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.
Digital health services can serve as scalable solutions to address the growing demand for mental health care. However, more research is needed to better understand the association between engagement with care and improvements in subclinical outcomes.
This study aims to fill this research gap by examining the relationship between members’ engagement with the Ginger platform and changes in their psychological resilience.
We conducted a retrospective observational study of 3272 members who accessed Ginger, an on-demand mental health service, between January 2021 and November 2021. Each member completed the 10-item Connor-Davidson Resilience Scale questionnaire, a measure of psychological resilience, at baseline and again during a 6- to 16-week follow-up window. Depression and anxiety symptoms (9-item Patient Health Questionnaire and 7-item Generalized Anxiety Disorder) were also measured. Linear regression was used to identify the association between engagement with Ginger’s multiple care modalities and changes in resilience. Moderator analysis was conducted to test whether clinical depression or anxiety at baseline moderated the relationship between engagement level and changes in resilience.
Of the 3272 members, 2683 (82%) reported low resilience at baseline. The mean change in resilience was 0.77 (SD 5.50) points. Linear regression models showed that age and census region did not predict changes in resilience; however, male members showed larger improvements (coefficient=0.58;
Engagement with Ginger services was associated with improvements in resilience. Members who engaged in coaching or clinical care had significantly larger improvements compared with those who only engaged in self-guided content, regardless of whether a member screened positive for clinical depression or anxiety at baseline.
Resilience is a multidimensional construct that may be viewed as one’s ability to cope with stress or represent personal qualities that enable one to thrive in the face of adversity [
Although many researchers’ conceptualization of resilience is somewhat related to the notions mentioned earlier, namely, individuals’ ability to cope with stress and adverse circumstances, there is no consensus on the operational definition of resilience according to a recent meta-analysis [
In light of the current pandemic, there has been increased focus on employee mental health and recognition of the importance of resilience in daily stress management and overall well-being [
Behavioral health coaching draws from several theoretical approaches that can effectively impact resilience or well-being. Coaching interventions address a variety of day-to-day challenges by identifying and working toward concrete and actionable goals. Resilience is seen as a proactive capability that supports the attainment of such goals and enhances overall mental health [
Previous studies have indicated a positive relationship between resilience and well-being, with higher resilience in the workplace setting associated with better mental health, reduced stress, and greater well-being [
In addition to resilience-focused interventions, studies have found that general coaching interventions have also demonstrated that coaching can support resilience, even in the absence of it being the focus or aim of services. For example, a randomized controlled study of executives in a public health agency found that individual coaching sessions enhanced goal attainment, increased resilience and workplace well-being, and decreased depression and stress compared with controls [
Ginger offers various types of care designed to provide mental health support, including self-guided content, text-based behavioral health coaching, teletherapy, and telepsychiatry. Theoretically, each modality of care has a different effect on resilience. This hypothesis was tested in this study. Furthermore, given the clinical focus of therapy and psychiatry, these modalities may have different impacts on resilience depending on whether a member presents with clinical symptoms. It could be that these modalities are more effective at impacting resilience for these members if interventions designed to impact clinical symptoms are more impactful on resilience. Alternatively, if addressing clinical symptoms is the focus of care before addressing subclinical outcomes such as resilience, we could expect to see that therapy and psychiatry have a smaller impact on resilience (but perhaps have an equal or larger impact on a time horizon beyond the scope of this study). Given the unique Ginger context that offers multiple care modalities, testing whether clinical symptoms moderate the impact of engagement on resilience is possible. We are not aware of any existing studies that explicitly test this moderator hypothesis.
Overall, literature supports the relationship between resilience and other mental health and well-being outcomes and the fact that interventions, including coaching, can bolster resilience. Given that most of these studies have been conducted in controlled research settings, it is important to supplement this knowledge to better understand what is happening in real-world settings, particularly when a global pandemic introduces unique challenges to resilience.
The purpose of this study was to examine changes in resilience among members seeking on-demand mental health treatment. We explicitly tested the following three hypotheses:
Change in resilience is associated with member characteristics at baseline, including demographic characteristics and baseline mental health outcomes (baseline resilience, depression symptoms, and anxiety symptoms).
Engagement with Ginger care is associated with larger improvements in resilience.
Baseline depression and anxiety symptoms moderate the association between engagement and changes in resilience.
Consistent with previous literature [
This study contributes to the literature on resilience in several ways. First, we present the results of one of the largest longitudinal studies of resilience. Our sample includes 3272 individuals. Second, to our knowledge, this is the first study to specifically examine resilience in the context of a digital mental health system that offers self-guided content, text-based behavioral health coaching, telepsychotherapy, and telepsychiatry. Third, by leveraging our rich data on Ginger members, we tested specific hypotheses that relate resilience to clinical depression and anxiety symptoms. In particular, we were able to test for the first time whether depression and anxiety symptoms moderate the impact of coaching and clinical interventions on resilience.
This was a retrospective observational study of Ginger members: individuals who joined Ginger, an on-demand mental health system. Data were collected between January 1, 2021, and November 13, 2021, from Ginger members residing in the United States. As part of its measurement-based care system, Ginger used the CD-RISC-10 as an indicator of resilience at intake as well as to track treatment progress beyond anxiety and depression symptom scores. By leveraging a retrospective design, this study contributes to the growing literature using real-world evidence. Although such studies often lack clear causal inference, they offer increased feasibility, larger samples, and robust external validity.
Study participants had access to the Ginger system as part of their employee or health plan benefits. Internal clinical protocols include the following exclusionary criteria, where self-directed telehealth is not likely appropriate and more specialized and urgent psychiatric services are required: (1) active suicidal ideation; (2) active high-risk self-harm behavior; (3) 2 or more hospitalizations within the past 6 months or 1 hospitalization in the past month for psychiatric reasons; (4) certain symptoms of psychosis that are poorly managed (eg, member is not medication compliant or symptoms are unresponsive to treatment) and are likely incompatible with telehealth; (5) a primary diagnosis of a substance use disorder or moderate to severe substance abuse issues, owing to the high complexity, severity, and risk frequently associated with such members, as well as the need for specialized care; (6) active eating disorders with symptoms considered high-risk; (7) ongoing grave disability, including certain patients who are bipolar with active mania or hypomania or mixed episodes who are unmedicated or have poor compliance with a medication regimen over time; and (8) two or more medical hospitalizations in the last month, owing to the high likelihood that the individual has a poorly controlled medical condition that requires close monitoring. For this study, we included Ginger users aged ≥18 years who downloaded the app during the data collection period.
Ginger provides virtual on-demand mental health services, primarily through employee or health plan benefits. Using a mobile app platform, Ginger members can access text-based behavioral health coaching, teletherapy, and telepsychiatry, as well as self-guided content and assessments. Individuals who are eligible for Ginger can download the mobile app, complete an onboarding process, and begin texting with a behavioral health coach within minutes. Members who are interested in or have been determined to be in need of a higher level of care can meet with a therapist or psychiatrist via video. All participants had access to self-care activities via mobile apps. Additional details regarding the Ginger system can be found in prior publications evaluating depression and anxiety outcomes as measured by the 9-item Patient Health Questionnaire (PHQ-9) and 7-item Generalized Anxiety Disorder (GAD-7) surveys [
Ginger uses various assessments including the PHQ-9 and GAD-7 surveys as part of its measurement-based care system. Since December 2020, Ginger has used the CD-RISC-10 survey (referred to as an adaptability check-in within the app) to track progress beyond depression and anxiety symptom scores. This is particularly relevant to understand the needs of
The 10-item Connor-Davidson Resilience Scale survey in the Ginger mobile app.
A common assessment tool for resilience is the CD-RISC-10. As mentioned earlier, Ginger uses the CD-RISC-10 as a proxy measure of an individual’s psychological resilience level. The original researchers initially developed a 25-item scale to measure resilience or how well one is able to adapt to change and
The PHQ-9 is a 9-item self-report questionnaire that assesses the frequency and severity of depression symptomatology within the previous 2 weeks. Each of the 9 items is based on the Diagnostic and Statistical Manual of Mental Disorders, 4th edition criteria for major depressive disorder and is scored on a 0 (not at all) to 3 (nearly every day) scale. Items include
The GAD-7 is a valid, brief self-report tool used to assess the frequency and severity of anxious thoughts and behaviors over the past 2 weeks. Each of the 7 items is based on the Diagnostic and Statistical Manual of Mental Disorders, 4th edition diagnostic criteria for generalized anxiety disorder and is scored on a 0 (not at all) to 3 (nearly every day) scale, with total scores ranging from 0 to 21. Items include
Coaching sessions were operationalized as the number of unique days on which members and coaches each exchanged at least five text messages, the minimum we believe is needed to capture a productive conversation between members and their coaches. Clinical sessions were operationalized as the number of video sessions completed with a clinician.
For this study, 5 different levels of engagement were considered based on members’ engagement with self-guided content, text-based coaching, and teletherapy sessions. Specifically, members who engaged only with self-guided content and did not complete any coaching or clinical sessions were categorized as the
Engagement levels.
Engagement level | Definition | Rationale |
Self-guided | Engagement only with self-guided content; 0 coaching and 0 clinical sessions | These are members who have engaged with the app’s self-guided content but have not interacted with any coaching or clinical care. This group serves as the primary reference group in the linear regression models. |
Low engagement | Between 0 and 3 total sessions comprising coaching or clinical care | Internal analyses suggest that 4 sessions are an inflection point for meaningful symptom reduction. These are members who have not reached this threshold. |
Coaching only | ≥4 coaching sessions and 0 clinical sessions | These are members who have completed at least the internally established threshold of 4 sessions but exclusively with coaching care. |
Clinical only | ≥4 clinical sessions and ≤1 coaching session | These are members who have completed at least the internally established threshold of 4 sessions but exclusively with clinical care. |
Hybrid care | >1 coaching session and ≥4 clinical sessions or ≥1 clinical session and ≥4 coaching sessions | These are members who have completed more than the internally established threshold of 4 sessions using a combination of coaching and clinical care. |
For each member, the following data were either collected at baseline or were fixed characteristics of members: age group, gender, geographic region, PHQ-9 score, and GAD-7 score. The demographic and location data were not self-reported. Instead, they were reported by a member’s parent organization, which is either their employer or health insurance plan. Baseline PHQ-9 and GAD-7 data were collected using the Ginger system. The baseline PHQ-9 and GAD-7 scores were selected within 1 week before and after a member’s baseline CD-RISC-10 score was collected, and the first PHQ-9 and GAD-7 scores were chosen. Members without baseline PHQ-9 and GAD-7 scores were excluded from analysis.
For many of our participants, the baseline characteristics were missing. The data were missing owing to 1 of 2 reasons. First, a member’s parent organization may not share members’ demographic information. Thus, missing demographic data are a signal of a member’s parent organization and not necessarily a signal of information specific to a given member. For example, of the 197 parent organizations represented in this study, 118 (59.9%) reported all their members’ gender information, 76 (38.6%) reported no gender information, and the remaining 3 (1.5%) organizations reported gender information for some but not all of their members.
The sample for this study included Ginger members residing in the United States who completed a baseline survey between January 1, 2021, and November 13, 2021. Members were excluded from the analysis if they satisfied any of the following criteria:
Engagement criterion: Members completed more than 1 coaching session or any number of clinical appointments before their baseline resilience scores. The members’ baseline survey was sent after their first coaching session.
Follow-up criterion: A member did not have a follow-up resilience score between 6 and 16 weeks from baseline.
PHQ-9 and GAD-7 criteria: Members without valid PHQ-9 and GAD-7 scores within a week of their baseline resilience score.
There were 17,654 members in the baseline resilience survey, of whom 6061 (34.33%) were excluded for meeting the engagement exclusion criteria. Of the remaining 11,593 participants, 3383 (29.18%) completed a follow-up survey between 6 and 16 weeks from their baseline survey. Of these, 3.28% (111/3383) did not have a valid PHQ-9 or GAD-7 score and were thus excluded. The resulting 3272 members comprised the full sample for this study and were used for our descriptive analysis. Of these 3272 members, 2674 (81.72%) had a low baseline score (CD-RISC-10 score<30) and comprised the low-resilience subsample for the study [
Sample flowchart. CD-RISC-10: 10-item Connor-Davidson Resilience Scale; GAD-7: 7-item Generalized Anxiety Disorder; PHQ-9: 9-item Patient Health Questionnaire.
Our descriptive analysis summarized the changes in resilience scores by presenting the mean, median, and SD of these change scores. To analyze differences across subgroups of members, we used a 2-tailed Welch 2-tailed
The subgroups of focus in this study were based on baseline resilience, depression, and anxiety symptom scores. In particular, members are grouped by their presence at baseline with low resilience (CD-RISC-10 score<30) and moderate to severe depression or anxiety (ie, PHQ-9 or GAD-7 score≥10) [
To understand the association between the demographic and baseline survey responses and changes in resilience, we estimated a multivariate linear ordinary least squares (OLS) regression. The dependent variable for this model was the change in resilience scores. The following categorical independent variables were included in the model: gender, age group, census region, and interacted indicators for whether a member’s baseline PHQ-9 or GAD-7 score was ≥10. For depression and anxiety at baseline, interacted indicators were included in the regression model to account for possible differences based on combinations of depression and anxiety clinical status. Given that missing demographic data are highly dependent on whether a member reports such data for any of their members, indicators for a member’s parent organization were included as independent variables. In addition, given the relatively wide follow-up period (6-16 weeks), indicator variables for the number of weeks between a member’s baseline and follow-up scores were included to account for secular time trends. The coefficients of the indicators for parent organization and weeks between scores have not been reported. For each of these independent variables, a category of members with missing data was included. Homoscedasticity was not assumed, and robust SEs were computed.
To understand the association between engagement with Ginger coaching and changes in resilience scores, we leveraged a moderator model with baseline depression and anxiety clinical status as the moderator and the level of engagement category (eg, self-guided, coaching only, and clinical only) as the independent variable. For our moderator categorization, members with either moderate to severe depression or anxiety at baseline were included in the clinical group (ie, PHQ-9 or GAD-7 score≥10), whereas all other members were included in the subclinical group. Clinical status at baseline was the hypothesized moderator of the association between the level of engagement and changes in resilience. We present the mean changes in resilience according to the clinical status for each engagement level. This analysis was restricted to members with low resilience at baseline (ie, CD-RISC-10 score<30).
To formally test whether clinical status at baseline was a moderator for engagement, we used a multivariate OLS regression model that included an indicator for engagement level interacting with an indicator of clinical status at baseline.
This study represents a secondary analysis of pre-existing deidentified data. The study team did not have access to participants or information to identify participants and did not intend to recontact participants. This study protocol was reviewed by Advarra institutional review board and determined to be exempt from institutional review board oversight, as deidentified secondary data analysis is generally not regarded as human subject research.
Distribution of baseline resilience (full sample). N=3272; mean=23.83; SD=6.47; median=24.
Distribution of changes in resilience at follow-up (full sample). N=3272; mean=.77; SD=5.5; median=1.
Resilience score characteristics by overall sample and subgroups group.
Characteristics | Participants, n (%) | Resilience score | Resilience change | |||
|
|
Mean (SD) | Mean (SD) | |||
All | 3272 (100) | 23.83 (6.47) | —a | 0.77 (5.50) | — | |
|
.19 |
|
.14 | |||
|
Female | 1377 (42.08) | 23.95 (6.49) |
|
0.67 (5.49) |
|
|
Male | 569 (17.38) | 24.39 (6.84) | 1.07 (5.56) | ||
|
Missing gender | 1326 (40.52) | 23.45 (6.26) | .005 | 0.74 (5.49) | .83 |
|
.01 |
|
.34 | |||
|
18 to 24 | 176 (5.37) | 22.84 (5.94) |
|
0.27 (5.16) |
|
|
25 to 34 | 844 (25.79) | 23.76 (6.10) |
|
0.54 (5.23) |
|
|
35 to 44 | 547 (16.71) | 24.54 (6.76) |
|
0.94 (5.37) |
|
|
45 to 64 | 461 (14.08) | 24.29 (6.49) |
|
0.89 (5.58) |
|
|
≥65 | 37 (1.13) | 25.03 (6.49) |
|
0.14 (5.77) |
|
|
Missing age | 1207 (36.88) | 23.48 (6.61) | .02 | 0.91 (5.75) | .28 |
|
.54 |
|
.40 | |||
|
West | 1052 (32.15) | 23.69 (6.06) |
|
0.64 (5.22) |
|
|
Midwest | 290 (8.86) | 23.93 (6.25) |
|
0.80 (5.55) |
|
|
South | 994 (30.37) | 23.94 (6.83) |
|
1.03 (5.81) |
|
|
Northeast | 397 (12.13) | 24.23 (6.46) |
|
0.65 (5.33) |
|
|
Missing region | 539 (16.47) | 23.53 (6.69) | .26 | 0.60 (5.56) | .45 |
|
<.001 |
|
<.001 | |||
|
High resilience (CD-RISC-10b≥30) | 598 (18.27) | 33.15 (2.84) |
|
–1.54 (5.11) |
|
|
Low resilience (CD-RISC-10<30) | 2674 (81.72) | 21.74 (5.06) | 1.28 (5.45) | ||
|
<.001 |
|
.31 | |||
|
PHQ-9c score≥10 | 1477 (45.14) | 21.71 (6.38) |
|
0.66 (5.78) |
|
|
PHQ-9 score<10 | 1795 (54.85) | 25.56 (6.01) | 0.86 (5.26) |
|
|
|
<.001 |
|
.55 | |||
|
GAD-7d score≥10 | 1327 (40.55) | 21.70 (6.44) |
|
0.70 (5.75) |
|
|
GAD-7 score<10 | 1945 (59.44) | 25.28 (6.07) | 0.82 (5.33) | ||
|
<.001 |
|
<.001 | |||
|
Self-guided | 499 (15.25) | 24.33 (6.66) |
|
–0.11 (5.64) |
|
|
Low engagement | 989 (30.22) | 23.87 (6.64) |
|
0.52 (5.60) |
|
|
Coaching only | 544 (16.62) | 24.61 (6.36) |
|
1.26 (5.03) |
|
|
Clinical only | 670 (20.47) | 23.45 (6.29) |
|
1.06 (5.35) |
|
|
Hybrid care | 570 (17.42) | 22.99 (6.18) |
|
1.15 (5.70) |
|
a
bCD-RISC-10: 10-item Connor-Davidson Resilience Scale.
cPHQ-9: 9-item Patient Health Questionnaire.
dGAD-7: 7-item Generalized Anxiety Disorder.
Demographic data were missing for a large portion of the sample because of irregular reporting by members’ employers or health plans. Of those without missing demographic data, most participants were female (1377/1946, 70.76%) and aged ≥35 years (1045/2065, 50.61%). Members were most likely to live in the West (1052/2733, 38.49%) and South (994/2733, 36.37%); however, all 4 census regions were represented in the baseline sample.
The baseline statistics and changes at follow-up are presented in
For categories based on gender and census region, neither the mean baseline resilience score nor the mean changes at follow-up were statistically different across groups (all
The vast majority (2674/3272, 81.72%) of members reported low resilience at baseline (ie, CD-RISC-10 score<30). On the basis of the PHQ-9 and GAD-7 scores at baseline, 45.14% (1477/3272) of the members screened positive for clinical depression at baseline (PHQ-9 score≥10) and 40.55% (1327/3272) for clinical anxiety at baseline (GAD-7 score≥10). Consistent with prior work [
By construction, members with low baseline resilience scored below those with high resilience (mean 21.74 vs 33.15). On an average, members with low resilience at baseline demonstrated an increase of 1.28 points at follow-up. Conversely, members with high resilience at baseline evidenced decreasing scores (−1.54 points on average, SD 5.11) at follow-up. The difference in mean resilience score changes between these groups was significant (
The most common engagement level was low engagement, with 30.22% (989/3272) of members meeting the criteria. There were 15.25% (499/3272) of members in the
Gender predicted changes in resilience scores when controlling for all baseline characteristics; male participants (mean 1.07, SD 5.56) had significantly larger mean improvements in resilience scores than females (mean 0.67, SD 5.49; coefficient=0.58;
Baseline resilience score was a strong predictor of changes in resilience scores at follow-up. Specifically, controlling for other variables, for each 1-point increase in baseline resilience score, the follow-up score decreased by 0.28 points, which was statistically significant at the 1% level. Baseline depression and anxiety were also strong predictors of changes at follow-up. Specifically, members without clinical depression or anxiety at baseline had mean resilience improvements of 1.44 points more than members with both clinical depression and anxiety (
Ordinary least squares regression of resilience change scores.
|
β (95% CI) | ||
|
|||
|
Female | Reference | Reference |
|
Male | 0.58 (0.02 to 1.14) | .04 |
|
Missing gender | –1.44 (–6.44 to 3.56) | .57 |
|
|||
|
18 to 24 | Reference | Reference |
|
25 to 34 | 0.10 (–0.74 to 0.95) | .81 |
|
35 to 44 | 0.55 (–0.36 to 1.46) | .24 |
|
45 to 64 | 0.59 (–0.36 to 1.53) | .22 |
|
≥65 | 0.28 (–1.78 to 2.35) | .79 |
|
Missing age | 0.22 (–2.46 to 2.89) | .87 |
|
|||
|
West | Reference | Reference |
|
Midwest | 0.08 (–0.91 to 1.07) | .88 |
|
South | 0.55 (–0.12 to 1.21) | .11 |
|
Northeast | –0.30 (–1.12 to 0.52) | .47 |
|
Missing Region | –0.30 (–1.53 to 0.94) | .64 |
Baseline resilience score | –0.28b (–0.31 to –0.25) | <.001 | |
|
|||
|
PHQ-9a score≥10; GAD-7b score≥10 | Reference | Reference |
|
PHQ-9 score≥10; GAD-7 score<10 | 0.07 (–0.57 to 0.71) | .83 |
|
PHQ-9 score<10; GAD-7 score ≥10 | 0.37 (–0.37 to 1.10) | .33 |
|
PHQ-9 score<10; GAD-7 score<10 | 1.44 (0.96 to 1.93) | <.001 |
R-squared | 0.1568234 (—c) | — | |
Adjusted R-squared | 0.0972076 (—) | — | |
Observations | 3272 (—) | — |
aPHQ-9: 9-item Patient Health Questionnaire.
bGAD-7: 7-item Generalized Anxiety Disorder.
cNot available.
Change in resilience by engagement level (low resilience sample).
Ordinary least squares regression of changes in resilience scores (low resilience sample).
|
β (95% CI) | ||||
|
|||||
|
Self-guided | Reference | Reference | ||
|
Low engagement | 0.91 (0.20-1.63) | .01 | ||
|
Coaching only | 1.82 (1.05-2.59) | <.001 | ||
|
Clinical only | 1.55 (0.78-2.32) | <.001 | ||
|
Hybrid care | 1.40 (0.61-2.19) | <.001 | ||
Subclinical, both | 1.30 (0.85-1.75) | <.001 | |||
R-squared | 0.1357372 (—a) | — | |||
Adjusted R-squared | 0.0658413 (—) | — | |||
Observations | 2674 (—) | — |
aNot available.
Change in resilience by engagement level and subclinical status (low resilience sample).
Ordinary least squares regression of changes in resilience scores, interacted model (low resilience sample).
Engagement level | β (95% CI) | ||
|
Reference | Reference | |
|
Interacted with subclinical | 1.39 (0.21 to 2.57) | .02 |
|
0.83 (–0.11 to 1.77) | .08 | |
|
Interacted with subclinical | 0.19 (–1.25 to 1.63) | .80 |
|
1.67 (0.55 to 2.79) | .003 | |
|
Interacted with subclinical | 0.21 (–1.33 to 1.76) | .79 |
|
1.74 (0.76 to 2.71) | <.001 | |
|
Interacted with subclinical | −0.51 (–1.97 to 0.95) | .49 |
|
1.53 (0.54 to 2.52) | .002 | |
|
Interacted with subclinical | −0.39 (–1.98 to 1.20) | .63 |
R-squared | 0.1364253 (—a) | — | |
Adjusted R-squared | 0.0650729 (—) | — | |
Observations | 2674 (—) | — |
aNot available.
The purpose of this study was to examine changes in resilience in members seeking on-demand mental health treatment as a function of both baseline symptoms of anxiety and depression, as well as engagement level. At baseline, most members (2674/3272, 81.72%) reported low resilience scores (ie, CD-RISC-10<30; mean 23.83, SD 6.47), which was well below the benchmarks of the US general population [
In subgroup analyses, members with low resilience at baseline, on average, demonstrated a 1.28-point improvement in resilience scores at follow-up. Resilience has been conceptualized as a dynamic process, and engagement with coaching or teletherapy may contribute to improvements in resilience, even in the absence of it being the focus of services [
The construct of resilience can vary as a function of time, context, gender, and age [
We found that members with symptom scores within the clinical range (ie, PHQ-9 or GAD-7 score≥10) demonstrated significantly lower baseline resilience scores. When controlling for other baseline characteristics, members without clinical depression or anxiety at baseline had improvements in resilience of 1.4 points more than members with both clinical depression and anxiety. Given the negative association between resilience and clinical symptom severity [
Regarding the association between care engagement level and both baseline and changes in resilience, there were significant differences in the mean baseline resilience and changes in resilience between those with and without meaningful engagement (ie, coaching only, clinical only, or hybrid care vs self-guided or minimal engagement). Specifically, members with self-guided engagement did not show any improvement in resilience, followed by those with minimal engagement having a 0.5-point improvement on average. Those with meaningful engagement (ie, coaching only, clinical only, or hybrid care) demonstrated improvements in resilience between 1.06 and 1.26 points. Within the Ginger context, these findings support the hypothesis that engagement with a human care provider (which is true for members in the coaching only, clinical only, and hybrid care groups) can lead to larger improvements in resilience than engagement with self-guided content alone. This is perhaps not surprising, given the more intensive nature of care delivered by a human provider. Behavioral health coaches, therapists, and psychiatrists tailor their care to the specific needs of a member, whereas it is incumbent on a member to find and engage with content that is applicable to their needs. This does not rule out that resilience-focused content could have an impact on resilience; however, given the wide variety of content available on the Ginger platform, this study was not designed to test whether resilience-specific content was associated with greater or lesser improvements in resilience than the care provided by a trained human provider. These findings are not surprising, as previous research has demonstrated that individuals who had greater engagement in a web-based resilience training program achieved the greatest improvements [
Controlling for baseline characteristics, members with subclinical status at baseline had larger improvements in resilience across all engagement levels; however, our moderation analysis found that the association between care engagement and changes in resilience did not significantly differ by baseline subclinical status.
From a research perspective, there is much ambiguity in approaching the concept of resilience and no current consensus on the operational definition of resilience [
The construct of resilience can vary as a function of time, context, gender, age, and cultural origin [
Another limitation of this study is that baseline surveys for resilience, depression, and anxiety symptoms were conducted at different times. This could have led to measurement errors, as both types of outcomes could evolve before the other is measured. The lag between collecting CD-RISC-10 and PHQ-9 and GAD-7 (the latter 2 were collected at the same time) was intentional in an effort to avoid survey fatigue. Therefore, the results of this study should be interpreted with this lag in mind.
This study examined changes in resilience over time among members of an on-demand virtual mental health system. Overall, members with low baseline resilience and subclinical symptoms of anxiety and depression demonstrated the largest improvement in resilience over time. Even for interventions that did not focus on resilience, members could demonstrate improvements in resilience over the course of treatment with virtual-based treatment for anxiety and depression. Future studies should examine symptom scores of anxiety and depression over time in relation to resilience. In addition, the inclusion of measures such as perceived social support might provide additional insight into treatment, given its association with both mental health and resilience [
Change in resilience by engagement level (full sample).
Additional results using the full sample.
Change in resilience by engagement level and subclinical status (full sample).
10-item Connor-Davidson Resilience Scale
7-item Generalized Anxiety Disorder
ordinary least squares
9-item Patient Health Questionnaire
All authors are employees of Ginger.