Suicide Risk in Adolescents During the COVID-19 Pandemic

Children’s Mercy Kansas City initiated universal suicide screening in 2018. We conducted over 100 000 screens per year in 2019 and 2020. Patients are eligible for screening if they are 12 years or older and do not have a medical contraindication (eg, urgent medical problems, severe pain, nonverbal, etc) to completing a screen.

Data on suicide risk were based on the “Ask Suicide-Screening Questions” (ASQ).^2,15  The ASQ is a freely available 4-item screening tool originally developed and validated for use in assessing suicide risk in pediatric emergency departments. The tool has since been further validated for use in pediatric inpatient and outpatient settings for ages 10 to 21, with sensitivity values ≥95% and specificity values ≥87%.^3,4  The ASQ is administered independently with the patient (ie, not with parents present), unless the patient’s parent or guardian is opposed to the patient being screened without their presence. The ASQ items inquire about wishes to be dead, feelings of oneself or others being better off if dead, thoughts of killing oneself, and previous attempts at killing oneself. If a patient’s response is “no” to all 4 items, this is considered a negative screen and no further questions are asked unless clinical judgement overrides the screening result. If a patient responds “yes” to any item or refuses to answer, the screen is considered “positive” and a fifth question is asked to determine acuity (“Are you having thoughts of killing yourself right now?”). A “yes” answer or refusal to answer this fifth question is defined as an “acute positive screen.” The ASQ has been validated for administration by a variety of clinicians in a wide range of clinical settings. We are unaware of a validation study specifically focused on telemedicine.

For any positive screen, further safety assessment and safety precautions matched to risk level were implemented by social workers according to a hospital protocol (eg, safety planning including means restriction education, referral to outpatient care or transfer to inpatient care). This occurred for both face-to-face (F2F) visits and for telehealth visits.

For patients who had a documented past attempt from a previous ASQ screen, the ASQ was modified to ask whether there had been a suicide attempt since their last screening. If not, then the screen results were classified as negative.

Suicide risk screening is administered by nurses or care assistants who ask the screening questions verbally. Screening results, along with key demographic and clinical variables in the electronic health record, are deidentified by an “honest broker”¹²  and stored in a data repository that is available for research inquires. Data from this repository were used for the current study. Data were analyzed from individuals ages 12 to 24 years old (inclusive) having encounters occurring between April to June of 2019 (T1) and April to June of 2020 (T2).

We excluded patients who were seen in mental health clinics because the suicide risk assessment in those clinics follows a different protocol. We also excluded patients who had encounters for which suicide screening was omitted due to medical contradictions.

Within the specified study windows, some patients had multiple encounters and, in some encounters (eg, inpatient visits), some patients had multiple suicide screens. For patients with multiple encounters and/or screens in T1 or T2, we selected the first suicide screening of the first encounter to keep the unit of analysis at the patient level. Fields that were extracted from the repository for study analyses included patient demographic variables (age at the encounter, self-reported sex and race), encounter characteristics including the type of visit (F2F or telemedicine), the patient’s insurance type (public, private, or self-pay), the month in which screening took place at the encounter, and the patient’s answers to the suicide risk assessment questions. The Children’s Mercy institutional review board has determined research involving these deidentified data to be “not human subjects research.”

Frequency counts and percentages were used to summarize data on prevalence of suicide risk across medical settings. Considering some patients were seen in both periods (T1 and T2), patient-specific random intercepts were applied in generalized linear mixed models (GLMM) to evaluate the association of suicide risk and the COVID-19 pandemic. These analyses were carried out adjusting for the following covariates: encounter month, encounter type, patient age in years, patient sex, race, and type of insurance the family had at the time of the given encounter (the main effects model hereinafter). Our intention to evaluate the pandemic-by-visit-type interaction was not feasible because of virtually no telehealth visits before the pandemic. Instead, we fitted a separate GLMM to compare suicide risk among the year-by-visit type combinations (2019 FTF, 2020 FTF, and 2020 telehealth), adjusting covariates of month, age, sex, race, and insurance type (the pandemic-by-visit combination model hereinafter). GLMM parameters were estimated by adaptive Gaussian Quadrature by using 11 quadrature points. Because GLMM provided patient-specific estimates, the fixed-effect parameters were marginalized to give population-averaged interpretation¹⁶  and presented in adjusted odds ratio (aOR) and the corresponding 95% confidence interval (CI). The analysis was conducted in R version 4.0.2 by using the GLMMadaptive package version 0.8-0.¹⁷ 

Wilcoxon rank test and χ² tests were used as applicable to evaluate the association of screening status (screened or not) with patient demographics with standardized mean differences (SMD) of at least 0.2 used as a threshold for identifying group differences.

The inclusion and exclusion criteria led to analyzing records of 22 896 patients at T1 and 14 635 patients at T2. Of this total number of patients across time points, 32 728 were unique and 4803 had encounters during both the T1 and T2 time periods. Of the patients analyzed for T1 and T2, 9 (0.0%) and 6359 (43.5%) were seen via telemedicine visits.

Suicide screening was conducted in 19 265 of 22 896 (84.1%) and 9984 of 14 635 (68.2%) patients in T1 and T2. The proportion of eligible patients who were screened in the two years was similar among F2F visits (84.1% vs 85.7%). For telehealth visits in T2, 2891 of 6359 (45.5%) were screened (Fig 1).

Demographic characteristics of screened patients were similar across the 2 years (Table 1), but White patients and patients with private insurance were overrepresented among patients screened by telehealth at T2 (Table 2).

The overall rate of positive screens was higher in T2 than in T1 (12.2% vs 11.1%, aOR, 1.24; 95% CI, 1.15–1.35; Fig 2, year: 2020 vs 2019). For patients seen F2F, the proportion of positive screens was similarly higher in T2 than in T1 (12.9% vs 11.1%, aOR, 1.24; 95% CI, 1.14–1.35; Fig 2, 2020 F2F vs 2019 F2F). There was a lower rate of positive screens among patients seen in telehealth visits than among patients seen in F2F visits (10.7% vs 12.9%, aOR, 0.75; 95% CI, 0.66–0.87; Fig 2, 2020 telehealth vs 2020 F2F) in T2. When we compared positivity on telehealth visits to those in outpatient visits during the previous year, the positivity rate was higher for 2020 telehealth visits than for 2019 outpatient visits (10.7% vs 9.7%, aOR, 1.11; 95% CI, 0.97–1.26).

The odds of a positive screen increased by 12% with each additional year increment in age (Fig 2, aOR, 1.12; 95% CI, 1.10–1.14). When dichotomized into <16 or ≥16 age groups, a significant difference in positive screen rates among older adolescents as compared to younger ones (15.3% vs 10.6% in 2020, Table 3) was observed. The odds of a positive screen were higher for female patients relative to male patients (aOR, 2.23; 95% CI, 2.00–2.48), lower in Black patients (aOR, 0.85; 95% CI, 0.77–0.95) and patients of other races (aOR, 0.82; 95% CI, 0.74–0.91) relative to White patients, and higher among patients with public insurance (aOR, 1.88; 95% CI, 1.72–2.07) relative to those with private insurance. These covariate effects were similar between the models that contrasted year and year-by-visit (Fig 2).

Rates of F2F positive screens in T1 and T2, also differed by clinical settings. For both T1 and T2, respectively, patients who were screened in the inpatient setting had the highest rate of positive suicide screens (18.6% vs 20.0%), followed by patients screened in the ED (13.7% vs 15.7%) and patients screened in outpatient clinics (9.7% vs 9.9%). These differences between clinical settings were statistically significant in the combined 2-year data and in each year (P < .01).

The early months of the COVID-19 pandemic were psychologically stressful for children and adolescents. Children and adolescents reportedly had higher rates of anxiety and depression during the pandemic than in the previous year.^7,8  Studies produced conflicting results regarding suicide risk level during the pandemic. Nevertheless, the clear evidence that more adolescent patients are experiencing increased anxiety and depression gives reason for concern about the possibility of more suicide attempts.^8,9,11,18  Results of these studies are difficult to interpret because so many things changed during the early months of the pandemic. Fewer patients were seen in emergency departments and outpatient clinics, more visits were done by telemedicine, and some patients had better access to broadband internet (and thus both telemedicine and remote schooling) than others.

Patients seen F2F had higher rates of positive screens for suicide risk in 2020 compared to those seen in 2019. Patients seen by telemedicine, however, had lower rates. Furthermore, screening was completed in a lower percentage of patients seen by telemedicine than in those seen F2F. These results could be interpreted in a number of ways. Perhaps, in the early months of the pandemic, teenagers with higher levels of psychological stress were more likely to seek a F2F health visit. It could also reflect the recommendation of a health care provider for a F2F visit when a patient is already known to be high-risk. Such patient selection could have affected the rates of positive screening if sicker patients were more likely to be seen F2F. It is also possible that patients with access to broadband and thus telemedicine were less socially isolated and thus less psychologically stressed than patients who sought F2F visits. Higher rates of positive screens during F2F visits could also reflect better communication and trust during F2F visits leading to an increased willingness to disclose suicidal thoughts during the screening process. Future quality improvement efforts should focus on assessing and comparing quality in F2F and telehealth visits and ensuring that quality is comparable for the two types of clinical encounters. We speculate but cannot prove from this data set that some of the difference in rates of positive screens between F2F and telehealth visits could reflect the fact that F2F visits were more likely in clinical encounters that took place in higher-risk clinical situations such as the ED or inpatient units.

There were more “missed” screening opportunities during 2020 for telehealth encounters. Transition to telehealth was somewhat abrupt, and many clinicians were unsure in those early months how to best incorporate screening into this new clinic flow or no longer had access to nursing support in the same way during these visits. We cannot determine if our results would have remained the same if similar proportions of patients had been screened via telehealth visits in 2020.

Previous studies have shown lower rates of suicidal ideation in Black patients than in White patients.^1,2,9,17  This is changing. Increased risk among Black youth of all ages has been reported, with not only increased thoughts and behaviors, but higher rates of suicide deaths for Black children and adolescents compared to their White peers.¹⁹  The COVID-19 pandemic has had a disproportionate negative effect on minority communities; therefore, we thought it important to examine the effect of race on suicide screening results. Our results showed a relatively lower rate of positive suicide screens for Black patients relative to White patients. Further research is needed to understand what factors may lead to this finding, examining factors such as trust, racial discordance between patient and provider, and stigma associated with mental health problems.

Our study has several limitations. As noted above, we do not know how patients chose or were directed to F2F versus telehealth visits so we can only speculate about the possibility of selection bias. In addition, this was a study from the first 3 months of the pandemic. Our hospital, along with the rest of the country, was in crisis mode. Thus, we may not have been as skilled at telemedicine as we should have been. This may have led to lower rates of screening in patients who sought care by telemedicine. Further study should examine whether the patterns seen in this study persisted in later months of the pandemic. Finally, as noted, the ASQ has not been validated for telemedicine encounters. The study also has strengths. It differs from most previous reports in that it includes data from inpatient, outpatient, and ED visits. Ours is also the first to compare results on F2F visits to those from telehealth visits. Another strength is that we used screening data rather than relying on discharge diagnoses or previous reports of behavioral problems. Thus, we were more likely to identify children without known risk factors or previous reports of suicidal thoughts. Our results suggest important factors to be considered in both implementing and evaluating suicide screening programs.

Further research should examine trends for suicide risk during the later months of the pandemic. Research should also determine whether overall rates of suicide among teenagers increased during these months. Such data should help assess the efficacy of screening programs and clarify the effect of the COVID-19 pandemic on suicide risk and suicide among teenagers. These data could guide efforts to address the mental health needs of adolescents in the aftermath of the pandemic or during future similar public health crises.

We thank the Suicide Risk Assessment Task Force at Children’s Mercy Kansas City and the following collaborating stakeholders at Children’s Mercy Kansas City for their contributions to the design and implementation of the original clinical data collection and/or data extraction procedures from which our study data were derived: Children’s Mercy Research Institute Population Health; Clinical Informatics; Department of Accreditation and Regulatory Readiness; Department of Social Work; Division of Developmental and Behavioral Sciences; Medical Informatics; Nursing and the Patient Care Services Department.

aOR

adjusted odds ratio

ASQ

Ask Suicide-Screening Questions

CI

confidence interval

COVID-19

coronavirus disease 2019

ED

emergency department

F2F

face-to-face

GLMM

generalized linear mixed models

SMD

standardized mean differences