1 The impact of a remote digital health intervention for anxiety and depression 2 on occupational and functional impairment: an observational, pre-post 3 intervention study 4

5 Marcos Economides, PhD a, Kristian Ranta MSc a, Outi Hilgert, MD a,

6 Dolores M. Kelleher, MS, DMHa,b, Patricia Arean, PhD c, Valerie L. Forman-Hoffman, PhD,

7 MPH a*

9 Author Note

10 a Meru Health Inc, San Mateo, CA, United States

11 b D Kelleher Consulting, Alameda, CA, United States

12 c Department of Psychiatry and Behavioral Sciences, University of Washington, Seattle, WA,

13 United States

15 This research did not receive any specific grant from funding agencies in the public,

16commercial, or not-for-profit sectors.

17 Correspondence concerning this article should be addressed to Valerie Hoffman.

18Contact: [email protected]

19 20

1 1 2 21Abstract 22Objectives 23The objective of this study was to assess whether the Meru Health Program (MHP), a novel, 24self-guided, structured smartphone-based intervention with asynchronous therapist 25support, is associated with improvements in occupational impairment and daily functioning 26in working adults with symptoms of depression and anxiety. 27Methods 28In this observational, pre-post study, occupational and functional impairment (assessed via 29the Workplace Performance and Activity Impairment [WPAI] questionnaire) was measured 30pre- and post-intervention in a real-world sample of employed patients (n = 64) receiving 31treatment via the MHP for symptoms of mild to moderate depression and/or anxiety 32(Patient Health Questionnaire-9 item scale [PHQ-9] > 4 and/or Generalized Anxiety Disorder- 337 item scale [GAD-7] > 4). 34Results 35The MHP was associated with improvements in presenteeism (mean decrease = 12.1, 95% 36CI = [4.17 – 20.0], p = 0.003; d = 0.56), overall work impairment (mean decrease = 13.4, 95% 37CI = [5.62 – 21.2], p < 0.001; d = 0.54), and impairment in daily functioning (mean decrease = 3814.8, 95% CI = [9.09 – 20.4], p < 0.001; d = 0.60), as well as smaller (non-significant) 39improvements in absenteeism (mean decrease = 4.32, 95% CI = [-1.00 to 9.64], p = 0.11; d = 400.25). A conservative estimate of annual productivity cost savings associated with the MHP 41is $6271 per employee, corresponding to a 30% relative improvement in overall work 42productivity. 43Conclusions 44Remote, smartphone-based interventions such as the MHP may help employees suffering 45with mild-to-moderate mental health issues function and perform better at work, resulting 46in cost savings for employers. 47 48 49 50Keywords: mhealth, digital therapeutics, depression, anxiety, productivity, absenteeism, 51presenteeism

3 2 4 52Introduction 53Depression and anxiety are both leading causes of disability and incur substantial costs 54associated with reduced work functioning, including absences, impaired productivity, and 55even decreased job retention worldwide [1-5]. A recent estimate across eight diverse 56countries suggests that depression in the workplace costs a collective $250 billion a year, 57and over $84 billion in the US alone [3]. Moreover, even minor or subthreshold symptoms of 58depression and anxiety have been shown to substantially decrease work performance [6-8]. 59Thus, there is an urgent need for cost-effective, evidence-based interventions that support 60employees with symptoms of depression and/or anxiety. Although effective treatments 61exist, interventions that can be delivered outside of normal working hours are urgently 62needed given the fact that working adults are particularly prone to citing time as a barrier to 63mental health care receipt. 64The advent of digital health technology has provided one such solution to counter these 65issues. Online and smartphone-based interventions for common mental health problems 66overcome access barriers to care [9,10] and have demonstrated clinical effectiveness 67comparable to face-to-face interventions [11-13]. However, the impact of such 68interventions remains primarily symptom-focused, with potential effects on work difficulties 69receiving considerably less attention. As a consequence, few studies have focused on the 70delivery and outcomes of these interventions to working adults, resulting in limited 71consensus regarding the effectiveness of both traditional treatment methods and newer 72digitally-delivered interventions in improving workplace outcomes such as productivity and 73attendance [14,15]. 74Previously, we described the design and feasibility of the Meru Health Program (MHP) – a 75scalable, smartphone-based, 8- to 12-week intervention, derived from several evidence- 76based treatments for depression and anxiety, such as mindfulness-based practices [16,17] 77and cognitive-behavioral therapy [18]. Individuals enrolled in the MHP complete weekly 78modules alongside an anonymous peer group (with whom they can interact) and have 79regular contact via messaging with a remote licensed therapist who monitors participant 80progress and offers support. Prior to and during the MHP, participants complete self- 81reported assessments of symptoms of depression and anxiety so that therapists can monitor 82their progress. 83Drawing from such data, we reported that the program is associated with clinically 84significant reductions in symptoms of depression and anxiety that persist for up to 12- 85months post-intervention [19,20]. The objective of this study was to assess whether 86completion of the MHP is also associated with improvements in occupational impairment 87and daily functioning (assessed via the Workplace Activity and Impairment Questionnaire 88[WPAI], included as a program outcome at a later date) in working adults with elevated 89symptoms of depression and/or anxiety. The WPAI, a long-standing and well-validated 90measure, produces four summary scores which in turn describe absenteeism, presenteeism, 91overall workplace impairment, and daily activity impairment. While the first three scores 92pertain to occupational functioning, the latter score pertains to day-to-day activities and 93thus does not require the respondent to be employed. Consistent with previous studies 94[3,7,21], we hypothesized that completion of the MHP would be associated with

5 3 6 95improvements in all four WPAI scores, and that the degree of improvement would be 96correlated with symptom change.

97Methods 98Study Design

99We used an observational, pre-post research design that included a single-arm pre- and 100post-intervention assessment of outcomes. Patient-reported outcomes were measured pre- 101intervention (“baseline”) and at the end of the MHP (“post-intervention”).

102Participants

103The present study included adult patients treated at the Meru Health online clinic (between 104March 2018 and May 2019), a national remote healthcare provider that currently operates 105in the United States and Finland. Participants, a majority of whom were female and from 106Finland, were either i) self-referrals, ii) referred by a healthcare provider, iii) enrolled via an 107employee wellness program, iv) or enrolled via a university health service (see Table 1). For 108inclusion, participants had to provide informed consent via the Meru Health app, own a 109smartphone, have at least mild symptoms of depression and/or anxiety (Patient Health 110Questionnaire-9 item scale [PHQ-9] score > 4 and/or Generalized Anxiety Disorder-7 item 111scale score [GAD-7] > 4 at baseline), and acknowledge/demonstrate the ability to commit to 112a minimum of 20 minutes of practice per day, for 6 days per week, across the intervention 113(as judged by both the participant and their assigned therapist). Exclusion criteria included a 114previous suicide attempt, severe active suicidal ideation with a specific plan, severe self- 115harm, active , or a history of psychosis. 116 117Data were collected as part of the standard MHP. The PHQ-9 and GAD-7 were administered 118before, during, and after the program in the Meru Health app, while the WPAI was 119administered via an online form (hosted by www.typeform.com) external to the app. All 120procedures used were reviewed by Pearl IRB, who granted institutional review board 121exemption for analyses of previously collected and de-identified data, having been 122performed in accordance with the 1964 Helsinki declaration and its later amendments or 123comparable ethical standards. All participants provided informed consent for their 124anonymized data to be used for research purposes prior to participation. 125 126Sample size 127 128Since this study involved analysis of previously collected data, we did not plan to recruit a 129specific sample size. However, a post-hoc power calculation (using G*power 3.1; ANOVA: 130Repeated measures, within-factor [22]) based on a per protocol analysis (n = 64) and the 131effect sizes calculated in our analysis indicated we achieved >90% power for all outcomes 132apart from absenteeism where we achieved only 43% power. 133 134

135

7 4 8 136

137

Per Protocol All Participants Analysis (n = 139) (n = 64)

Age (mean, SD) 36.4 (9.7) 36.9 (9.0)

Gender Female (n, %) 105 (75.5) 53 (82.8) Male (n, %) 34 (24.5) 11 (17.2) Antidepressants Yes (n, %) 53 (38.1) 25 (39.1) No (n, %) 86 (61.9) 39 (60.9) Country Finland (n, %) 110 (79.1) 54 (84.4) US (n, %) 29 (20.9) 10 (15.6) Source Employee 43 (30.9) 25 (39.1) Wellness Program University Health 12 (8.6) 9 (14.1) System Self-referral 74 (53.2) 29 (45.3) Healthcare 10 (7.2) 1 (1.6) referral Intervention 8-week MHP 50 (36.0) 22 (34.4) 8-week MHP 48 (34.5) 20 (31.3) HRV-B 12-week MHP 41 (29.5) 22 (34.4) Baseline PHQ-9 13.0 (5.2) 11.9 (4.6) (mean, SD)

Baseline GAD-7 10.7 (4.3) 10.6 (4.2) (mean, SD)

138GAD-7=Generalized Anxiety Disorder-7 item scale; MHP=Meru Health Program; 139PHQ=Patient Health Questionnaire-9 item scale; SD=standard deviation 140Table 1. Participant demographics and baseline symptom severity. 141

9 5 10 142Intervention 143The MHP has been described in detail previously [19,20]. Briefly, the original 8-week 144intervention includes daily content (typically 10-30 minutes) derived from evidence-based 145practices such as Mindfulness-Based Stress Reduction (MBSR) [17], Cognitive-Behavioral 146Therapy (CBT) [18], and Behavioral Activation Therapy (BAT) [23]. The content includes text, 147video, audio-guided mindfulness exercises, CBT infographics, and journal prompts. The 148intervention also includes anonymous peer support via a group discussion board, and 149regular 1:1 messaging with a remote licensed therapist. 150 151As part of making continual improvements to the program, Meru Health have recently 152piloted two variations of the original intervention. The first involved adding a new, 153promising self-directed component to the 8-week program - heart-rate variability 154biofeedback (HRV-B). HRV-B is a breathing technique that involves maximising heart-rate 155variability (HRV) via real-time visual feedback of one’s own heart-rate trace during training 156[27]. The HRV-B component of the program has been described extensively in a previous 157publication [28]. 158The second variation involved extending the original 8-week MHP into a 12-week version of 159the program that includes additional content regarding the link between food and mood, as 160well as content derived from principles of mindful eating [24], CBT-based sleep therapy [25] 161and sleep hygiene [26]. Since this study used data from real-world Meru Health patients, a 162proportion of participants were enrolled in one of these two program variations (see Table 1631). 164Outcome Measures

165The primary outcome was work function and performance, assessed via the Work 166Productivity and Activity Impairment (WPAI) questionnaire [29]. The WPAI is a six-item 167questionnaire used to measure the number of work hours attended and missed, and 168impairment resulting from a specific health condition (here, symptoms of depression and/or 169anxiety) while working or performing usual daily activities over the past week. The WPAI 170produces four scores expressed as percentages (where higher scores indicate greater 171impairments) that correspond to i) absenteeism, ii) presenteeism, iii) overall work 172impairment, iv) impairments in day-to-day activities. The WPAI has good construct validity 173and test-retest reliability [29].

174Secondary outcomes included two widely used, reliable, and validated measures of 175depression and anxiety with diagnostic sensitivity for severity: the PHQ-9, which was used 176to measure the severity of depression symptoms [30], and the GAD-7, which was used to 177measure the severity of anxiety symptoms [31]. These were included to validate (in a 178working population) our previous finding that the MHP is associated with reductions in 179symptoms of depression and anxiety, and to test an exploratory hypothesis that 180improvements in workplace productivity would be related to symptom improvement.

181Statistical Analyses

182Descriptive statistics were used to summarize program engagement, attrition, and baseline 183participant characteristics. Engagement metrics included ‘total active days’ (defined as any 11 6 12 184day in which a participant completes any active component of the MHP via the app, 185including completing >3 minutes of meditation practice), the total volume of meditation 186practice (in hours), and the total number of days in which participants had contact with their 187therapist across the intervention. Summary engagement metrics are presented separately 188for participants on the 8-week versions and 12-week version of the MHP.

189We performed a per protocol (PP) analysis whereby only participants with complete WPAI 190data at baseline and post-intervention were included. Note, we chose not to use missing 191data techniques as this can produce erroneous findings when the proportion of missing data 192is particularly high. We used linear mixed effects models to investigate the impact of the 193MHP on primary and secondary outcomes, running separate models for the four WPAI 194scores, the PHQ-9, and the GAD-7. In each case, “time” (baseline vs post-intervention) was 195included as a categorical fixed effect, with a separate baseline for each participant (random- 196intercept model). Age, gender, antidepressant status, program version, country, participant 197source, baseline PHQ-9 severity, total meditation hours, and total active days were included 198as covariates. We report the contrast estimate, 95% CI and p-value associated with the main 199effect of time for each. We calculated a Cohen’s d effect size and 95% CI using equations 1, 2002, 15 and 18 from Nakagawa and Cuthill [32].

201We used multivariate logistic regression to investigate whether any demographic or baseline 202characteristics were predictive of the presence/absence of complete WPAI data at post- 203intervention. Further, we used multiple regression to test whether participant demographics 204and baseline PHQ-9 and GAD-7 symptom severity was predictive of baseline levels of 205workplace/activity impairment, for each of the four WPAI scores.

206We also ran exploratory multiple regressions to investigate whether program engagement 207or change in PHQ-9 (or GAD-7) symptoms was predictive of change in the four WPAI scores 208from pre- to post-intervention. In each case, predictor variables included participant 209demographics, program version (dummy coded as two variables), baseline PHQ-9 severity, 210pre-post change in PHQ-9 scores, baseline WPAI score (for the score being examined), the 211total volume of meditation practice in hours (used as an indicator of program engagement), 212and the total days in which participants had contact with their therapist. To reduce 213multicollinearity, we repeated each multiple regression using either PHQ-9 or GAD-7 as an 214indicator of symptomatology / symptom improvement, which produced equivalent results, 215and we thus report PHQ-9 for simplicity. For the daily activity impairment outcome, we 216included data from an additional n=45 participants that were not employed at baseline or 217post-intervention, as this measure does not require the respondent to be working.

218To estimate indirect average worker productivity cost savings resulting from completion of 219the MHP, we multiplied average (unadjusted) change in the WPAI overall work impairment 220score by the latest US annual average income for full-time workers ($46,800) [33].

221Results

222Participant characteristics and engagement

13 7 14 223Out of 139 participants with complete WPAI data at baseline, 12 (8.6%) participants were 224inactive for more than half of the MHP, whilst 33 (23.7%) were inactive on the last week of 225the MHP. Sixty-four (46.7%) out of 137 participants completed all four WPAI scores at post- 226intervention, excluding 2 participants that were no longer working (see Figure 1). However, 227110 (79.1%) out of the original 139 participants completed the PHQ-9 at post-intervention 228(which unlike the WPAI, was collected in-app). Those with higher PHQ-9 symptoms at 229baseline were less likely to complete the WPAI post-intervention (b = 0.09, p = 0.04), but not 230less likely to complete the PHQ-9 post-intervention (b = 0.03, p = 0.54). In addition, females 231were marginally more likely to complete the WPAI post-intervention than males (b = -0.83, p 232= 0.08), though this did not reach statistical significance.

233Considering the 64 employed participants with complete WPAI data at baseline and post- 234intervention, those on the 8-week MHPs (total n = 42) engaged on 42.5 (SD = 12.1) separate 235days, completed a total of 8.3 (SD = 3.9) hours of meditation, and had contact with their 236therapist on 12.9 (SD = 5.9) separate days. Participants on the 12-week MHP (n = 22) 237engaged on 60.7 (SD = 16.4) separate days, completed a total of 14.5 (SD = 6.8) hours of 238meditation, and had contact with their therapist on 19.4 (SD = 7.6) separate days.

239Participants in the PP analysis were on average 36.9 (SD = 9.0) years of age, were 240predominantly female (82.8%), and reported symptoms above the cut-off for moderate 241levels of depression and anxiety at baseline (PHQ-9 and GAD-7 > 10; see Table 1). In 242addition, participants reported high levels of presenteeism (mean = 45.5%, SD = 26.2) and 243daily activity impairment (mean = 52.7%, SD = 23.9), but comparatively low levels of 244absenteeism (mean = 12.8%, SD = 26.6) at baseline (see Figure 2), with 42 (65.6%) of 64 245participants reporting no absenteeism at all. Participants taking antidepressants reported 246higher levels of absenteeism at baseline (b = 16.4, p = 0.01). Moreover, participants with 247higher levels of depression and/or anxiety at baseline reported higher levels of 248presenteeism (b = 1.95, p = 0.003), overall work impairment (b = 1.95, p = 0.01), and daily 249activity impairment (b = 2.07, p < 0.001), but not absenteeism (b = 0.79, p = 0.28), after 250controlling for demographics and use of antidepressants.

15 8 16 251

252 Figure 1. Participant flow through the study for the primary outcome (WPAI).

253Worker Productivity Outcomes (Presenteeism, Workplace and Daily Activity Impairment, 254Absenteeism)

255Figure 1 shows mean (unadjusted) pre- and post-intervention WPAI scores among 256participants reporting employment at baseline and program end (n = 64). Using linear mixed 257effects modelling to adjust for covariates revealed significant end-of-treatment 258improvements in presenteeism (mean decrease = 12.1, 95% CI = [4.17 – 20.0], p = 0.003; d = 2590.56, 95% CI = [0.23 to 0.89]), overall workplace impairment (mean decrease = 13.4, 95% CI 260= [5.62 – 21.2], p < 0.001; d = 0.54, 95% CI = [0.22 to 0.85]), and daily activity impairment 261(mean decrease = 14.8, 95% CI = [9.09 – 20.4], p < 0.001; d = 0.60, 95% CI = [0.34 to 0.85]) 262following completion of the MHP. Absenteeism decreased following the MHP but did not 263reach statistical significance (mean decrease = 4.32, 95% CI = [-1.00 to 9.64], p = 0.11; d = 2640.25, 95% CI = [-0.03 to 0.52]).

265Cost Outcomes

266Based on an average salary of $46800 in the US [33], an adjusted improvement in overall 267workplace productivity of 13.4 percent corresponds to an average annual cost-saving of 268$6271 per employee, and a ~30% relative increase in work productivity.

17 9 18 269

270Figure 2. Mean workplace productivity and activity impairment (WPAI) scores for the 271participants in the PP analysis at baseline (blue) and post-intervention (yellow). Error bars 272represent SEM.

273Mental Health Outcomes

274Consistent with previous reports, the MHP was associated with significant post-intervention 275improvements in both symptoms of depression (mean PHQ-9 decrease = 4.25, 95% CI = 276[2.83 – 5.66], p < 0.001; d = 0.98, 95% CI = [0.61 to 1.35]) and anxiety (mean GAD-7 decrease 277= 3.53, 95% CI = [2.40 – 4.65], p < 0.001; d = 0.98, 95% CI = [0.63 to 1.32]) in participants 278who were employed at baseline and post-intervention. 279Predictors of change in work and activity impairment 280Multiple regression analysis revealed that participants with larger improvements in 281symptoms of PHQ-9 also experienced larger decreases in presenteeism (b = 2.46, p = 0.002), 282overall workplace impairment (b = 2.29, p = 0.01), and daily activity impairment (b = 3.18, p 283< 0.001), but not absenteeism (b = 0.36, p = 0.67; see Table S1 for full regression results). 284This relationship remained robust after adjusting for participant demographics and 285productivity impairment / symptom severity at baseline, and in the case where GAD-7 was 286used in lieu of PHQ-9. The total volume of meditation practice (used as an indicator of 287program engagement) did not influence pre-post change in any of the four WPAI outcomes, 288apart from a trend towards a higher volume of meditation practice predicting greater 19 10 20 289improvement in daily activity impairment (b = 0.82, p = 0.08; see Table S1). Finally, 290participants with higher PHQ-9 severity at baseline were likely to report smaller 291improvements in daily activity impairment (b = -2.31, p < 0.001), overall work impairment (b 292= -1.99, p = 0.06) and presenteeism (b = -1.74, p = 0.08), though this was only marginally 293significant for the latter two outcomes (see Table S1).

294Supplementary Material 295

Absenteeism Presenteeism Overall Impairment Activity Impairment

b SE p b SE P b SE p b SE p

Intercept 0.10 23.6 .99 -8.95 23.7 .71 15.9 25.6 .54 15.2 14.0 .28

Age 0.07 0.40 .86 0.51 0.38 .19 0.23 0.42 .56 -0.22 0.22 .32

Gender -3.55 9.16 .70 -5.99 8.42 .48 -7.34 9.73 .45 -8.68 5.17 .10

Antidepressants -1.16 7.36 .88 6.16 6.58 .35 1.93 7.37 .79 -0.37 3.80 .93

8-week MHP 1.62 9.88 .87 -7.40 9.51 .44 -11.1 10.5 .30 -4.31 6.37 .50

8-week MHP HRV- -4.59 10.2 .66 4.91 9.97 .62 -4.94 10.8 .65 -2.45 6.77 .72 B

Baseline PHQ-9 -0.49 0.98 .62 -1.74 0.96 .08 -1.99 1.05 0.06 -2.31 0.63 <.001

Change in PHQ-9 0.36 0.83 .67 2.46 0.76 .002 2.29 0.87 .01 3.18 0.48 <.001

Baseline WPAI score 0.59 0.15 <.001 0.60 0.13 <.001 0.47 0.13 <.001 0.58 0.10 <.001

Total meditation hours -0.16 0.82 .85 -0.14 0.77 .86 -0.73 0.83 .38 0.82 0.46 .08

Days with therapist contact 0.17 0.60 .78 -0.64 0.53 .23 -0.30 0.61 .62 -0.55 0.30 .07

296Table S1. Multiple regression analysis. Summary results from exploratory multiple 297regression analyses showing unstandardized beta coefficients (b) and their standard error 298(SE) and p-value, for each individual WPAI score. P-values <.05 are shown in bold, and p- 299values <.1 are shown in italics.

300

21 11 22 301Discussion

302The aim of the present study was to evaluate whether the Meru Health Program (MHP), 303previously associated with reductions in symptoms of depression and anxiety, is also 304associated with improvements in workplace productivity and daily functioning in employees 305with elevated symptoms of depression and/or anxiety. Consistent with our hypothesis, we 306found that participants who completed the MHP reported statistically significant 307improvements in presenteeism, overall work impairment, and daily activity impairment, as 308well as smaller (non-significant) reductions in absenteeism. As with previous studies, the 309MHP also reduced symptoms of both depression and anxiety. 310 311Participants reported far greater levels of presenteeism (45.5%) than absenteeism (12.8%) 312at baseline, which is consistent with multiple studies showing that presenteeism is a greater 313problem in individuals with depression and/or anxiety [3,7,34,35]. Indeed, baseline 314presenteeism in this study was higher than norms associated with common health problems 315such as (20%) [36], (13%) [37], rheumatic (20%) [38], and migraine 316(35%) [39], but consistent with other reports of depression- and anxiety-related 317presenteeism [3,6,7,21]. Higher levels of presenteeism may reflect limits on sick days 318available for employees, stigma associated with reporting mental health issues [40], or the 319perception that one's own symptoms are not severe enough to warrant time off work. 320Participants taking antidepressants (and who therefore had already sought help from a 321healthcare professional) were more likely to report absenteeism at baseline, even after 322adjusting for baseline symptom severity. 323 324Participants also reported larger average percent change in presenteeism and daily activity 325impairment than absenteeism following completion of the MHP. This likely reflects the low 326level of absenteeism at baseline, with 65% of participants reporting no absenteeism 327compared to just 5% for presenteeism. Moreover, participants that reported the largest 328improvements in depression or anxiety symptoms also reported larger improvements in 329workplace productivity, overall work impairment, and daily activity impairment. Specifically, 330each additional 1-point decrease in PHQ-9 or GAD-7 score was on average associated with 3312-4% further improvement in workplace productivity, overall work impairment, and activity 332impairment. This is consistent with previous research reporting a somewhat linear 333relationship between the severity of depression or anxiety symptoms (or symptom 334improvement over time) and productivity loss (or gain) [6,7,41]. 335 336Several meta-analyses have evaluated the impact of digital interventions for depression and 337anxiety on symptom reduction [11,12,42], but few studies have included work-related 338outcomes. The medium effect size (d ~ 0.55) reported in the present study for presenteeism 339and overall work impairment is generally consistent with (and in some cases larger than) 340previous studies evaluating online or smartphone-based interventions in working 341populations [43-49], and those reporting on changes in productivity following 342antidepressant treatment [21,50]. In addition, a recent meta-analysis of web-based 343psychological interventions delivered in the workplace reported a more modest effect (d = 3440.25) on workplace functioning than the present study [51], though we note this comparison 345may be confounded by the lack of a control group in our study. The present results are

23 12 24 346encouraging given that work-related outcomes do not always show robust improvement, 347even after the successful reduction of symptoms of depression and/or anxiety [15,52]. 348 349Although the present findings are largely positive, there is still limited evidence regarding 350which treatments are the most effective for employees with depression and anxiety, 351specifically with regards to improving work-related impairments [14,15]. The impact of 352previous digital mental health interventions on workplace outcomes is highly 353heterogeneous [51], and while one study reported increased workplace productivity 354following antidepressant treatment [53], recent systematic reviews report limited or 355inconsistent evidence that antidepressants can reduce work disability in depressed 356employees [54,55]. Comprehensive, smartphone-based interventions such as the MHP may 357provide an alternative treatment for individuals who fail to respond to traditional 358pharmacotherapy, as patients who demonstrate some degree of treatment-resistance are 359more prone to persistent impairment in workplace productivity [49]. Moreover, there is 360growing evidence that incorporating multiple therapeutic approaches into a single 361intervention may provide the most benefit for employees with depression [56]. However, 362given these gaps in knowledge, the authors strongly advocate routine collection of work- 363related outcomes if we are to further understand the occupational effectiveness of different 364mental health interventions. 365 366Our results suggest that on average, completion of the MHP results in a cost-saving of $6271 367per employee each year, based on the average (median) wage across all full-time workers in 368the US. This estimate increases to an average of $9048 for employees in management, 369professional, and related occupations, and up to $16202 for individuals with advanced 370degrees and in the upper earning quartile. This is reasonably consistent with previous 371estimates of cost-savings resulting from antidepressant treatment ($7508) [50], an in- 372person work-focused care program ($6042) [47], a telephone-based depression intervention 373($6048) [57], and an online mindfulness-based workplace intervention ($9360) [58]. This 374suggests that remote digital health interventions such as the MHP may be as effective as 375other (non-digital) occupational interventions, whilst also being widely accessible and 376affordable. However, this needs to be substantiated in larger, controlled trials that include a 377more detailed cost-benefit analysis. 378 379Limitations 380 381This study should be interpreted with a full understanding of its limitation. Our study used 382self-reported data from real-world patients of the Meru Health Clinic, with no comparison 383group, and thus we cannot causally link the improved workplace outcomes to the MHP. 384Moreover, the effect sizes reported here may be an overestimate of the true treatment 385effect (relative to a control), especially considering that a portion of participants with 386symptoms of depression or anxiety will spontaneously recover [59,60]. 387 388Similarly, ~40% of participants were on antidepressants at baseline, so we cannot preclude 389the possibility that a portion of the benefit was caused by antidepressants, and not the 390MHP. However, there was no difference in benefit to workplace outcomes reported by 391participants on antidepressants versus those not on antidepressants. 392

25 13 26 393In addition, we used a per protocol analysis without taking into account participants with 394missing data. This is largely because the proportion of participants with missing WPAI data 395at post-intervention (54% missing) was greater than what is generally considered 396appropriate for missing data techniques. Thus, while our results show a positive association 397between completion of the MHP and workplace outcomes, they should not be interpreted 398as reflecting the overall efficacy of the MHP, and further work is needed in order to evaluate 399the MHP in an intention-to-treat analysis and compared to usual care. One possibility for the 400large portion of missing data is that (unlike the secondary study outcomes) the WPAI was 401collected outside of the intervention app. Indeed, completion of routine in-app outcomes 402was far higher than the WPAI (79.1% completion versus 46%). Further, only 23.7% of 403participants were inactive on the last week of the program, whilst only 8.6% were inactive 404for the majority of the program, suggesting that the missing WPAI data was not a 405consequence of poor program engagement. 406 407Further, our estimate of employer cost savings resulting from the MHP was based on the 408average US national wage, and not employee-specific earnings. While this approach is 409arguably the most conservative, a more accurate estimate could be derived from individual 410earnings. The study also included self-selected participants, the majority of whom were 411female and based in Finland, making it difficult to generalise these findings to wider groups. 412Future studies should strive to include more diverse study samples from different 413recruitment sources and occupational settings. 414 415Finally, since this study used data from real-world patients being treated at Meru Health 416Online Clinic, we included participants who were enrolled in three different variations of the 417MHP. Whilst we attempted to control for this in our analyses, it remains unknown which 418version of the MHP is the most effective or which program components are the most 419important for addressing work-related difficulties. In addition, the frequency of assessing 420depressive and anxiety symptoms (e.g., weekly, biweekly, every three weeks) changed with 421the development of program versions, precluding the reliable testing of any mediation of 422the impact of the program on worker productivity through early improvements in 423depression and anxiety symptoms. Future studies should ensure all participants are 424enrolled in the same program or be powered to test for differences between program 425versions. 426 427Conclusions 428 429Depression and anxiety are extremely common in the working population and result in 430substantially impaired workplace functioning. There is still limited consensus regarding 431whether psychological interventions can improve workplace outcomes, and which 432interventions are likely to be the most effective in the workplace. Preliminary evidence from 433this study suggests that affordable and widely accessible interventions such as the MHP 434have the potential, in addition to reducing symptoms of depression and anxiety, to improve 435work productivity and thus provide cost savings to employers. 436 437 438 439

27 14 28 440References 441

4421. Knudsen AK, Harvey SB, Mykletun A, Overland S. Common mental disorders and long- 443 term sickness absence in a general working population. The Hordaland Health Study. 444 Acta Psychiatr Scand. John Wiley & Sons, Ltd (10.1111); 2013;127: 287–297. 445 doi:10.1111/j.1600-0447.2012.01902.x

4462. Katon W. The impact of depression on workplace functioning and disability costs. Am J 447 Manag Care. 2009;15: S322–7.

4483. Evans-Lacko S, Knapp M. Global patterns of workplace productivity for people with 449 depression: absenteeism and presenteeism costs across eight diverse countries. Soc 450 Psychiatry Psychiatr Epidemiol. Springer Berlin Heidelberg; 2016;51: 1525–1537. 451 doi:10.1007/s00127-016-1278-4

4524. Henderson M, Harvey SB, Overland S, Mykletun A, Hotopf M. Work and common 453 psychiatric disorders. J R Soc Med. SAGE PublicationsSage UK: London, England; 454 2011;104: 198–207. doi:10.1258/jrsm.2011.100231

4555. Lerner D, Adler DA, Chang H, Lapitsky L, Hood MY, Perissinotto C, et al. Unemployment, 456 job retention, and productivity loss among employees with depression. Psychiatr Serv. 457 American Psychiatric Publishing; 2004;55: 1371–1378. doi:10.1176/appi.ps.55.12.1371

4586. Erickson SR, Guthrie S, Vanetten-Lee M, Himle J, Hoffman J, Santos SF, et al. Severity of 459 anxiety and work-related outcomes of patients with anxiety disorders. Depression and 460 Anxiety. John Wiley & Sons, Ltd; 2009;26: 1165–1171. doi:10.1002/da.20624

4617. Beck A, Crain AL, Solberg LI, Unützer J, Glasgow RE, Maciosek MV, et al. Severity of 462 depression and magnitude of productivity loss. Ann Fam Med. American Academy of 463 Family Physicians; 2011;9: 305–311. doi:10.1370/afm.1260

4648. Jain G, Roy A, Harikrishnan V, Yu S, Dabbous O, Lawrence C. Patient-reported 465 depression severity measured by the PHQ-9 and impact on work productivity: results 466 from a survey of full-time employees in the United States. J Occup Environ Med. 467 2013;55: 252–258. doi:10.1097/JOM.0b013e31828349c9

4689. Bakker D, Kazantzis N, Rickwood D, Rickard N. Mental Health Smartphone Apps: Review 469 and Evidence-Based Recommendations for Future Developments. JMIR Ment Health. 470 2016;3: e7. doi:10.2196/mental.4984

47110. Weil TP. Insufficient dollars and qualified personnel to meet United States mental 472 health needs. J Nerv Ment Dis. The Journal of Nervous and Mental Disease; 2015;203: 473 233–240. doi:10.1097/NMD.0000000000000271

47411. Firth J, Torous J, Nicholas J, Carney R, Pratap A, Rosenbaum S, et al. The efficacy of 475 smartphone-based mental health interventions for depressive symptoms: a meta- 476 analysis of randomized controlled trials. World Psychiatry. John Wiley & Sons, Ltd; 477 2017;16: 287–298. doi:10.1002/wps.20472

47812. Firth J, Torous J, Nicholas J, Carney R, Rosenbaum S, Sarris J. Can smartphone mental 479 health interventions reduce symptoms of anxiety? A meta-analysis of randomized

29 15 30 480 controlled trials. Journal of Affective Disorders. Elsevier; 2017;218: 15–22. doi:10.1016/ 481 j.jad.2017.04.046

48213. Carlbring P, Andersson G, Cuijpers P, Riper H, Hedman-Lagerlöf E. Internet-based vs. 483 face-to-face cognitive behavior therapy for psychiatric and somatic disorders: an 484 updated systematic review and meta-analysis. Cogn Behav Ther. Routledge; 2018;47: 485 1–18. doi:10.1080/16506073.2017.1401115

48614. Joyce S, Modini M, Christensen H, Mykletun A, Bryant R, Mitchell PB, et al. Workplace 487 interventions for common mental disorders: a systematic meta-review. Psychological 488 Medicine. Cambridge University Press; 2016;46: 683–697. 489 doi:10.1017/S0033291715002408

49015. Furlan AD, Gnam WH, Carnide N, Irvin E, Amick BC, DeRango K, et al. Systematic review 491 of intervention practices for depression in the workplace. J Occup Rehabil. Springer US; 492 2012;22: 312–321. doi:10.1007/s10926-011-9340-2

49316. Morgan D. Mindfulness-Based Cognitive Therapy for Depression: A New Approach to 494 Preventing Relapse. Psychotherapy Research. Taylor & Francis Group; 2003;13: 123– 495 125. doi:10.1093/ptr/kpg004

49617. Kabat-Zinn J. Wherever You Go, There You Are: Mindfulness meditation for everyday 497 life. Hachette Books; 1994.

49818. Beck AT. Cognitive Therapy and the Emotional Disorders. Penguin; 1979.

49919. Economides M, Ranta K, Nazander A, Hilgert O. One-year outcomes of a therapist- 500 supported, smartphone-based intervention for elevated symptoms of depression and 501 anxiety. PsyArXiv. 2019.

50220. Goldin PR, Lindholm R, Ranta K, Hilgert O, Helteenvuori T, Raevuori A. Feasibility of a 503 Therapist-Supported, Mobile Phone–Delivered Online Intervention for Depression: 504 Longitudinal Observational Study. JMIR Formative Research. JMIR Publications Inc., 505 Toronto, Canada; 2019;3: e11509. doi:10.2196/11509

50621. Hammer-Helmich L, Haro JM, Jönsson B, Tanguy Melac A, Di Nicola S, Chollet J, et al. 507 Functional impairment in patients with major depressive disorder: the 2-year 508 PERFORM study. Neuropsychiatr Dis Treat. Dove Press; 2018;14: 239–249. doi:10.2147/ 509 NDT.S146098

51022. Faul F, Erdfelder E, Lang A-G, Buchner A. G*Power 3: a flexible statistical power analysis 511 program for the social, behavioral, and biomedical sciences. Behav Res Methods. 512 2007;39: 175–191.

51323. Jacobson NS, Martell CR, Dimidjian S. Behavioral Activation Treatment for Depression: 514 Returning to Contextual Roots. Clinical Psychology: Science and Practice. John Wiley & 515 Sons, Ltd (10.1111); 2001;8: 255–270. doi:10.1093/clipsy.8.3.255

51624. Nelson JB. Mindful Eating: The Art of Presence While You Eat. Diabetes Spectr. 517 2017;30: 171–174. doi:10.2337/ds17-0015

51825. Carney CE, Edinger JD, Kuchibhatla M, Lachowski AM, Bogouslavsky O, Krystal AD, et al. 519 Cognitive Behavioral Therapy for Those With Insomnia and Depression: A

31 16 32 520 Randomized Controlled Clinical Trial. Sleep. 2017;40: e1001547. 521 doi:10.1093/sleep/zsx019

52226. Riemann D. Sleep hygiene, insomnia and mental health. J Sleep Res. John Wiley & Sons, 523 Ltd (10.1111); 2018;27: 3–3. doi:10.1111/jsr.12661

52427. Lehrer PM, Gevirtz R. Heart rate variability biofeedback: how and why does it work? 525 Front Psychol. Frontiers; 2014;5: 756. doi:10.3389/fpsyg.2014.00756

52628. Economides M, Lehrer P, Ranta K, Nazander A, Hilgert O, Raevuori A, et al. Feasibility 527 and efficacy of the addition of heart rate variability biofeedback to a remote digital 528 health intervention for depression. PsyArXiv. 2019. 529 doi:https://doi.org/10.31234/osf.io/8wpq7

53029. Reilly MC, Zbrozek AS, Dukes EM. The Validity and Reproducibility of a Work 531 Productivity and Activity Impairment Instrument. PharmacoEconomics. Springer 532 International Publishing; 1993;4: 353–365. doi:10.2165/00019053-199304050-00006

53330. Kroenke K, Spitzer RL, Williams JB. The PHQ-9: validity of a brief depression severity 534 measure. J Gen Intern Med. 2nd ed. Springer; 2001;16: 606–613. doi:10.1046/j.1525- 535 1497.2001.016009606.x

53631. Spitzer RL, Kroenke K, Williams JBW, Löwe B. A brief measure for assessing generalized 537 anxiety disorder: the GAD-7. Arch Intern Med. 2006;166: 1092–1097. 538 doi:10.1001/archinte.166.10.1092

53932. Nakagawa S, Cuthill IC. Effect size, confidence interval and statistical significance: a 540 practical guide for biologists. Biological Reviews. 5 ed. Wiley/Blackwell (10.1111); 541 2007;82: 591–605. doi:10.1111/j.1469-185X.2007.00027.x

54233. Usual Weekly Earnings of Wage And Salary Workers Second Quarter 2019. In: Bureau 543 of Labor Statistics [Internet]. 2019 [cited 24 Jul 2019]. Available: 544 https://www.bls.gov/news.release/pdf/wkyeng.pdf

54534. Stewart WF, Ricci JA, Chee E, Hahn SR, Morganstein D. Cost of lost productive work 546 time among US workers with depression. JAMA. American Medical Association; 547 2003;289: 3135–3144. doi:10.1001/jama.289.23.3135

54835. Sanderson K, Andrews G. Common mental disorders in the workforce: recent findings 549 from descriptive and social epidemiology. Can J Psychiatry. 2006;51: 63–75. 550 doi:10.1177/070674370605100202

55136. Chen H, Blanc PD, Hayden ML, Bleecker ER, Chawla A, Lee JH, et al. Assessing 552 productivity loss and activity impairment in severe or difficult-to-treat asthma. Value 553 Health. 2008;11: 231–239. doi:10.1111/j.1524-4733.2007.00229.x

55437. Rodbard HW, Fox KM, Grandy S, Shield Study Group. Impact of obesity on work 555 productivity and role disability in individuals with and at risk for diabetes mellitus. Am J 556 Health Promot. 2009;23: 353–360. doi:10.4278/ajhp.081010-QUAN-243

55738. Braakman-Jansen LMA, Taal E, Kuper IH, van de Laar MAFJ. Productivity loss due to 558 absenteeism and presenteeism by different instruments in patients with RA and

33 17 34 559 subjects without RA. Rheumatology (Oxford). 2012;51: 354–361. 560 doi:10.1093/rheumatology/ker371

56139. Vo P, Fang J, Bilitou A, Laflamme AK, Gupta S. Patients' perspective on the burden of 562 migraine in Europe: a cross-sectional analysis of survey data in France, Germany, Italy, 563 Spain, and the United Kingdom. J Headache Pain. BioMed Central; 2018;19: 82. 564 doi:10.1186/s10194-018-0907-6

56540. Haslam C, Atkinson S, Brown SS, Haslam RA. Anxiety and depression in the workplace: 566 effects on the individual and organisation (a focus group investigation). Journal of 567 Affective Disorders. 2005;88: 209–215. doi:10.1016/j.jad.2005.07.009

56841. Beck A, Crain LA, Solberg LI, Unützer J, Maciosek MV, Whitebird RR, et al. The effect of 569 depression treatment on work productivity. Am J Manag Care. NIH Public Access; 570 2014;20: e294–301.

57142. Stratton E, Lampit A, Choi I, Calvo RA, Harvey SB, Glozier N. Effectiveness of eHealth 572 interventions for reducing mental health conditions in employees: A systematic review 573 and meta-analysis. Reed P, editor. PLoS ONE. Public Library of Science; 2017;12: 574 e0189904. doi:10.1371/journal.pone.0189904

57543. Birney AJ, Gunn R, Russell JK, Ary DV. MoodHacker Mobile Web App With Email for 576 Adults to Self-Manage Mild-to-Moderate Depression: Randomized Controlled Trial. 577 JMIR Mhealth Uhealth. JMIR Publications Inc., Toronto, Canada; 2016;4: e8. 578 doi:10.2196/mhealth.4231

57944. Geraedts AS, Kleiboer AM, Wiezer NM, van Mechelen W, Cuijpers P. Short-term effects 580 of a web-based guided self-help intervention for employees with depressive 581 symptoms: randomized controlled trial. J Med Internet Res. JMIR Publications Inc., 582 Toronto, Canada; 2014;16: e121. doi:10.2196/jmir.3185

58345. DellaCrosse M, Mahan K, Hull TD. The Effect of Messaging Therapy for Depression and 584 Anxiety on Employee Productivity. J technol behav sci. Springer International 585 Publishing; 2018;4: 1–5. doi:10.1007/s41347-018-0064-4

58646. Phillips R, Schneider J, Molosankwe I, Leese M, Foroushani PS, Grime P, et al. 587 Randomized controlled trial of computerized cognitive behavioural therapy for 588 depressive symptoms: effectiveness and costs of a workplace intervention. 589 Psychological Medicine. Cambridge University Press; 2014;44: 741–752. 590 doi:10.1017/S0033291713001323

59147. Lerner D, Adler D, Hermann RC, Chang H, Ludman EJ, Greenhill A, et al. Impact of a 592 work-focused intervention on the productivity and symptoms of employees with 593 depression. J Occup Environ Med. 2012;54: 128–135. 594 doi:10.1097/JOM.0b013e31824409d8

59548. Deady M, Johnston D, Milne D, Glozier N, Peters D, Calvo R, et al. Preliminary 596 Effectiveness of a Smartphone App to Reduce Depressive Symptoms in the Workplace: 597 Feasibility and Acceptability Study. JMIR Mhealth Uhealth. JMIR Publications Inc., 598 Toronto, Canada; 2018;6: e11661. doi:10.2196/11661

59949. Trivedi MH, Morris DW, Wisniewski SR, Lesser I, Nierenberg AA, Daly E, et al. Increase 600 in work productivity of depressed individuals with improvement in depressive 35 18 36 601 symptom severity. Am J Psychiatry. 2013;170: 633–641. 602 doi:10.1176/appi.ajp.2012.12020250

60350. Woo J-M, Kim W, Hwang T-Y, Frick KD, Choi BH, Seo Y-J, et al. Impact of depression on 604 work productivity and its improvement after outpatient treatment with 605 antidepressants. Value Health. 2011;14: 475–482. doi:10.1016/j.jval.2010.11.006

60651. Carolan S, Harris PR, Cavanagh K. Improving Employee Well-Being and Effectiveness: 607 Systematic Review and Meta-Analysis of Web-Based Psychological Interventions 608 Delivered in the Workplace. J Med Internet Res. 2017;19: e271. doi:10.2196/jmir.7583

60952. Timbie JW, Horvitz-Lennon M, Frank RG, Normand S-LT. A meta-analysis of labor supply 610 effects of interventions for major depressive disorder. Psychiatr Serv. American 611 Psychiatric Publishing; 2006;57: 212–218. doi:10.1176/appi.ps.57.2.212

61253. Aikens JE, Kroenke K, Nease DE, Klinkman MS, Sen A. Trajectories of improvement for 613 six depression-related outcomes. General Hospital Psychiatry. 2008;30: 26–31. 614 doi:10.1016/j.genhosppsych.2007.10.003

61554. Nieuwenhuijsen K, Bültmann U, Neumeyer-Gromen A, Verhoeven AC, Verbeek JHAM, 616 van der Feltz-Cornelis CM. Interventions to improve occupational health in depressed 617 people. Nieuwenhuijsen K, editor. Cochrane Database Syst Rev. Chichester, UK: John 618 Wiley & Sons, Ltd; 2008;23: CD006237. doi:10.1002/14651858.CD006237.pub2

61955. Nieuwenhuijsen K, Faber B, Verbeek JH, Neumeyer-Gromen A, Hees HL, Verhoeven AC, 620 et al. Interventions to improve return to work in depressed people. Cochrane Work 621 Group, editor. Cochrane Database Syst Rev. John Wiley & Sons, Ltd; 2014;23: 622 CD006237. doi:10.1002/14651858.CD006237.pub3

62356. Wan Mohd Yunus WMA, Musiat P, Brown JSL. Systematic review of universal and 624 targeted workplace interventions for depression. Occup Environ Med. BMJ Publishing 625 Group Ltd; 2018;75: 66–75. doi:10.1136/oemed-2017-104532

62657. Lerner D, Adler DA, Rogers WH, Chang H, Greenhill A, Cymerman E, et al. A randomized 627 clinical trial of a telephone depression intervention to reduce employee presenteeism 628 and absenteeism. Psychiatr Serv. American Psychiatric AssociationArlington, VA; 629 2015;66: 570–577. doi:10.1176/appi.ps.201400350

63058. Aikens KA, Astin J, Pelletier KR, Levanovich K, Baase CM, Park YY, et al. Mindfulness 631 goes to work: impact of an online workplace intervention. J Occup Environ Med. 632 2014;56: 721–731. doi:10.1097/JOM.0000000000000209

63359. Whiteford HA, Harris MG, McKeon G, Baxter A, Pennell C, Barendregt JJ, et al. 634 Estimating remission from untreated major depression: a systematic review and meta- 635 analysis. Psychological Medicine. Cambridge University Press; 2013;43: 1569–1585. 636 doi:10.1017/S0033291712001717

63760. Yonkers KA, Bruce SE, Dyck IR, Keller MB. Chronicity, relapse, and illness--course of 638 panic disorder, social phobia, and generalized anxiety disorder: findings in men and 639 women from 8 years of follow-up. Depression and Anxiety. John Wiley & Sons, Ltd; 640 2003;17: 173–179. doi:10.1002/da.10106

37 19 38