Anxiety and depressive disorders affect one in three people in their lifetime,[1] are commonly comorbid with a range of diagnoses,[2] and impose multibillion-dollar cost burdens on national healthcare economies.[3] Despite their impacts, these disorders remain under-treated, largely because of behavioral healthcare (BH) shortages. In the United States, 75% of counties are so-called “mental health deserts,” where BH professionals are critically understaffed; 50% have no BH professionals whatsoever.[4] Accordingly, anxiety and depressive disorders are commonly treated in ambulatory settings, where medical staff may lack robust BH expertise.[5] Innovative approaches are needed to improve this dynamic.
Digital behavioral health (dBH) platforms promise to improve BH access and outcomes by extending tailored treatment of anxiety and depressive disorders beyond the clinic. With US internet penetration at 95%, dBH platforms are well-poised to deliver computerized equivalents of highly effective treatment approaches, such as cognitive behavioral therapy (CBT).[6] Here, we describe the findings of a retrospective database study to evaluate the suitability of dBH platforms to deliver tailored anxiety- and depression-focused digital CBT (dCBT) interventions.
This study was conducted in accordance with the principles of the Declaration of Helsinki and was exempted from oversight and waived for informed consent following a review by Advarra’s Center for Institutional Review Board (IRB) Intelligence. The IRB determined that the study constituted secondary research as defined in 45 CFR 46.104(d)(4) and posed minimal risk to participants, who were irreversibly anonymized and who consented to such research per NeuroFlow’s terms of service.
We hypothesized that greater engagement with dCBT content would predict significant reductions in anxious and depressive symptom severity. To test this hypothesis, we retrospectively queried preintervention Generalized Anxiety Disorder-7 item (GAD-7)[7] scores for an anxiety cohort (N = 142; 101 women [W], 38 men [M], 3 other; mean [SD] age = 48.35 [12.21] years) and preintervention Patient Health Questionaire-9 (PHQ-9)[7] scores for a depression cohort (N = 104; 70 W, 32 M, 2 other; mean [SD] age = 48.54 [10.72] years). Cohorts comprised adult members covered by Magellan Behavioral Health (MBH) services. All covered members had access to a self-service dBH platform that began with initial GAD-7 and PHQ-9 assessments. Members who scored ≥ 5 (i.e., those with mild or greater symptom severity) on their initial GAD-7 and/or PHQ-9 assessment automatically received recommended dCBT interventions.
The start date for each study participant was the date of the initial assessment score that triggered the dCBT content recommendation. We retrospectively queried each participant’s activity on the dBH platform, including dCBT engagement and assessment scores, for 180 days from the start date. Data collection occurred between Jan 8, 2022, and Mar 8, 2023. All adults covered by MBH who triggered a dCBT content recommendation for anxiety or depression within those dates were included in our analysis. Of note, because dBH platforms rely on design features, behavioral economics, and gamification principles to drive engagement, our retrospective design was well-suited to capture “real-world” treatment-outcome relationships: Absent study-imposed pressure to engage (i.e., from investigators), participants were free to complete as much or as little dCBT content as they preferred. The dCBT interventions were designed using MBH’s CBT principles. The anxiety dCBT intervention, FearFighter, consists of nine 45-minute sessions designed to build skills that help participants manage generalized anxiety, panic attacks, and phobias. The depression intervention, MoodCalmer, consists of four 30-minute sessions designed to help participants recognize and manage unhelpful automatic thoughts. All dCBT interventions were previously validated[8,9] and administered via NeuroFlow, a HIPAA–compliant web- and mobile-based dBH platform that provides remote assessment capabilities, features a library of self-directed BH content and exercises, and leverages gamification principles to promote user engagement.[10,11]
Our first analysis evaluated pre–post GAD-7 and PHQ-9 scores via paired sample t-tests and effect-size calculations (Cohen’s d). We observed significant reductions in mean pre-post GAD-7 (pre = 12.15, post = 9.98, t = 4.65, p < 0.001, d = 0.46) and PHQ-9 scores (pre = 10.5, post = 8.74, t = 5.22, p < 0.001, d = 0.45) (Table 1). As hypothesized, each cohort saw reduced symptom severity during the study, irrespective of more nuanced aspects of engagement. The medium effect sizes observed suggest modest clinical significance.
Next, to evaluate engagement-outcome relationships, we defined subcohorts of participants who completed 75% or more of the recommended dCBT content and those who completed less than 75%. We calculated paired samples t-statistics and effect sizes for pre–post results within each subcohort, as well as independent sample t-statistics to evaluate changes in pre–post assessment scores between subcohorts. In the anxiety cohort, highly engaged participants (n = 25) saw a 41% reduction in GAD-7 scores (pre = 11.2, post = 6.6, t = 5.09, p < 0.001, d = 1.15) compared with less-engaged peers’ (n = 117) 11% reduction (pre = 10.39, post = 9.23, t = 2.92, p < 0.01, d = 0.31) (Table 1). The between-groups difference in pre–post deltas was significant (t = 3.59, p < 0.001). In the depression cohort, highly engaged participants (n = 49) saw a 24% reduction in PHQ-9 scores (pre = 11.61, post = 8.82, t = 4.35, p < 0.001, d = 0.56) compared with peers’ (n = 55) 13% reduction (pre = 12.64, post = 11.02, t = 3.03, p < 0.01, d = 0.36) (Table 1). The between-groups difference was not significant, although the trend was in the expected direction (t = 1.42, p = 0.16). These findings suggest that individuals who engage with more of their recommended dCBT content experience enhanced clinical outcomes. This was particularly evident with anxiety-focused dCBT interventions: the large effect size associated with high engagement indicated strong clinical significance.
Finally, we tested the hypothesis that initial symptom severity would predict enhanced outcomes, such that higher-acuity individuals would see the greatest benefits. We first established each participant’s initial categorical severity (e.g., mild, moderate, severe, etc.) using established thresholds for the GAD-7 and PHQ-9. We then computed F statistics via one-way ANOVA to examine relationships between preintervention severity and outcomes. Initial severity was a significant predictor of outcomes for the anxiety cohort, where severely anxious participants saw the greatest benefits (F = 9.88, p < 0.001). In the depression cohort, a nonsignificant trend emerged in the same direction (F = 1.99, p = 0.12). These results suggest a promising role for dCBT interventions as an adjunctive therapy for individuals who experience severe anxiety.
Our findings (Table 1) highlight the potential for dBH-administered dCBT content to augment existing treatments for anxiety and depression. Our cohorts saw significant symptom reductions in the 180 days after the first engagement with dCBT content; this aligns with findings from previous dCBT randomized controlled trials (RCTs).[12] In our study, the anxiety cohort saw the most pronounced relationships, with high engagement and initial severity predicting enhanced outcomes. While this suggests the possibility that dCBT interventions have greater potency in the treatment of anxiety versus depression, well-powered RCTs will be needed to test causal hypotheses. Furthermore, our lack of a dBH-engaged, non-dCBT control group limits our correlational inference and should be addressed in future studies. While our study focused on engagement metrics for dCBT interventions, future research should compare other delivery methods, such as self-guided modules with therapist support and SMS-augmented interventions. These alternatives may enhance accessibility and personalization, potentially improving patient outcomes. Nevertheless, our findings underscore the potential for dBH tools to extend, standardize, and complement clinicians’ work beyond the practice. Such approaches hold great promise in expanding behavioral healthcare access by leveraging technology to meet people where they are.
.png?versionId=20811)
