FIGURE 2. Case Summaries
The case summaries reflect key clinical details from 4 fathers’ narratives (see accompanying article for more details from these parents). In each case, the decedent had recent contact with a clinician and denied suicidal ideation, after which death by suicide occurred. These narratives have been published with permission from the fathers.
Case 1: Eric, college student, male
Family and social history: College freshman with supportive family
Psychiatric history: Three days’ hospitalization after reporting ideation; discharged with referral to psychiatrist; diagnosed with bipolar disorder and started on complex medication regimen and recommended outpatient therapy
History of illness preceding death: Returned to college; continued treatment; voices under control with medication; appeared stable during follow-ups; psychiatrist and therapist reported patient was safe
Outcome: Died by suicide (intentional overdose of prescribed medication during a solo road trip)
Case 2: Alyssa, female, aged 24 years
Family and social history: High achiever; perfectionistic
Psychiatric history: Depression beginning in late high school; multiple therapists and psychiatrists with different diagnoses; multiple medications and residential programs, resulting in repeated withdrawals from college
History of illness preceding death: Developed severe insomnia, intense rumination, and isolation; patient’s father requested hospitalization, but the psychiatrist did not believe it was necessary, as patient denied suicidal ideation
Died by suicide by self-asphyxiation days later
Case 3: Kiele, adult female
Family and social history: High-functioning professional; meticulous, perfectionistic, brilliant
Psychiatric history: Recent diagnosis of generalized anxiety disorder
History of illness preceding death: Weeks of severe anxiety and extreme insomnia; telemedicine visit with a psychologist 2 days prior, and documented racing thoughts but no suicidal ideation; cognitive behavioral therapy was planned; following day, patient expressed inability to control negative thoughts
Outcome: Died by suicide the following day; note described overwhelming anxiety
Case 4: Maddy, male, aged 26 years
Family and social history: Outgoing, funny, and compassionate; mechanical engineer; relocated a week ago for a new job
Psychiatric history: None reported
History of illness preceding death: Rapid onset of symptoms in 6 days; insomnia, agitation, racing and irrational thoughts, intense fear; psychiatrist conducted a standard phone screening; denied suicidal ideation and was considered safe until scheduled appointment next day
Outcome: Died by suicide before dawn the following morning
Stage 2. Trigger: Stressful Life Events.
Stressful life events (SLEs), which are distinct from traumatic events, are events that can cause a significant change or adjustment in an individual’s life, accompanied by severe psychological distress.18 Examples in adults include final romantic rejection; catastrophic financial failure, including job loss, eviction, or foreclosure; threat to one’s core identity, such as academic failure or collapse of a career or of a family, or incarceration; and onset of a life-threatening medical illness or serious mental illness, bullying, and bereavement.13 Evidence indicates that both the timing and cumulative number of SLEs are associated with suicidality, with proximal events and multiple events demonstrating stronger associations.18
According to NCM, when an individual with predisposing vulnerabilities experiences one or more SLEs, it may trigger the next (subacute) stage of the suicidal mental process called suicidal narrative.1 It is important to note that SLEs do not directly trigger suicidal behavior but rather act as a catalyst that enables the mental state of the individual to progress to the next stage in the NCM.
Case Reflections
Returning to the 4 case examples (Figure 2), Eric first disclosed suicidal ideation during his freshman-year vacation, suggesting that the stress of college adjustment may have triggered symptom onset; subsequent concerns the following spring further support this interpretation. Likewise, Alyssa’s burnout during her senior year, compounded by her perfectionistic self-expectations, may have contributed to a perception of personal failure, triggering her progression into the suicidal narrative stage.
Stage 3. Subacute: Suicidal Narrative.
During the suicidal narrative (SN) stage, the individual experiences a shift in their cognitive-affective processes, which leads them to adopt a self-defeating internal narrative.13 In this narrative, they experience a distorted sense of self, diminished self-worth, and interpersonal alienation, which leads them to perceive life as intolerable and their future as unimaginable. The SN framework identifies 8 interrelated components (Figure 3) that collectively capture the psychological and social processes underlying this subacute stage.1
The first 3 components of SN are grounded in goal disengagement theory, which posits that failure to disengage from unattainable goals, especially when coupled with feelings of entitlement to happiness, can lead to severe distress.1 The components (ie, social defeat and fear of humiliation, drawn from the integrated motivational-volitional model [IMV] and the cry of pain model) reflect experiences of entrapment, social rejection, and shame, which are critical predictors of suicidality. Originating from the interpersonal theory of suicide, the final components (ie, thwarted belongingness and perceived burdensomeness) reflect feelings of isolation and self-perceived liability to others.
The SN stage may last for months, marked by persistent cognitive inflexibility, perceived social disconnection, and internalized beliefs of worthlessness, which are experienced as fixed and unchangeable.13 When this narrative reaches a critical level of intensity, it precipitates the next and most acute stage of the NCM, the suicide crisis syndrome.
Case Reflections
Retrospectively, the fathers’ stories illustrate how the SN may have unfolded, narrowing perception and constricting cognition and affect. Eric’s withdrawal from school may have been internalized as personal failure, eroding identity and self-worth. Returning to college, although signaling recovery, may have unintentionally reexposed him to prior stressors, reinforcing a narrative of entrapment and hopelessness. In Alyssa’s case, repeated attempts to return to New York University suggest goal fixation and inability to disengage from idealized self-expectations. Each perceived failure likely reinforced humiliation and diminished her self-worth, consolidating a narrative in which life felt intolerable. For Kiele, the sudden onset of severe anxiety and insomnia may have accelerated her SN, producing rapid internalization of entrapment and loss of cognitive control. Her perception of suicide as the only escape reflects the tightening grip of a self-defeating narrative. Maddy’s abrupt cognitive and affective dysregulation illustrates how quickly the SN can collapse into acute crisis. Within days or hours, his narrative fragmented, leaving the perception of suicide as the only solution.
Stage 4. Acute: Suicide Crisis Syndrome.
Suicide crisis syndrome (SCS) represents an acute, transient suicidal mental state that is predictive of imminent suicidal behavior.15 Originally conceptualized more than a decade ago as the suicide trigger state, SCS has since been refined through empirical findings, reflecting the iterative nature of rigorous scientific research and efforts to better understand acute suicidal states.1 SCS is characterized by a constellation of cognitive, affective, physiological, and behavioral symptoms that typically emerge within hours to days and culminate in a suicide attempt or death by suicide.
SCS is characterized by 5 interrelated symptom domains, grouped under 2 diagnostic criteria (Figure 4).1 Building on the concept of the suicide crisis—a short-term suicidal state characterized by intense desperation, self-hatred, rage, anxiety, and loneliness—the NCM proposes that during SCS the individual experiences an overwhelming sense of entrapment with an urgent need to escape.1 This is accompanied by heightened affective disturbances, extreme loss of cognitive control, hyperarousal, and social withdrawal. Combined, these symptoms create a cascading crisis that propels the individual toward suicidal action. Crucially, recognition of SCS warrants immediate high-intensity intervention as it indicates imminent danger of suicide even in the absence of disclosed SI.2
The NCM proposes that SCS can manifest even in the absence of SI, allowing for more objective identification of imminent suicidality without relying on SI disclosure. This feature enhances its potential utility as a clinical diagnostic tool.1 Current clinical practices typically use clinical interviewing to determine treatment and next steps for individuals stratified as moderate or high risk based on the presence of SI, yet a substantial number of those who die by suicide are classified as low risk.4,5 In contrast, the NCM proposes that evaluation using a brief but structured clinical interview to identify SCS (an acute suicidal state) can reduce the likelihood of missed opportunities for intervention.2 Supporting this approach, both self-report and clinician-rated measures of SCS, adapted and implemented across clinical settings worldwide, have demonstrated strong clinical utility and predictive validity for the prediction of near-term suicide attempts, with Area Under the Curve (AUC) scores ranging between 0.733 and 0.883.19-29
Furthermore, longitudinal studies show that SCS predicts prospective suicidal ideation and attempts, with the NCM explaining only 10.8% of variance in ideation as opposed to 40.7% in attempts at 1-month follow-up.1 Empirical studies to date examining NCM have also found significant associations between chronic risk factors such as perfectionism, impulsivity, and substance use and suicidal ideation and behaviors, with effects largely explained by SCS.17 Additionally, SLEs have been shown to mediate progression from the chronic to subacute stage, thereby offering evidence supporting the staged progression of suicidal mental states.17
SCS is being evaluated as a suicide-specific diagnosis and is under review by the DSM-5 Steering Committee as a potential predictor of imminent suicide risk for inclusion in future revisions. More research is being conducted in response to the committee’s feedback.19-29
Case Reflections and Clinical Takeaways
The 4 cases (Figure 2) illustrate the potential utility of the NCM and SCS. For example, Eric’s story exemplifies the invisibility of the acute SCS state during therapy or standard screening. Although Eric’s care team tracked symptoms and self-reports, the internal experience of frantic hopelessness and cognitive dysregulation was missed. His withdrawal, flat reassurances, and diminished engagement in therapy suggested loss of cognitive control and social withdrawal, hallmarks of the SCS stage. The father noted that relying on patients to tell us how they are doing is insufficient and highlights the need for structured, objective recognition of acute states beyond ideation-based checklists.
Alyssa’s presentation demonstrates the full dimensionality of SCS: persistent emotional pain, cognitive flooding, exhaustion and hyperarousal from insomnia, and profound social withdrawal. Her ruminative self-monitoring (“metacognition”) exemplifies loss of cognitive control. Her denial of suicidality was not a contradiction but a symptom of her cognitive rigidity and inability to perceive alternatives. This case powerfully demonstrates that absence of suicidal ideation does not equate to absence of imminent risk.
Similarly, Kiele’s case vividly illustrates hyperarousal, affective disturbances, and loss of cognitive control within SCS. Severe insomnia and agitation correspond to hyperarousal; pervasive anxiety and emotional pain represent affective disturbance; racing thoughts and her belief that she had no ability to control her negative thinking demonstrate loss of cognitive control. Her note’s reference to “anxiety hell” and “peace of mind” through death reflect a state of entrapment and frantic hopelessness, perceiving suicide as the only escape from intolerable anxiety. Kiele’s case underscores the clinical need for dynamic risk assessment that recognizes affective and cognitive dysregulation as imminent warning signs, even in high-functioning patients who deny ideation.
Maddy’s case demonstrates an acute-onset SCS presentation with striking temporal proximity to suicide, symptoms evolving within days. The abrupt onset of insomnia, escalating anxiety, and expressed fear of “what is happening to my brain” signal cognitive and affective dysregulation. His denial of ideation was truthful yet misleading within the standard assessment protocol. For clinicians, this underscores the critical importance of identifying SCS symptom clusters as markers of imminent suicide risk, even in the absence of stated intent.
Across these 4 cases, SCS emerged as an acute, rapidly evolving state characterized by entrapment, cognitive-affective dysregulation, hyperarousal, and social withdrawal. Consistent with this, research in adult psychiatric populations shows that 4 of the 5 SCS dimensions significantly predict future suicide attempts, with AUC values above 0.84; only the social withdrawal dimension demonstrated a lower AUC (0.633). Overall, at optimal cut points, the SCS showed 67% sensitivity and 78% specificity for predicting suicidal behavior, and 100% sensitivity with 82% specificity for predicting suicide attempts.21 Each case underscores that imminent suicide risk may be present even in the absence of expressed SI, and that traditional, ideation-based assessments may fail to detect this crisis state. Early recognition by screening for abrupt onset of severe insomnia, agitation, loss of cognitive control, or feelings of entrapment followed by prompt, high-intensity intervention is critical to preventing suicide during this acute stage.2,4
It is critical to note that NCM and SCS are suggested as useful clinical guidelines to be incorporated alongside existing clinical protocols. Until SCS is adopted as an official DSM diagnosis with a diagnostic code and as a potential reason for involuntary hospitalization, it should be used as an adjunct to existing SRAs to inform clinicians’ overall judgment. Comprehensive clinical judgment should continue to guide any high-intensity interventions, ensuring a delicate balance between prevention and patient rights.4
Comparison With Other Theoretical Frameworks
Distinctive Features. The NCM represents a paradigm shift in how suicide is conceptualized.13 NCM incorporates well-established long-term vulnerabilities and SLEs already included in current risk assessment frameworks. It also integrates key elements from existing models in a streamlined manner to conceptualize a subacute suicidal narrative stage and introduces a novel acute stage that captures the suicidal crisis state.2 Rather than stratifying individuals based on the presence or intensity of SI, it focuses on the underlying mechanisms that drive the transition from chronic vulnerability to acute crisis.1
Although NCM shares overlapping constructs with other established theories, it differs in key ways. Traditional ideation-to-action models, such as the interpersonal theory of suicide, the IMV, and the three-step theory (3ST), conceptualize SI as a necessary precursor to suicidal behavior. In contrast, the NCM proposes that SI may be absent, transient, or appear only during the final stages of crisis.2
Central to the NCM is the SCS, a discrete and acute mental state characterized by cognitive and emotional dysregulation that can precipitate suicidal behavior independent of conscious ideation or intent.13 This distinguishes it from IMV, which emphasizes a gradual transition from ideation to intent through motivational and volitional phases, and from the 3ST, which frames suicidal behavior as a progression from psychological pain and hopelessness to active ideation and, ultimately, action.1 Similarly, while both the acute suicidal affective disturbance (ASAD) model and the NCM describe acute suicidal crisis states, ASAD’s emphasis on a sudden escalation of SI distinguishes it from the NCM.
Another prominent ideation-based framework is the C-SSRS, primarily an assessment tool that stratifies risk based on past and present SI, intent, plan, and behavior.5 In contrast, the NCM is a theoretical framework describing the suicidal mental process beyond SI by integrating distal and proximal factors. Notably, SCS has demonstrated strong predictive validity for near-term suicidal behavior, including incremental predictive validity over C-SSRS.19
Shared Conceptual Foundations. Despite its unique structure and emphasis, NCM shares several conceptual foundations with earlier frameworks, such as diathesis-stress models of suicide, which propose that enduring vulnerabilities like impulsivity or hopelessness interact with stressors to produce suicidal outcomes.1 NCM extends this perspective by introducing 2 intermediary stages, the SN and the SCS, which together trace the progression from enduring vulnerability to acute suicidality.
Across models, recurring constructs such as trait vulnerability, emotional dysregulation, entrapment, hopelessness, burdensomeness, and defeat reflect a shared understanding of critical predisposing factors.1 The NCM reorganizes and contextualizes these elements within a temporally sensitive, layered framework that emphasizes acute, suicide-specific psychopathology. Through this integrative approach, the NCM accounts for both long-term predispositions and acute clinical states, offering a more nuanced understanding of the transition from trait vulnerability to suicidal action state.
Clinical Significance of NCM
The structured approach may help clinicians create personalized case formulations and strengthen therapeutic alliance through shared insight into the patient’s experiences. The model’s ability to incorporate both distal factors and proximal indicators allows clinicians to assess imminent threat more effectively.20,28,29 Although skilled clinicians may intuitively consider these factors, the NCM systematizes the assessment of distal vulnerabilities, proximal triggers, and acute suicidal states, offering a structured framework that ensures critical risk indicators are not overlooked.13 Importantly, screening for SCS symptom domains allows clinicians to identify acute suicide risk objectively, even in the absence of disclosed SI, as illustrated by the case narratives.
The model also supports a staged treatment approach, with interventions tailored to each stage.13 For example, clinicians can address enduring trait vulnerabilities using long-term therapy (stage 1) and response to SLEs using stress-reduction strategies (stage 2). Cognitive restructuring and intensive crisis management can be used to target the subacute suicidal narrative (stage 3). For individuals experiencing SCS, the model’s most acute stage (stage 4), interventions may include aggressive means restriction and fast-acting targeted crisis pharmacotherapy for stabilization, which are standard, routinely administered treatments in clinical practice.
Beyond clinical intervention, the NCM serves as a psychoeducational tool, enabling patients to better understand the psychological processes that contribute to suicidal thoughts and behavior as well as recognize when they are at acute stage, thereby encouraging timely help-seeking.1
Limitations and Critiques
Although the NCM and SCS have gained empirical support, as is often the case with new diagnostic categories, several limitations and critiques of the proposal to formalize SCS as a psychiatric diagnosis have been raised regarding its diagnostic structure, predictive scope, and potential medicalization.30 Some scholars note that requiring at least 1 symptom from each domain of SCS may be overly restrictive, potentially excluding individuals experiencing an acute crisis who present with only a subset of features. Accordingly, studies are ongoing to assess clinical significance of mental states meeting 3 or 4 SCS criteria.19
Others have highlighted risk factors not explicitly included in the current criteria, such as recent psychiatric hospitalization, acute substance use escalation, or broader sleep disturbances.30 Therefore, it is important to reiterate that clinicians must continue to consider additional risk factors during the broader SRA evaluation while using SCS to detect acute suicidal mental states that may occur even in the absence of disclosed SI.4,13
Predictive limitations also exist. Like all suicide research, SCS faces the low base rate problem, and time frames in research studies (often 0-60 days) may not fully align with the clinical definition of imminence, typically the first week post discharge.19 Finally, concerns have been raised about the potential medicalization of acute distress, with possible implications for insurance eligibility or employment.30 Proponents argue, however, that the conceptualization of SCS is not intended to pathologize transient distress, but rather provide a structured approach for identifying and responding to individuals experiencing an acute suicidal crisis that might otherwise go undetected and treated inadequately, for example, with antidepressants, which may instead exacerbate symptoms associated with acute suicidality in the short term.13,31,32
These considerations underscore the need for ongoing empirical validation and careful clinical implementation to optimize SCS’s utility within existing risk assessment frameworks.
Concluding Thoughts
Empirical findings underscore the limitations of ideation-based models and risk assessment, revealing high rates of nondisclosure and suicide attempts occurring in individuals who denied SI during clinical encounters.1,9 This gap has prompted a shift toward understanding acute crisis states that may precede action, even in the absence of explicit ideation. NCM and SCS represent novel ways of thinking and approaches to better support clinicians in identifying patients at risk for suicide. Although more time and data are needed to further support the reliability and validity of the model and tools, they can provide additional support. Clinicians are reminded to always use evidence-based medicine and to consider malpractice and liability issues when supporting patients.
Ms Balasubramanian is a clinical research coordinator at the Icahn School of Medicine at Mount Sinai, New York.
Dr El Hayek is a postdoctoral research fellow at the Icahn School of Medicine at Mount Sinai, New York.
Dr Galynker is a clinical professor of psychiatry and the director of the Suicide Research and Prevention Laboratory at the Icahn School of Medicine at Mount Sinai, New York. He is the founder and director of the Galynker Family Center for Bipolar Disorder and the Galynker Family Center for Suicide Prevention.
Acknowledgments
The authors acknowledge the support of the American Foundation for Suicide Prevention, the Eric Masinter Memorial Fund, and the Kiele Miller Memorial Fund.
References
1. Bloch-Elkouby S, Rogers ML, Goncearenco I, et al. The narrative crisis model of suicide: a review of empirical evidence for an innovative dynamic model of suicide and a comparison with other theoretical frameworks. Pers Med Psychiatry. 2024;45-46:100131.
2. Galynker I. The Suicidal Crisis: A Clinical Guide to the Assessment of Imminent Suicide Risk. 2nd ed. Oxford University Press; 2023.
3. Ryan EP, Oquendo MA. Suicide risk assessment and prevention: challenges and opportunities. Focus (Am Psychiatr Publ). 2020;18(2):88-99.
4. Berman AL. Risk factors proximate to suicide and suicide risk assessment in the context of denied suicide ideation. Suicide Life Threat Behav. 2018;48(3):340-352.
5. Richardson JS, Brabbs S, Szymanski BR, et al. Predictive properties of the C-SSRS suicide screen among individuals in Veterans Health Administration care. Arch Suicide Res. Published online February 27, 2026.
6. Mérelle S, Foppen E, Gilissen R, et al. Characteristics associated with non-disclosure of suicidal ideation in adults. Int J Environ Res Public Health. 2018;15(5):943.
7. Blanchard M, Farber BA. “It is never okay to talk about suicide”: patients’ reasons for concealing suicidal ideation in psychotherapy. Psychother Res. 2020;30(1):124-136.
8. Calear AL, Batterham PJ. Suicidal ideation disclosure: patterns, correlates and outcome. Psychiatry Res. 2019;278:1-6.
9. Husky MM, Zablith I, Alvarez Fernandez V, Kovess-Masfety V. Factors associated with suicidal ideation disclosure: results from a large population-based study. J Affect Disord. 2016;205:36-43.
10. Kleiman EM, Turner BJ, Fedor S, et al. Examination of real-time fluctuations in suicidal ideation and its risk factors: results from two ecological momentary assessment studies. J Abnorm Psychol. 2017;126(6):726-738.
11. Hallford DJ, Rusanov D, Winestone B, et al. Disclosure of suicidal ideation and behaviours: a systematic review and meta-analysis of prevalence. Clin Psychol Rev. 2023;101:102272.
12. Jones JD, Boyd RC, Calkins ME, et al. Parent-adolescent agreement about adolescents’ suicidal thoughts. Pediatrics. 2019;143(2):e20181771.
13. Rogers ML, Bloch-Elkouby S, Goncearenco I, et al. The narrative crisis model of suicide as a framework for suicide prevention. Pers Med Psychiatry. 2024;45-46:100130.
14. Franklin JC, Ribeiro JD, Fox KR, et al. Risk factors for suicidal thoughts and behaviors: a meta-analysis of 50 years of research. Psychol Bull. 2017;143(2):187-232.
15. O’Connor RC. The relations between perfectionism and suicidality: a systematic review. Suicide Life Threat Behav. 2007;37(6):698-714.
16. Sommerfeld E, Malek S. Perfectionism moderates the relationship between thwarted belongingness and perceived burdensomeness and suicide ideation in adolescents. Psychiatr Q. 2019;90(4):671-681.
17. Cohen LJ, Ardalan F, Yaseen Z, Galynker I. Suicide crisis syndrome mediates the relationship between long-term risk factors and lifetime suicidal phenomena. Suicide Life Threat Behav. 2018;48(5):613-623.
18. Howarth EJ, O’Connor DB, Panagioti M, et al. Are stressful life events prospectively associated with increased suicidal ideation and behaviour? A systematic review and meta-analysis. J Affect Disord. 2020;266:731-742.
19. Melzer L, Forkmann T, Teismann T. Suicide crisis syndrome: a systematic review. Suicide Life Threat Behav. 2024;54(3):556-574.
20. Bloch-Elkouby S, El-Hayek R, Cohen L, et al. The clinician rated suicide crisis syndrome checklist (SCS-C): structure, reliability, and concurrent validity among adult psychiatric inpatients. J Affect Disord. 2026;402:121341.
21. Bloch-Elkouby S, Barzilay S, Gorman BS, et al. The revised suicide crisis inventory (SCI-2): validation and assessment of prospective suicidal outcomes at one month follow-up. J Affect Disord. 2021;295:1280-1291.
22. Menon V. The narrative-crisis model of suicide as an emerging framework for suicide risk stratification, management, and prevention. Indian J Psychol Med. 2024;46(4):285-288.
23. Barzilay S. Suicide crisis syndrome in youth. Presented at: IASR/AFSP International Summit on Suicide Research; 2021; virtual.
24. Chistopolskaya KA, Rogers ML, Enikolopov SN, et al. Adaptation of the suicide crisis inventory (SCI-2) in a Russian internet sample during the COVID-19 epidemic. Suicidology. 2022;13(1):88-104.
25. Otte S, Lutz M, Streb J, et al. Analyzing suicidality in German forensic patients by means of the German version of the suicide crisis inventory (SCI-G). J Forensic Psychiatry Psychol. 2020;31(5):731-746.
26. Park JY, Rogers ML, Bloch-Elkouby S, et al. Factor structure and validation of the revised suicide crisis inventory in a Korean population. Psychiatry Investig. 2023;20(2):162-173.
27. Wu CY, Lee JI, Lee MB, et al. Identification of suicidality by five-item suicide crisis scale (SCS-5) in an online general population survey in Taiwan. J Suicidol. 2022;17(4):342-351.
28. Mitelman AM, Bafna A, Rogers ML, et al. Perceived clinical utility of the suicide crisis syndrome: a multisite pilot study. Psychol Serv. 2023;20(suppl 2):61-67.
29. Osváth P, Molnár C, Lovig C, et al. Az akut szuicid veszély értékelése: A komplex rizikóbecslés lehetőségei a klinikai gyakorlatban. Orvosi Hetilap. 2025;166(4):146-153.
30. Berman AL, Silverman MM. A suicide-specific diagnosis – the case against. Crisis. 2023;44(3):183-188.
31. Pompili M, Serafini G, Innamorati M, et al. Antidepressants and suicide risk: a comprehensive overview. Pharmaceuticals (Basel). 2010;3(9):2861-2883.
32. Jick H, Kaye JA, Jick SS. Antidepressants and the risk of suicidal behaviors. JAMA. 2004;292(3):338-343.