This study was a 2-arm, parallel-assignment randomized controlled trial (RCT). Job-search behaviors and other outcomes were captured at baseline, during each of the 6 weeks of the intervention, and again at the 8- and 16-week postintervention follow-ups. Additional details about the study design, study protocol, and statistical analysis plan can be found in our protocol study [14]. The study was funded by the National Institute of Mental Health; the sponsor had no role in the design, conduct, analysis, or reporting of the trial. Throughout the study, an advisory board was maintained, comprising community members and experts in human resources, business, clinical psychology, and community engagement. A Data Safety Monitoring Board comprising a clinical psychologist with expertise in digital interventions and 2 statisticians was maintained; it met midway through the trial and approved continuation based on its review of data and procedures. The first participant was enrolled on August 7, 2023. The study was retrospectively registered with ClinicalTrials.gov on August 30, 2023, and the registration was publicly visible on September 7, 2023.

Data were stored in REDCap, which allowed data to be exported for analysis without identifying information, such as names and addresses [15]. Access to the REDCap database was available only to approved study team members, and data files used for analysis were kept on secure computers. Participants received compensation for the completion of study assessments of up to US $350 (US $50 for the baseline assessment; US $20 for each of the first 3 weeks of questionnaires; US $30, US $40, and US $50 for the fourth, fifth, and sixth weeks of questionnaires, respectively; and US $60 for each of the 2 follow-ups). Compensation was not provided for time spent on the intervention procedures in either of the 2 groups. Advarra served as the Ethics Review Board (Pro00062697) and approved the study; all participants provided written informed consent.

A randomization list, stratified by gender, was created by JWG using the blockrand package for R (R Foundation for Statistical Computing) [16], which used randomly varying block sizes and then equally allocated participants to groups within the block (ie, the study used a 1:1 allocation ratio). Block sizes were 2, 4, 6, and 8. A seed was set for reproducibility. The randomization list was uploaded by a study coordinator to REDCap [15] and then used for sequential participants, within gender. After randomization, participants were onboarded by study coordinators (who had access to the randomization in REDCap) to their respective arms of the study. Participants were told in the consent form “You will be randomized to one of two approaches that are designed to give you the tools to help you (a) cope with the stress of being unemployed and having to search for work and (b) navigate your job search most effectively.” Thus, participants did not receive information about which group was the DRIVEN group vs the control group.

Table 1 summarizes the content of the DRIVEN intervention as well as that of the control group.

R statistical software (version 4.5.3) was used for data analyses [23]. A mixed-model regression analysis, implemented in the nlme package of R [24], was used to compare job outcomes in the intervention group vs the control group. The baseline value was included in the model as a covariate, time was included in the model as a continuous variable (days since baseline), and a subject-specific random intercept was included in the model. A group × time interaction effect tested whether the intervention group influenced the trajectory of outcomes over time, and quadratic components were explored in the modeling to test for possible nonlinear changes over time. Separate models were fitted for follow-up data. Mixed model regression was chosen for the analysis, in part, because of the ability to include cases in the analyses even in the context of missing data, assuming that the data were missing at random. The proportion of people who obtained a job was compared using a chi-square test; survival analyses assessed whether time-to-job-finding was different between groups (survival analyses are presented in the Supplemental Results in Multimedia Appendix 1). All randomized participants were included for intent-to-treat analyses. A type I error rate of 0.05 was used for 2 primary outcomes without adjustment for familywise error rate.

Descriptive statistics are presented in Table 2, which show that participants were roughly evenly divided between men and women and were diverse with respect to race and ethnicity. Table S2 in Multimedia Appendix 1 presents participants’ engagement with the mobile app based on 3 activity types—cognitive behavioral exercises designed to help cope with unemployment-related distress, mood self-assessments, and instructional and/or motivational videos. Engagement was high (84.4%‐86.7%) among participants randomized to the DRIVEN condition. A 2025 CONSORT (Consolidated Standards of Reporting Trials) diagram is presented in Figure 1 [25], which shows the flow of participants through the study from screening to completion (Checklist 1). Table 3 shows the marginal values of the means, SEs, and between-group standardized mean differences (and corresponding 95% CIs) for preparatory job-search behavior, as well as for active job-search behavior, depression, and positive affect.

At the start of the trial, participants were asked to rate the credibility and expectations of the intervention to which they were randomized. For intervention credibility, DRIVEN participants had a mean of 22.5 (SD 4.0), while control participants had a mean of 23.7 (SD 3.6); 2-tailed Welch t_122.77=1.82; P=.07. For expectations of change, DRIVEN participants had a mean of 15.2 (SD 3.1), while control participants had a mean of 15.5 (SD 3.0); Welch t_123.18=0.66; P=.51. In sum, there were no differences in expectations or credibility between the groups.

The rate of missing data was low in this sample, with less than 5% at each time point for the primary outcomes. In a test of our primary hypothesis, we found that participants receiving the DRIVEN intervention exhibited significantly more job-search behavior. In the DRIVEN group, preparatory job-search behavior increased significantly by 0.053 (95% CI 0.031-0.076) units per day since baseline. In the control condition, the increase in job-search behavior was smaller (0.01-unit increase for each day, 95% CI –0.011 to 0.033; not significant). Importantly, the contrast of these 2 quantities was statistically significant: Welch t₆₀₄=2.658; P=.008. A test of whether time (continuous) had a nonlinear trajectory over the course of the study was not significant, as inclusion of a time-squared effect did not significantly improve the model (likelihood ratio test=0.11; P=.95). Results of this hypothesis are shown in Figure 2. Importantly, there was a high level of individual differences across trajectories over time, suggesting caution in interpreting summary statistics (Figure 2A).

For active job-search behavior, a similar pattern was evident (Figure 3). The increase in active job-search behavior for those in DRIVEN was 0.061 (95% CI 0.038-0.084), whereas the control group experienced a smaller increase over time in active job search (point estimate: 0.025, 95% CI 0.001-0.047); Welch t₆₀₈=2.226; P=.03. Collectively, Figures 2 and 3 show the difference in trajectory between the DRIVEN and control groups in terms of job seeking, as well as a high degree of person-to-person variability.

As shown in Table 3, the groups were neither statistically nor substantively different at the 8- and 16-week follow-ups on the 2 job-search measures. However, the groups differed on our secondary outcome of tangible job-attainment. In the DRIVEN group, 68.3% (43/63) of participants found a job compared with 50.8% (32/63) in the control group, a statistically significant and meaningful difference (χ₁²=4.0; P=.046). The groups did not differ on time to job-finding (also a secondary outcome), with the difference between the groups emerging at the last time point of the study (Supplemental Results and Figure S1 in Multimedia Appendix 1).

We also examined differences in change over time (measured using continuous time since baseline) between the 2 groups on PROMIS Depression T-scores. The DRIVEN group showed a decrease in depression (point estimate=–0.214, 95% CI –0.255 to –0.172) over the course of the study, a decrease that was significantly larger than that observed in the control group (point estimate=–0.122, 95% CI –0.163 to –0.081); Welch t₆₀₄=3.113; P=.002.

A similar pattern was observed for PROMIS positive affect T-scores. The DRIVEN group showed an increase in positive affect (point estimate=0.306, 95% CI 0.260-0.352) over the course of the study, an increase that was significantly larger than that observed in the control group (point estimate=0.202, 95% CI 0.156-0.248); Welch t₆₀₄=3.132; P=.002.

We found that a 6-week app-based intervention that integrated teaching job-search behaviors and distress-reduction strategies was efficacious for (1) increasing job-search behavior, (2) increasing the odds of finding a job during the study period, and (3) reducing depression. These effects were statistically significant, even relative to an active control group in which people received standard job-seeking advice by email. We observed small-to-medium-sized effects, and these were not sustained in the 8- and 16-week postintervention periods, where the 2 groups no longer differed in job-search behavior, depression, or positive affect.

These results are relevant to psychiatric care. Mental disorders often co-occur with, and can maintain, joblessness [26], but there are few scalable treatments that focus specifically on helping job seekers return to the workforce that also take their mental health into account. Pharmacotherapy and behavioral interventions that are commonly used for depression and anxiety focus on treating symptoms, which in turn are expected to help improve functional domains such as being able to attain and keep a job. Although interventions exist for unemployment and the intersection with mental health, they are not in wide use and mixed findings on their efficacy have led to calls for more research [27]. Moreover, a recent Cochrane review and meta-analysis found that the few employment interventions that included “health-improving” components (eg, CBT and exercise) improved employment rates, but had mixed results, preventing broad and definitive conclusions about employment interventions due to low-quality evidence [28].

DRIVEN, therefore, fills a public health need as it is scalable, being largely delivered via a smartphone app, including 3 sessions with a job coach to support engagement with the intervention as well as the participant’s job search. Moreover, engagement with the DRIVEN app was high (>84%; Table S2 in Multimedia Appendix 1), suggesting a high degree of acceptability. However, more research is needed on the durability of the intervention, as the effects were no longer present during follow-up periods. In addition, given the heterogeneity of the needs of job seekers, it is likely that a “precision medicine” approach may be needed [29], where more attention and resources are allocated to higher-need participants, and specific techniques are matched to the needs of the job seeker.

To increase access to interventions for unemployed job seekers, interventions such as DRIVEN can be delivered in venues beyond traditional mental health clinics. First, federal, state, and local government agencies typically offer basic career training, job fairs, online resources, and resume guidance [30]; app-based interventions like DRIVEN can be easily implemented in these contexts. Second, private companies sometimes offer outplacement services prior to employee dismissal (eg, career counseling resources to support transitions to a new job) and are often interested in addressing the mental health needs of their recently laid-off employees, suggesting another venue for DRIVEN [31]. Partnering with these organizations could expand the possible mental health and economic impact of interventions like DRIVEN.

This is the first RCT of DRIVEN, a cognitive behavioral intervention designed for job seekers that addresses the emotional and practical needs of unemployed individuals. Subsequent work should seek to replicate these findings, as well as to determine the generalizability to other populations (eg, in a mental health setting where all participants have a mental health diagnosis). The variability in change over time suggests that tailoring the intervention may be needed (eg, certain individuals may benefit from a more intensive intervention or more coaching sessions). Modifications to the program may be needed for specific subpopulations, as well as for underemployed individuals seeking to improve work-related occupational functioning.

This study focuses on whether DRIVEN works, and subsequent analyses will focus on the mechanisms of change for this intervention (ie, why it works). In our protocol, we proposed to analyze which changes in psychological constructs lead to changes in job-seeking behavior, and vice versa. We intend to publish these results in the future as a separate journal article. The hypothesized mechanisms of DRIVEN focus on thinking flexibly and taking action, two aspects of CBT. Other employment-related interventions focus on additional mechanisms of change, such as interpersonal competence and the reduction of social anxiety [32]. Specific mechanisms of job-support interventions remain an important direction for future studies.

Although these initial results are encouraging, some limitations must be noted. First, participants were recruited online from major US metropolitan areas and completed the intervention remotely. Although this approach is advantageous for increasing access, it may not generalize well to certain locales (eg, rural areas) and other health care settings (eg, a medical center environment). In addition, the use of Craigslist for recruitment limits generalizability to the broad population of job seekers. Second, we excluded people with suicidal ideation. This was done for reasons of participant safety. However, the association between suicidality and unemployment is well documented [33]. Third, there were limitations in scope; the DRIVEN program was only available in English, and only available to those with access to a smartphone. Although this represents a large population, we hope to remedy this in future iterations of the program. Fourth, in trials of behavioral interventions, it is difficult to maintain blinding because patients are aware of the procedures in their received intervention (for a review of issues related to blinding in behavioral intervention trials, Juul et al [34]). Study staff were not blind to assigned treatment group because research team members needed to set up participants for the study based on their assigned condition. That said, knowledge of the active vs control group was masked from participants, and the 2 groups did not differ in terms of expectations. Fifth, participants in both groups of this study were paid for participation and could earn up to US $350 over the course of the study for completing study assessments. In real-world implementations of interventions, participants are typically not paid. Thus, whether DRIVEN is effective in future implementations of the program remains to be seen and will be the subject of future studies. Finally, groups did not significantly differ on self-reported outcomes during the follow-up periods, or on time taken to find a job. Future studies may need to consider a slightly longer intervention, especially for people who have been unemployed for a long time and need intensive support.

In sum, compared with a control condition designed to reflect job seeking as usual, a scalable, smartphone-based intervention that integrates mental health with job-seeking strategies (DRIVEN) was more efficacious in improving job-search behaviors, reducing depression, and enhancing positive affect from baseline to the 6-week follow-up, ultimately increasing the likelihood of obtaining a job. Given the robust impact of unemployment on mental health across mental disorders [35], these results further point to the importance of targeting employment status as a public health outcome. As rates of depression may be increasing [36], combined with economic uncertainties, interventions that acknowledge the intersection of employment and mental health may be increasingly needed.