All adolescents (ages 13-17 years) who met the following inclusion criteria were eligible for inclusion in the study (N=689): (1) started mental health care (first synchronous event) with Bend Health Inc between January 1, 2023, and October 1, 2023 (9 months); (2) had at least 1 synchronous session with a Bend Health Inc coach or therapist during the study time frame; and (3) completed the assessment of eating behaviors before the start of care (baseline).

Bend Health Inc is a DMHI that provides behavioral care for adolescents (aged 13-17 years), using a whole-family approach (ie, caregivers are closely involved in care), via a web-based platform. Bend Health Inc’s behavioral care has been described elsewhere [24,25]. Members enroll via referral from a health care provider, or they use insurance, employer benefits, or self-pay. Members are assigned a BCM, who conducts an initial evaluation of the member’s mental health concerns and circumstances, and they continue to monitor the member’s care while they are enrolled in the program. BCMs assign members a coach, and also a therapist in some cases, based on their mental health symptoms (eg, type and acuity), goals for treatment, and insurance coverage; therapists tend to be assigned only to members with more severe symptoms and conditions, whereas nearly all members are assigned a coach. Members with higher symptom acuity or a psychiatric referral may also be assigned a prescribing psychiatric practitioner (eg, a medical doctor or psychiatric nurse practitioner). In synchronous video-based coaching and therapy sessions, Bend Health Inc’s practitioners guide members and their caregivers through structured care programs. These care programs are designed to deliver evidence-based tools and techniques that target common mental and behavioral health issues such as anxiety, depression, and body image. Most care programs are designed to be completed in approximately 3 months, and some care programs (eg, the body image program) are designed to be completed in a shorter amount of time. Once a month, caregivers and adolescent members are asked to complete validated web-based assessments of the adolescent’s mental health symptoms (see Measures section). Caregivers are required to be in the same general location for their adolescents’ synchronous sessions with a coach or therapist for safety purposes.

Demographic and health information of adolescent participants is gathered during enrollment with Bend Health Inc. Caregivers respond to basic demographic questions, providing their adolescent member’s date of birth, sex at birth (male, female, or others), gender identity (male, female, transgender, nonbinary, or others), and race or ethnicity. Details on the race or ethnicity response options are included in Multimedia Appendix 1.

In addition, at enrollment, caregivers and adolescent members complete a series of symptom screening questions and validated assessments to identify common mental and behavioral health concerns. To assess eating behaviors, all adolescent members complete the SCOFF (Cronbach α=.48) [27,28]. The SCOFF is a validated questionnaire, in which the member responds “yes” or “no” to 5 items about disordered eating behaviors that they might have. The 5 items, which each correspond with a letter of the SCOFF name, are as follows: (1) Do you make yourself s ick because you feel uncomfortably full? (2) Do you worry that you have lost c ontrol over how much you eat? (3) Have you recently lost more than o ne stone (14 lb) in a 3-month period? (4) Do you believe yourself to be f at when others say you are too thin? and (5) Would you say that f ood dominates your life?

To identify members with elevated anxiety and depressive symptoms, caregivers of adolescent members respond to 5 screening questions drawn from the Diagnostic and Statistical Manual of Mental Disorder, Fifth Edition, Text Revision (DSM-5-TR) Cross-Cutting Symptom Measure [29]. For anxiety, the caregivers and adolescents together respond to the following questions: Over the last 2 weeks, how often have you been bothered by any of the following problems? (1) Feeling nervous, anxious, or on edge? and (2) Not being able to stop or control worrying? For depressive symptoms, caregivers and teens responded to the items: Over the last 2 weeks, how often have you been bothered by any of the following problems? (1) Had less fun doing things than you used to? and (2) Felt sad or depressed for several hours? Best-fit responses are selected from a 4-item Likert-type scale, with responses ranging from “not at all” (item score=0) to “nearly every day” (item score=3). Screener scores are calculated by aggregating all item scores for each screener.

Members who have an anxiety screener score of 2 or more are prompted to complete the Generalized Anxiety Disorder-7 (GAD-7) assessment (Cronbach α=.91) [30,31], and members who have a depressive screener score of 2 or more are prompted to take the Patient Health Questionnaire-9 for Adolescents (PHQ-9A; Cronbach α=.85) [32,33]. The GAD-7 has 7 questions regarding symptoms of anxiety in the prior 2 weeks with the same Likert-type scale as used in the screener questions (“not at all” to “nearly every day”). The PHQ-9A is a modified version of the PHQ-9 for adolescents aged 11-17 years. The original measure has 9 questions, but we omit a question regarding suicide and self-harm (ie, the PHQ-9A here includes 8 questions). The PHQ-9A asks adolescents about depressive symptoms in the prior week using the same Likert-type scale as in the GAD-7 and screener questions. Adolescents are asked to report their own symptoms for both GAD-7 and PHQ-9A.

At enrollment with Bend Health Inc, all study participants provided informed consent for primary data collection (required for participation in care) and use of their data in further analyses. Given the retrospective observational nature of the study, participants were not compensated for their participation in the study. Procedures for this study were approved by the Biomedical Research Alliance of New York (Study 23-12-034-1374; approved on June 5, 2023). To ensure the privacy and confidentiality of the human participants in this study, all data (eg, from electronic health records) were deidentified prior to analysis.

The distribution of baseline SCOFF scores is reported in Table 1. Ultimately, 75.5% (n=520) had nonelevated SCOFF scores, and 24.5% (n=169) had elevated SCOFF scores. For members with nonelevated SCOFF scores, 13.7% (n=71) of members responded “yes” to the item about control (item 2), and 9.2% (n=48) responded “yes” to the item about believing you are fat (item 4; Table 2). For members with elevated SCOFF scores, the most commonly reported items were loss of control (item 2: n=151, 89.3%), believing you are fat (item 4: n=113, 66.9%), and food dominating life (item 5: n=92, 54.4%).

Compared to the nonelevated SCOFF score group, the elevated SCOFF score group was more predominantly female (χ²₁=24.2; P<.001) and also had a higher rate of diagnoses with depressive disorders (χ²₁=9.5; P=.005; Table 3). Rates of elevated anxiety and depressive symptoms were higher for members with elevated SCOFF scores than members with nonelevated SCOFF scores (anxiety: χ²₁=31.9; P<.001 and depression: χ²₁=63.2; P<.001). Age, gender-sex conformity, race or ethnicity, and rates of anxiety disorder diagnoses did not differ between groups.

Members were in care for a median of 2.60 (IQR 1.27-4.23) months. In terms of participation in behavioral care, 98.8% (n=681) of all members were in coaching, and 39.8% (n=274) of all members were in therapy. Rates of participation in coaching and therapy, as well as months in care, did not differ between groups. Approximately 1 in 2 members participated in the anxiety care program, and this was similar between groups. However, participation in the depression and body image care programs was higher for members with elevated SCOFF scores versus members with nonelevated SCOFF scores (depression: χ²₁=15.0; P<.001).

SCOFF assessment counts during the study time frame are reported for each group in Table 4. For all members with a baseline and postcare assessment (ie, 2 or more total assessments), baseline assessments were completed at a median of 0.27 (IQR 0.5-0.13) months before the start of care, and the last assessments were completed at a median of 2.23 (IQR 1.1-3.6) months after the start of care. The timing of baseline and last assessments did not differ between groups (baseline: z=0.70; P=.70 and last assessment: z=–0.38; P=.70).

For members in the nonelevated SCOFF score group, 16.4% (n=56) had a decrease in SCOFF score from baseline to their last assessment, 13.7% (n=47) had an increase in SCOFF score, and 69.9% (n=239) had no change (Table 5). For members in the elevated SCOFF score group, on the other hand, 61.4% (n=70) had a decrease in SCOFF score from baseline to their last assessment, 16.7% (n=19) had an increase, and 21.1% (n=25) had no change. While SCOFF scores remained stable for the nonelevated SCOFF score group (z=–0.47; P=.70), median SCOFF scores decreased from 2 (IQR 2-3) at baseline to 1 (IQR 0-3) at the last assessment for the elevated SCOFF score group (z=–6.39; P<.001). Individual item responses at the last assessment for members who responded “yes” to the item at baseline are reported in Table 6.

In the linear mixed effects model of SCOFF score over time in care for members with elevated SCOFF scores, the following predictors were included in the final model: elevated anxiety symptoms at baseline (χ²₁=7.6; P=.006), elevated depressive symptoms at baseline (χ²₁=8.6; P=.003), and participation in therapy (χ²₁=4.4; P=.03). SCOFF score decreased by 0.26 points for each month in care (F_1,233=72.82; P<.001; Figure 1). Members with elevated anxiety symptoms at baseline had SCOFF scores 0.40 points higher than those with nonelevated anxiety symptoms (F_1,101=9.87; P=.005). Similarly, members with elevated depressive symptoms at baseline had SCOFF scores 0.43 points higher than those with nonelevated depressive symptoms (F_1,101=5.40; P=.044). Participation in therapy did not relate to SCOFF score (F_1,101=2.87; P=.16).

Using retrospective analyses of data collected from adolescents participating in care with Bend Health Inc, the purpose of this study was to determine whether mental health care with a collaborative care DMHI is associated with a reduction in disordered eating behaviors while also accounting for demographic and mental health symptom covariates. Disordered eating scores decreased throughout care with the DMHI for 61.4% (n=70) of adolescents with elevated disordered eating scores at baseline, and longer participation was associated with greater reductions in scores. Elevated disordered eating behaviors at baseline were associated with female sex and elevated mental health symptoms. This study provides preliminary evidence that mental health care with a DMHI may positively affect disordered eating behaviors.

Over the course of care with the DMHI, 61.4% (n=70) of adolescents with disordered eating behaviors exhibited fewer problematic eating behaviors throughout participation. Further, each additional month in care was associated with larger improvements in disordered eating behaviors, even while controlling for elevated mental health symptoms. These findings, which suggest the effectiveness of a pediatric DMHI in mitigating symptoms of disordered eating, are especially timely, as many youths with EDs and disordered eating symptoms are unable to access adequate care via in-person modalities—a lingering effect of the COVID-19 pandemic and current mental health crisis in youths [34,35]. Moreover, these results suggest that collaborative care DMHIs may mitigate disordered eating behaviors in a relatively short time frame, with the change in disordered eating behaviors assessed within this study sample after just 2.23 (IQR 1.1-3.6) months in care. Given the complex and chronic nature of disordered eating behaviors and EDs, it is understandable that health care professionals prefer team-based, integrated behavioral care (eg, the collaborative care model used by the DMHI in this study) for the identification and treatment of EDs [36-39].

Compared to those without disordered eating behaviors, adolescent members with disordered eating behaviors were more likely to be female than male. Extant literature suggests a similar trend, namely, that females tend to report more disordered eating behaviors and higher rates of ED diagnosis than their male peers [40,41]. However, this does not mean that disordered eating is not a problem among males. In this study, 19.5% (n=33) of adolescents flagged with problematic eating behaviors were male. Disordered eating behaviors may go underreported and unrecognized in males due to stigma [42], and ED diagnostic criteria may not accurately capture disordered eating behaviors that are more prevalent in males than females, such as preoccupation with gaining muscle mass and fear of losing weight [43,44]. Mental health providers and practitioners in both traditional and web-based modalities should continue to screen for disordered eating behaviors among all clients, regardless of sex and gender, while paying particular attention to muscle dysmorphia and excessive exercise [41].

We also found that adolescents with disordered eating behaviors also had higher rates of elevated anxiety and depressive symptoms than their peers who were not flagged with disordered eating behaviors. Moreover, having elevated anxiety and depressive symptoms at baseline was a significant predictor of more severe disordered eating behaviors among youths flagged with disordered eating behaviors. In children, adolescents, and adults, internalizing disorders such as anxiety and depression frequently co-occur with disordered eating behaviors and EDs [45-47]. This study sample exhibited higher rates of anxiety and depressive symptom comorbidity than previously reported. This is to be expected, given that all children and adolescents in this study were treatment-seeking, whereas previous estimates have included both treatment- and nontreatment-seeking individuals [45,46]. Notably, 95.9% (n=162) of members flagged with disordered eating behaviors participated in care programs other than the body image care program (namely, the anxiety and depression programs), and this group exhibited improvements in disordered eating behaviors, nonetheless. This finding adds to the body of literature suggesting that the overlap between disordered eating behaviors and internalizing problems such as anxiety and depression may have similar underlying constructs and thus may warrant similar treatment [45,48]. Further research is necessary to study the effectiveness of DMHIs in directly addressing disordered eating behaviors as well as treating comorbid EDs and internalizing problems among young people.

While our findings provide compelling evidence that participation in a collaborative care DMHI may be associated with improvements in disordered eating behaviors, there are several limitations. The SCOFF assessment has attracted criticism for lack of sensitivity and accuracy and other questionnaires. The Eating Attitudes Test [49], for instance, may indeed allow for more fine-grained assessment and diagnosis of eating behaviors. However, the SCOFF has been used and validated extensively as a screener and assessment of common EDs among adolescents [27,50,51]. Moreover, the SCOFF has a high correlation with other validated disordered eating surveys [52] and has also shown acceptable convergent validity when compared to clinical interviews [53].

It should be noted that the SCOFF questionnaire includes an item that may not be appropriate for adolescents. Adolescence is a time of puberty and rapid growth, and therefore, losing 14 pounds may indicate extreme weight loss and a significant health issue (more so than the other SCOFF items). Further, if an adolescent responds “yes” to this item at one time point, it is unlikely that they will respond “yes” in the next several months. Therefore, we investigated whether our findings were driven or altered by the inclusion of this potentially problematic item. We found that few participants in the study sample responded “yes” to this item (n=44, 7.3% at baseline), and follow-up analyses revealed that our longitudinal results did not substantively change when this item was removed from the calculation of SCOFF scores (Multimedia Appendix 2). Thus, the inclusion of this item did not confound the primary findings reported here.

Although we evaluated cross-sectional associations between improvements in disordered eating behaviors and internalizing problems at baseline, future research should explore longitudinal and bidirectional associations between disordered eating symptoms and changes in mental health symptoms among youths involved in collaborative care DMHIs. Extant research regarding causal associations between disordered eating behaviors and other psychiatric symptoms among those engaged in treatment has yielded heterogeneous results [54-56], but no relevant studies to date have been conducted among youths involved in a collaborative care DMHI.

Young people today are exhibiting increasing rates of disordered eating behaviors and EDs alongside other mental and behavioral problems such as anxiety and depression. Collaborative care DMHIs have the capacity to mitigate the growing mental health crisis by providing holistic and evidence-based care that is more accessible and scalable than traditional modalities. The findings from this study suggest that participation with a collaborative care DMHI such as Bend Health Inc may be beneficial in the reduction of disordered eating symptoms in adolescents. Future studies, particularly those bolstered by improved measurement of eating behaviors over time and with a larger and more diverse cohort of youths, are paramount to establishing the effectiveness of collaborative care DMHIs as an evidence-based provider of care for those with problematic eating behaviors and EDs.