Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Aug 1.
Published in final edited form as: J Affect Disord. 2023 Apr 20;334:43–49. doi: 10.1016/j.jad.2023.04.038

A 50-state survey study of thoughts of suicide and social isolation among older adults in the United States

Nili Solomonov 1, Jon Green 2, Alexi Quintana 2, Jennifer Lin 3, Katherine Ognyanova 5, Mauricio Santillana 6,7, James Druckman 3, Matthew A Baum 8, David Lazer 3, Faith M Gunning 1, Roy H Perlis 4,6,8
PMCID: PMC10751855  NIHMSID: NIHMS1898625  PMID: 37086804

Abstract

Background:

We aimed to characterize the prevalence of social disconnection and thoughts of suicide among older adults in the United States, and examine the association between them in a large naturalistic study.

Methods:

We analyzed data from 6 waves of a fifty-state non-probability survey among US adults conducted between February and December 2021. The internet-based survey collected the PHQ-9, as well as multiple measures of social connectedness. We applied multiple logistic regression to analyze the association between presence of thoughts of suicide and social disconnection. Exploratory analysis examined heterogeneity of effects across sociodemographic groups.

Results:

Of 16,164 survey respondents age 65 and older, mean age was 70.9 (SD 5.0); the cohort was 61.4% female and 29.6% male; 2.0% Asian, 6.7% Black, 2.2% Hispanic, and 86.8% White. A total of 1144 (7.1%) reported thoughts of suicide at least several days in the prior 2 week period. In models adjusted for sociodemographic features, households with 3 or more additional members (adjusted OR 1.73, 95% CI 1.28–2.33) and lack of social supports, particularly emotional supports (adjusted OR 2.60, 95% CI 2.09 – 3.23), were independently associated with greater likelihood of reporting such thoughts, as was greater reported loneliness (adjusted OR 1.75, 95% CI 1.64 −1.87). The effects of emotional support varied significantly across sociodemographic groups.

Conclusions:

Thoughts of suicide are common among older adults in the US, and associated with lack of social support, but not with living alone.

Keywords: suicide, suicidality, survey, cohort, older adults, geriatric psychiatry, social network, social support, loneliness

Older adults represent a particularly high-risk group for suicide in the US and internationally: the 2016 Global Burden of Disease study found that age-standardized mortality rates from suicide were greatest in this group worldwide, regardless of region and gender (“Global, Regional, and National Burden of Suicide Mortality 1990 to 2016,” 2019). Rates of suicide among older men in particular approach those of young men, the highest-risk group (FastStats, 2022). Older adults are substantially more likely to die if they make a suicide attempt, consistent with the understanding that such attempts may involve more planning and more lethal means, on average(Cai et al., 2022; Oh et al., 2014). Yet there are few interventions specifically targeting suicidality – i.e., suicidal thoughts and behaviors - and even fewer in this population (Kiosses et al., 2018) (for a review, see Van Orden (Van Orden & Deming, 2018)), although psychosocial treatment for late-life depression has shown evidence of benefit (Alexopoulos et al., 2021).

Social disconnection is characterized by social isolation and lack of emotional social support(The Major Health Implications of Social Connection - Julianne Holt-Lunstad, 2021, n.d.). Disconnection increases with age (Kotwal, Cenzer, et al., 2021). It occurs at its highest levels in mid- and late life (Kotwal, Cenzer, et al., 2021), and predicts increased risk of thoughts of suicide, suicidal behavior, and suicide (Fässberg et al., 2012). Older suicide attempters have fewer closer friends and are less likely to speak with their children or engage in volunteer activities (Szanto et al., 2012).

The COVID-19 pandemic has placed unprecedented stresses on social networks, leading to increased loneliness and social isolation (Kotwal, Holt-Lunstad, et al., 2021) (van Tilburg et al., 2021). Older adults, at higher risk of COVID-19 infection and serious illness, were disproportionally impacted by social distancing guidelines(Hwang et al., n.d.). The closure of community and senior centers and the transition of social services to telehealth intensified age-related access disparities for elders who have low technological literacy and have limited social networks. These circumstances provide a unique opportunity to investigate the relationship between social disconnection and thoughts of suicide in late life while accounting for these stressors over time. Clarifying these mechanisms could facilitate development of novel interventions, or better targeting of existing interventions.

While there has been prior investigation of thoughts of suicide and social factors, most has focused on clinical trial populations (Kiosses et al., 2017) or treatment-seeking populations (Barak-Corren et al., 2020, p.; McCoy et al., 2016, 2019), which may not generalize to broader more diverse older adult population. Thus, beyond a lack of understanding of mechanisms, an additional gap is the lack of estimates of the effects of social disconnection in more generalizable cohorts.

To address these gaps, we used data from a large and nationally-representative, 50-state survey conducted over multiple waves, during the COVID-19 pandemic, between February 2021 and December 2021. We aimed to investigate the prevalence of social disconnection among older adults and estimate the extent to which it correlates with thoughts of suicide in a naturalistic setting.

Method

Study Design

We examined data from a multi-university consortium, the COVID States Project (covidstates.org), that has surveyed approximately 20,000 United States adults every four to six weeks since April 2020 on a range of topics related to COVID attitudes and behaviors, and more broadly examined aspects of social connectedness, including social interaction, social support, and loneliness. The surveys apply representative quotas and nonprobability sampling (Kennedy & Caumont, n.d.) to approximate the state-level distribution of age, gender, and race and ethnicity for each of the 50 states as well as the District of Columbia, a lower-cost approach to large-scale sampling that has been validated in similar contexts (Coppock & McClellan, 2019; Berinsky et al., 2012). For the present study, we drew on data from the six waves conducted between February and December 2021 that included consistent assessment of social connectedness; for individuals participating in more than one wave, the index survey was included. All sociodemographic variables were collected by self-report; race and ethnicity were drawn from 5 United States Census categories to confirm representativeness of the US population. The study design was determined to be exempt by the Institutional Review Board of Harvard University; survey participants signed consent online prior to survey access. We report results in accordance with AAPOR guidelines (Survey Disclosure Checklist - AAPOR, n.d.).

Outcomes and assessments

Survey participants completed the PHQ-9, a well-validated measure of major depressive symptoms in outpatient settings that reflects the prior 2 weeks (Kroenke & Spitzer, 2002). A value of 10 or greater reflects a minimum of moderate major depression with a sensitivity and specificity of 88% (Kroenke & Spitzer, 2002), supported by meta-analysis (Levis et al., 2019). Thoughts of suicide was defined on the basis of item 9 of the PHQ-9, which inquires about “Thoughts that you would be better off dead, or thoughts of hurting yourself in some way”, with values of Not at all (0), Several days (1), More than half the days (2), and Nearly every day (3). Scores above 0 have been robustly associated with suicide risk, including in older individuals (Louzon et al., 2016; Rossom et al., 2017; Simon et al., 2013).

To characterize aspects of social connectedness, we applied a framework adopted by an NIMH conference report (Lutz et al., 2021) that distinguishes dimensions reflecting social interaction (nature of social network, including household characteristics); social support (instrumental and emotional); and social quality (how the network is perceived, and how much satisfaction or distress it yields, most often reflected in loneliness).

To quantify social interaction, we examined number of household members, as well as number of face-to-face meetings with individuals outside of the household over the past 24 hours. To characterize social support, respondents were asked to identify the number of people who could care for them if they became ill (caregiving support), lend them money (financial support), serve as someone to talk to if they felt down or depressed (emotional support), or help them find employment (employment support); these validated questions were drawn from a prior examination of social networks (Lubbers et al., 2019). For example, the emotional support question asks how many people are available “to talk to if you had a problem, felt sad or depressed”. Finally, to characterize social quality, individuals in 2 of the 6 survey waves were asked the 3-item loneliness scale (Hughes et al., 2004), specifically developed for application in large-scale surveys and highly correlated with the Revised UCLA Loneliness Scale (Russell et al., 1980).

Analysis

We used multiple logistic regression to estimate odds ratios for the relationship between individual social characteristics and thoughts of suicide, alone and then with adjustment for sociodemographic features.

Survey results were re-weighted for regression models using interlocking national weights for age, sex, race/ethnicity, education, urbanicity (urban, suburban, or rural), and region, based on the 2019 US Census American Community Survey (Bureau, n.d.) (see Table 1), using the „survey‟ package in R 4.0 (R Core Team, 2019), as has been demonstrated to generate reliable estimates for non-probability samples (Valliant, 2020).

Table 1.

Demographics and Clinical Characteristics of Participants

No thoughts of suicide (N=15020) Thoughts of suicide (N=1144) Total (N=16164) P value
Age (in years) 70.9 (5.0) 70.4 (4.9) 70.9 (5.0) < 0.004
Gender
 Female 9274 (61.7%) 657 (57.4%) 9931
 Male 5746 (38.3%) 487 (42.6%) 6233
Household income a < 0.001
 Less than 2764 (18.5%) 320 (28.0%) 3084
 $25k to less than 4667 (31.2%) 343 (30.0%) 5010
 $50k to less than 5148 (34.4%) 325 (28.5%) 5473
 $100k/year or more 2387 (15.9%) 154 (13.5%) 2541
Race and ethnicity < 0.001
 African American 1041 (6.9%) 46 (4.0%) 1087 (6.7%)
 Asian American 293 (2.0%) 30 (2.6%) 323 (2.0%)
 Hispanic 327 (2.2%) 34 (3.0%) 361 (2.2%)
 Native American 44 (0.3%) 2 (0.2%) 46 (0.3%)
 Other 221 (1.5%) 21 (1.8%) 242 (1.5%)
 Pacific Islander 65 (0.4%) 9 (0.8%) 74 (0.5%)
 White 13029 (86.7%) 1002 (87.6%) 14031
Setting 0.003
 Rural 2718 (18.1%) 227 (19.8%) 2945
 Suburban 9154 (60.9%) 639 (55.9%) 9793
 Urban 3148 (21.0%) 278 (24.3%) 3426
Education 0.384
 Graduate Degree 2943 (19.6%) 219 (19.1%) 3162
 College Degree 5533 (36.8%) 423 (37.0%) 5956
 Some College 3754 (25.0%) 265 (23.2%) 4019
 High School 2642 (17.6%) 224 (19.6%) 2866
 Some High School 148 (1.0%) 13 (1.1%) 161 (1.0%)
PHQ-9 Total 2.9 (3.7) 13.0 (6.2) 3.6 (4.7) < 0.001
Moderate or greater depression 951 (6.3%) 730 (63.8%) 1681 (10.4%) < 0.001
PHQ-9 thoughts of suicide score n/a
 0 15020 (100.0%) 0 (0.0%) 15020
 1 0 (0.0%) 758 (66.3%) 758 (4.7%)
 2 0 (0.0%) 239 (20.9%) 239 (1.5%)
 3 0 (0.0%) 147 (12.8%) 147 (0.9%)
Open in a new tab
(a)

household income missing for n=54 without thoughts of suicide, n=2 with thoughts of suicide

In exploratory analysis, we also sought to identify subgroup effects, testing for variables that moderate associations between social disconnection and suicidality using generalized random forests (Athey et al., 2019). The generalized random forest is an extension of the widely-used random forest algorithm (Breiman, 2001), adapted specifically for identifying conditional effects that are robust to out-of-sample prediction. While the standard random forest builds decision trees based on which split, for which variable, generates the biggest difference in outcomes, the generalized random forest identifies which split, for which variable, generates the biggest differences in a given association (such as a treatment effect in an experiment). Crucially, each tree in the generalize random forest is split based on one randomly-assigned portion of the sample data. Actual predicted effects are evaluated by then applying those splits to observations from the held-out, “estimating” subsample that was not used to fit the tree. Building held-out estimation into the algorithm allows researchers to test a wide range of potential moderating variables at once in a systematic manner without overfitting to statistical noise, reducing the potential for type 1 error and offering important advantages over traditional approaches to testing for heterogeneity. (Here, trees were estimated in a randomly-selected 70%, and then effects estimated in the held-out 30%.)

As in our prior work (e.g., Green(Green et al., 2020) and Green(Green et al., in press)) we applied the open-source grf package (version 1.2) in R to implement this set of analyses, considering a range of potential moderating variables. These included additional sociodemographic variables (e.g., income, education, urbanicity, US region). We used the default hyperparameters specified in the package‟s causal_forest() function, with the exception that we generated 5000 trees to minimize overfitting risk.

Results

Of 16,164 survey respondents age 65 and older, mean age was 70.9 (SD 5.0); the cohort was 61.4% female and 29.6% male; 2.0% Asian, 6.7% Black, 2.2% Hispanic, and 86.8% White (Table 1). A total of 1144 (7.1%) reported thoughts of suicide at least several days in the prior 2 week period, including 386 (2.4%) who reported such symptoms more than half the time.

In the cohort as a whole, 4139 (25.6%) reported that they lived alone, while 7868 (48.7%) lived with 1 other person (Table 2). Respondents reported an average of 5.1 (SD 11.1) face-to-face interactions in the prior 24 hours with non-household members; 5022 (31.2%) reported no such interactions. Additional characteristics of social disconnection are summarized in Table 2.

Table 2.

Social characteristics of adults age 65 years and older who did or did not report thoughts of suicide on PHQ-9 Item 9.

No thoughts of suicide (N=15020) Thoughts of suicide (N=1144) Total (N=16164) P value
Health supports b 2.9 (2.4) 2.0 (2.1) 2.8 (2.4) < 0.001
Financial supports b 2.0 (2.1) 1.3 (1.6) 1.9 (2.1) < 0.001
Emotional 3.6 (2.9) 2.4 (2.3) 3.5 (2.9) < 0.001
Employment 1.7 (2.7) 1.0 (2.0) 1.7 (2.6) < 0.001
No health supports b 1298 (8.7%) 235 (20.6%) 1533 (9.5%) < 0.001
No financial 3511 (23.6%) 414 (36.4%) 3925 < 0.001
No emotional 1046 (7.0%) 229 (20.1%) 1275 (8.0%) < 0.001
No employment 7446 (50.4%) 697 (61.8%) 8143 < 0.001
Face-to-face 5.2 (11.1) 4.7 (11.1) 5.1 (11.1) 0.179
No face-to-face 4634 (31.0%) 388 (34.0%) 5022 0.031
Additional household < 0.001
 None 3823 (25.5%) 316 (27.6%) 4139
 1 7383 (49.2%) 485 (42.4%) 7868
 2 2513 (16.7%) 203 (17.7%) 2716
 3+ 1301 (8.7%) 140 (12.2%) 1441 (8.9%) < 0.001
Loneliness score d 4.3 (1.6) 6.5 (2.0) 4.5 (1.8)
Open in a new tab
(b)

instrumental social support data missing for n=46 (41 without thoughts of suicide, 5 with thoughts of suicide) for health support, n=137 (129 without, 8 with) for financial support, n=129 (122 without, 7 with) for emotional support, and 268 (251 without, 17 with) for employment support.

(c)

face-to-face meeting data not available for n=55 (51 without thoughts of suicide, 4 with thoughts of suicide).

(d)

loneliness data not available for n=10045 (9373 without thoughts of suicide, 672 with thoughts of suicide).

In logistic regression models, living in a household with 2 or more people was associated with significantly greater likelihood of thoughts of suicide in adjusted models (OR 1.45, 95% CI 1.11–1.90), as was living with 3 or more people (OR 1.73, 95% CI 1.28–2.33). (Table 3) Among non-household interactions, lack of any face-to-face meetings in the prior 24 hours was not associated with significantly greater likelihood of thoughts of suicide (adjusted OR 1.09, 95% CI 0.92–1.30).

Table 3.

Association between individual social network features and thoughts of suicide, crude and with adjustment for sociodemographic features

Crude OR Adjusted OR
2 household members (vs 1) 1.02 [0.83 – 1.25] 1.20 [0.96 – 1.48]
3 household members (vs 1) 1.22 [0.94 – 1.57] 1.45 ** [1.11 – 1.90]
4+ household members (vs 1) 1.48 ** [1.11 – 1.96] 1.73 *** [1.28 – 2.33]
No face-to-face meetings 1.13 [0.96 – 1.34] 1.09 [0.92 – 1.30]
No health supports 2.48 *** [2.02 – 3.05] 2.24 *** [1.81 – 2.78]
No financial supports 1.74 *** [1.47 – 2.06] 1.69 *** [1.42 – 2.00]
No emotional supports 2.86 *** [2.31 – 3.54] 2.60 *** [2.09 – 3.23]
No employment supports 1.58 *** [1.33 – 1.88] 1.54 *** [1.29 – 1.83]
Loneliness score 1.72 *** [1.62 – 1.84] 1.75*** [1.64 – 1.87]
Open in a new tab
***

p < 0.001

**

p < 0.01

*

p < 0.05.

OR, odds ratio

We next examined associations between thoughts of suicide and absence of social support. Individually, lack of each of the 4 categories of social support was significantly associated with suicidality in both crude and adjusted models (Table 3). In models incorporating all categories simultaneously (Figure 1), the largest independent effect was observed for lack of emotional support (adjusted OR 1.84, 95% CI 1.38–2.46), followed by health care (caregiving) support (i.e., whether there was someone who could care for them if they became ill) (adjusted OR 1.54, 95% CI 1.17–2.03).

Figure 1.

Figure 1.

Open in a new tab

Logistic regression model for thoughts of suicide incorporating sociodemographic and social connection and support features

To characterize effects of loneliness, we then examined the association between thoughts of suicide and the 3-item loneliness scale and suicidality, available for survey participants in 2 of 6 waves (n=5611). Adjusted models demonstrated significant association (OR 1.75, 95% CI 1.64–1.87); that is, a 1-point increase on this 6-point scale was associated with a 75% increase in odds of reporting thoughts of suicide.

Finally, in exploratory analysis, we examined differential associations between emotional support and thoughts of suicide across sociodemographic subgroups using generalized random forests. We focused on emotional support in this analysis as it demonstrated the largest effect on thoughts of suicide in univariate models, in line with previous studies demonstrating it is more predictive of persistent distress (Solomonov et al., 2021). Analysis of individual variability in predicted effect of emotional support suggests marked heterogeneity (Supplemental Figure 1). The sociodemographic predictors with greatest importance in identifying differential treatment effects (i.e., most frequently selected as branches in random forests) included age followed by gender, household size, and income, with lesser effects for race and ethnicity and urbanicity (Supplemental Figure 2). The marginal effects of age, by quantile, are illustrated in Supplemental Figure 3, indicating that the effect size of association between emotional support and thoughts of suicide among older adults is largest among younger individuals in this group (i.e., the younger-old), and least among older individuals (i.e., the older-old). For household size, emotional support is associated with effect size among the largest households (Supplemental Figure 4). Conversely, for household income, the effect of lack of social support is greatest among those in the highestincome quartile (Supplemental Figure 5).

Discussion

Using national survey data from 16,164 respondents aged 65 and older, we found that 7.1% reported thoughts of suicide at least several days in the prior 2-week period, including 2.4% more than half the days in this period. While 25.6% of older adults reported living alone, living alone was not associated with greater risk of suicidality; in fact, suicidality was less likely among those who lived alone or with a partner compared to larger households. However, thoughts of suicide was more likely among individuals who identified fewer social supports, particularly emotional supports, and greater levels of loneliness.

Social disconnection increases the risk of thoughts of suicide in late life (Joiner, 2007). Our results are broadly consistent with prior studies supporting such an association (Fässberg et al., 2012; Cui et al., 2022). In a case-control design, older adults with thoughts of suicide described lower levels of perceived social support (Harrison et al., 2010); a similar relationship was observed in a small community sample of older adults (Vanderhorst & McLaren, 2005). Poorer social integration – whether with family or community – was also associated with suicide death in older adults (Duberstein et al., 2004). Conversely, a greater sense of belonging, reflecting greater social engagement, correlated with diminished risk for suicidality in this age group in a community sample (McLaren et al., 2007).

Our results underscore the importance of precision in describing social networks: living alone in and of itself is not a risk factor for thoughts of suicide, for example. They are aligned with previous work demonstrating that low perceived social support and lack of social interactions with significant others serve as critical risk factors in late life (Solomonov et al., 2021; Solomonov et al., 2019). These latter factors are modifiable and can be targeted using psychosocial interventions that increase social engagement and alter negative perceptions of social supports (Solomonov et al., 2019; Solomonov & Phan, 2021).

We also demonstrate the application of a machine learning-based methodology for characterizing heterogeneity in the relationship between social connectedness and thoughts of suicide. We show, first, that there is strong evidence of heterogeneity, indicating the potential utility of more targeted interventions, since the risk conferred by a perceived lack of emotional support network varies by population subgroup. Further, we show that, in the case of age, the effect is greatest among younger rather than older individuals in the 65 and older age group. Others reported that the impact of social isolation and loneliness on mental health outcomes during COVID-19 was smaller among the oldest older adults, compared to other age groups, including younger older adults(McDonald et al., 2022). One possible explanation is that the oldest adults show high resilience and often perceive events as less stressful than their younger counterparts, due to increased emotional regulation and life experiences with past public crises(Knepple Carney et al., 2021; Neubauer et al., 2019). Effects also vary by income and household size, although the nature of these effects is more complex and requires further investigation.

Social isolation is not in and of itself simply a proxy for depression. Strategies that reduce depressive symptoms in older adults do diminish suicidality (Alexopoulos et al., 2021), but recent work suggests that the relationship between social connectedness and depression is more complex - and, in particular, highlights the role of positive affect. For example, in a survey of 429 older adults, positive affect was associated with lower levels of loneliness and greater social satisfaction (Davidson et al., 2021). Conversely, loneliness was common in that cohort, despite generally low levels of depressive symptoms. These findings highlight the importance of more nuanced investigation of social connectedness, rather than attributing it solely to depression or negative valence (Solomonov & Phan, 2021). In our sample, depression severity was relatively low, yet approximately a third of older adults reported absence of recent social interactions. Thus, strategies to increase social connectedness could be beneficial to older adults broadly, even those who may not experience depression. For example, psychotherapies based on behavioral activation principles that can increase engagement in rewarding social interactions may be particularly helpful(Solomonov et al., 2019; Solomonov & Phan, 2021).

Though this study employed only a single item to assess thoughts of suicide, that item has been demonstrated to be strongly associated with risk for suicide. Most notably, a prior study found that individuals who endorsed near-daily suicidality were between 5 and 8 times more likely to make a suicide attempt, and between 3 and 11 times more likely to die by suicide within 30 days, than individuals without such frequent ideation (Rossom et al., 2017). Rates of suicide attempt, and death by suicide, remained elevated within 365 days, and elevated risk was also observed in analyses limited to those 65 and older. As such, while we did not assess other suicide risk factors, or employ more detailed measures of suicide, we nonetheless are able to apply a measure with high predictive validity.

We note several additional caveats in interpreting our results. First, while our cohort is likely more generalizable to unselected populations of older adults than clinical/care-seeking samples, it nonetheless remains vulnerable to sources of ascertainment bias. As the survey is delivered via internet, we undersampled groups that are less apt to use such technology, while possibly oversampling those who use the internet for a greater proportion of their social interactions. Depressed older adults often suffer from comorbid medical conditions, disability, and low technological literacy, and may be less likely to participate in an online survey. The use of nonprobability sampling generally allows sampling a much larger cohort, balancing risks to ecological validity inherent in internet sampling (for a discussion of these tradeoffs, see Kennedy & Caumont, n.d.); this approach has been validated in other contexts (Coppock & McClellan, 2019; Berinsky et al., 2012) as well as in our prior work (Perlis et al., in press). On the other hand, as a general opinion survey that is not targeted to individuals with interest in mental health, we avoid other biases that might lead to inflated estimates of risk prevalence.

A second limitation is that we were unable to include more detailed assessments of suicidal behavior or risk – in particular, measures of comorbidity – given the need to keep the survey sufficiently brief. More targeted sub-surveys – for example, to those who endorse thoughts of suicide on initial screening - could help to address this limitation in future work. A final limitation to consider is that the surveys occurred during the COVID-19 pandemic, a source of both acute and chronic stressors that we and others have shown is associated with elevated prevalence of depression and anxiety (Ettman et al., 2020; Rossi et al., 2020). After the pandemic subsides, it will be important to revisit these results to examine the extent to which they will persist absent the high ambient levels of pandemic-associated stress.

Conclusion

In aggregate, our results demonstrate that both social disconnection and thoughts of suicide are common among older adults in the United States. Emotional social support was the strongest predictor of thoughts of suicide, with more pronounced effects in younger older adults. Our findings are correlational in nature, and we cannot assess causality. However, results highlight the need for future longitudinal studies to investigate the bi-directional relationships between social disconnection and thoughts of suicide. Further, our findings may guide development of targeted psychosocial interventions for suicidality, such as streamlined psychotherapies, that modify social risk factors in this vulnerable population.

Supplementary Material

1

Highlights.

  • Older adults are disproportionally impacted by social disconnection. Such disconnection is a leading risk factor for suicide in late life, a period when suicide rates are high

  • In our 50-States survey, conducted during COVID-19, we found that 7.1% of older adults reported suicidality. Lack of social supports, and particularly emotional supports, were independently associated with greater likelihood of reporting suicidality

  • Social support is linked with suicidality in late life. Interventions targeting social disconnection may be beneficial for older adults at risk of suicide

Acknowledgments:

This study was supported by the National Science Foundation (SES-2029292 and SES-2029792; Drs. Baum, Ognyanova, and Lazer), the National Institute of Mental Health (R01MH116270, Dr. Perlis; RF1MH132335, Drs. Perlis and Lazer; K23MH123864, Dr. Solomonov), Northeastern University, Harvard Kennedy School of Government, and Rutgers University.

Footnotes

Trial Registration: NA

Ethical standards: The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.

Financial Disclosures

Dr. Perlis has received consulting fees from Burrage Capital, Genomind, RID Ventures, Belle Artificial Intelligence, Vault Health, and Takeda. He holds equity in Belle Artificial Intelligence, Circular Genomics, Psy Therapeutics, and Vault Health. He receives patent royalties from Massachusetts General Hospital. The other authors report no financial disclosures.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Alexopoulos GS, Raue PJ, Banerjee S, Mauer E, Marino P, Soliman M, Kanellopoulos D, Solomonov N, Adeagbo A, Sirey JA, Hull TD, Kiosses DN, & Areán PA (2021). Modifiable predictors of suicidal ideation during psychotherapy for late-life major depression. A machine learning approach. Translational Psychiatry, 11, 536. 10.1038/s41398-021-01656-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Athey S, Tibshirani J, & Wager S. (2019). Generalized random forests. The Annals of Statistics, 47(2). 10.1214/18-AOS1709 [DOI] [Google Scholar]
  3. Barak-Corren Y, Castro VM, Nock MK, Mandl KD, Madsen EM, Seiger A, Adams WG, Applegate RJ, Bernstam EV, Klann JG, McCarthy EP, Murphy SN, Natter M, Ostasiewski B, Patibandla N, Rosenthal GE, Silva GS, Wei K, Weber GM, … Smoller JW (2020). Validation of an Electronic Health Record–Based Suicide Risk Prediction Modeling Approach Across Multiple Health Care Systems. JAMA Network Open, 3(3), e201262. 10.1001/jamanetworkopen.2020.1262 [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Berinsky AJ, Huber GA, & Lenz GS (2012). Evaluating Online Labor Markets for Experimental Research: Amazon.com’s Mechanical Turk. Political Analysis, 20(3), 351–368. 10.1093/pan/mpr057 [DOI] [Google Scholar]
  5. Breiman L. (2001). Random Forests. Machine Learning, 45(1), 5–32. 10.1023/A:1010933404324 [DOI] [Google Scholar]
  6. Bureau, U. C. (n.d.). American Community Survey (ACS). Census.Gov. Retrieved November 28, 2021, from https://www.census.gov/programs-surveys/acs [Google Scholar]
  7. Cai Z, Junus A, Chang Q, & Yip PSF (2022). The lethality of suicide methods: A systematic review and meta-analysis. Journal of Affective Disorders, 300, 121–129. 10.1016/j.jad.2021.12.054 [DOI] [PubMed] [Google Scholar]
  8. Coppock A, & McClellan OA (2019). Validating the demographic, political, psychological, and experimental results obtained from a new source of online survey respondents. Research & Politics, 6(1), 2053168018822174. 10.1177/2053168018822174 [DOI] [Google Scholar]
  9. Cui R, Gujral S, Galfalvy H, & Szanto K. (2022). The Role of Perceived and Objective Social Connectedness on Risk for Suicidal Thoughts and Behavior in Late-Life and Their Moderating Effect on Cognitive Deficits. The American Journal of Geriatric Psychiatry: Official Journal of the American Association for Geriatric Psychiatry, 30(4), 527–532. 10.1016/j.jagp.2021.08.014 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Davidson EJ, Taylor CT, Ayers CR, Quach NE, Tu XM, & Lee EE (2021). The Relationship Between Loneliness and Positive Affect in Older Adults. The American Journal of Geriatric Psychiatry: Official Journal of the American Association for Geriatric Psychiatry, S1064–7481(21)00530–3. 10.1016/j.jagp.2021.11.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Duberstein PR, Conwell Y, Conner KR, Eberly S, Evinger JS, & Caine ED (2004). Poor social integration and suicide: Fact or artifact? A case-control study. Psychological Medicine, 34(7), 1331–1337. 10.1017/s0033291704002600 [DOI] [PubMed] [Google Scholar]
  12. Ettman CK, Abdalla SM, Cohen GH, Sampson L, Vivier PM, & Galea S. (2020). Prevalence of Depression Symptoms in US Adults Before and During the COVID-19 Pandemic. JAMA Network Open, 3(9), e2019686. 10.1001/jamanetworkopen.2020.19686 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Fässberg MM, van Orden KA, Duberstein P, Erlangsen A, Lapierre S, Bodner E, Canetto SS, Leo DD, Szanto K, & Waern M. (2012). A Systematic Review of Social Factors and Suicidal Behavior in Older Adulthood. International Journal of Environmental Research and Public Health, 9(3), 722–745. 10.3390/ijerph9030722 [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. FastStats. (2022, January 6). https://www.cdc.gov/nchs/fastats/suicide.htm
  15. Global, regional, and national burden of suicide mortality 1990 to 2016: Systematic analysis for the Global Burden of Disease Study 2016. (2019). The BMJ, 364, l94. 10.1136/bmj.l94 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Green J, Baum M, Druckman J, Lazer D, Ognyanova K, Simonson MD, Perlis R, & Santillana M. (2020). Identifying and Measuring Conditional Policy Preferences: The Case of Opening Schools During a Pandemic. OSF Preprints. 10.31219/osf.io/zsh8u [DOI] [Google Scholar]
  17. Green J, Druckman JN, Baum MA, Lazer D, Ognyanova K, Simonson M, Lin J, Santillana M, & Perlis RH (in press). Using General Messages to Persuade on a Politicized Scientific Issue. British Journal of Political Science. [Google Scholar]
  18. Harrison KE, Dombrovski AY, Morse JQ, Houck P, Schlernitzauer M, Reynolds CF, & Szanto K. (2010). ALONE? PERCEIVED SOCIAL SUPPORT AND CHRONIC INTERPERSONAL DIFFICULTIES IN SUICIDAL ELDERS. International Psychogeriatrics / IPA, 22(3), 445–454. 10.1017/S1041610209991463 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Hughes ME, Waite LJ, Hawkley LC, & Cacioppo JT (2004). A Short Scale for Measuring Loneliness in Large Surveys. Research on Aging, 26(6), 655–672. 10.1177/0164027504268574 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Hwang T-J, Rabheru K, Peisah C, Reichman W, & Ikeda M. (n.d.). Loneliness and social isolation during the COVID-19 pandemic. International Psychogeriatrics, 1–4. 10.1017/S1041610220000988 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Joiner TE (2007). Why People Die by Suicide. Harvard University Press. [Google Scholar]
  22. Kennedy C, & Caumont R. (n.d.). What we learned about online nonprobability polls. Pew Research Center. Retrieved November 27, 2021, from https://www.pewresearch.org/fact-tank/2016/05/02/q-a-online-nonprobability-polls/ [Google Scholar]
  23. Kiosses DN, Alexopoulos GS, Hajcak G, Apfeldorf W, Duberstein PR, Putrino D, & Gross JJ (2018). Cognitive Reappraisal Intervention for Suicide Prevention (CRISP) for Middle-Aged and Older Adults Hospitalized for Suicidality. The American Journal of Geriatric Psychiatry: Official Journal of the American Association for Geriatric Psychiatry, 26(4), 494–503. 10.1016/j.jagp.2017.11.009 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Kiosses DN, Gross JJ, Banerjee S, Duberstein PR, Putrino D, & Alexopoulos GS (2017). Negative Emotions and Suicidal Ideation during Psychosocial Treatments in Older Adults with Major Depression and Cognitive Impairment. The American Journal of Geriatric Psychiatry: Official Journal of the American Association for Geriatric Psychiatry, 25(6), 620–629. 10.1016/j.jagp.2017.01.011 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Knepple Carney A, Graf AS, Hudson G, & Wilson E. (2021). Age Moderates Perceived COVID-19 Disruption on Well-Being. The Gerontologist, 61(1), 30–35. 10.1093/geront/gnaa106 [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Kotwal AA, Cenzer IS, Waite LJ, Covinsky KE, Perissinotto CM, Boscardin WJ, Hawkley LC, Dale W, & Smith AK (2021). The epidemiology of social isolation and loneliness among older adults during the last years of life. Journal of the American Geriatrics Society, 69(11), 3081–3091. 10.1111/jgs.17366 [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Kotwal AA, Holt-Lunstad J, Newmark RL, Cenzer I, Smith AK, Covinsky KE, Escueta DP, Lee JM, & Perissinotto CM (2021). Social Isolation and Loneliness Among San Francisco Bay Area Older Adults During the COVID-19 Shelter-in-Place Orders. Journal of the American Geriatrics Society, 69(1), 20–29. 10.1111/jgs.16865 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Kroenke K, & Spitzer RL (2002). The PHQ-9: A new depression diagnostic and severity measure. Psychiatric Annals, 32(9), 509–515. 10.3928/0048-5713-20020901-06 [DOI] [Google Scholar]
  29. Levis B, Benedetti A, Thombs BD, & DEPRESsion Screening Data (DEPRESSD) Collaboration. (2019). Accuracy of Patient Health Questionnaire-9 (PHQ-9) for screening to detect major depression: Individual participant data meta-analysis. BMJ (Clinical Research Ed.), 365, l1476. 10.1136/bmj.l1476 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Louzon SA, Bossarte R, McCarthy JF, & Katz IR (2016). Does Suicidal Ideation as Measured by the PHQ-9 Predict Suicide Among VA Patients? Psychiatric Services (Washington, D.C.), 67(5), 517–522. 10.1176/appi.ps.201500149 [DOI] [PubMed] [Google Scholar]
  31. Lubbers MJ, Molina JL, & Valenzuela-García H. (2019). When networks speak volumes: Variation in the size of broader acquaintanceship networks. Social Networks, 56, 55–69. 10.1016/j.socnet.2018.08.004 [DOI] [Google Scholar]
  32. Lutz J, Van Orden KA, Bruce ML, Conwell Y, & Members of the NIMH Workshop on Social Disconnection in Late Life Suicide. (2021). Social Disconnection in Late Life Suicide: An NIMH Workshop on State of the Research in Identifying Mechanisms, Treatment Targets, and Interventions. The American Journal of Geriatric Psychiatry: Official Journal of the American Association for Geriatric Psychiatry, 29(8), 731–744. 10.1016/j.jagp.2021.01.137 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. McCoy TH, Castro VM, Roberson AM, Snapper LA, & Perlis RH (2016). Improving Prediction of Suicide and Accidental Death After Discharge From General Hospitals With Natural Language Processing. JAMA Psychiatry, 73(10), 1064. 10.1001/jamapsychiatry.2016.2172 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. McCoy TH, Pellegrini AM, & Perlis RH (2019). Research Domain Criteria scores estimated through natural language processing are associated with risk for suicide and accidental death. Depression and Anxiety, 36(5), 392–399. 10.1002/da.22882 [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. McDonald AJ, Wickens CM, Bondy SJ, Elton-Marshall T, Wells S, Nigatu YT, Jankowicz D, & Hamilton HA (2022). Age differences in the association between loneliness and anxiety symptoms during the COVID-19 pandemic. Psychiatry Research, 310, 114446. 10.1016/j.psychres.2022.114446 [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. McLaren S, Gomez R, Bailey M, & Van Der Horst RK (2007). The association of depression and sense of belonging with suicidal ideation among older adults: Applicability of resiliency models. Suicide & Life-Threatening Behavior, 37(1), 89–102. 10.1521/suli.2007.37.1.89 [DOI] [PubMed] [Google Scholar]
  37. Neubauer AB, Smyth JM, & Sliwinski MJ (2019). Age Differences in Proactive Coping With Minor Hassles in Daily Life. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 74(1), 7–16. 10.1093/geronb/gby061 [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Oh SH, Lee KU, Kim SH, Park KN, Kim YM, & Kim HJ (2014). Factors associated with choice of high lethality methods in suicide attempters: A cross-sectional study. International Journal of Mental Health Systems, 8(1), 43. 10.1186/1752-4458-8-43 [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Perlis RH, Simonson MD, Green J, Lin J, Safarpour A, Lazer D, & Baum M. (in press). Firearm ownership among individuals with major depressive symptoms. JAMA Network Open. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. R Core Team. (2019). R: A Language and Environment for Statistical Computing (3.6.0). R Foundation for Statistical Computing. www.R-project.org [Google Scholar]
  41. Rossi R, Socci V, Pacitti F, Di Lorenzo G, Di Marco A, Siracusano A, & Rossi A. (2020). Mental Health Outcomes Among Frontline and Second-Line Health Care Workers During the Coronavirus Disease 2019 (COVID-19) Pandemic in Italy. JAMA Network Open, 3(5), e2010185. 10.1001/jamanetworkopen.2020.10185 [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Rossom RC, Coleman KJ, Ahmedani BK, Beck A, Johnson E, Oliver M, & Simon GE (2017). Suicidal Ideation Reported on the PHQ9 and Risk of Suicidal Behavior across Age Groups. Journal of Affective Disorders, 215, 77–84. 10.1016/j.jad.2017.03.037 [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Russell D, Peplau LA, & Cutrona CE (1980). The revised UCLA Loneliness Scale: Concurrent and discriminant validity evidence. Journal of Personality and Social Psychology, 39(3), 472–480. 10.1037//0022-3514.39.3.472 [DOI] [PubMed] [Google Scholar]
  44. Simon GE, Rutter CM, Peterson D, Oliver M, Whiteside U, Operskalski B, & Ludman EJ (2013). Does response on the PHQ-9 Depression Questionnaire predict subsequent suicide attempt or suicide death? Psychiatric Services (Washington, D.C.), 64(12), 1195–1202. 10.1176/appi.ps.201200587 [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Solomonov N, Bress JN, Sirey JA, Gunning FM, Flückiger C, Raue PJ, Areán PA, & Alexopoulos GS (2019). Engagement in Socially and Interpersonally Rewarding Activities as a Predictor of Outcome in “Engage” Behavioral Activation Therapy for Late-Life Depression. The American Journal of Geriatric Psychiatry: Official Journal of the American Association for Geriatric Psychiatry, 27(6), 571–578. 10.1016/j.jagp.2018.12.033 [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Solomonov N, Lee J, Banerjee S, Flückiger C, Kanellopoulos D, Gunning FM, Sirey JA, Liston C, Raue PJ, Hull TD, Areán PA, & Alexopoulos GS (2021). Modifiable predictors of nonresponse to psychotherapies for late-life depression with executive dysfunction: A machine learning approach. Molecular Psychiatry, 26(9), Article 9. 10.1038/s41380-020-0836-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Solomonov N, & Phan DK (2021). Improving Interventions for Loneliness and Social Isolation in Late Life: The Potential Role of Positive Affect. The American Journal of Geriatric Psychiatry: Official Journal of the American Association for Geriatric Psychiatry, S1064–7481(21)00551–0. 10.1016/j.jagp.2021.11.012 [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Survey Disclosure Checklist—AAPOR. (n.d.). Retrieved January 25, 2022, from https://www.aapor.org/Standards-Ethics/AAPOR-Code-of-Ethics/Survey-Disclosure-Checklist.aspx
  49. Szanto K, Dombrovski AY, Sahakian BJ, Mulsant BH, Houck PR, Reynolds CF, & Clark L. (2012). Social emotion recognition, social functioning, and attempted suicide in late-life depression. The American Journal of Geriatric Psychiatry: Official Journal of the American Association for Geriatric Psychiatry, 20(3), 257–265. 10.1097/JGP.0b013e31820eea0c [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. The Major Health Implications of Social Connection—Julianne Holt-Lunstad, 2021. (n.d.). Retrieved March 26, 2023, from https://journals.sagepub.com/doi/abs/10.1177/0963721421999630
  51. Valliant R. (2020). Comparing Alternatives for Estimation from Nonprobability Samples. Journal of Survey Statistics and Methodology, 8(2), 231–263. 10.1093/jssam/smz003 [DOI] [Google Scholar]
  52. Van Orden K, & Deming C. (2018). Late-life suicide prevention strategies: Current status and future directions. Current Opinion in Psychology, 22, 79–83. 10.1016/j.copsyc.2017.08.033 [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. van Tilburg TG, Steinmetz S, Stolte E, van der Roest H, & de Vries DH (2021). Loneliness and Mental Health During the COVID-19 Pandemic: A Study Among Dutch Older Adults. The Journals of Gerontology. Series B, Psychological Sciences and Social Sciences, 76(7), e249–e255. 10.1093/geronb/gbaa111 [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Vanderhorst RK, & McLaren S. (2005). Social relationships as predictors of depression and suicidal ideation in older adults. Aging & Mental Health, 9(6), 517–525. 10.1080/13607860500193062 [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1898625-supplement-1.docx (125.5KB, docx)

RESOURCES