Suicide, Firearms, and Terminal Illness: A Latent Class Analysis Using Data From the National Violent Death Reporting System
Abstract
Objective:
Firearms are highly lethal when used for suicide and are used more frequently as a suicide method by persons of older age. Individuals with terminal illness are at high risk for suicide, yet little research has explored how firearms may be used for self-harm in this population. The authors sought to understand the patterns of psychiatric diagnoses, substance use disorders diagnoses, and suicide mechanisms for individuals with terminal illness who died by suicide as well as their demographic and circumstantial characteristics.
Methods:
A latent class analysis using data from the National Violent Death Reporting System was undertaken to better understand typologies of individuals with terminal illness who died by suicide in 2003–2018 (N=3,072). To develop the classes, the authors considered diagnoses of mental illness and of alcohol or substance use disorders, suicidal thoughts and behaviors, and mechanism of suicide (firearm or no firearm). Demographic and circumstantial variables were examined across classes.
Results:
The analysis revealed four classes of persons with terminal illness who died from suicide: depression and nonfirearm methods (N=375, 12%), suicidal intent and firearm use (N=922, 30%), alcohol or substance use disorder and nonfirearm methods (N=70, 2%), and firearm use only (N=1,705, 56%).
Conclusions:
Firearm access is an important consideration for terminally ill persons at risk for suicide. Screening for psychiatric and substance use disorders may not identify terminally ill persons who are at increased suicide risk because of the presence of a firearm in the home. This population may benefit from tailored interventions in specialty care settings to address firearm safety.
HIGHLIGHTS
Firearm safety may be an important suicide prevention strategy for individuals with a terminal diagnosis.
Screening for mental illness and substance use disorders as a suicide prevention strategy would likely fail to identify many individuals who are at risk for suicide.
Broadly implementing a public health approach that emphasizes lethal means assessment and restriction in specialty care settings, where many terminally ill persons receive treatment, may help address suicide in this high-risk group.
Suicide is among the leading causes of death for Americans, constituting a public health crisis (1). Suicide prevention has proven elusive, as has correctly predicting who is in need of intervention (2). Recent commentors have noted that suicide may be optimally viewed as a socially determined public health problem (3). Whereas the traditional approach to suicide prevention emphasizes the detection of individuals with suicide risk factors and their subsequent referral for behavioral health treatments, the social determinants perspective has elicited new research trajectories (3). In addition to behavioral interventions and clinical screening, a focus on policy solutions to reduce access to lethal means used in suicide has become of interest to stakeholders (4, 5). Firearm use is the most lethal method for suicide, with 80%–90% of attempts being fatal (6). It is also the most common method; 50% of all suicides include use of a firearm, increasing to 70% among older adults (1).
Individuals who develop terminal illnesses (e.g., advanced and untreatable cancer) are likely to have feelings of hopelessness and to worry about becoming a burden—both well-documented risk factors for suicide (7–9). Diagnosed mental illness is less common among individuals who die by firearm suicide than among those who die by nonfirearm suicide (10), and having a diagnosed mental illness is more common among women than among men who die by suicide (11). It is not clear whether this trend persists among individuals with terminal illness, who may have an increased risk for depression overall (12–15). From a perspective that considers the social determinants of public health, individuals with terminal illness and firearm access may constitute a high-risk, yet understudied, population with regard to suicide.
Legislative barriers in the United States have made scientific inquiry into prevention strategies for firearm injury challenging to implement, resulting in a dearth of studies informing public health policy and clinical practice (16). Few examples exist of large-scale secondary data analyses of public data sets that allow for assessment of terminal illness and firearm-related suicide (17). The National Violent Death Reporting System (NVDRS) is a population-based reporting system for violent deaths occurring in reporting states and associated territories and allows for descriptive assessments of firearm-related suicide among individuals with a terminal illness.
Our primary aim was to explore the patterns of psychiatric diagnoses, substance use disorder diagnoses, and means or methods of suicide for individuals with terminal illnesses who died by suicide. Our secondary aim was to investigate the distribution of demographic and circumstantial characteristics across each of these groups of individuals. Understanding typologies of individuals with terminal illness and their demographic and circumstantial characteristics can inform screening and intervention procedures as well as policies to address suicides, particularly those undertaken with firearms in the vulnerable and understudied population of patients with terminal illness.
Methods
Data Source
For this study, we used the NVDRS Restricted Access Database (18) from the Centers for Disease Control and Prevention to examine firearm suicides among individuals with terminal illness. The NVDRS contains incident-level data on violent deaths, linking information from the death certificate, coroner or medical examiner report, and law enforcement report, as well as optional secondary sources such as hospital records and crime lab data. Hundreds of variables are coded by trained abstractors as they review the incident documentation from the primary sources. Much of the data—and especially general medical and mental health variables—are from investigator interviews of persons close to the victim, which are noted in the coroner, medical examiner, and law enforcement reports on the incident. Over the study period, participation increased from seven to 41 states and territories (including the District of Columbia and Puerto Rico) reporting at least 1 full year of data (18). (A full list with the first year each state reported is available in the online supplement to this article.) JMP (19) and Stata, version 17.0 (20), were used for case selection and data analysis, respectively. Because the data are anonymized and publicly available through a restricted-access process, this study was exempt from institutional review board approval.
Case Selection
In a previous study (21), our research group has described the use of an efficient text-mining procedure and subsequent manual review process to derive additional circumstances and potential risk factors from NVDRS narrative reports. Advances in accessible software, such as JMP (19), have made language-processing technology widely accessible. Our initial population included all individuals ages ≥50 years who died by suicide in the 2003–2018 period. We chose individuals in this age group because they have higher rates of suicide relative to other age groups and are more likely to be diagnosed as having or to die from serious health problems, such as cancer (22, 23).
We iteratively identified cases of suicide in two steps: text mining for identification and manual narrative review. First, we used JMP software’s text explorer function (19) to report all root terms in the law enforcement and coroner or medical examiner narratives for each individual in our initial population. To create a document term matrix, we used JMP text explorer (24) to create tokens, or basic terms and phrases. We used the built-in “stop word” list provided in JMP to remove common and unimportant words (e.g., a, an, and the). We also included misspellings, because JMP language-processing software reports these in the text explorer function. Tokenizing was performed by using the Regex approach in JMP, which uses regular expressions to identify patterns in the text, resulting in term and phrase lists as tokens from our narratives. We collated these tokens, and two researchers (L.C.P., A.C.) reviewed them and agreed on a list of terms to define “terminal illness.” In the case of advanced cancer, which is not always terminal, we applied the definition of terminal cancer by the National Cancer Institute (NCI) during manual review (25). NCI terms included hospice, incurable, inoperable, comfort care, palliative, terminal, end stage, untreatable, metastasize, metastatic, metastasis, metastases, metastasized, malignant, and malignancy. Second, after creating a document term matrix, we conducted a manual review of all narratives identified as meeting criteria to ensure that they were referring to a terminal illness of the individual and not a corollary person mentioned in the narrative (e.g., spouse) and to ensure the term was not referring to an individual after a suicide attempt (e.g., a near-lethal suicide attempt warranting “comfort care” until eventual death).
Variables
The NVDRS “weapon type” variable was used to identify firearm versus nonfirearm (e.g., sharp instrument, blunt instrument, poisoning, hanging, fall, explosive, or drowning) causes of death. The firearm variable includes a variety of makes and models, including handguns, submachine guns, rifles, shotguns, and long guns.
The following demographic and circumstantial variables, identified as potential risk factors for suicide, were coded by NVDRS abstractors and were analyzed in this study: sex, age, race-ethnicity, marital status, military status, education level, psychiatric and substance use disorder diagnoses, current mental health and substance use disorder treatment, history of suicidal thoughts, history of suicide attempts, intimate partner problem, recent suicide by friend or family, other death of friend or family, job problem, financial problem, eviction or loss of home, traumatic anniversary, exposure to disaster, and recent criminal or legal problem (26, 27). (Details on how these terms are defined are available in the online supplement.)
Latent Class Analysis
To pursue the primary aim for this study, we used probabilistic unrestricted parameterization of the latent class model to create subgroups of terminally ill suicide decedents, constructed with our set of observed variables (28). The observed variables used to create our classes included a diagnosis of mental illness (e.g., depression or anxiety), diagnosis of alcohol or substance use disorders, suicidal ideation or attempts, and mechanism for suicide (firearm or other mechanism). Our process included first identifying groups or classes on the basis of the manifest variables identified above. Then, to explore our secondary aim, we examined demographic, suicide-related, and circumstantial variables across the classes. In determining the appropriate number of classes, we assessed the Bayesian information criterion (BIC), Akaike information criterion (AIC), and entropy values, while also balancing our preference for a parsimonious model for practical interpretability. We used the gsem command in Stata, version 17.0 (20), to conduct the latent class analysis.
Demographic and circumstantial characteristics for the full sample are reported as proportions or means with standard deviations. We reported the demographic, suicide-related, and circumstantial characteristics as proportions or means with standard deviations across each class. Analysis of variance and chi-square tests of significance were performed across classes for continuous and dichotomous variables, respectively.
Results
Using the NVDRS, we identified 3,072 individuals with a terminal illness who died by suicide from 2003 to 2018. Demographic characteristics are shown in Table 1. Of this group, 2,346 (76%) used a firearm as a mechanism for suicide, and 726 (24%) used another mechanism (e.g., hanging, overdose). Two hundred individuals (7%) had alcohol use disorder, and 68 (2%) had other substance use disorders. Overall, 693 (23%) had a documented current mental health problem.
Characteristic | N | % |
---|---|---|
Sex | ||
Male | 2,669 | 87 |
Female | 403 | 13 |
Age (M±SD years) | 72±10 | |
Race-ethnicity | ||
White | 2,915 | 95 |
Black | 72 | 2 |
Asian | 42 | 1 |
American Indian or Alaska Native | 16 | 1 |
All other races | 27 | 1 |
Hispanic | 54 | 2 |
Non-Hispanic or unknown | 3,018 | 98 |
Marital status | ||
Single, divorced, separated | 1,608 | 52 |
Married, civil union, domestic partnership | 1,464 | 48 |
Ever served in military | 1,382 | 45 |
Education level | ||
High school or greater | 1,949 | 63 |
Less than high school | 379 | 12 |
Unknown | 744 | 24 |
Alcohol use disorder | 200 | 7 |
Other substance use disorders | 68 | 2 |
Current mental health problem | 693 | 23 |
Diagnosisa | ||
Depression | 527 | 17 |
Anxiety | 127 | 4 |
Bipolar disorder | 28 | 1 |
PTSD | 13 | <1 |
Schizophrenia | 8 | <1 |
Unknown | 68 | 2 |
Other | 57 | 2 |
Undergoing mental health treatment | 240 | 8 |
Suicidal thoughts or actions | ||
History of suicidal thoughts | 655 | 21 |
History of suicide attempts | 183 | 6 |
Intimate partner problem | 90 | 3 |
Recent suicide by friend or family | 29 | 1 |
Other death of friend or family | 149 | 5 |
Job problem | 23 | 1 |
Financial problem | 59 | 2 |
Eviction or loss of home | 48 | 2 |
Recent criminal or legal problem | 19 | <1 |
Means or method of suicide | ||
Firearm | 2,346 | 76 |
Poisoning | 376 | 12 |
Hanging or strangulation | 208 | 7 |
All other means | 142 | 5 |
Characteristics of suicide decedents ages ≥50 years with terminal illness, 2003–2018 (N=3,072)
We chose a four-class model on the basis of the analysis-of-fit statistics displayed in Table 2. We aimed to minimize both BIC and AIC, while choosing the entropy value closest to one. Our latent class analysis uncovered the four-class model as the best-fitting and most clinically relevant model, while maintaining parsimony (29, 30). The proportion of respondents in each class and the endorsement probabilities for variables used to generate the classes are shown in Table 3. Overall, 12% (N=375) of the suicide decedents were assigned to class 1, 30% (N=922) to class 2, 2% (N=70) to class 3, and 56% (N=1,705) to class 4. The distribution of demographic and circumstantial variables across classes is shown in Table 4.
Model | Akaike information criterion | Bayesian information criterion | Entropy |
---|---|---|---|
2 classes | 17,432.2 | 17,558.9 | .94 |
3 classes | 17,268.6 | 17,449.5 | .93 |
4 classes | 17,083.6 | 17,324.8 | .93 |
5 classes | 17,065.4 | 17,385.0 | .90 |
6 classes | 17,040.6 | 17,425.6 | .88 |
Analysis-of-fit statistics for five different class models
Marginal probability | ||||
---|---|---|---|---|
Variable | Class 1 (N=375) | Class 2 (N=922) | Class 3 (N=70) | Class 4 (N=1,705) |
Mental illness | ||||
Bipolar disorder | .06 | — | .04 | — |
Anxiety | .25 | .01 | .13 | — |
Depression | .70 | .12 | .42 | .08 |
PTSD | .01 | — | .01 | — |
Alcohol use disorder | .13 | .03 | 1.00 | .02 |
Substance use disorder | — | — | .80 | — |
Suicidal behavior | ||||
Thoughts | .32 | .86 | .40 | .11 |
Intent | .33 | .90 | .43 | .20 |
Attempt | .21 | .08 | .28 | .24 |
Firearm used in suicide | .54 | .73 | .37 | .82 |
Class 1 (N=375) | Class 2 (N=922) | Class 3 (N=70) | Class 4 (N=1,705) | ||||||
---|---|---|---|---|---|---|---|---|---|
Characteristic | N | % | N | % | N | % | N | % | pb |
Sex | <.001 | ||||||||
Male | 263 | 70 | 811 | 88 | 49 | 70 | 1,546 | 91 | |
Female | 112 | 30 | 111 | 12 | 21 | 30 | 159 | 9 | |
Age (M±SD years) | 69±11 | 72±10 | 61±9 | 72±10 | <.001 | ||||
Race | |||||||||
White | 354 | 94 | 882 | 96 | 67 | 96 | 1,612 | 95 | |
Black | 6 | 2 | 13 | 1 | 51 | 3 | |||
Asian | 16 | 2 | 22 | 1 | |||||
American Indian or Alaska Native | 6 | 2 | 6 | <1 | |||||
≥2 races | 11 | 1 | |||||||
Ethnicity | |||||||||
Hispanic | 14 | 2 | 36 | 2 | |||||
Non-Hispanic | 370 | 99 | 901 | 98 | 70 | 100 | 1,654 | 97 | |
Unknown/missing | 7 | 1 | 15 | 1 | |||||
Marital status | <.001 | ||||||||
Single | 223 | 60 | 467 | 51 | 53 | 76 | 865 | 51 | |
Married, civil union, domestic partnership | 152 | 41 | 455 | 49 | 17 | 24 | 840 | 49 | |
Ever served in military | 107 | 29 | 439 | 48 | 15 | 21 | 821 | 48 | |
Education level | .040 | ||||||||
Less than high school | 46 | 12 | 111 | 12 | 17 | 24 | 205 | 12 | |
Circumstantial characteristics | |||||||||
Undergoing mental health treatment | 234 | 62 | 41 | 5 | 23 | 33 | 104 | 6 | <.001 |
History of suicide attempts | 75 | 20 | 37 | 4 | 22 | 31 | 49 | 3 | <.001 |
Intimate partner problem | 27 | 7 | 24 | 3 | 4 | 6 | 35 | 2 | <.001 |
Recent suicide by friend or family | 12 | 1 | 12 | 1 | |||||
Other death of friend or family | 38 | 10 | 41 | 5 | 8 | 11 | 62 | 4 | <.001 |
Job problem | 7 | 2 | 6 | 1 | 10 | 1 | ns | ||
Financial problem | 16 | 4 | 13 | 1 | 27 | 2 | .003 | ||
Eviction or loss of home | 13 | 4 | 13 | 1 | 20 | 1 | .010 | ||
Recent criminal or legal problem | 8 | 1 | 6 | <1 |
The four classes that emerged were as follows (Table 3). Individuals in class 1 were characterized by a diagnosis of depression. Decedents in class 2 were highly likely to use a firearm and to have had suicidal thoughts and previously stated suicidal intent. In class 3, alcohol and substance use disorders were common as was a history of suicide attempts. Individuals in class 4 were characterized by a high proportion of firearm use with low proportions of mental illness, substance use disorders, or previous suicidal thoughts or behaviors.
Table 4 displays the demographic characteristics of individuals across the classes. In the firearm-only group (class 4), the proportion of men was relatively high, at 91%. Classes 1 and 3 had relatively higher proportions of women (both 30%). The marital status of decedents ranged from 52% to 60% in classes 1, 2, and 4, but in class 3 (in which alcohol and substance use disorders were common), 76% of decedents were reported as being “single” (Table 4). The proportion of people who did not complete high school was around 12% in classes 1, 2, and 4, but was 24% in class 3. Classes 2 and 4 had relatively few individuals undergoing mental health treatment (5% and 6%, respectively), whereas treatment was more common in classes 1 and 3 (62% and 33%, respectively). Class 1 (characterized by depression diagnoses) and class 3 (characterized by alcohol and substance use disorders) had greater proportions of individuals with previous suicide attempts (20% and 31%, respectively) than did classes 2 and 4 (4% and 3%, respectively).
Discussion
Very little research has been done to better understand suicide among individuals with terminal illness, despite general medical health problems being a risk factor for suicide. Advancing our primary research aim, we identified four distinct groups of individuals with terminal illness, ages ≥50 years, who died by suicide. Previous work has shown that individuals who die by firearm suicide are less likely to have a diagnosed mental illness or substance use disorder than are those who use other means for suicide (10). Our results reiterate this point in the distinct groups we identified: the two classes that had firearm use as an emergent characteristic also had low proportions of mental illness or substance use disorders. Notably, the lack of reported mental illness does not mean that such illness was not present, because most of mental illness diagnoses come from next-of-kin reports (31). More than three-quarters of our overall sample used a firearm as a mechanism for suicide, which is higher than estimates in other studies of all older adult suicides (1).
Our secondary goal was to examine demographic characteristics across classes. Classes 1 and 3 tended to mirror each other demographically and were also the smaller of the four classes. Relative to classes 2 and 4, there were higher proportions of women in classes 1 and 3 had, which also showed a lower marginal probability of firearm use for suicide; despite recent increases in firearm use as a means for suicide among females, they traditionally use less lethal methods than males (1). Despite these similarities, class 3 (characterized by higher levels of alcohol and substance use disorders) tended to have lower proportions of married persons, poorer education, and fewer individuals undergoing mental health treatment compared with class 1. In class 1 (characterized by a higher proportion of depression diagnoses) individuals were most likely to be receiving mental health treatment. Individuals in classes 2 and 4 had a high proportion of individuals having served in the military. Universal suicide screenings are implemented throughout the U.S. Department of Veterans Affairs (VA) and are easily accessible in the medical record. Our results might indicate gaps in VA service use for at-risk veterans or reluctance of veterans to report suicidal thoughts (32–34). Classes 2 and 4 also had the highest probability of firearm use, indicating the potential for situational suicidal behaviors, where suicide screening might be less effective.
More rigorous case-control or cohort studies are needed to assess whether, and if so, by how much, a terminal illness diagnosis increases a person’s risk for suicide. The results of our study, however, support a mechanism for such an assessment. Thomas Joiner’s interpersonal-psychological theory on suicide (IPS) states that serious attempts or deaths by suicide typically happen among those who both desire suicide and are capable of the act (35). Individuals who desire suicide either perceive themselves as a burden or have a sense of failed belongingness; capability is acquired through exposure to painful or provocative events (36). For firearm owners with a terminal illness, the presence or use of the firearm over time may desensitize the person to using this method for suicide. Similar to the IPS, the Three-Step Theory (3ST) of suicide suggests that hopelessness and pain can overwhelm feelings of connectedness to produce suicidal desire—suicidal desire is not enough to initiate action; instead, the capability (i.e., fearlessness) to make a suicide attempt must also be present (37). Firearm owners with a terminal illness may feel physical pain associated with their illness and reduced ability to form social connections and have the practical capability fostered by firearm access. Both theories support the mechanisms through which persons with a terminal illness and access to a firearm have increased risk for moving from suicidal desire to attempt.
Previous work has indicated that the diagnosis of medical problems presents a period of acute risk for the person (7) by increasing identification with the “perceived burdensomeness” construct in Joiner’s theory (38–41). Our results suggest that elevated risk may persist as symptoms progress and prognosis becomes one of terminal illness. Of note, diagnosed mental illness and substance use disorders were not common among the classes of individuals most likely to use firearms in our study. We believe this finding underscores the importance of building a public health approach beyond treatment environments for psychiatric or substance use disorders and beyond evidence-based behavioral interventions. Because firearms constitute such a high proportion of deaths from suicide in this population of individuals not seeking mental health care, traditional approaches to suicide prevention would likely not reach many individuals before a lethal attempt. Clinically, intervention points could include the periods of transition from curative to palliative care or to conversations about hospice care. The focus needs to shift from targeted interventions for those with the most plausible risk to broad interventions aimed at all individuals with a terminal diagnosis.
In light of our findings, a socially determined public health approach to firearm suicide prevention for terminally ill persons might suggest the development of screening, intervention, and policy for care settings where terminally ill patients are most likely to receive treatment. Lethal means assessment, when focused on firearms (i.e., asking “Do you have access to a firearm?”) may result in a reduction in suicide attempts and deaths (42). Safe storage interventions for firearm owners have been effective, particularly if they include a safety device as part of the intervention (43–46). Although our study did not show a causal link between terminal illness and firearm suicide risk, the cost of implementing a lethal means assessment or firearm safety intervention when it is clear that a person’s prognosis is terminal is a relatively low-cost intervention.
This descriptive characterization of terminal illness and suicide had several limitations. First, although we observed a pattern of suicide due to use of a firearm among individuals with a terminal illness, we make no claims of a causal link between firearm use for suicide and the diagnosis of a medical illness. Our identification of terminally ill persons and their diagnosis of mental illness or substance use disorders was limited to information provided in narrative case reports from law enforcement and coroner or medical examiner reports, and individuals with a terminal illness were likely undercounted. Reporting of mental illness in NVDRS studies, however, appears to mirror other literature on suicide and psychiatric diagnoses (47). Furthermore, this information was often derived from interviews with next of kin, a method that is subject to recall bias. Our definition of terminal illness likely underestimated the number of those with a terminal illness because it is unlikely that all cases of general medical illness were described in the narrative case reports. In addition, suicides, particularly poisoning (48), are sometimes misclassified as accidental death. Finally, the early years in the NVDRS have fewer reporting states, incrementally increasing to 41 states and territories in 2018, limiting the generalizability of these data (49).
Conclusions
Our study documents typologies of individuals with a terminal illness who died by suicide and reveals that the overwhelming majority used a firearm. Relatively few persons had a diagnosed mental illness or substance use disorder, which are factors traditionally used to identify someone as high suicide risk in a clinical setting. Screening, intervention, and referral procedures shown to be effective (42) for individuals without a terminal illness, such as lethal means assessment and firearm safe storage programs, could be tested in clinical trials. Firearm suicide is a particularly traumatic and violent way to end one’s life, and clinicians who frequently deliver terminal diagnoses or refer patients to palliative care should take care to address firearm safety as part of care for their patients, in collaboration with their family care partners.
1. WISQARS—Web-Based Injury Statistics Query and Reporting System. Atlanta, Centers for Disease Control and Prevention, National Center for Injury Prevention and Control,2005. www.cdc.gov/injury/wisqars. Accessed May 1, 2022 Google Scholar
2. : Hard truths about suicide prevention. JAMA Netw Open 2020; 3:e2022713Crossref, Medline, Google Scholar
3. : Preventing suicide through better firearm safety policy in the United States. Psychiatr Serv 2021; 72:174–179Link, Google Scholar
4. : Implementation and effectiveness of Connecticut’s risk-based gun removal law: does it prevent suicides? Law Contemp Probl 2017; 80:179–208 Google Scholar
5. : Understanding the research on extreme risk protection orders: varying results, same message. Psychiatr Serv 2019; 70:953–954Link, Google Scholar
6. : Lethality of firearms relative to other suicide methods: a population based study. J Epidemiol Community Health 2003; 57:120–124Crossref, Medline, Google Scholar
7. : Perceived burdensomeness and suicidality: two studies on the suicide notes of those attempting and those completing suicide. J Soc Clin Psychol 2002; 21:531–545Crossref, Google Scholar
8. : Suicides among Danish patients with cancer: 1971 to 1986. Cancer 1992; 69:1507–1512Crossref, Medline, Google Scholar
9. : Suicidal behavior in patients diagnosed with cancer of the cervix. Crisis 2003; 24:168–172Google Scholar
10. : The association of firearm suicide with mental illness, substance use conditions, and previous suicide attempts. Ann Intern Med 2017; 167:287–288Crossref, Medline, Google Scholar
11. : Attributable risk of psychiatric and socio-economic factors for suicide from individual-level, population-based studies: a systematic review. Soc Sci Med 2011; 72:608–616Crossref, Medline, Google Scholar
12. : End-of-life care: issues relevant to the geriatric psychiatrist. Am J Geriatr Psychiatry 2004; 12:457–472Crossref, Medline, Google Scholar
13. : Psychological distress in cancer patients with advanced disease. Radiother Oncol 1993; 27:193–197Crossref, Medline, Google Scholar
14. : Psychological distress in patients with advanced cancer. Clin Med 2006; 6:148–150Crossref, Medline, Google Scholar
15. : Assessing and managing depression in the terminally ill patient. ACP-ASIM End-of-Life Care Consensus Panel. American College of Physicians—American Society of Internal Medicine. Ann Intern Med 2000; 132:209–218Crossref, Medline, Google Scholar
16. : Priorities for Research to Reduce the Threat of Firearm-Related Violence. Washington, DC, National Academies Press, 2013. Google Scholar
17. : Firearm suicide among persons with terminal illness. J Pain Symptom Manage 2022; 63:e260–e263Crossref, Medline, Google Scholar
18. National Violent Death Reporting System. Atlanta, Centers for Disease Control and Prevention, n.d. https://www.cdc.gov/violenceprevention/pdf/NVDRS-factsheet508.pdf Google Scholar
19. JMP, Version 15.2.0. Cary, NC, SAS Institute, 2019 Google Scholar
20. Stata Statistical Software, Release 17. College Station, TX, StataCorp, 2021 Google Scholar
21. : Risk factors for child death during an intimate partner homicide: a case-control study. Child Maltreat 2021; 26:356–362Crossref, Medline, Google Scholar
22. : Changes in suicide rates—United States, 2019 and 2020. MMWR Morb Mortal Wkly Rep 2022; 71:306–312Crossref, Medline, Google Scholar
23. Cancer Stat Facts: Cancer of Any Site. Bethesda, National Institutes of Health, National Cancer Institute, 2022. https://seer.cancer.gov/statfacts/html/all.html. Accessed May 1, 2022 Google Scholar
24. : Fundamentals of Predictive Analytics With JMP. Cary, NC, SAS Institute, 2016 Google Scholar
25. Terminal Cancer. Bethesda, National Institutes of Health, National Cancer Institute, 2021. https://www.cancer.gov/publications/dictionaries/cancer-terms/def/terminal-cancer. Accessed Nov 23, 2021 Google Scholar
26. : Risk factors for suicide in individuals with depression: a systematic review. J Affect Disord 2013; 147:17–28Crossref, Medline, Google Scholar
27. : Life events and suicidal ideation and behavior: a systematic review. Clin Psychol Rev 2014; 34:181–192Crossref, Medline, Google Scholar
28. : Applied latent class analysis. Can J Sociol 2003; 28:584–586 Crossref, Google Scholar
29. : Latent class analysis; in Latent Class Analysis. Edited by McCutcheon A. Newbury Park, CA, SAGE, 2011 Google Scholar
30. : A typology of civilians shot and killed by US police: a latent class analysis of firearm legal intervention homicide in the 2014–2015 National Violent Death Reporting System. J Urban Health 2020; 97:317–328Crossref, Medline, Google Scholar
31. : Do undiagnosed suicide decedents have symptoms of a mental disorder? J Clin Psychol 2017; 73:1744–1752Crossref, Medline, Google Scholar
32. : Impact of attitudes and rurality on veterans’ use of Veterans Health Administration mental health services. Psychiatr Serv 2021; 72:521–529Link, Google Scholar
33. : Combat duty in Iraq and Afghanistan, mental health problems, and barriers to care. N Engl J Med 2004; 351:13–22Crossref, Medline, Google Scholar
34. : Suicide risk assessment and content of VA health care contacts before suicide completion by veterans in Oregon. Psychiatr Serv 2010; 61:1192–1197Link, Google Scholar
35. : Why People Die by Suicide. Cambridge, MA, Harvard University Press, 2005 Google Scholar
36. : Capable of suicide: a functional model of the acquired capability component of the interpersonal-psychological theory of suicide. Suicide Life Threat Behav 2010; 40:266–275Crossref, Medline, Google Scholar
37. : The three-step theory of suicide: description, evidence, and some useful points of clarification. Prev Med 2021; 152:106549Crossref, Medline, Google Scholar
38. : Characteristics of terminal cancer patients who committed suicide during a home palliative care program. J Pain Symptom Manage 2001; 22:544–553Crossref, Medline, Google Scholar
39. : Chronic dyspnea and suicide in elderly men. Hosp Community Psychiatry 1992; 43:1198–1203Abstract, Google Scholar
40. : Cancer and suicide. Ann Oncol 1991; 2:19–23Crossref, Medline, Google Scholar
41. : Late-life suicide in terminal cancer: a rational act or underdiagnosed depression? J Pain Symptom Manage 2017; 54:835–842Crossref, Medline, Google Scholar
42. : A quasi-experimental analysis of lethal means assessment and risk for subsequent suicide attempts and deaths. J Gen Intern Med 2020; 35:1709–1714Crossref, Medline, Google Scholar
43. : Effectiveness of interventions to promote safe firearm storage. Epidemiol Rev 2016; 38:111–124Medline, Google Scholar
44. : Effectiveness of gun-safety counseling and a gun lock giveaway in a Hispanic community. Arch Pediatr Adolesc Med 2005; 159:1049–1054Crossref, Medline, Google Scholar
45. : Is office-based counseling about media use, timeouts, and firearm storage effective? Results from a cluster-randomized, controlled trial. Pediatrics 2008; 122:e15–e25Crossref, Medline, Google Scholar
46. : Improving firearm storage in Alaska native villages: a randomized trial of household gun cabinets. Am J Public Health 2012; 102(suppl 2):S291–S297Crossref, Medline, Google Scholar
47. : Race/ethnicity, substance abuse, and mental illness among suicide victims in 13 US states: 2004 data from the National Violent Death Reporting System. Inj Prev 2006; 12(suppl 2):ii22–ii27Medline, Google Scholar
48. : Classifying undetermined poisoning deaths. Inj Prev 2006; 12:338–343Crossref, Medline, Google Scholar
49. National Violent Death Reporting System Web Coding Manual—Version 5.4.1. Atlanta, Centers for Disease Control and Prevention, 2021. https://www.cdc.gov/violenceprevention/pdf/nvdrs/nvdrsCodingManual.pdf Google Scholar