Research on the Tufts Be Well at Work Program for Employees With Depression: 2005–2020

Tufts Be Well at Work is aimed at addressing this large unmet need for services with integrated and tailored mental health and vocational strategies. Although the program gives primacy to occupational functioning, the knowledge and skills that employees gain often generalize to other life domains.

Tufts Be Well at Work is a telephone-based program and, therefore, can be provided by employers; service suppliers, such as contracted EAPs; behavioral health provider organizations; health insurers; and health care systems. Care is provided by counselors who have a master’s degree or higher during eight approximately 50-minute biweekly telephone sessions. The total duration is 16 weeks. Although the program is compatible with videoconferencing, we have tested the telephonic version only.

The program is supported by an electronic data system that is public facing (for employee screening, enrollment, informed consent, and data collection) and counselor facing (for scheduling, care documentation, standardized assessment, self-management planning, monitoring of counselor fidelity to the protocol, clinical supervision, and outcome reporting). The websites are privacy protected, password protected, and HIPAA compliant.

The program has four integrated components, each addressing a specific barrier to recovering occupational functioning. They are private electronic screening, accessible from a computer or mobile device to improve participation; work coaching and modification strategies targeting work-related barriers to functioning; work-focused CBT strategies to address maladaptive thoughts, feelings, and behaviors interfering with working; and psychoeducational and care coordination interventions to strengthen linkages to primary care physicians or other providers. The program also makes use of motivational interviewing, monthly reassessment, and self-management of depression principles.

Screening uses the self-administered nine-item Patient Health Questionnaire (PHQ-9) (28, 29) to assess depression symptoms and the Work Limitations Questionnaire (WLQ) (30, 31) to assess deficits in occupational functioning. Both employee and counselor receive the results. Work coaching and modification addresses barriers to effective functioning through acquiring new work strategies. The intervention includes a detailed discussion of the employee’s job demands, job control, stressors, and supports. With methods applied in vocational rehabilitation, supported employment, and management, the counselor and employee develop a work modification plan. This plan, which is designed to be self-administered and does not require employer involvement, promotes improved functioning through adopting changes in work behavior (e.g., new time management techniques), work routines (e.g., collaborating versus staying isolated), or the work environment (e.g., reducing distractions).

Work-focused CBT addresses psychological barriers to functional improvement, promoting strategies to change behaviors and cognitions that interfere with effective functioning (20, 32). Employees learn to identify the maladaptive patterns of thinking, feelings, and acting that are eroding work functioning and to substitute more adaptive habits. A series of work-related experiments, customized to the employee, are carried out.

The program also employs some of the principles of high-quality care associated with the more extensive collaborative care model (18, 33). This core component supports patient activation through motivational interviewing, psychoeducation, care coordination, self-management support, and monthly assessments. With the employee’s permission, monthly assessments of symptoms and occupational functioning are sent to the employee’s regular care provider. More detailed description of the intervention is available elsewhere (3, 7, 15–17).

Over the course of 15 years, we have conducted six studies (Tables 1–3). (A table in an online supplement to this article presents further details.) These six studies were conducted in different settings and study populations, partly by design and partly opportunistically. Because we are interested in presenting both the journey and the results, we report on each study, including the nuances of design and methods, results, and conclusions, rather than conducting a meta-analysis of these studies. A recent and comprehensive systematic review, which includes our three RCTs, is available (27).

The primary objective of each of the six studies was to test the effectiveness of the program and secondarily to determine whether a return on investment (ROI) was obtained. Our research objectives, as well as the study designs and methods, reflect our original vision of the intervention as a program for employees delivered through the workplace directly by employers or their service providers. Because employers as well as employees were considered key stakeholders in the outcome of the research, most of the study sites were connected to the workplace (e.g., private-sector companies and public-sector government employers and benefits administrators). Over time, as health care systems diversified their services and telephonic care became more common, we conducted two studies (studies 4 [17] and 6) within health care systems. By 2020, we had conducted one relatively large multisite, multiemployer RCT (study 3 [16]); three single-employer studies, each involving multiple decentralized work locations (studies 1 [15], 2 [7], and 5); and two single-site studies (studies 4 [17] and 6) (Table 1 presents a detailed overview).

From the outset, we were committed to both preserving the applied, real-world nature of our research, which was important to employer stakeholders, and maintaining scientific rigor, and studies were designed to have a high degree of internal validity. We placed a priority on high-quality, uniform methods of screening, randomly assigning, and sampling employees, as well as data collection, measurement, and statistical analysis.

Each study included employed adults (based in the United States) who met criteria for major depression or dysthymia or (with the adoption of DSM-5) persistent depressive disorder, had productivity loss of 5% or higher in the past 2 weeks, could read and understand English, worked at least 15 hours per week, and had no history of psychosis or bipolar disorder. Five of the six studies had a minimum age requirement of 18 (study 3 enrolled middle-aged and older employees). Eligibility criteria related to health plan or source of care were included in three studies (studies 4–6).

In five of the six studies, screening and eligibility procedures were fully web based and self-administered; study 4 required in-person and telephone screening. Several employers had specific preferences related to the intensity, timing, frequency, and content of study advertising and the use of incentives—some required us to offer screening to all employees rather than directing advertising to those with symptoms. Such open screening increased the volume of website hits and decreased site eligibility rates (Table 2) while increasing variation in rates across studies. Ultimately, we completed 26,356 screenings, 2,251 individuals met study eligibility criteria, and 1,037 enrolled. The enrollment rate for the RCTs ranged from 35% to 88%. The lower rate reflects the variety in the study involving 24 workplaces (study 3). The higher rate reflects the benefits of enrolling in an academic medical center (study 4).

Among the studies, similar screening and data collection tools were used. Every study had a core set of screening, baseline, and follow-up questionnaires, which included validated and internationally used tools, such as the PHQ-9 for depression, the WLQ for presenteeism, and the WLQ time loss module for absenteeism. The timing of study observations was consistent with baseline at preintervention and follow-up at immediately postintervention (month 4). In the study at the Veterans Affairs medical center (VAMC) (study 4 [17]), a second follow-up at study month 8 was included, and all follow-up data were collected by telephone interviewers blinded to participant treatment group.

For the RCTs (studies 2–4 [7, 16, 17]), we used simple random sampling generally, assigning participants to the intervention or to usual care. Prior to intervening, we tested for treatment group equivalency on multiple clinical and demographic variables at baseline (preintervention). We also went the extra step of testing for group equivalence on occupational variables, such as psychosocial job demands and control, by using validated assessment metrics (34). Bias related to randomization, enrollment, and attrition was explored in detail throughout each study. In all studies, counselors were not blinded to treatment group assignment. To minimize treatment contamination, program counselors did not provide usual care.

In all RCTs, all participants could participate in usual care, although the nature of the usual care was not specified in advance and not paid for by the study. Subsequently, across studies, the usual care (whatever usual employer-sponsored health care, EAP, or other services were offered by the employer health insurer and EAP) obtained by study enrollees was highly variable, even within study sites. This variability reflects the realities of service access and use and employee preferences.

Five of the six studies used identical intervention models and each of the four components. In the VAMC study (17), care coordination was already available, and only work coaching and modification and CBT were provided as the intervention. Consistent with our general approach to real-world testing, three RCTs (studies 2–4 [7, 16, 17]) and one pilot implementation study (study 5) involved collaborations with the EAP contracted to the main collaborating organization. For example, study 2 involved the employer’s EAP in providing counselors, whereas in study 3, which involved many employers, Optum EAP supplied the study counselors.

Every study used our counselor-facing website to standardize intervention procedures. Because the counselors have some discretion in choosing specific intervention strategies, such as designing experiments for the CBT and work coaching portions, the website included several reference materials, including “Tools and Tips” containing strategies to improve functioning.

Except for the VAMC study (study 4), most participants were female (Table 2). Differences in age, race-ethnicity, and income were also reflective of the composition of the study sites. Although not shown, many different occupations were represented across the six studies.

Mean baseline PHQ-9 scores in the six studies ranged from 12.3 to 17.5 (on a scale of 0–27), indicating moderate levels of depression severity. In the RCTs, 25%−44% of participants had major depressive disorder only, 16%−50% had dysthymia or persistent depressive disorder, and 25%−40% had double depression (i.e., both). More than 50% had a history of antidepressant treatment, except in the VAMC sample (study 4), in which the rate was 27% (Table 2).

Regarding baseline occupational functioning (Table 2), the mean amount of time in the prior 2 weeks that enrollees had difficulty with time management ranged from 41.0% to 52.0%. Limitations performing physical job tasks ranged from 15.4% to 41.3%. Limitations in performing mental and interpersonal job tasks ranged from 32.4% to 46.8%. Difficulty performing output-related tasks ranged from 36.2% to 46.1%, Assuming 40 weekly work hours (80 hours over 2 weeks), occupational functioning was impaired approximately 32 to 40 hours in a 2-week period. These scores are typical of those we have observed in our other studies. On the basis of the WLQ productivity loss score algorithm, these scores translate into an estimated productivity loss at work of between 9.3% and 12.3% across studies. The mean number of days missed in the 2 weeks prior to baseline ranged from 1.1 to 2.7 days. Related productivity losses were between 15.0% and 25.9%.

Table 3 highlights the change from baseline to follow-up in outcomes and, for the RCTs, the degree to which the change differed between the intervention and the usual care groups. The specific statistical techniques included multiple regression, analysis of covariance to test for group differences, and mixed models for confirmation of effects and reporting of group differences. The non-RCTs analyzed changes from baseline to follow-up without adjustment (see table in online supplement).

In the RCTs, the adjusted mean changes in the percentage of at-work productivity loss in the intervention group were –3.5 for study 2, –4.5 for study 3, and –2.1 for study 4. In the usual care group, the mean changes in this primary endpoint were –0.3, –1.4, and –0.1, respectively (Table 3). In each RCT, the between-group differences in the changes were statistically significant. In the intervention group, the productivity gains were between 2% and 6%. In the usual care group, gains were minimal (0.3% to 1.4%).

For depression symptom severity, the adjusted mean changes in the PHQ-9 score for the intervention group were –5.4, –7.3, and –3.8 in each RCT, and in the usual care group, they were 0.6, –3.7, and –1.6, respectively (Table 3). Each of the differences in changes was statistically significant in the intervention group, demonstrating substantial improvement in PHQ-9 symptom severity (4–10 points).

In the non-RCTs (the original pilot and two implementation studies), the unadjusted mean changes in the percentage of at-work productivity loss ranged from –3.0 to –6.9. The unadjusted mean changes in depression symptom severity ranged from –6.1 to –10.6 (Table 3).

In the VAMC study (study 4), we conducted a second follow-up 4 additional months from the immediate postintervention period. Using mixed models, we found that the intervention group’s initial gain in at-work productivity loss at first follow-up did not erode at second follow-up (mean difference between the two follow-ups, −0.5; 95% confidence interval [CI]=−1.9 to 0.9, p=0.46). No significant loss was noted in the initial improvement in depression severity (mean difference between the two follow-ups, 0.6; 95% CI, −0.9 to 2.1, p=0.44).

As to whether the changes from baseline to 4 months are meaningful clinically or financially, several guideposts are useful for putting these into perspective. The improvement in symptom severity observed in the intervention groups in all studies (VAMC study 4 had smaller gains) is comparable to the reductions seen in antidepressant RCT trials that compared treatment to placebo (18, 35). For the presenteeism and absenteeism productivity parameters, a 1–percentage point productivity improvement is meaningful at group, company, and national levels. For example, a 2%−3% change in output nationally is regarded as an important indicator (36). On an employer level, if 6% of employees in a 10,000-person company with mean annual earnings of $45,000 were to have depression, a 1% improvement in productivity would generate productivity savings of $450 per employee with depression, or $270,000 companywide.

Finally, to assess ROI, we computed the intervention’s marginal benefits over those of usual care, minus the marginal program cost, divided by the marginal program cost. ROI is annualized. Marginal cost excluded website maintenance and research administrative costs. The benefit was the monetized value of the reduction in productivity loss due to presenteeism and absenteeism. Incremental benefits in productivity loss were then multiplied by median earnings to convert savings into dollars. In the three RCTs, the ROI was 5.4:1, 5.2:1, and 1.6:1. The lowest ROI was from the VAMC study (study 4), which had a smaller effect on productivity loss because of work absences and a relatively higher counselor cost (see table in online supplement). The bottom line is that the Tufts Be Well at Work program was both effective and worth the investment, particularly from the employer perspective, with a cost savings range of $671–$979 per treated employee.

Much of the focus around psychiatric services delivery has been on individuals with identified mental disorders treated in the mental health system. We used a population-based approach, including universal screening for depression, to address the greater need of employed individuals, who seldom see mental health professionals, through a work-focused intervention (37). Our experience with the entire set of studies suggests that the Tufts Be Well at Work program is practical to implement and certainly has a place in the health care delivery system, which increasingly relies on telemental health services. As reviewed in the recent Cochrane Collaborative systematic review (27), results from the three RCTs suggest that the intervention is effective for employees with depression and deficits in occupational functioning and provides an ROI both in the workplace and in a health care system. We also have some evidence, although limited, that effects remained after the program ended. We are optimistic that we will have further studies, which will facilitate further evidence synthesis and economic analysis.

From the very beginning, we considered the importance of translating the Tufts Be Well at Work program to the workplace and working people. We were committed to moving care outside the clinical setting, making it accessible to busy people who may not have been willing or able to leave work and go to appointments. We were also aware that employer adoption of the intervention depended on making both a clinical and a business case. The business case is mainly based on cost considerations, although both payers and purchasers are becoming more interested in evidence of value that includes both treatment benefits and costs for employees and their employed dependents.

In the years between the first and last study, the employee health services landscape and patient care have changed. Workplace service offerings have become quite diversified, and companies frequently have contracts with multiple vendors, whose services have become more specialized. As a result, the EAP is no longer the only place for Tufts Be Well at Work to reach the workplace. Employers and payers have become more reliant on Internet-based mental health services, which are lower cost and relatively easy to implement and, at least in theory, offer access to care. These changes suggest that there is more opportunity for Tufts Be Well at Work to become part of a comprehensive workplace mental health care strategy. However, although many payers and purchasers can appreciate the importance of addressing occupational functioning and productivity loss, they are still bogged down by trying to reduce barriers to primary care and specialty care.

We also deliberately sought to address the concerns of a clinical audience interested in high-quality care by developing a manual-based, protocol-driven program that relies on developments in CBT, psychoeducation, patient activation, telemedicine, and, importantly, measurement-based care delivery and outcome assessment. We understood that for Tufts Be Well at Work to be disseminated and implemented, health care providers will need to appreciate both the need for such services, the effectiveness of the program for their patients, and the ease of using it. In the overburdened world of health care providers and systems generally, and during the COVID-19 pandemic specifically, this is still a challenge. However, when we started testing the program, there was no reimbursement mechanism to cover the care. Telehealth as a successful mode of behavioral treatments has slowly emerged, but until the pandemic, reimbursement for counseling was restricted by both federal and state regulations. This situation has changed, possibly making it more feasible financially to integrate Be Well at Work with patient care delivery systems.

The ability to move beyond the research arena into pilot implementation programs was strengthened by the conceptual model and study design, as well as by a primary emphasis on work outcomes as the clinical goal delivered in multiple real-world employment settings. The studies of Tufts Be Well at Work used extensively validated self-report measures of depression and work outcomes and an accessible protocol-driven intervention with careful monitoring and documentation of treatment and systematic efforts to follow up enrollees. Limitations included the reliance on self-report measures of outcomes; lack of administrative, independent work-functioning data; and the generally brief follow-up periods that did not permit observation of whether effects are long lasting. In addition, no head-to-head comparisons of our program with other programs have been conducted. Finally, we were unable to perform a formal cost-effectiveness analysis.

Clinical and translational research and development of Tufts Be Well at Work was not entirely a linear process of phases, with one study building upon another through several implementation phases. By necessity, we responded to available funding and collaborative opportunities in order to advance the science and practice of functional improvement for people with depression. We have learned that we can improve occupational functioning with an evidenced-based, short-term, low-cost program and provide care that is highly convenient, acceptable, and accessible to working people. The rationale and need for this program were strong before the COVID-19 pandemic. Now, with mental distress and job insecurity skyrocketing and creating additional pressures on working people, the rationale and need are even stronger. Accelerating the dissemination and sustainable implementation of high-quality, efficient care is imperative to improve the employment outlook of adults with depression.