Table 3 Summary of study characteristics and findings for DHD (n = 14)
Study/Setting | Sample size and characteristics (Total sample size, % female, mean age) | Study design, patient population & recruitment procedure | Social connectedness measure | DHD measure | Summary of whether exposure/outcome measures were validated (✔ or X) | Association between exposure and outcome (✔ or X) | Statistical analytic approach | Main findings | Quality assessment and notable methodological limitations |
|---|---|---|---|---|---|---|---|---|---|
Loneliness (n = 1) | |||||||||
Stutzki et al. (2014) [77] Germany/ Switzerland | N = 66; Mean age 62 41% female Majority were Catholic, married, and had lower secondary education | Longitudinal study of patients with Amyotrophic Lateral Sclerosis (ALS) recruited from ALS clinics and their caregivers, surveyed at two points: T0 conducted as soon as the patients had been informed about the option of life-sustaining measures, and T1 conducted when the patients’ scores on the Amyotrophic Lateral Sclerosis Functional Rating Scale (ALSFRS) had deteriorated by > = 5 points (but not later than 15 months after baseline; average 13.2 months) | Loneliness measured using a question developed for the purpose of this study, capturing feelings of loneliness using a 10-point rating scale (unvalidated measure) | Desire for hastened death (DHD) measured using a 10-point Likert-type scale asking how strong the patient’s current wish was to hasten their death (face validity established only) Also captured descriptive measure of caregivers’ views on whether they could imagine helping the patient to hasten death by means of suicide assistance or euthanasia (yes/no) (face validity established only) | Exposure X DHD outcome X | ✔ | Generalised linear mixed regression models with no adjustment for confounders | Loneliness was significantly associated with desire to hasten death (OR = 1.20, 95% CI = 1.02–1.38; p = .021), and text mentioned that this was the case at both time points | High risk of bias some recruitment via caregivers, no adjustment for confounders or for multiple testing, unvalidated measures |
Satisfaction with social relationships (n = 1) | |||||||||
Cheung et al. 2020 [78] (New Zealand) NB: study conducted before EAS became legal in NZ | N = 771 palliative care patients; mean age 76.0 years (SD 11.6; range 20–100). About half (50.1%) of the sample were female, and most (87.0%) were European (Maori 8.7%, Asian 1.7%, Pacific people 1.7%, and other 0.9%) | Cross-sectional study of palliative care patients who had received a Resident Assessment Instrument for Palliative Care (RAI-Palliative Care) needs assessment tool anywhere in New Zealand between January 1 and December 31, 2018 | Satisfaction with family relationships rated by researchers using multiple sources of information (e.g., referral note, face-to-face interview, observation, discussion with family, carers, or health professionals) to rate degree to which patient had a strong and supportive relationship with family (unvalidated measure) | Desire to hasten death (DHD) captured using responses to an item on the RAI-Palliative Care questionnaire exploring the person’s ‘wish of wanting to die now’, with responses coded as Yes/No/Unable to determine, as derived from interview with the patient and family about advanced directives and end-of-life wishes (9.3% of the sample responded yes, 59.8% reported no, and for 30.9% assessors were unable to determine a response) (unvalidated measure) | Exposure X DHD outcome X | X | Bivariate associations of individual characteristic with wanting to died now; only performed logistic regression on those variables that were significant on chi-square testing | No association between measure of strong and supportive relationship with family and wanting to die now (on Chi2 testing; not entered into multivariable model) | High risk of bias unvalidated measure of satisfaction with family relationships relying on multiple sources; no adjustment for multiple testing |
Social support (n = 24) | |||||||||
Breitbart et al. (2000) [22] USA (New York) | N = 92; Mean age 66; 60% female; majority were Catholic, separated and had more than a high school education; mixed ethnicities | Cross-sectional survey of patients with terminal cancer in a palliative care hospital providing responses via interviews | Social support measured using the Duke University of North Carolina Functional Support Questionnaire (FSSQ) (validated measure) | Desire for hastened death (DHD) measured using the Schedule of Attitudes Towards Hastened Death (SAHD) (validated measure) | Exposure ✔ DHD outcome ✔ | X | Correlational analysis; stepwise multiple regression analysis to identify variables predicting DHD (multivariable analysis) | Social support was not correlated with desire for hastened death. The final multiple regression model was significantly associated with DHD (overall F = 18.79; p = < 0.001) with the inclusion of social support (partial F = 4.35; p = 0.05; note borderline significance), as well as hopelessness, depression, and overall physical functioning There was no significant association between perceived quality of social support and DHD (p = 0.64) | Low risk of bias |
Chochinov et al. (1995) [79] Canada | N = 199; mean age 71; 52% female;majority were Protestant, married, living with family and friends, and had less than a high school education | Cross-sectional survey of patients with terminal illness in hospital palliative care units, using diagnostic interviews and self-reported scales (mainly administered orally) | Perceived social support (quantity and perceived quality) measured by asking patients to indicate the number of family members and friends they have weekly contact with and rate the supportiveness (perceived support score) of these contacts out of 100 using a visual analogue scale (unvalidated measures) Patients also rated the perceived supportiveness of the nursing staff out of 100 (unvalidated measure) (unvalidated measures) | Desire for hastened death (DHD) measured using a set of questions created by the authors for this study, now termed the Desire for Death Rating Scale (DDRS), and since used widely as a set of standardised questions, including by other authors represented in this review, but not formally validated) (unvalidated measure) | Exposure X DHD outcome X | X | Correlations followed by t-tests and chi-square tests for those variables shown to be correlated with DHD; then used a stepwise multiple logistic regression procedure to identify the conjoint predictive value of the individual variables found to be associated with DHD (multivariable analysis) | No correlation of quantity of family and friend support with DHD, or of perceived quality of family and friend support with DHD (reported to be of marginal significance but p = 0.06) t-tests showed no association of family weekly contact with DHD, or of friends weekly contact with DHD, but a significant association of quality of family support with DHD (p = 0.01) and no association of quality of friend support with DHD t = -.25; p-value = 0.8 (family weekly contact) t = .32; p-value = 0.75 (friends weekly contact) t = 2.57; p-value = 0 .01 (family support score) t = 0.92; p-value = 0.36 (friends support score) Three variables reported to be significant correlates of the DHD (depression, pain, and quality of family support) were entered into the multiple logistic regression model, but only depression was a significant predictor of desire for death, and was significantly collinear with family support (r = -0.25, N = 196, p < O.OO1). Family support did not make a unique contribution to the model | Medium risk of bias used unvalidated measures of exposure and outcome, no correction for multiple testing |
Ganzini et al. (2006) [57] USA (Oregon) (also see Table 1) | N = 161; Mean age 62; 27% female; majority were white and married; mean of 13.5 years in education | Longitudinal study of patients with advanced cancer recruited from oncology clinics in Oregon (where physician assisted death has been legal since 1997), interviewed at baseline in 1998 and followed-up every three months for up to two years | Social support measured using the Duke University of North Carolina Functional Support Questionnaire (FSSQ) (validated measure) | Desire for hastened death (DHD) measured using Likert-style scale rating i) the degree to which they had considered requesting a legal lethal prescription in the previous two weeks (main outcome), ii) their desire for death to come sooner in the previous two weeks, used as an ordinal scale (unvalidated measures) Insufficient power to investigate the following outcome: Objective request for EAS measured as follows; those who indicated they might consider or were planning to request PAS were followed up every three months for up to two years; once referred to hospice, patients were followed monthly, as tolerated; after the patient’s death, each oncologist was asked if the patient had initiated discussion about PAS and any details of discussion (objective measures based on routine records) | Exposure ✔ DHD outcome X | X (association between increasing social support and increasing interest in obtaining a lethal prescription, but likely to be a spurious association in view of small sample size) | Bivariate models to estimate associations at baseline using a proportional odds regression model For 42/161 patients followed, 9 patient factors were regressed against whether interest in obtaining a lethal prescription (ordinal outcome) changed over time, using a random effects model | Social support was not associated with likelihood of considering requesting a lethal prescription at baseline (OR = 0.96, 95% CI = 0.92–1.00; p = .06) in a bivariate model; only three variables (hopelessness, importance of religion, and quality of life) were entered into the final multivariable model For the 42 subjects followed up longitudinally, increasing social support was associated with increasing interest in obtaining a lethal prescription (coefficient = -0.119; SE = 0.037; p = 0.002), as was declining functional status, increasing depression, increasing hopelessness, increasing sense of burden to family, increasing degree to which poor health limited quality of life, increasing suffering Only 2 patients requested PAS so formal statistical analysis was not possible | High risk of bias small sample size for longitudinal analysis; association with social support not adjusted for potential confounders; findings on social support in longitudinal analysis seem anomalous particularly in the context of other variables identified reflecting negative experiences |
Kelly et al. (2003) [80] Australia | N = 256; Mean age 66 48% female; majority were Protestant, married, and living with their spouse | Cross-sectional survey of patients with terminal cancer (completed with the assistance of a carer if required), recruited from hospices and palliative care services in Brisbane from 1998–2001 | Social support measured using the Social Support Questionnaire, 12-item measure capturing both the total number of social supports, and satisfaction with social support (validated measure) Perceptions of family interaction using the Family Relationships Index (FRI) to capture family cohesion (validated measure) | Desire for hastened death (DHD) measured using a modified version of the Desire for Death Rating Scale (DDRS), a set of standardised questions developed by Chochinov et al. 1995 [79] (unvalidated measure) In the current study the authors substituted one item on whether they had discussed this wish to die with two items in order to specify whether it was with family/friends or with a health professional | Exposure✔ DHD outcome X | ✔ | Discriminant function analysis used to predict membership of the group expressing DHD | Discriminant function analysis identified that the following variables were significantly associated with DHD: lower family cohesion scores (on FRI), lower number of social supports, less satisfaction with social supports Standardised correlation coefficients were: -0.29 (family cohesion) -0.13 (number of social supports), -0.10 (satisfaction with social supports) Within-group correlation coefficients were: -0.32 (family cohesion)-0.21 (number of social supports), -0.40 (satisfaction with social supports) Other variables also showing a significant association included: higher depression scores, hospice treatment setting, greater perceived burden on others, higher anxiety, and higher physical symptom scores | Low risk of bias some risk of social desirability bias where carers had assisted patients |
N = 93; Mean age 59; 58% female; majority were married; mean of 10.98 years in education | Longitudinal study of patients with Amyotrophic Lateral Sclerosis (ALS) recruited at an outpatient neurology clinic, interviewed at 3 timepoints: T1 baseline interview, T2 6 months, T3 12 months | Perceived social support measured using the emotional scale of the 14-item German version of the Social Support Questionnaire (F-SozU K-14) (validated for a German sample) | Desire for hastened death (DHD) measured using the Schedule of Attitudes towards Hastened Death (SAHD) (validated measure) | Exposure ✔ DHD outcome ✔ | X | Repeated measures ANOVAs were conducted for comparisons of all measures at each of the three timepoints and for both outcomes (DHD and attitudes to EAS); Kruskal–Wallis ANOVA to estimate associations of six factors and attitudes to life-prolonging treatments; multivariate logistic regression to examine which variables predicted desire for hastened death, mutually adjusted for quality of life, depression, feeling of being a burden, physical function, age and perceived social support | No association of perceived social support with desire for hastened death (but no test statistics reported) | Medium risk of bias unvalidated measures of attitudes to EAS | |
O’Mahony et al. (2005) [23] USA (New York) (same sample as O’Mahony et al. (2010) below) | N = 64; mean age 54; 52% female; majority were married and had a college education; ethnicity was mixed | Longitudinal (prospective observational) study of patients with terminal cancer recruited from cancer hospitals and palliative care services, T0 baseline and T1 four weeks later (± one week) | Perceived social support measured using the Bottomley Cancer Social Support Scale (BCSSS) at baseline and follow-up; a 9-item, cancer-specific, social support scale assessing perceived adequacy of social support in patients with cancer (validated measure) also measured living alone at baseline (objective measure) | Desire for hastened death (DHD) measured using a modified version of the Desire for Death Rating Scale (DDRS), a set of standardised questions developed by Chochinov et al. 1995) [79] (unvalidated measure) In the current study the authors agreed on their own cut-offs | Exposure ✔ DHD outcome X | ✔ for perceived social support ✔ for living alone | Linear regression was planned to measure the association between variables and desire for death, but social support was not entered into the model; only chi-square tests and correlation analyses were reported for social support variable | Low social support moderately correlated with desire to hasten death at baseline (r = 0.38, p < .01); text indicated that social support scores were higher for those without DHD, and that social support score categories (decreasing, stable, increasing) differed significantly by DHD trajectory but these chi-square tests for changes in social support were not interpretable as direction not stated; DHD scores increased significantly over follow-up (p = 0.03) but social support scores did not change significantly (p-0.051), although the association was not formally tested DHD at both baseline and follow-up was associated with living alone (chi-square = 5.98, P = < 0.02) | Medium risk of bias Only 53 patients completed the social support measure at baseline; unvalidated measure of DHD |
O’Mahony et al. (2010) [81] USA (New York) (same sample as O’Mahony et al. (2005) above, but linked to mortality data) | N = 64; Mean age 54;52% female | Longitudinal (prospective observational) study of patients with terminal cancer recruited from cancer hospitals and palliative care services, linked to mortality data for survival analysis (analysed data from above study by O’Mahony et al. (2005) but linked data to mortality outcomes) | Perceived social support measured using the Bottomley Cancer Social Support Scale (BCSSS) at baseline and at 4 week follow-up; a nine-item, cancer-specific, social support scale assessing perceived adequacy of social support in patients with cancer (validated measure) | Desire for hastened death (DHD) measured at baseline and at 4 week follow-up using a modified version of the Desire for Death Rating Scale (DDRS), a set of standardised questions developed by Chochinov et al. 1995 [79] (unvalidated measure) In the current study the authors agreed on their own cut-offs | Exposure ✔ DHD outcome X | ✔ | Generalized linear model with DHD as dependent variable and the other variables as predictors (multivariable analysis) | Social support was associated with DHD in adjusted (but not unadjusted) analyses Unadjusted effect size r = 0.25; p = 0.068 Adjusted effect size B = 0.496; p = 0.004 Over follow-up, mean DHD was not significantly lower for persons with improvement in perceived social support versus worsened social support (1.13 vs. 0.77, p = .39) In a model including depression, pain, social support, physical functioning, and existential well-being, two variables (physical functioning and existential well-being) were associated with lower group DHD. When an interaction term was included in this model between social support and physical functioning, lower perceived social support was significantly associated with DHD (β = 0.496, p = .004), meaning that the relationship between perceived social support and DHD was related to physical performance status [lower social support was also associated with shorter survival time, which was the primary research question] | Medium risk of bias Small sample size (as above); unvalidated measure of DHD |
Rodin et al. (2007) [82] Canada (Toronto) | N = 326; median age 61.9; 43% female | Cross-sectional study of patients with metastatic cancer recruited from consecutive patients attending their outpatient medical and/or radiation oncology clinic appointments with a treating oncologist at a network of acute care cancer centres in Toronto | Social support measured using the 20-item MOS-SSS, with items capturing 1) emotional/informational support; 2) tangible support; 3) affectionate support; 4) positive social interactions; and 5) global social support (validated measure) | Desire for hastened death (DHD) captured using the Schedule of Attitudes toward Hastened Death (SAHD) (validated measure) | Exposure ✔ DHD outcome ✔ | X | Stepwise, backward elimination regressions analysis was used to build a multivariate model using socio- demographic variables, physical and illness-related factors, and psychosocial and psychological variables Repeated in the subset of 251 patients who had died by the time of the analysis | Once all variables were entered into the model, social support did not contribute to an association with DHD in either the full sample or the subset who had died. Hopelessness, depression and poor physical functioning were the main predictors (but not physical functioning for the subset model) Overall, 79% reported good social support | Low risk of bias |
Rosenfeld et al. (2000) [83] USA (New York State) | N = 92; mean age 66; 60% female; majority Catholic; mean years of education was 12.7 years; mixed ethnicity | Cross-sectional interview study of patients with terminal cancer recruited from a palliative care hospital | Social support measured using the Duke University of North Carolina Functional Support Questionnaire (FSSQ) (validated measure) | Desire for hastened death (DHD) captured using the Schedule of Attitudes toward Hastened Death (SAHD) (validated measure) Also measured Desire for Death Rating Scale (DDRS) (but only for validation of SAHD) | Exposure ✔ DHD outcome ✔ | X | Spearman correlation coefficients | No significant correlation between social support and wish to hasten death (r = -0.06; non-significant) | Low risk of bias |
Rosenfeld et al. (2006) [84] USA (New York State) | N = 372; mean age 44; 25% female; majority were Catholic and had less than a high school education; mixed ethnicity. Religiosity was measured but not reported | Cross-sectional study of patients with advanced AIDS recruited from nursing facilities and medical centres in New York City | Social support measured using the Duke University of North Carolina Functional Support Questionnaire (FSSQ) (validated measure) | Desire for hastened death (DHD) captured using two measures; Schedule of Attitudes toward Hastened Death (SAHD) (validated measure) Desire for Death Rating Scale (DDRS) (unvalidated measure) Scores on each measure were highly correlated but it was unclear which measure was used in the main analyses to create a binary measure of low versus high DHD | Exposure ✔ DHD outcomes ✔ and X | X | Correlations to identify predictors of DHD; variables identified as significant on univariate analysis were entered into a stepwise logistic regression model to identify the most parsimonious set of predictors (multivariable analysis) | Social support was significantly correlated with DHD ( r = –0.26 (p = 0.0001) but social support did not make a unique contribution to a model testing the association between a range of independent variables (sociodemographic and clinical) and DHD once depression and hopelessness taken into account (only these two variables provided significant unique contributions) | Low risk of bias |
Rosenfeld et al. (2014) [85] USA (New York State) | N = 128; mean age 66; 52% female; majority were Catholic; mean years of education were 13.3 years; mixed ethnicity | Longitudinal study of patients with terminal cancer recruited shortly after admission to a palliative care hospital, questionnaire delivered at 2 time-points: T1 = baseline, T2 = 2–4 weeks later (originally a third data collection point was planned during this 42 month study but data were combined with T2 due to high attrition as many patients were too ill to complete the second or third assessment) | Social support measured using the Duke University of North Carolina Functional Support Questionnaire (FSSQ) (validated measure) | Desire for hastened death (DHD) measured using the Schedule of Attitudes Towards Hastened Death (SAHD) (validated measure) | Exposure ✔ DHD outcome ✔ | X | Univariate analyses (ANOVA) to describe association between social support and predictors of low versus high DHD at baseline; variables found to be significant were entered into a stepwise logistic regression model to predict membership of four DHD trajectories (low DHD; rising DHD; falling DHD; high DHD) (multivariable analysis) | No association between mean social support score and the four trajectories of DHD (F = 1.12; p = 0.34) so not entered into final model | Low risk of bias |
Schroepfer (2008) [62] USA (Michigan) | N = 96; mean age 74; 56% female; majority were Protestant, married, and had low religiosity; mean years in education was 12.1; mixed ethnicity | Cross-sectional interview study of terminally ill patients recruited purposively from palliative care hospitals and outpatient clinics | Social support measured using a binary measure of perceived support derived from discussion with patients (unvalidated measure) | Desire for hastened death (DHD) captured using a binary measure derived from asking if patients were considering hastening their death “Have you ever given serious thought to hastening the end of your life in any way?” (unvalidated measure) | Exposure X DHD outcome X | ✔ | Bivariate associations between DHD and predictors; Logistic regression, adjusted for years of education, religiosity, depression, pain intensity, marital status, parental status, direct social control (relating to support from caregivers to safeguard health), and social support (multivariable analysis) | Bivariate associations showed that conflictual social support was a significant predictor of DHD (p = < 0.001). In a fully adjusted model conflictual social support was significantly associated with DHD (adjusted OR = 25.46; p-value < 0.05) | High risk of bias purposive rather than representative sampling; unvalidated measures |