Parent-Adolescent Agreement About Adolescents’ Suicidal Thoughts

The Philadelphia Neurodevelopmental Cohort (PNC)¹² includes 9498 youth between the ages of 8 and 21 years recruited from a large pediatric health care network. Importantly, participants were not recruited from mental health treatment programs; thus, the PNC is a community-based sample not enriched for individuals seeking psychiatric help. Potential participants were excluded from the cohort if they were not proficient in English, had significant developmental delays, or had other physical or cognitive conditions that interfered with study participation. In this study, we focused on PNC participants between the ages of 11 and 17 years (n = 5137) for whom information about suicidal thoughts was collected from both adolescents and a collateral informant. See Table 1 for sample demographics. Given that 97.2% of collaterals were parents or served in some type of caregiving capacity, we refer to the collateral informants as parents.

After receiving a description of study procedures, adolescents and their guardians provided written assent and informed consent, respectively. The Institutional Review Boards of the University of Pennsylvania and Children’s Hospital of Philadelphia approved study procedures.

Adolescents and parents completed GOASSESS,¹² a computerized, structured clinical interview designed to screen for symptoms of major psychiatric disorders. Trained interviewers assessed adolescents’ treatment history as well as the prevalence, frequency and duration, and distress and impairment associated with symptoms of major domains of psychopathology. Psychometric evaluation of the factor structure of GOASSESS revealed an overall psychopathology factor score and 4 subfactor scores (T.M.M., M.E.C., T.D. Satterthwaite, et al., unpublished observations).¹² In the current study, we used the overall psychopathology factor score as an index of the severity of adolescents’ psychological difficulties. GOASSESS is based on the Kiddie Schedule for Affective Disorders and Schizophrenia and has been shown to be a valid screener of adolescent psychopathology in multiple studies.^12,14,–17 

During the GOASSESS interview, adolescents were asked about lifetime suicidal ideation: “Have you ever thought about killing yourself?” In addition, adolescents were asked about thoughts of death: “Have you ever thought a lot about death or dying?” Although not necessarily suicidal ideation, recurrent thoughts of death or dying are clinically meaningful when evaluating suicide risk, are associated with elevated distress,¹⁸ and are included as a symptom of major depressive disorder in the Diagnostic and Statistical Manual of Mental Disorders, 5th Edition.¹⁹ In addition, previous studies in which reporter agreement about suicidal thoughts was examined have included questions about thoughts of death or dying.^10,20,–24 Therefore, we considered it useful and important to also examine reporter agreement about thoughts of death or dying. Parents answered the same questions about their adolescents’ lifetime suicidal thoughts and thoughts of death. Responses to these 2 questions were coded dichotomously as yes (1) or no (0).

Interviewers were trained in a standardized protocol for responding to reports of suicidal thoughts. This included asking follow-up questions about current thoughts and consulting with a study clinician. If necessary, a doctoral-level clinician would conduct a formal suicide evaluation and safety check and make a determination about the appropriate course of action (eg, referral, emergency procedures). Any participant that requested a referral was provided with information about mental health resources.

Parents were administered an abbreviated version of the Family Interview for Genetic Studies.²⁵ Family history of suicide attempts or completions among first-degree biological relatives of the adolescent were coded from responses to open-ended screening questions.

Eight items from the posttraumatic stress disorder section of the GOASSESS interview provided an index of trauma exposure. We created a 3-level categorical variable, reflecting assaultive trauma exposure, nonassaultive trauma exposure, or no trauma exposure.¹⁷ 

During the GOASSESS interview, parents provided information about the adolescents’ history of mental health treatment, psychiatric hospitalizations, and psychoactive medication use.

Demographic information was obtained during the interview. In addition, participants’ addresses were linked to census and crime data, enabling us to obtain neighborhood-level estimates of the adolescents’ socioeconomic status (SES).¹³ 

First, we examined agreement between adolescent and parent reports of adolescents’ suicidal thoughts. We used κ as our primary agreement index. In addition to κ, we calculated sensitivity and specificity estimates by using the adolescents’ responses as the standard for comparison. Second, we examined the directionality of disagreements by calculating the proportion of dyads in which (1) the adolescent reported suicidal thoughts the parent was unaware of (“parental unawareness”) and (2) the parent reported suicidal thoughts that the adolescent denied (“adolescent denial”). Third, we performed logistic regressions to examine factors that might be associated with reporter disagreements. The first model included demographic characteristics. Given sex differences in the prevalence and developmental progression of suicidal thoughts across adolescence,²⁶ we included the age-by-sex interaction term in the model. The second model included clinical characteristics.

The amount of missing data were minimal (≤3.5%). Given that missing data were <5%^27,28 and because data from both parent and adolescent is required to evaluate agreement, listwise deletion was used. As a result, the sample size varies across analyses.

We report endorsement of suicidal thoughts by reporter and agreement rates broken down by demographic characteristics in Tables 2 and 3. Landis and Koch²⁹ recommended the following cutoffs for interpreting agreement on dichotomous data: ≤0.20 = slight, 0.21 to 0.40 = fair, and 0.41 to 0.60 = moderate. By using these commonly cited guidelines, reporter agreement on thoughts of killing self was moderate (κ = 0.466), whereas agreement on thoughts of death was slight (κ = 0.171). The sensitivity and specificity estimates indicate substantial parental unawareness of adolescents’ suicidal thoughts. When adolescents provided an affirmative response, 49.9% of parents were unaware of adolescents’ thoughts of killing themselves, and 75.6% of parents were unaware of thoughts of death. We also identified dyads in which the parent reported suicidal thoughts that the adolescent denied. When parents provided an affirmative response, 48.4% of adolescents denied thoughts of killing themselves and 67.5% of adolescents denied thoughts of death. Specificity was high for both thoughts of killing self (0.959) and thoughts of death (0.909).

Age emerged as an important factor across all the models (Tables 4 and 5). Older adolescent age was associated with better agreement: decreased odds of parental unawareness and decreased odds of adolescent denial. In addition, for parental unawareness, the age-by-sex interaction was significant for both thoughts of killing self and thoughts of death. We examined the slopes for boys and girls to further explore these interactions (Figs 1 and 2). For thoughts of killing self, neither slope was significant. For thoughts of death, the slope was negative and significant for girls (b = −0.28, P < .05) and positive and marginally significant for boys (b = 0.24, P = .057). This suggests that, for adolescent girls, parental unawareness of thoughts of death is high at younger ages and then decreases at older ages. The opposite pattern emerged for adolescent boys, with unawareness increasing slightly in older boys. Race and ethnicity were associated with reporter discrepancies. The odds of parental unawareness and adolescent denial of thoughts of death were higher for racial minority adolescents relative to white adolescents. Hispanic or Latino ethnicity was associated with decreased odds of parental unawareness of thoughts of death. Finally, fathers were more likely to be unaware of adolescents’ thoughts of killing themselves than mothers.

Adolescents’ psychiatric treatment history (eg, psychotherapy) and family history of suicide were associated with decreased odds of parental unawareness (Tables 4 and 5). For adolescent denial, there was a large effect of psychopathology. More severe psychopathology was associated with decreased odds of adolescent denial of both thoughts of killing self and thoughts of death. In addition, a history of psychiatric hospitalization was associated with decreased odds of adolescent denial of thoughts of killing self.

Research to date has revealed substantial reporter discrepancies in the assessment of suicidal thoughts,^8,9 indicating that a large number of adolescents at risk for suicide may go undetected in real-world health care settings. In general, this research has focused mostly on clinical or high-risk youth, the samples have been relatively small, and there has been limited exploration of factors associated with agreement. Therefore, our goal in the current study was to conduct the largest investigation to date of agreement between parent and adolescent reports of adolescents’ suicidal thoughts in a well-characterized, community-based sample and to identify demographic and clinical factors associated with agreement.

Similar to previous research, rates of agreement in this study were low to moderate,^21,30 with better agreement about thoughts of killing self than about thoughts of death. Alarmingly, ∼75% of parents were unaware of thoughts about death, and 50% of parents were unaware of thoughts about suicide that were reported by their adolescents. Interestingly, we also identified numerous dyads in which the parent reported suicidal thoughts that the adolescent did not endorse. Overall, there has been a paucity of research examining adolescents’ “denial” of suicidal thoughts endorsed by parents, and this study advances our understanding of this phenomenon.

Adolescent age emerged as an important demographic factor: older age was associated with decreased odds of both parental unawareness and adolescent denial. This indicates that younger adolescents may be more likely to go unnoticed and not receive services either because their parents are unaware of their suicidal thoughts or because they deny suicidal thoughts that their parents think they are having. It is also possible that younger adolescents have interpretative difficulties with questions about suicidal ideation and thoughts of death, resulting in low reporter agreement. If this is the case, extra attention may be warranted in research and clinical settings to ensure that younger adolescents fully understand questions about suicidal thoughts. These age findings are particularly noteworthy in light of recent evidence that deaths by suicide have increased among younger adolescents.³¹ We also found a significant age-by-sex interaction. Follow-up analyses revealed that parental unawareness of thoughts of death are decreased in older adolescent girls but increased in older boys. This interaction could reflect sex differences in the trajectory of depressive symptoms in which symptoms spike among midadolescent girls relative to boys.³² The surge in symptoms in girls may make thoughts of death more salient to their parents. The above findings suggest a need for age- and sex-specific strategies for facilitating adolescent disclosure and parental awareness of suicidal thoughts.

Parents of racial minority adolescents were more likely to be unaware of thoughts of death and to report thoughts of death that the adolescent denied. However, Hispanic parents were more aware of their adolescents’ thoughts of death. These results are consistent with a recent study that revealed greater parent-adolescent discrepancies in reports of suicidal ideation among African American and Asian American families compared with European American and Hispanic Americans families.²⁴ These findings, combined with results indicating that racial minorities are less likely to use health care services before a suicide attempt,³³ highlight the importance of taking racial and cultural differences into account when assessing suicide risk and implementing prevention programs.

Fathers were less aware of adolescents’ thoughts of killing themselves than mothers. These findings are consistent with research revealing that adolescents are less likely to disclose personal information to fathers than mothers³⁴ and that fathers possess less “parental knowledge” about their adolescents than mothers.^35,36 

In general, our findings suggest that a history of youth behavioral health treatment is associated with greater parental awareness of suicidal thoughts. For instance, parents were more likely to be aware of their adolescents’ thoughts of both death and killing self when the adolescents had a history of being in therapy or psychiatric hospitalization. Additionally, history of psychotropic medication usage was associated with increased parental awareness of adolescents’ thoughts of suicide. Given the cross-sectional design, it is not possible to definitively determine the direction of effects. It is possible that parents become aware of suicidal thoughts, which may lead them to seek treatment of their adolescents. Conversely, involvement in behavioral health care could uncover previously unknown suicidal thoughts or increase parents’ knowledge about and attention to adolescents’ suicidal thoughts.

Family history of suicide attempts or completions was associated with increased parental awareness of thoughts of death. To our knowledge, there has not been any previous research examining how family history of suicide relates to reporter agreement about suicidal thoughts. However, studies have examined the role of family history of mood disorders in reporter agreement about suicidal thoughts.^9,10,24 The results of these studies have been mixed. Additional attention to family history of suicide is warranted to corroborate the present findings.

Regarding adolescent denial of suicidal thoughts, we found a large effect of adolescent psychopathology severity. Specifically, greater psychopathology was associated with less denial of both thoughts of death and thoughts of killing self. A study of a sample of adolescents who were hospitalized also revealed that youth with greater externalizing symptoms were less likely to deny suicidal ideation reported by a parent.⁹ Yet, in a large study of children who were maltreated, aggression and social withdrawal were associated with increased child denial of suicidal ideation reported by a parent.¹¹ Differences in sample characteristics and the assessment of suicidal thoughts could account for the discrepant findings. In the present community-based sample of adolescents, it is possible that those adolescents with more severe psychological difficulties have been involved in behavioral health treatment and are more familiar with clinical assessment and, as a result, are more comfortable disclosing suicidal thoughts than adolescents with less severe psychopathology.

First, the PNC includes families presenting in pediatric settings in one geographic area. It is unclear whether the present findings would generalize to the larger population or to clinical samples. Second, to assess thousands of families in a short period of time, our assessment of suicidal thoughts was brief. In addition, the questions we asked were about lifetime suicidal thoughts and could be susceptible to recall bias. Future research could include a more detailed, multimethod assessment that covers both current and lifetime suicidal thoughts as well as plans and attempts. Third, the prevalence of suicidal ideation reported in this study (8%) is lower than that reported in a national survey of youth (18%).³ This could possibly be due to our inclusion of younger adolescents (11–17 years), whereas the national survey was focused on high school students. Another possibility is that the mode of assessment used in this study (in-person interview) affected the disclosure of suicidal thoughts. Research indicates that the prevalence of suicidal thoughts is higher in studies that use youth self-report questionnaires (as in the national survey) relative to face-to-face interviews.¹⁰ Fourth, thoughts of death or dying may encompass a wide range of youth ideation, some of which may not be related to suicide risk. Fifth, this study was focused on parent-adolescent agreement about adolescents’ suicidal thoughts. However, there are other individuals that adolescents might confide in, such as peers or other trusted adults. Further examination of the role of nonparental figures is warranted and could reveal other important collateral informants that should be included in suicide screenings.

The findings from the current study have important clinical implications for pediatric settings, especially in light of growing endorsement of depression screening initiatives in pediatric primary care.^37,38 Given the high prevalence of parental unawareness and adolescent denial of suicidal thoughts found in this study, it is possible that a large number of adolescents with suicide risk may not be detected by brief screens at routine check-ups. This highlights the urgent need for continued training of pediatric primary care physicians in the evaluation and management of suicidal ideation and the importance of collecting information from multiple informants and rectifying discrepant reports. In addition, the current study identified several demographic and clinical characteristics that may increase or decrease the risk of parental unawareness and adolescent denial of suicidal thoughts. Increasing awareness of these factors could aid clinicians in creating a more personalized approach to suicide risk assessment. The early identification and treatment of youth at risk for suicide depends on reliable and valid assessment of suicide risk. The present findings indicate the need for thorough, multi-informant screening of suicide risk in pediatric populations that is informed by research on factors that might increase or decrease the reliability of assessments.

PNC

Philadelphia Neurodevelopmental Cohort

SES

socioeconomic status

We thank the participants of this study and all the members of the recruitment, assessment, and data teams whose individual contributions collectively made this work possible.