This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Formative Research, is properly cited. The complete bibliographic information, a link to the original publication on https://formative.jmir.org, as well as this copyright and license information must be included.
Suicide is a leading cause of death in the United States, and suicidal ideation (SI) is a significant precursor and risk factor for suicide.
This study aimed to examine the impact of a telepsychiatric care platform on changes in SI over time and remission, as well as to investigate the relationship between various demographic and medical factors on SI and SI remission.
Participants included 8581 US-based adults (8366 in the treatment group and 215 in the control group) seeking treatment for depression, anxiety, or both. The treatment group included patients who had completed at least 12 weeks of treatment and had received a prescription for at least one psychiatric medication during the study period. Providers prescribed psychiatric medications for each patient during their first session and received regular data on participants. They also received decision support at treatment onset via the digital platform, which leveraged an empirically derived proprietary precision-prescribing algorithm to give providers real-time care guidelines. Participants in the control group consisted of individuals who completed the initial enrollment data and completed surveys at baseline and 12 weeks but did not receive care.
Greater feelings of hopelessness, anhedonia, and feeling bad about oneself were most significantly correlated (
The results highlight the efficacy of an antidepressant intervention in reducing SI, in this case administered via a telehealth platform and with decision support, as well as the importance of considering covariates, or subpopulations, when considering SI. Further research and refinement, ideally via randomized controlled trials, are needed.
Suicide is a leading cause of death in the United States, claiming the lives of >47,000 people in 2019 [
Amid the current global COVID-19 pandemic, concerns arose about increases in SI and suicide, with a study suggesting a particularly heightened risk at the intersection of patient vulnerability, risk, resources, and mental health status [
Although 90% of those who commit suicide have a psychiatric diagnosis [
Several studies have demonstrated that depression is the most common psychiatric disorder among people who die by suicide, with an estimated 50% to 75% diagnostic prevalence in suicide cases [
In addition to the mental health conditions associated with an increased risk of suicide, certain physical health conditions such as chronic pain and chronic medical conditions have also been shown to be associated with increased SI and suicide attempts [
Various demographic variables have been investigated as potential risk factors for SI and suicide. In general, factors such as sex (male), ethnicity (White, American Indian, or Alaska Native individuals), education level (high school or less), and economic factors (unemployment) are associated with higher rates of suicide [
It has been relatively well established that suicide has a strong association with psychiatric disorders, especially major depressive disorder, and that pharmacological and nonpharmacological methods are often indicated for patients expressing SI as part of depressive symptomology. The course of treatment may commonly include prescribing antidepressants, such as selective serotonin reuptake inhibitors, serotonin-norepinephrine reuptake inhibitors, more modern antidepressants such as bupropion, older tricyclic antidepressants, and monoamine oxidase inhibitor antidepressants. Although antidepressants are a common treatment route, overall, there are conflicting findings regarding whether they reduce SI or suicide, or both [
Given the limited and inconsistent understanding of the effects of psychotropic treatment on SI, this study seeks to add to the literature by investigating the impact of psychiatric care, delivered via a telehealth platform, on SI. The objective of this study was, therefore, to examine the impact of this psychiatric care platform on SI, change in SI over time, and remission, as well as to investigate the relationship between various demographic and medical factors on SI and SI remission.
Participant data used in this study were obtained from a national mental health telehealth company (ie, Brightside) and consisted of 8581 US-based patients receiving psychiatric care for depression or anxiety, or both between October 2018 and April 2021 (treatment, n=8366; control, n=215). Participants were eligible if they (1) completed surveys at baseline and at 12 weeks; (2) denied any history of psychosis, schizophrenia, or bipolar I disorder; and (3) denied any history of chronic liver or kidney disease. Participants in the control group met the same criteria and signed up initially for Brightside but did not receive care. Brightside uses a free self-care product that sends emails requesting the completion of survey data over a period of 14 weeks even with no sign-up. The control group therefore consisted of individuals who completed the initial enrollment data and completed surveys at baseline and at 12 weeks. The treatment group included individuals who engaged in treatment with Brightside for at least 12 weeks. The demographic and clinical characteristics of the 2 groups are shown in
Demographic and clinical characteristics of the sample by group.
Characteristic | Treatment (n=8366) | Control (n=215) | |||||
Suicidal ideationa (baseline), mean (SD) | 0.77 (0.98) | 0.80 (1.04) | .63 | ||||
Patient Health Questionnaire-8 score, mean (SD) | 16.92 (4.38) | 16.15 (5.06) | .01 | ||||
Generalized Anxiety Disorder-7 score, mean (SD) | 14.81 (4.52) | 14.69 (4.82) | .69 | ||||
Age (years), mean (SD) | 32.02 (8.70) | 31.97 (10.42) | .94 | ||||
Sex (female), n (%) | 5928 (70.86) | 122 (74.39) | .34 | ||||
|
|||||||
|
Not White | 1727 (20.64) | 38 (25.33) | .16 | |||
|
White | 6639 (79.36) | 112 (74.67) | .16 | |||
|
Black or African American | 296 (3.54) | 9 (6) | .16 | |||
|
Asian | 286 (3.42) | 4 (2.67) | .16 | |||
|
Hispanic | 671 (8.02) | 16 (10.67) | .16 | |||
|
Other | 474 (5.67) | 9 (6) | .16 | |||
Education (beyond high school), n (%) | 5727 (68) | 78 (52) | <.001 | ||||
|
|||||||
|
Full-time | 5738 (68.59) | 135 (63.08) | .02 | |||
|
Part-time | 975 (11.65) | 19 (8.89) | .02 | |||
|
Unemployed by choice | 808 (9.66) | 28 (13.08) | .02 | |||
|
Unemployed | 845 (10.10) | 32 (14.95) | .02 | |||
|
|||||||
|
0 | 6261 (75.82) | 110 (73.83) | .68 | |||
|
1 | 1738 (21.05) | 32 (21.48) | .68 | |||
|
2 | 235 (2.85) | 6 (4.03) | .68 | |||
|
3 | 24 (0.01) | 1 (0.01) | .68 | |||
Presence of chronic painc, n (%) | 851 (10.30) | 28 (13.08) | .19 | ||||
Panic attacks, n (%) | 5712 (69.16) | 147 (69.01) | .96 | ||||
History of illicit drug use, n (%) | 650 (7.87) | 15 (7.04) | .80 | ||||
History of suicide attempts, n (%) | 195 (2.36) | 8 (3.72) | .18 |
aSuicidal ideation was measured using item 9 of the Patient Health Questionnaire-9.
bChronic pain status was missing in 1.28% (107/8366) of the patients in the treatment group.
cParticipants (65/215, 30.2%) of the control group had missing data on racial minorities, gender, and education.
During a patient’s first session, a licensed professional prescribed psychiatric medications for the patient in the treatment group. Enrolled Brightside patients completed an initial digital intake that included clinically validated measures of depression and anxiety, as well as questions about clinical presentation, medical history, and demographics. Brightside’s proprietary precision-prescribing platform analyzes these data points using an empirically derived algorithm to provide real-time care guidelines and clinical decision support to providers via the digital platform. Brightside platform provides this decision support via a computerized symptom cluster analysis at treatment intake. On the basis of the analysis of presenting symptom clusters, as well as decision support based on the empirical literature, treatment recommendations are provided [
The Patient Health Questionnaire (PHQ-9) is a 9-item self-report measure used to assess the severity of depressive symptoms present within the prior 2 weeks, as outlined by the
The Generalized Anxiety Disorder-7 (GAD-7) scale is a 7-item self-report instrument used to assess the severity of anxiety symptoms present within the prior 2 weeks as outlined by
Basic demographic variables such as age, education, sex, and employment status were collected at baseline. In addition, individuals were asked if they had the following chronic health conditions: asthma, cancer, Crohn disease, irritable bowel syndrome, heart condition, obesity, or diabetes. A simple count variable was created with the number of chronic medical conditions endorsed. In addition, endorsement of either chronic pain or fibromyalgia was considered a variable representing chronic pain. Respondents were asked whether they used illicit substances and whether they currently experienced panic attacks. Finally, the patients were asked if they had ever attempted suicide in the past. Some participants (65/215, 30.2%) in the control group were missing the information about race, ethnicity, sex, and education level, whereas 1.28% (107/8366) of the participants in the treatment group were missing chronic pain status.
Data analyses were performed using SPSS (version 28; IBM Corp) to assemble the patient data sample, apply inclusion and exclusion criteria, and establish baseline versus follow-up survey outcomes. Brightside maintains deidentified databases for analytics that facilitate granular insights into clinical decisions, interactions, and outcomes. Assumptions for conducting regression models were assessed using visual inspection of distributions, a scatter plot of the residuals, and variance inflation factor values, as well as by examining potential multicollinearity among predictors. For the logistic regression models, the Box-Tidwell test was used to test whether the logit transform was a linear function of the predictor. In univariate general linear modeling of baseline SI severity, checking the assumptions revealed heteroscedasticity. As such, models were run using Box-Cox transformations [
First, we examined zero-order correlations between item 9, the SI item, on the PHQ-9 and change over time, as well as with various demographic, clinical, and medical variables. We also examined the correlations between item 9 and the other PHQ items. Pearson or point biserial correlation coefficients were calculated based on the variables included. Univariate general linear modeling was used to explore the independent effects of demographic and clinical variables on the baseline SI severity.
We then examined the relative rates of SI at baseline and 12 weeks, as well as the percentage change over time. Chi-square analyses compared relative proportions between groups. An analysis of covariance examined the treatment effect by group, on changes in SI over time, controlling for baseline age and PHQ-8 scores. Bonferroni corrections were used in all follow-up 2-tailed
Next, using only those who endorsed SI at baseline (a score of ≥1), we calculated the proportion of people who
The WCG Institutional Review Board, Ethics Committee Panel 1, approved the retrospective research analysis of clinical data obtained by Brightside as part of routine clinical care (#1308524). The data were drawn from a deidentified clinical database.
Correlations between the SI item on the PHQ-9 and all other items were examined (
A univariate general linear modeling was used to explore the independent effects of these demographic and clinical variables on baseline SI severity (
Zero-order correlations between suicidal ideation (SI) and demographic, medical, and clinical variables (N=8581).
|
SI at baseline | SI at 12 weeks | Change in SIa | ||||
|
|||||||
|
Age | −0.16b | −0.10b | −0.12b | |||
|
Sex (female) | −0.00 | –0.03c | 0.01 | |||
|
Racial minority | 0.07b | −0.03c | 0.04b | |||
|
Education (beyond high school) | −0.16b | −0.10b | −0.11b | |||
|
Employment (less than full-time) | 0.05b | 0.05b | 0.03b | |||
|
|||||||
|
Anhedonia | 0.24a | 0.12a | 0.18a | |||
|
Feeling down, depressed, or hopeless | 0.37a | 0.15a | 0.30a | |||
|
Sleep issues | 0.13a | 0.07a | 0.10a | |||
|
Tired or low energy | 0.14a | 0.06a | 0.11a | |||
|
Appetite issues | 0.18a | 0.09a | 0.14a | |||
|
Feeling bad about self | 0.36a | 0.14a | 0.30a | |||
|
Trouble concentrating | 0.18a | 0.08a | 0.15a | |||
|
Psychomotor retardation or restless | 0.23a | 0.09a | 0.19a | |||
|
|||||||
|
PHQ-8 | 0.38a | 0.14a | 0.31a | |||
|
Generalized Anxiety Disorder-7 | 0.17a | 0.06a | 0.14a | |||
|
|||||||
|
Number of chronic medical conditions | 0.01 | 0.02 | –0.01 | |||
|
Chronic pain or fibromyalgia | 0.03b | 0.04a | 0.01 | |||
|
Panic attacks | 0.08a | 0.06a | 0.07a | |||
|
History of illicit drug use | 0.08a | 0.04a | 0.05a | |||
|
History of suicide attempts | 0.10a | 0.04a | 0.04a |
aChanges in SI represent changes from baseline to 12 weeks, with higher numbers representing decreased symptom severity over time on the PHQ-9 SI item.
b
c
dPHQ: Patient Health Questionnaire. PHQ-9 items are not the exact wording of the item. SI was assessed by responses to item 9 of the PHQ-9.
Independent effects of demographic and clinical characteristics on baseline suicidal ideation (N=8581)a.
Factor | Sum squares |
|
||
Age | 38.06 | 107.54 (1,7733) | <.001 | 0.02 |
Sex | 0.21 | 0.60 (1,7733) | .44 | 0.00 |
Racial minority | 2.44 | 6.90 (1,7733) | .01 | 0.00 |
Education | 18.14 | 51.24 (1,7733) | <.001 | 0.02 |
Employment | 0.56 | 1.51 (1,7733) | .22 | 0.00 |
Baseline PHQ-8b | 373.39 | 1054.98 (1,7733) | <.001 | 0.12 |
Chronic pain | 0.59 | 1.68 (1,7733) | .20 | 0.00 |
Number of medical conditions | 1.06 | 1.00 (1,7733) | .39 | 0.00 |
Illicit drug use | 8.12 | 22.94 (1,7733) | <.001 | 0.00 |
Suicide attempt | 5.60 | 15.82 (1,7733) | <.001 | 0.00 |
aThe dependent measure, baseline SI score, was transformed using the Box-Cox correction.
bPHQ-8: Patient Health Questionnaire-8.
At baseline, 46.5% (100/215) of the participants in the control group and 47.12% (3942/8366) of the participants in the treatment group expressed SI (
In terms of emergence of SI in those who did not initially endorse it (n=4539), the control group had 15.6% (18/115) emergence, whereas the treatment group had only 2.98% (132/4424; χ21=56.2;
Average suicidal ideation scores over time by group.
Mixed model analyses, controlling for age and PHQ-8, investigating differences in SI over time by treatment group, education level, and employment status revealed significant effects for the treatment group (F
Interaction among group, time, and education level. Covariates appearing in the model were evaluated at age=32.03 years and PHQ-8 score=16.87. Error bars represent 95% CIs.
Among those who endorsed SI at baseline but then endorsed none at 12 weeks, 76.37% (3087/4042) of the total sample manifested this complete remission. Complete remission was observed in 77.19% (3043/3942) of the treatment group and 44% (44/100) of the control group (χ21=59.5;
On the basis of bivariate relationships, the following factors were significantly associated with SI remission: being in the treatment group, older age, being female, being White (as opposed to a racial minority), obtaining education beyond high school, and lower depression severity (as measured by PHQ-8 scores) at baseline (
Factors predicting suicidal ideation (SI) remission.
Predictor | No SI remission (n=955) | SI remission (n=3087) | Odds ratio (95% CI) | ||||||
|
4.31 (2.88-6.44)a | ||||||||
|
Treatment group (n=3942) | 899 (22.81) | 3043 (77.19) |
|
|||||
|
Control group (Refb) (n=100) | 56 (56) | 44 (44) |
|
|||||
|
|||||||||
|
Age (years), mean (SD) | 29.43 (8.20) | 30.81 (8.28) | 1.02 (1.01-1.03)a | |||||
|
|
1.38 (1.18-1.62)a | |||||||
|
|
Female (n=2875) | 626 (21.77) | 2249 (78.23) |
|
||||
|
|
Male (n=1144) | 318 (27.79) | 826 (72.2) |
|
||||
|
|
0.79 (0.67-0.94)a | |||||||
|
|
Minority (n=934) | 250 (26.77) | 684 (73.23) |
|
||||
|
|
White (Ref) (n=3079) | 690 (22.41) | 2389 (77.59) |
|
||||
|
|
1.37 (1.18-1.59)a | |||||||
|
|
Beyond high school (n=2466) | 523 (21.21) | 1943 (78.79) |
|
||||
|
|
High school or less (Ref) (n=1547) | 417 (26.96) | 1130 (73.04) |
|
||||
|
|
0.95 (0.76-1.19) | |||||||
|
|
Unemployed (n=466) | 114 (24.46) | 352 (75.54) |
|
||||
|
|
Employed (Ref) (n=3575) | 840 (23.5) | 2735 (76.5) |
|
||||
|
|||||||||
|
PHQ-8c | 21.01 (3.98) | 19.81 (4.10) | 0.94 (0.92-0.96)a | |||||
|
GAD-7d | 15.57 (4.42) | 15.28 (4.46) | 0.99 (0.97-1.00) | |||||
|
|||||||||
|
|
0.81 (0.65-1.02) | |||||||
|
|
Chronic pain (n=456) | 123 (26.97) | 333 (73.03) |
|
||||
|
|
No chronic pain (Ref) (n=3537) | 818 (23.13) | 2719 (76.87) |
|
||||
|
|
0.91 (0.79-1.05) | |||||||
|
|
0 (Ref) (n=2993) | 695 (23.22) | 2298 (76.78) |
|
||||
|
|
1 (n=853) | 193 (22.63) | 660 (77.37) |
|
||||
|
|
2 (n=106) | 34 (32.08) | 72 (67.92) |
|
||||
|
|
3 (n=12) | 5 (41.67) | 7 (58.33) |
|
||||
|
|
0.79 (0.67-0.94) | |||||||
|
|
Yes (n=2939) | 724 (24.63) | 2215 (75.37) |
|
||||
|
|
No (Ref) (n=1053) | 217 (20.61) | 836 (79.39) |
|
||||
|
|
0.86 (0.68-1.08) | |||||||
|
|
Yes (n=405) | 106 (26.17) | 299 (73.83) |
|
||||
|
|
No (Ref) (n=3587) | 835 (23.28) | 2752 (76.72) |
|
||||
|
|
0.85 (0.57-1.27) | |||||||
|
|
Yes (n=128) | 34 (26.56) | 94 (73.44) |
|
||||
|
|
No (Ref) (n=3867) | 908 (23.48) | 2959 (76.52) |
|
aValues indicate that the predictor significantly predicts SI remission at the 95% CI.
bRef. represents the reference group.
cPHQ-8: Patient Health Questionnaire-8.
dGAD-7: Generalized Anxiety Disorder-7.
Multivariate model predicting suicidal ideation remission.
Factor | Odds ratioa (95% CI) |
Treatment group | 5.10 (3.03-8.43) |
Age | 1.02 (1.01-1.03) |
Sex | 1.49 (1.26-1.77) |
PHQ-8b | 0.94 (0.92-0.96) |
aAll odds ratios are
bPHQ-8: Patient Health Questionnaire-8.
The objective of this study was to examine the impact of psychiatric care on SI, change in SI over time, and remission, as well as to investigate the relationship between various demographic and medical factors on SI and SI remission.
Greater feelings of hopelessness, anhedonia, and feeling bad about oneself were most significantly correlated with SI at baseline. Sleep issues and feeling tired or having low energy, although significant, had lower correlations with SI. These patterns of associations between the SI item and other items of the PHQ-9 mirror those found in other studies with primary care patients who had depression or chronic pain and had completed the PHQ-9 [
Associations between greater SI severity and younger age are consistent with national survey data finding that younger adults more frequently endorse SI than older adults [
Not surprisingly, both the PHQ-8 and, to a lesser extent, the GAD-7 were significantly associated with SI (much more so at baseline) and with change over time. The baseline PHQ-8 score was the most prominent independent predictor of SI, accounting for 12% of the variance. This reflects a consistent finding in the literature that depression severity is highly associated with SI [
The number of chronic medical conditions was not significantly associated with SI or change over time. Endorsing either chronic pain or fibromyalgia was significantly associated with SI at baseline and week 12 but did not change over time, although the overall correlations were quite small. This is counter to prior research [
Although the groups were similar at baseline in terms of the presence of SI (47%) at baseline, after 12 weeks of treatment, only 12.32% (1031/8366) of the participants in the treatment group expressed any SI, compared with 34% (74/215) of the participants in the control group. These numbers are similar to a much smaller study investigating psychotherapy for depression [
Although the treatment group improved over time regardless of various demographic variables, in the control group, after controlling for age and depression severity, those with less education worsened over time. Greater education levels are protective against many adverse outcomes, including SI [
Those in the treatment group were 4.3 times more likely to remit than those in the control group (OR 4.31, 95% CI 2.88-6.44). Zisook et al [
Treatment was clearly the biggest predictor of SI remission. Other factors found to be significantly associated with SI remission were older age, being female, being White, obtaining education beyond high school, and having lower depression severity at baseline. Women and those with advanced education beyond high school were about 1.4 times (OR 1.38, 95% CI 1.18-1.62) more likely to remit than men and those without advanced education (OR 1.37, 95% CI 1.18-1.59). In the STAR*D trial, remission of depressive symptoms was more likely in those who were White, female, employed, or had higher levels of education or income [
The primary limitation of this study is that there was no random assignment to treatment, and alternative explanations for any observed treatment effects are possible. Although comparison to a control group that completed assessments on the same schedule as the treatment group, and which was largely equivalent to the treatment group at baseline, reduced the likelihood that any effects were because of engagement on line or other Hawthorne effects, the control group did not engage with providers. As the participants were not randomized, there is potential for confounding. For example, participants in the control group were more likely to be unemployed and have no education beyond high school. As it is unclear why the control group participants did not pursue treatment, one possibility is that they had less scheduling flexibility or less ability to take time off to attend treatment.
Another limitation is the inability to directly compare different medications, as this was a clinical sample with >400 different combinations of medications. Although this prevents generalizability to specific medication groups, it does speak to the ability of antidepressant treatment, as rendered in this novel virtual treatment regimen, to positively affect SI. An additional limitation, however, is that patients may have had other treatments such as psychotherapy that were not assessed. In addition, a specific measure of SI, such as the Beck Scale for Suicide Ideation or the Scale for Suicide Ideation [
Depression severity is the primary driver of SI, relative to demographic and other clinical or medical factors. Certainly, many clinicians may be reluctant to prescribe antidepressants in those with SI because of the perceived risk of working with patients who are suicidal. These results address, at least to some degree, these concerns. The results of this study, as well as those of others, are consistent with the efficacy of psychiatric care administered via a telehealth platform with decision support. In antidepressant trials, depression severity mediates the effect of antidepressant medication on suicide risk [
Finally, these results align with a growing body of literature demonstrating the effectiveness of using a telehealth platform for providing mental health services [
general anxiety disorder
Generalized Anxiety Disorder-7
odds ratio
Patient Health Questionnaire
suicidal ideation
Sequenced Treatment Alternatives to Relieve Depression
EOC, HGB, SS, and MW hold stocks in Brightside Health Inc.