tests were 2-tailed, with alpha set at 0.05.
The sample was derived from a larger study of late-life depression and suicide ideation in which 250 patients with suspected depression were recruited from a geriatric psychiatry hospital inpatient unit and an older adult mental health outpatient clinic.20 For our study we included only patients who were willing to identify a family member, friend, colleague, or other social support person who was available to complete proxy measures on their behalf. Collateral data were available for 109 of the larger study’s sample, 35 (32%) of whom were friends or caretakers of study patients, 29 (27%) spouses or significant others, 21 (19%) children, 12 (11%) siblings, 1 (1%) parent, 10 (9%) other relatives, and 1 (1%) other relation. We investigated the potential association between patient demographic and clinical variables for patients with (n = 109) or without (n = 87) a proxy social support. Only patient marital status differentiated the 2 groups; married patients were significantly more likely to have nominated a proxy social support (67%) than unmarried patients (50%; Pearson’s χ2 = 4.93, df = 1, P < 0.05). Demographic data for our study patients and their social supports are presented in Table 1. Diagnostic interviews indicated that 108 study patients had an active mood disorder and 1 had a remitted mood disorder. Sixty-nine patients met diagnostic criteria for multiple mental disorders. Three patients were diagnosed with a cognitive disorder, 4 with an additional comorbid mood disorder (for example, major depressive disorder and dysthymic disorder), 2 with a delusional disorder, 1 with a dissociative disorder, 1 with hypochondriasis, 1 with a somatoform disorder, 1 with anorexia nervosa, 36 with an anxiety disorder, and 68 with a substance misuse disorder.
Data on the prevalence and concordance of patient suicide ideation, as reported by the patients and their social supports, are presented in Table 2. Collapsing the study’s suicide ideation measures into a composite, any suicide ideation, variable yielded a higher bounds estimate of the presence of patient suicide ideation, as reported by 80% of patients (n = 87) and 73% of their social supports (n = 80).
We investigated self–proxy concordance in ratings of patient suicide ideation based on the patients’ and social supports’ categorical responses on the any suicide ideation variable. Eighty-four (77%) pairs of patients and their social supports were concordant regarding the presence (Y/Y, n = 71 pairs or 65%) or absence of patient suicide ideation (N/N, n = 13 pairs or 12%). Twenty-five (23%) pairs were nonconcordant, including 16, in which the patient endorsed suicide ideation but the social support denied detection of patient suicide ideation (Y/N, 15%), and 9, in which the patient denied suicide ideation but the social support endorsed detecting patient suicide ideation (N/Y, 8%).
Next, we tested the hypothesis that concordance in reports of patient suicide ideation would be positively associated with patient depressive symptom severity, given the strong association between depression and suicide ideation. These analyses employed patient- and proxy-rated BDI-II and HDRS totals, excluding their suicide ideation items (Table 3). Findings revealed significant omnibus effects for BDI-II (F = 9.00, df = 3,91, P < 0.001) and HDRS totals (F = 6.63, df = 3,105, P < 0.001) for patient self-reports. Post hoc tests revealed significant between-group differences for both depression scales between the 2 concordant groups. BDI-II totals self-reported by patients were significantly higher for the agreement–suicide ideation group (Y/Y; mean 27.3, SD 13.3, n = 64) than for the agreement–no suicide ideation group (N/N; mean 6.6, SD 5.3, n = 10). Patient-reported HDRS totals were also significantly higher for the agreement–suicide ideation group (Y/Y; mean 27.7, SD 7.7, n = 71) than for the agreement–no suicide ideation group (N/N; mean 18.7, SD 9.0, n = 13). An unexpected significant between-group difference emerged for BDI-II totals; patients in the disagreement–patient denies group (N/Y) endorsed significantly higher depressive symptom severity (mean 25.6, SD 15.9, n = 7) than did patients in the agreement–no suicide ideation group (N/N; mean 6.6, SD 5.3, n = 10).
Findings also revealed significant omnibus effects for patient BDI-II (F = 4.60, df = 3,74, P = 0.005) and HDRS totals (F = 10.73, df = 3,96, P < 0.001) as reported by their social supports. BDI-II totals were significantly higher for the agreement–suicide ideation group (Y/Y; mean 27.4, SD 13.8, n = 57) than for the disagreement–proxy denies group (Y/N; mean 11.4, SD 8.2, n = 9), only partly supporting our study hypothesis. HDRS totals were similarly significantly higher for the agreement–suicide ideation group (Y/Y; mean 27.9, SD 7.4, n = 65) than for the disagreement–proxy denies group (Y/N; mean 16.9, SD 8.5, n = 15).
We investigated the prevalence of concordance in patient suicide ideation based on self-reports of a clinical sample 50 years of age or older and the proxy reports of their social supports, and potential associations between patient–proxy concordance and reports of patient depressive symptom severity. Contrary to past findings that cast doubt on the usefulness of reports from social supports about older adult suicide ideation,18,19 our findings suggest that reports gathered independently from social supports matched patients’ own reports reasonably well, although the degree of concordance varied across study measures.
One of our study’s main findings is that patients who themselves report, and are reported by others, to be thinking of suicide endorse significantly greater depressive symptom severity than do those who are concordantly viewed as not contemplating suicide, supporting our study hypothesis. This finding was expected, given the strong associations among depression, both as a clinical diagnosis and as a symptom, and suicide ideation, suicidal behaviour, and death by suicide.1,3,9,11,18,28–32 However, not all depressed people are suicidal, and not all suicidal people are depressed. Hence the finding that patients who deny suicide ideation, but are reported by their social supports to be contemplating suicide (N/Y), endorse significantly greater depressive symptom severity (BDI-II) than patients concordantly rated as not suicidal (N/N) suggests that social supports may be attuned to patients’ negative mood state, yet might confound depressive symptom severity with presence of suicide ideation. An alternative interpretation is that some patients with severe depression may acknowledge suicide ideation to their social supports but deny it to professionals. Both interpretations may have merit. The finding that the social supports’ ratings of patient depression scores were lower for patients who endorse suicide ideation but are not reported by their supports to be suicidal (Y/N) than for patients who endorse suicide ideation and are reported to be contemplating suicide according to their supports (Y/Y) offers some support for the interpretation that proxies may be conflating depressive symptom severity with presence of suicide ideation. In these situations, the main difference between the patients is not in whether they endorse suicide ideation, but rather in whether their social supports believe that they are contemplating suicide. As such, it follows that social supports of patients contemplating suicide who do not see evidence of severe depressive symptomatology may fail to recognize that patients might be contemplating suicide. Research is needed to assess the effectiveness of educational and gatekeeper intervention programs aimed at teaching social supports to identify suicide risk among older adults. A recent guide33 has been developed for this purpose, as have various gatekeeper training programs, but these have yet to be evaluated regarding detection of suicide risk among older adults. Because concordance and detection of hidden suicide ideation is likely never to be perfect, prevention efforts must also focus on increasing self-disclosure of suicide ideation among older adults.
Our study findings also suggest a need for research that focuses on understanding and influencing the relational and cognitive processes that govern disclosure and detection among older adults and those in a position to help them. Knowledge gained in our study about the importance of depression in concordance of symptom reporting can inform future studies that examine how observations and intuitions about depressive symptoms and suicide risk are processed and acted on by concerned others.
Our study’s findings should be considered in the context of its limitations, which include exclusive reliance on data collected from patients receiving mental health care and from their social supports. Interviews were not concurrent. Findings might thus not generalize to community-residing and (or) primary care samples of middle-aged and older adults. The study sample included adults 50 years of age or older, and, as such, does not exclusively comprise an older adult group. Fifty years of age was selected as the minimum age for study inclusion, both given that it is the minimum age for acceptance into the geriatric mental health services from which patients were recruited and so as to allow for assessment of suicide risk factors across birth cohorts with elevated suicide risk.1 Data were collected by trained interviewers using validated assessment tools in the context of a clinical research study. Although multiple measures of suicide ideation were collected, none of these measures was developed or validated exclusively with an older adult sample. Future research is needed investigating patient–proxy concordance with older adult suicide ideation assessment scales. We did not use a gold standard measure of suicide ideation, recognizing that all assessment methods and measures at best assess the expression of suicide ideation, rather than its actual presence. A subtler point, and one that is not the main focus of our study, is that only the patient ultimately knows whether he or she has been contemplating suicide; and yet patients do not always directly acknowledge suicide ideation, even when present. Future researchers might consider exploring additional potential correlates of patient–proxy concordance, including personality, mental health treatment history, and history of suicidal behaviour. Research is needed assessing whether incorporating systematic collection of collateral source data on patient suicide ideation helps to improve clinical risk detection and treatment delivery with older patients at risk for suicide.
Our findings suggest that social supports can generally detect suicide ideation in clearly depressed middle-aged and older adults. General support is thus provided for collection of collateral-source information on suicide risk. Nevertheless, findings suggest that social supports are most effective at detecting patient suicide ideation when they detect greater depressive symptom severity, and not necessarily the presence of suicide ideation. Family, friends, and acquaintances of middle-aged and older adults might thus serve as useful collateral informants regarding the presence of suicide ideation among patients clearly displaying depressive symptoms and contemplating suicide. The ability of these potential proxies to effectively inform clinical care may decline when people at risk for suicide do not appear to them to be obviously depressed; additional research is needed addressing this issue. Clinicians should thus consider collecting patient and collateral reports of patient suicide ideation, and treat any such report as a potential indicator of elevated risk necessitating further clinical evaluation. However, they should not simply take collateral reports of the absence of patient suicide ideation at face value, but rather follow up directly with patients while sensitively attempting to elicit thoughts of suicide directly from them.17,34
Our paper was supported in part by a Canadian Institutes of Health Research New Investigator Award, an Early Researcher Award through the Ontario Ministry of Research and Innovation, and a Leonard F Salzman Research Award of the Department of Psychiatry, University of Rochester Medical Center to Dr Heisel. United States Public Health Service grants were received (R01MH064579 and 5K24MH072712) by Dr Duberstein, and (T32MH020061) by Dr Conwell (Dr Pisani received support from this grant). Conference support was generously provided by The University of Western Ontario Department of Psychiatry’s Finance Management Committee.
The authors express their appreciation to Anthony Beckman, Dr Madalina Chirieac, Nathan Franus, Dr Stephanie Gamble, Dr Jameson Hirsch, Dr J David Useda, Holly Wadkins, and Patrick Walsh.
An earlier version of this paper was presented at the Fourth Aeschi Conference on Suicide Prevention, Aeschi, Switzerland, March 2006, and at the Academic Research Day of The University of Western Ontario Department of Psychiatry, June 2006.