Future Directions in Understanding and Interpreting Discrepant Reports of Suicidal Thoughts and Behaviors among Youth

Abstract

Both the quality and utility of youth suicide research depend on how we assess our outcomes of interest: suicidal thoughts and behaviors (STBs). We now have access to more STB assessments than ever before, with measures for youth that vary in what exact experiences are asked about, how such measures elicit information, when and how frequently measures are administered, and who the informants are. This growing armamentarium of assessments has the potential to improve the study and treatment of STBs among youth, but it hinges on meaningful interpretation of assessment responses. Interpretation can be especially challenging when different STB assessments yield conflicting information. Determining how to manage discrepant reports of STBs is a pivotal step toward achieving meaningfully comprehensive STB assessment batteries. Here, we outline several discrepant reporting patterns that have been detected, discuss the potential significance of these observed discrepancies, and present initial steps to formally investigate discrepant reports of STBs among youth. Developing coherent, interpretable, and comprehensive batteries assessing STBs among youth would address a fundamental step to uncovering etiology, improving clinical decision-making and case management, informing intervention development, and tracking prognosis.

The idea of a young person wanting or trying to end their own life raises unquestionable concern and alarm. Consensus around the importance of helping children and adolescents, hereafter referred to as youth, who have experienced suicidal thoughts and behaviors (STBs) is evidenced by multiple calls to improve both research and treatment of suicidal youth in recent years (Ayer et al., 2020; Esposito-Smythers et al., 2021; Riley et al., 2021). This is aligned with a global movement to reduce suicide rates within the next decade through concrete goals set within the United States (e.g., reducing suicide rates by 20 percent by 2025; American Foundation of Suicide Prevention, 2018) and around the world (e.g., reducing suicide rates by one-third by 2030; World Health Organization, 2021).

Despite the widely-accepted sense of urgency around addressing youth suicide risk, efforts to both understand and treat suicidal youth have been steady and slow. Prior research efforts testing risk factors of STBs (Franklin et al., 2017) and interventions targeting STBs (Fox et al., 2020) have yielded modest effects. Moreover, suicide death has continued to rise over the past two decades, with disproportionate increases in suicide rates among specific populations such as Black elementary school-aged children (Bridge et al., 2015) and early adolescent girls (Luby & Kertz, 2019). These rising suicide rates further amplify these calls for tangible reductions in youth suicide risk and call for a re-evaluation of how we achieve corresponding future goals.

Suicide research initiatives often overlook the central importance of rigorously assessing STBs. Most proposed solutions to addressing suicide risk pertain to the study around STBs, rather than the study of STBs themselves. In the past, researchers have predominantly pursued questions tied to what leads to STBs (Franklin et al., 2017) or how to mitigate them over time (Fox et al., 2020). Looking forward, current research agendas similarly promote the importance of advancing the prediction and prevention of suicide (Centers for Disease Control and Prevention [CDC], 2021; Gordon et al., 2020). As critical as these research pursuits may be, they rest on the assumption that we as a field have a full grasp of how to harness assessment batteries capturing STBs. With some exceptions (e.g., National Action Alliance for Suicide Prevention [NAASP], 2014), there is often little to no mention of improving assessments of these outcomes themselves. This presents an unsettling gap in foundational knowledge, considering that all efforts to understand and reduce STBs are only as good as our measurement of them. In some ways the field of STB research and treatment is no exception, as the benefits of evidence-based assessment apply to the field of youth mental health at large (e.g., Stewart & Hamza, 2017). But the intensifying concern around youth suicide risk designates STB assessments as an especially timely consideration.

Comprehensive STB assessment batteries, which are made possible by the wide array of available measures, are commonly accepted as a key step towards researching and treating suicidal youth. Numerous guidelines encourage the use of multiple assessments that vary across content, format, time interval, and informants to yield a comprehensive assessment battery and increase the likelihood of accurately detecting clinically-relevant STBs (e.g., Berk & Asarnow, 2015; Posner et al., 2007; Shaffer & Pfeffer, 2001). Despite the potential benefits of comprehensive STB assessment batteries, a critical consideration is that different STB measures may produce conflicting information. Separate questions about suicide attempt, for instance, could yield different reports from the same person depending on how they are worded or administered. Similarly, when collecting information from multiple informants about STBs among youth, parent- and self-report measures could yield very different responses. Indeed, there is emerging evidence in suggesting that assessments of youth STBs that vary in content, format, time interval, and informants are not interchangeable (e.g., Esposito et al., 2021; Gratch et al., 2021, 2022; Klimes-Dougan et al., 1998, 2022). The lack of formal guidance on how to integrate information across discrepant reports poses a potential barrier to addressing suicide risk and prevention. The utility of a comprehensive and integrated assessment battery¹ lies in the meaningful interpretation across its component measures.

In the present paper, we highlight discrepancies that may emerge among assessments that capture STBs in youth. To be clear, our scrutiny of reporting discrepancies should not be confused with rigid demand for concurrence. Formal study of discrepancies across youth-based measures of STBs is sorely needed to determine exactly which discrepancy patterns signal flawed measurement (i.e., ought to be concurrent), and which carry clinical utility (i.e., are meaningfully discrepant). Below, we organize reporting discrepancies across a number of assessment parameters: (1) what exactly is asked when assessing STBs (i.e., content); (2) how assessments are administered (i.e., format); (3) when and how frequently measures are administered (i.e., time interval); and (4) who is asked (i.e., informant). We introduce observed reporting discrepancies tied to each parameter below, along with unique implications tied to each discrepancy pattern.

What: Discrepancies Across Assessment Content

Assessments vary in the types of STBs they measure. Some focus solely on suicidal ideation (e.g., Reynolds & Mazza, 2019), whereas most STB assessments capture a range of thoughts and behaviors, including suicidal ideation, suicide planning, suicide gesture/threat, and suicide attempt (Fox et al., 2020; Posner et al., 2011). Here, we focus on the two most commonly studied STB outcomes among youth: suicidal ideation and suicide attempt.

Two measures assessing the same purported clinical outcome—so long as they match along format, frequency and referenced time frames, and informant—would presumably elicit the same set of reports. However, this is not always the case. Young respondents do not necessarily provide consistent reports to separate questions asking about lifetime history of suicidal ideation (Gratch et al., 2022) or a recent suicide attempt (Berk & Asarnow, 2015). Similar discrepancies have been reported among adults when asked about prior instances of suicidal ideation (Ammerman et al., 2021; Millner et al., 2015) and suicide attempt (Kidger et al., 2012). These reporting discrepancies are not to be taken lightly; without guidelines on how to interpret them, such discrepancies could result in misclassification of research participants or alter clinicians’ characterizations of their patients’ symptoms in ways that affect the course of treatment (or lack thereof).

A closer look at the content of these assessments may help explain these reporting discrepancies. Assessments that are readily accepted as capturing suicidal ideation feature content that varies to a surprising degree. Questions can range from inquiring about active suicidal ideation alone (e.g., Brener et al., 1995; Wittchen, 1994), to inquiring about active and passive suicidal ideation separately (e.g., Posner et al., 2011; Wood et al., 1995), to presenting them in a combined question (Johnson et al., 2002). Assessments accepted as capturing suicide attempt feature a similar degree of variability, with content ranging from referring to a suicide attempt in ways that are unaccompanied by a definition (e.g., How many times did you actually attempt suicide?; Brener et al., 1995; Wittchen, 1994), to including descriptive language referring to a suicide attempt (e.g., Have you ever tried to kill yourself? In other words, have you ever purposely hurt yourself with some intent to die?; Fox et al., 2020), to distinguishing across different types of suicide attempts (i.e., aborted vs. interrupted suicide attempt; Fox et al., 2020; Posner et al., 2011).

When reporting discrepancies exist across purported measures of suicidal ideation and suicide attempt, they may reflect nuanced yet consequential differences in assessment content, thereby uncovering two key obstacles hindering progress within youth suicide research. First, the research community lacks consistent agreement and communication regarding acceptable definitions of suicidal ideation and attempt. Despite repeated efforts to clarify nomenclature over the past 30 years (Goodfellow et al., 2017; O’Carroll et al., 1996; Silverman et al., 2007), researchers still do not agree over which nuances of STBs are meaningful enough to measure and label as ‘suicide attempt’ or even ‘suicide death’ (DeLeo et al., 2021). The challenge of achieving consensus in part reflects the complex and seemingly elusive nature of suicide-related experiences; there is still much to learn about the lived experience of suicidal individuals. As an example, only recently has it been discovered that youth frequently, and in some cases solely, experience suicidal ideation in the form of mental images (Lawrence et al., 2021a, 2021b). This highlights an aspect of suicidal ideation that has been entirely overlooked by nearly all STB assessments, which presume that ideation is experienced as verbal thoughts.

The second obstacle is poor face validity. Even if there were consensus around terms and definitions among researchers, common terms are prone to misinterpretation by youth (Velting et al., 1998) and remain unaccompanied by definitions in some of the most prominent survey items used to capture STBs (Brener et al., 1995; Wittchen, 1994). These obstacles uncover a fundamental challenge in suicide research, as defining the construct to be measured is the very first step of scale development (DeVellis, 2021).

How: Discrepancies Across Assessment Formats

Assessments of STBs are often delivered through a variety of formats, including in-person interviews and self-report measures, as well as and online surveys. These different formats facilitate versatile and widespread use of STB assessments—such as the use of self-report measures to frequently assess STB history in research settings, in contrast to a potential preference for interviews in treatment contexts. However, it remains unclear how comparable these different formats are to one another. Potential reporting discrepancies across self-report and interview formats could, for instance, jeopardize the extent to which research findings could influence clinical decision-making. Identifying and understanding cross-format discrepancies has important implications within both research and practice, as well as the translation from one to the other.

Prior research comparing interviews vs. self-report measures among youth has revealed that the same adolescent will provide inconsistent reports across these formats (Velting et al., 1998)—tending to endorse STBs more often via self-report (Klimes-Dougan, 1998). This pattern is likely not explained by overly sensitive self-report measures, as these prior investigations compared formats (self-report vs. interview) containing identical content. Findings from adult-based research mirror these discrepancy patterns, with greater endorsement via self-reports than via interviews (Deming et al., 2021; Kaplan et al., 1994; Terrill et al., 2021).

A separate consideration involves potential discrepancies across online vs. in-person administration of self-report measures. The evidence here is mixed. On the one hand, assessments of STBs in clinical settings indicate that youth may prefer to seek help and share their STB-related experiences online rather than in-person (Harris et al., 2009; Power et al., 2020). On the other hand, prior work comparing web-based and paper-and-pencil surveys pertaining to emotional and behavioral symptoms, including STBs, indicate that the two formats yield comparable reports (Denniston et al., 2010; Eaton et al., 2010; Van De Looij-Jansen & Jan De Wilde, 2008). Although the impact of online vs. in-person assessments is inconclusive, this remains a timely consideration to monitor given the increasing use of online surveys (Smith et al., 2021) and ecological momentary assessment (EMA) studies that can be conducted remotely (e.g., via smartphones; Kleiman et al., 2019).

One possible explanation for these reporting discrepancies may be the different degrees of confidentiality that respondents associate with each assessment format. Confidentiality concerns may be particularly relevant to disclosing STBs, due to both the surrounding stigma (McGilivary et al., 2022; Sheehan et al., 2019) and the communicated potential for involving third parties (e.g., parents, emergency services; Hom et al., 2017; McGilvary et al., 2022). Indeed, prior work has shown that inaccurate disclosure (i.e., denial) of suicidal ideation among youth is often driven by stigma-related concerns (Hom et al., 2017) and that the most common explanation adolescents and young adults provide for not disclosing suicidal ideation is fear that confidentiality will be breached (McGillivray et al., 2022). This is consistent with patterns of disclosing other sensitive health risk behaviors (e.g., tobacco use), where disclosure may be accompanied by fear of reprisal (Brener et al., 2003). Self-report measures of STBs, through their absence of immediate witnesses to their reports, may help remedy adolescents’ concerns about confidentiality breaches or potential judgment from others. This is consistent with adult-based research as well, which has shown that self-report measures perceived to be more confidential tend to yield greater disclosure of STBs (Anestis & Green, 2015; Deming et al., 2021) and that fear of hospitalization is cited as a reason to not disclose suicide-related experiences (Fulginiti & Frey, 2019). Of note, it remains unclear the extent to which confidentiality concerns would impact discrepancies across web-based vs. paper-and-pencil measures, as youth do not necessarily perceive one mode to be more confidential than the other (Van De Looij-Jansen & Jan De Wilde, 2008).

When: Discrepancies Across Time Intervals

Assessments of youth STBs also vary in time intervals, or the frequency of administration and time frames referenced. While some assessments involve a one-time query for the presence of STBs over an extended prior period (e.g., over the past two weeks, past year, lifetime), others may be deployed repeatedly and refer to time frames that are more proximal and sometimes immediate (e.g., today, right now). The latter assessment schedule is becoming increasingly common through research harnessing mobile technology to facilitate the real-time measurement (i.e., EMA) of psychological phenomena (Trull & Ebner-Priemer, 2020), including STBs (Kivela et al., 2022; Kleiman et al., 2019). Capturing and understanding cross-timescale discrepancies has important implications for both empirical and clinical work, ranging from the accurate classification of research participants (e.g., suicidal vs. nonsuicidal) to the real-time determination of risk made by a clinician.

This recent evolution of suicide research has prompted comparisons of longer-term retrospective vs. immediate recall abilities, which have begun to suggest that reports of recent suicidal ideation may vary across assessment schedules and time frames referenced. A month-long EMA study with adolescents who were recently discharged from acute psychiatric care revealed that more adolescents reported suicidal ideation and nonsuicidal self-injury throughout the month when asked in real-time (via EMA) as opposed to a retrospective interview at the month’s end (Esposito et al., 2022). Another study assessing feasibility of EMA for use with adolescents post-hospitalization found that more teenagers (70.6%) endorsed ideation through a daily diary over the course of a month than in a retrospective assessment at the end of the month (45.2%; Czyz et al., 2018). This is consistent with findings from an adult-based study in which 58% of respondents who endorsed suicidal ideation via EMA at some point during the week denied experiencing any ideation at all when asked retrospectively at the week’s end, even when report scales and questions were identical in all other ways (Gratch et al., 2021). These studies portray a similar pattern as observed in the psychopathology literature more broadly, with respondents appearing more likely to endorse symptoms with a repeat assessment schedule referencing shorter time frames relative to single time point assessments referencing extended time frames (e.g., Schuler et al., 2021; Shiffman et al., 2009; Solhan et al., 2009; Torous et al., 2015).

We present two potential explanations accounting for these reporting discrepancies across assessment time intervals. First, these discrepancies may result from a broader phenomenon of recall bias affecting all respondents. Cognitive and affective science studies may shed light on possible recall biases in the recollection of suffering more generally; Kahneman has written on the so-called ‘peak-end rule,’ wherein retrospective reporting of pain is most influenced by pain at the peak-moment and the most recent moment (Redelmeier & Kahneman, 1996). The same may hold true with psychological suffering, in which participants’ retrospective reporting may map on more closely to participants’ experiences at the peak-moment (positive or negative) and most recent moment, rather than some internally computed average (e.g., Schuler et al., 2019). Similarly, researchers have pointed to Positive Illusions Theory – a bias towards unrealistic optimism and enhanced self-evaluations – as one mechanism underlying recall bias more generally (Colombo et al., 2020). Both biases could explain, in part, these cross-timescale discrepancies. Separately, the field of EMA research is actively considering the possibility that elevated reports across repeated EMA survey prompts may reflect habitual responding (Napa Scollon et al., 2009).

Second, these discrepancies may be driven by a subset of respondents. Real-time monitoring research has revealed distinct phenotypes of suicidal thinking based on differences in frequency, intensity, and variability (Kleiman et al., 2018). These phenotypes demonstrate how these characteristics of ideation vary across individuals. It may be that some people – perhaps belonging to certain phenotypes of ideation – have a harder time characterizing their experiences over a longer period of time, as opposed to in the present moment. For example, a study comparing reporting of mood changes as measured in real-time vs. retrospectively, both after one week and after one month, found participants were more accurate in their recollection of days without extreme mood changes than days with extreme mood changes (Solhan, 2009). It is thus possible that those participants with more variable affect states – including greater within-person variability in suicidal ideation – demonstrate a failure in self-reflection in retrospective assessments, perhaps making up some percentage of participants whose retrospective reports and real-time reports do not converge. Alternatively, it may be that some participants recall experiencing low-intensity and/or low-frequency ideation that does not, in their minds, rise to the level of something warranting reporting when assessed retrospectively, during which they may attempt to compute a more global representation of their experience. This may be especially true in suicide assessments, where the perceived risks of endorsing suicidal ideation can be high (Fulginiti & Frey, 2018; Sheehan et al., 2019). Importantly, it remains unclear whether real-time monitoring approaches are picking up on undetected and clinically meaningful instances of suicidal thinking, or whether such tools are in fact too sensitive, indicating risk when there is little or none. In other words, we still do not know which tool or timescale will result in the most ‘accurate’ empirical work or clinical decision-making.

Who: Discrepancies Across Informants

Assessments of STBs could prompt different informants, ranging from youth themselves to other informants such as parents or teachers. Collecting cross-informant data is common and recommended practice tied to the assessment and treatment of suicidal youth (Berk & Asarnow, 2015; Shaffer & Pfeffer, 2001). Most of these cross-informant studies have involved self- and parent-reports (Deville et al., 2020; Gratch et al., 2021; Jones et al., 2019; Klaus et al., 2009; Walker et al., 1990), but select youth-based studies have also involved teacher-reports (Commisso et al., 2020; Connell et al., 2019; Thompson et al., 2006).

Cross-informant assessments of suicidal thoughts, plans, and attempts among youth tend to result in low agreement or moderate to high disagreement (Connell et al., 2019; Deville et al., 2020; Gratch et al., 2021; Jones et al., 2019; Klaus et al., 2009; Thompson et al., 2006). Most youth who report presence of suicidal thoughts have a parent informant who does not endorse youth suicidal thoughts (Deville et al., 2020). Moreover, research examining these cross-informant discrepancies has also found that significantly more suicide plans and attempts are reported by adolescents than their parents (Gratch et al., 2021; Klaus et al., 2009). Regarding other informants, modest correlations of child-teacher reports (Thomas et al., 2006) and parent- teacher-reports have been observed (Connell et al., 2019).

We present three possible explanations accounting for these cross-informant discrepancies. First, these discrepancies may be driven by characteristics of the existing relationship between informants. Potential interpersonal strain, for instance, between young respondents and their parents may increase parental unawareness of their child’s internal experiences such as STBs. Indeed, adolescents’ lower perceived family support has been associated with greater discordance across parent- and self-reported suicide plans among youth (Klaus et al., 2009). This is consistent with prior work suggesting that poor family communication and familial stress are associated with greater discrepancies across other parent- and self-reported psychiatric symptoms among youth (Grills & Ollendick, 2003; Kolko & Kazdin, 1993).

Second, these discrepancies may be associated with characteristics of the informants themselves.² Potential characteristics of parents to consider include their role in relation to their child (e.g., greater discrepancies with father- vs. mother-report; Jones et al., 2019) and parental psychopathology (e.g., lower discrepancies with parents with more severe psychiatric history vs. parents with no history of depression; Klaus et al., 2009; Walker et al., 1990). Potential characteristics of adolescents to consider include age (e.g., greater discrepancies involving younger vs. older adolescents; Jones et al., 2019; Klaus et al., 2009; Walker et al., 1990), race (e.g., greater discrepancies involving racial minority youth vs. White youth; Bell et al., 2021; Jones et al., 2019), and history of suicide attempt (e.g., greater discrepancies involving youth with multiple attempts; Klaus et al., 2009; Walker et al., 1990).

Third, these discrepancies may be driven by characteristics of STBs experienced by young respondents. For instance, parent-child reports tend to be more discrepant on reports of suicidal ideation compared to suicide attempt (Klaus et al., 2009). This may be due to the fact that internal experiences tied to considering suicide are simply less observable by third-party informants than the potential physical act of a suicide attempt. Alternatively, it may be the case that suicidal ideation is not often experienced when youth are with their parents and therefore not observed or reported in such contexts. In fact, prior work has shown that youth experience suicidal thoughts less than 20% of the time when with their parents/guardians and instead typically experience suicidal thoughts when alone or with friends (Nock et al., 2009).

Recommendations for Future Research

Nearly all research on youth suicide risk and prevention relies on well-informed, interpretable assessment of STBs. There is an urgent need to intensify and expedite the formal study of discrepancies across these assessments varying in content, format, time intervals, and informants. Below, we recommend five key steps to unpacking such assessment discrepancies in future research.

1. Improve Definitions.

The first step in developing any assessment is to clearly define the construct intended to be measured (DeVellis, 2021). Accordingly, a foundational obstacle to improving STB assessments is the lack of clear, precise, standard definitions of STBs (Berman & Silverman, 2017; Frey et al., 2020; Goodfellow et al., 2019). The field of suicide research has wrestled with such challenges for decades, with repeated attempts to refine the definitions and taxonomy of STBs (for a comprehensive review, see Goodfellow et al., 2017). This obstacle hinders the establishment of epidemiological, etiological, and treatment/prevention knowledge, as well as accurate decision-making, across both clinical and research settings.

Defining a psychological construct requires thorough, phenomenological understanding of the construct: how youth, in reality, experience STBs. Phenomenological understanding of STBs can be improved in two ways. The first way is to increase stakeholder input from those with lived experiences of STBs. This could be achieved through collecting children’s and adolescents’ narrative accounts of STBs (e.g., Hennefield et al., 2022), for instance, or soliciting direct feedback from youth on current assessments (e.g., phrasing of suicidal ideation definitions; Berman & Silverman, 2017; DeVellis, 2021). Mixed methods study designs as well as community-based participatory research conducted by emerging organizations (e.g., Recognizing and Enhancing Adolescent and Community Health [REACH] Youth Advisory Board; Hamilton et al., 2022) may help bridge the gap between research and the lived experiences of STBs. This bottom-up approach could uncover new and different ways in which STBs are experienced and expressed across diverse populations (e.g., somatic/emotional suicidal distress experienced by Latina adolescents, hidden suicidal ideation among racial and ethnic minority youth; Choi et al., 2009; Gulbas et al., 2021; Merchant et al., 2009; Morrison & Downey, 2000; Politano et al., 1986). Moreover, the use of less verbally-dependent, play-based measures may help discover new ways in which younger respondents experience suicidal ideation prior to adolescence (e.g., Luby et al., 2019; Whalen et al., 2021).

Another way to improve phenomenological understanding is to increase the precision and granularity of quantitative research when observing properties of STBs. Recent calls for more rigorous descriptive research suggest ways to clarify ‘inherently fuzzy’ constructs such as ‘suicidal thinking’ computationally (e.g., based on average observed duration of thoughts; Millner et al., 2020). Introducing and carefully measuring more features of STB experiences can then inform the development of phenomenologically-grounded STB assessments.

2. Use Psychometrically Strong Assessments.

Reporting discrepancies can only be meaningfully interpreted if they are systematic in nature. Theoretical frameworks guiding interpretation of cross-informant discrepancies, for instance, distinguish discrepancy patterns that may emerge due to methodological reasons from those that yield clinically meaningful information about the reported symptoms (e.g., Operations Triad Model for cross-informant discrepancies; De Los Reyes et al., 2013). Establishing strong psychometric properties of individual assessments for youth therefore marks a key step to interpreting discrepancy patterns. This includes taking developmental considerations into account when determining how to measure STBs. Thus, as we continue to establish the psychometric properties of these assessments, it is important to recognize that children in different age groups may respond differently to different assessment content, format, and timescale.

Although many youth-based STB assessments have been well-validated (Cwik et al., 2019), there are particular research areas where measurement error has not yet been ruled out definitively. Most EMA studies, for example, typically employ adaptations of validated retrospective questionnaires, but at present, most EMA survey items themselves have little to no direct psychometric support. It remains possible that the repeated administration schedule with references to short-term time frames is overly sensitive and may be detecting overly transient instances of suicidal ideation relative to more clinically significant, memorable episodes of suicidal ideation that are recalled through longer-term retrospective report. To our knowledge, only one study to date (Forkmann et al., 2018) has offered a comprehensive look into the psychometric properties of a suicide assessment created explicitly for EMA. Although this investigation reports satisfactory reliability and convergent validity of EMA items, psychometric properties were tested in adults, leaving these and other suicide-related EMA items yet to be validated among youth. Testing for sufficient item variability, validity, as well as between- and within-person reliability of EMA items with younger respondents would be a critical step prior to interpreting real-time vs. retrospective reporting discrepancies (e.g., Esposito et al., 2021).

Relatedly, localizing sources of potential measurement error would be further expedited by avoiding the confounding of parameters. An EMA survey and a longer-term, retrospective self-report measure, for example, may not only differ in time interval (e.g., right now vs. past 2 weeks) but also in delivery format associated with varying degrees of perceived confidentiality (e.g., self-report vs. interview). Indeed, challenges pinpointing possible reporting biases in STB assessments have been attributed to the extremely variable array of methods featured across studies (Klimes-Dougan et al., 2022).

3. Identify Baseline Correlates.

Among those psychometrically sound STB assessments, there remains the challenge of meaningfully interpreting observed discrepancies. Are there certain profiles of respondents who are more likely to yield discrepant reports? Testing potential correlates of discrepancy patterns can begin to inform such interpretations. As an example, prior work on cross-informant assessments involving youth psychopathology suggests that more discrepant parent-youth reporting of symptoms may be expected if the child has never utilized mental health services before (e.g., Jones et al., 2019; Lerner et al., 2017; Makol & Polo, 2018).

Within the suicide literature, respondent characteristics (e.g., demographic or clinical background) and features of the reported STBs themselves (e.g., suicidal ideation vs. suicide attempt) have been tested as potential correlates of discrepancy patterns (Bell et al., 2021; Jones et al., 2019; Klaus et al., 2009; Walker et al., 1990). However, such investigations are sparse and have been pursued predominantly in the context of cross-informant discrepancies. More formal, expansive attempts to identify correlates—in tandem with the aforementioned future directions (e.g., ruling out measurement error)—could ultimately help inform whether, and if so how, to meaningfully interpret discrepancy patterns varying in magnitude and directionality. For instance, researchers could seek to uncover clinical factors which may be tied to discrepant reporting patterns between retrospective recall and more immediate reports of STBs, such as introspection abilities or within-person variability in suicidal ideation.

4. Test Potential for Informing Prognosis and Treatment Personalization.

Beyond correlating with prior history and baseline characteristics, cross-assessment discrepancies may also carry prognostic value and help guide clinical decision-making. Results from other areas of psychopathology show that disagreements between youth and other informants are stable over time (Lippold et al., 2015) and predict poor child outcomes in a unique way that individual assessments do not (Al Ghriwati et al., 2018; Makol et al., 2019). This has begun to be explored among suicidal youth, in which patterns of youth reporting higher depressive symptoms than their parents predict greater youth suicidal ideation in the future (Augenstein et al., 2021). Identification of cross-informant disagreement patterns (e.g., parent low, adolescent high) may allow for diagnosis of youth who are at especially high risk for future STBs. Accordingly, there has been greater attention paid toward developing composite approaches that integrate information from multiple sources, which may improve the prediction of suicide attempts beyond one of these methods alone (i.e., risk scores, self-report, clinician-report, peer-report; Hawes et al., 2017; Nock et al., 2022). This approach could be extended to the examination of discrepancies along other assessment parameters, such as time interval. Specifically, longitudinal studies could test whether the participants for whom ideation goes undetected retrospectively carry the same level of future risk as those whose cross-timescale reports converge.

Patterns of discrepant STB reports could also be used to predict and potentially personalize the course of treatment. Multiple studies in youth psychopathology have found that the direction of agreement (i.e., agreement vs. disagreement) and magnitude of disagreement (i.e., small or large disagreement) can predict future symptoms (Goolsby et al., 2018; Zilcha-Mano et al., 2020). The pattern of children and adolescents reporting lower degrees of anxiety symptoms than their parents, for example, may signal their low treatment engagement as fewer of them report being diagnosis-free after cognitive behavioral therapy (Becker-Haimes et al., 2018). Such an approach has yet to be tested within the suicide treatment literature.

5. Explore New Assessment Parameters.

While the aforementioned discrepancies cover a range of commonly used STB assessments, the field of suicide research is fast-evolving. There is an increasing variety of potential STB assessment available for consideration, ranging from behavioral performance tasks (e.g., Death Implicit Association Task; Nock et al., 2010), to real-time monitoring (e.g., passively detected autonomic arousal; Kleiman et al., 2021), to electronic health records and social media activity (Roy et al., 2020; Su et al., 2020). In keeping with the observed trend in discrepancies across and within assessment methods, we expect these innovations to be prone to discrepancies with each other and with more established methods as well. It is therefore wise to tread carefully with the conclusion that any single novel approach is the ‘gold standard’ or ‘silver bullet’ for all purposes and all individuals.

What Can We Do Right Now?

Youth suicide research can be strengthened not only by the formal study of STB reporting discrepancies in the long term, but also by immediate practices applied to any ongoing investigation involving suicidal youth. This can occur at different points in the research process.

When planning your next study involving STB data collection.

Consider defining your STB outcomes clearly, deciding what function you would like your psychometrically strong STB assessments to serve, and applying the current knowledge base surrounding each parameter (what, how, when, who) when selecting assessments. Selecting multiple STB assessments that vary across frequency and referenced time frames (e.g., daily experience of suicidal thoughts, recall of month-long suicidal thoughts), for instance, may complement the pursuit of research questions pertaining to STB recurrence over time. Selecting multiple assessments that vary in format (e.g., interview and confidential exit survey) may strengthen the quality of STB data in clinical settings where respondents may be concerned with the consequences of their reports (e.g., hospitalization vs. discharge from emergency room). As a final example, selecting different informants (e.g., teacher-, peer-, parent-, and self-report) may enhance the study of how youth experience STBs across different settings.

When analyzing data involving multiple STB assessments.

Ideally prior to analysis, anticipate STB reporting discrepancies and document a decision tree. Similar to planning for missing data (Wood et al., 2021), researchers can preemptively plan for the possibility of inconsistent STB reports. Documented plans could range from prioritizing one set of assessment reports over another for empirically justified reasons to formally investigating the nature of discrepancy patterns and their potential clinical utility. As an example of the latter, several data analytic approaches have been adopted to identify discrete patterns of cross-informant discrepancies (e.g., latent class analysis; Lippold et al., 2013, 2014) and test their potential clinical utility (e.g., via polynomial regression, response surface analyses; Humberg et al., 2019; Laird & De Los Reyes, 2013).

Conclusion

Given the intensifying crisis and concern around youth suicide risk and the reporting discrepancies outlined in this paper, it is clear that urgent action is needed in order to advance youth suicide research. Findings across the field—whether in the context of etiology, risk prediction, or intervention—will only be as valid and reliable as the tools we use to arrive at them, underscoring the need for evidence-based development and use of assessments. Capitalizing on the rich research base accumulated thus far, there is a clear path toward both illuminating the existent discrepancies and innovating approaches that may more fully account for the complexities in experiencing, reporting on, and interpreting STBs. The importance of this foundational work cannot be understated considering that findings are ultimately translated into real-world contexts to address not just adverse outcomes but also the lived experience of youth and families with STBs.