BackgroundCognitive Behavioral Therapy (CBT) for depression is efficacious, but effectiveness is limited when implemented in low-income settings due to engagement difficulties including nonadherence with skill-building homework and early discontinuation of treatment. Automated messaging can be used in clinical settings to increase dosage of depression treatment and encourage sustained engagement with psychotherapy.ObjectivesThe aim of this study was to test whether a text messaging adjunct (mood monitoring text messages, treatment-related text messages, and a clinician dashboard to display patient data) increases engagement and improves clinical outcomes in a group CBT treatment for depression. Specifically, we aim to assess whether the text messaging adjunct led to an increase in group therapy sessions attended, an increase in duration of therapy attended, and reductions in Patient Health Questionnaire-9 item (PHQ-9) symptoms compared with the control condition of standard group CBT in a sample of low-income Spanish speaking Latino patients.MethodsPatients in an outpatient behavioral health clinic were assigned to standard group CBT for depression (control condition; n=40) or the same treatment with the addition of a text messaging adjunct (n=45). The adjunct consisted of a daily mood monitoring message, a daily message reiterating the theme of that week’s content, and medication and appointment reminders. Mood data and qualitative responses were sent to a Web-based platform (HealthySMS) for review by the therapist and displayed in session as a tool for teaching CBT skills.ResultsIntent-to-treat analyses on therapy attendance during 16 sessions of weekly therapy found that patients assigned to the text messaging adjunct stayed in therapy significantly longer (median of 13.5 weeks before dropping out) than patients assigned to the control condition (median of 3 weeks before dropping out; Wilcoxon-Mann-Whitney z=−2.21, P=.03). Patients assigned to the text messaging adjunct also generally attended more sessions (median=6 sessions) during this period than patients assigned to the control condition (median =2.5 sessions), but the effect was not significant (Wilcoxon-Mann-Whitney z=−1.65, P=.10). Both patients assigned to the text messaging adjunct (B=−.29, 95% CI −0.38 to −0.19, z=−5.80, P<.001) and patients assigned to the control conditions (B=−.20, 95% CI −0.32 to −0.07, z=−3.12, P=.002) experienced significant decreases in depressive symptom severity over the course of treatment; however, the conditions did not significantly differ in their degree of symptom reduction.ConclusionsThis study provides support for automated text messaging as a tool to sustain engagement in CBT for depression over time. There were no differences in depression outcomes between conditions, but this may be influenced by low follow-up rates of patients who dropped out of treatment.

Emotional support is a crucial ability for many conversation scenarios, including social interactions, mental health support, and customer service chats. Following reasonable procedures and using various support skills can help to effectively provide support. However, due to the lack of a well-designed task and corpora of effective emotional support conversations, research on building emotional support into dialog systems remains untouched. In this paper, we define the Emotional Support Conversation (ESC) task and propose an ESC Framework, which is grounded on the Helping Skills Theory (Hill, 2009). We construct an Emotion Support Conversation dataset (ESConv) with rich annotation (especially support strategy) in a help-seeker and supporter mode. To ensure a corpus of high-quality conversations that provide examples of effective emotional support, we take extensive effort to design training tutorials for supporters and several mechanisms for quality control during data collection. Finally, we evaluate state-of-the-art dialog models with respect to the ability to provide emotional support. Our results show the importance of support strategies in providing effective emotional support and the utility of ES-Conv in training more emotional support systems 1 . Strategies Stages Examples Lexical Features Question Can you talk more about your feelings at that time? do you (15.0), are you (13.8), how (13.7), what (12.3), do (11.5) Restatement or ParaphrasingIt sounds that you feel like everyone is ignoring you. Is it correct?is that (8.2), so you (8.2), it sounds (7.1), correct (7.1), so (6.6) Reflection of FeelingsI understand how anxious you are. can tell (7.4), understand how (5.8), are feeling (5.1), tell (5.1), understand (4.9)Self-disclosure I feel the same way! I also don't know what to say to strangers. my (15.3), was (10.5), me (10.2), had (9.7), myself (7.8) Affirmation and ReassuranceYou've done your best and I believe you will get it! its (5.7), thats (5.6), will (5.4), through this (5.1), you will (4.7) Providing SuggestionsDeep breaths can help people calm down. Could you try to take a few deep breaths? maybe (7.3), if (6.5), have you (6.4), talk to (5.8), suggest (5.8) InformationApparently, lots of research has found that getting enough sleep before an exam can help students perform better.there are (4.4), will (3.8), available (3.7), seen (3.3), possible (3.3) OthersI am glad to help you! welcome (9.6), hope (9.6), glad (7.3), thank (7.0), hope you (6.9)

A new trend in medicine is the use of algorithms to analyze big datasets, e.g. using everything your phone measures about you for diagnostics or monitoring. However, these algorithms are commonly compared against weak baselines, which may contribute to excessive optimism. To assess how well an algorithm works, scientists typically ask how well its output correlates with medically assigned scores. Here we perform a meta-analysis to quantify how the literature evaluates their algorithms for monitoring mental wellbeing. We find that the bulk of the literature (∼77%) uses meaningless comparisons that ignore patient baseline state. For example, having an algorithm that uses phone data to diagnose mood disorders would be useful. However, it is possible to explain over 80% of the variance of some mood measures in the population by simply guessing that each patient has their own average mood—the patient-specific baseline. Thus, an algorithm that just predicts that our mood is like it usually is can explain the majority of variance, but is, obviously, entirely useless. Comparing to the wrong (population) baseline has a massive effect on the perceived quality of algorithms and produces baseless optimism in the field. To solve this problem we propose “user lift” that reduces these systematic errors in the evaluation of personalized medical monitoring.

BackgroundAutomatically tracking mental well-being could facilitate personalization of treatments for mood disorders such as depression and bipolar disorder. Smartphones present a novel and ubiquitous opportunity to track individuals’ behavior and may be useful for inferring and automatically monitoring mental well-being.ObjectiveThe aim of this study was to assess the extent to which activity and sleep tracking with a smartphone can be used for monitoring individuals’ mental well-being.MethodsA cohort of 106 individuals was recruited to install an app on their smartphone that would track their well-being with daily surveys and track their behavior with activity inferences from their phone’s accelerometer data. Of the participants recruited, 53 had sufficient data to infer activity and sleep measures. For this subset of individuals, we related measures of activity and sleep to the individuals’ well-being and used these measures to predict their well-being.ResultsWe found that smartphone-measured approximations for daily physical activity were positively correlated with both mood (P=.004) and perceived energy level (P<.001). Sleep duration was positively correlated with mood (P=.02) but not energy. Our measure for sleep disturbance was not found to be significantly related to either mood or energy, which could imply too much noise in the measurement. Models predicting the well-being measures from the activity and sleep measures were found to be significantly better than naive baselines (P<.01), despite modest overall improvements.ConclusionsMeasures of activity and sleep inferred from smartphone activity were strongly related to and somewhat predictive of participants’ well-being. Whereas the improvement over naive models was modest, it reaffirms the importance of considering physical activity and sleep for predicting mood and for making automatic mood monitoring a reality.

Emotional support is a crucial ability for many conversation scenarios, including social interactions, mental health support, and customer service chats. Following reasonable procedures and using various support skills can help to effectively provide support. However, due to the lack of a well-designed task and corpora of effective emotional support conversations, research on building emotional support into dialog systems remains untouched. In this paper, we define the Emotional Support Conversation (ESC) task and propose an ESC Framework, which is grounded on the Helping Skills Theory (Hill, 2009). We construct an Emotion Support Conversation dataset (ESConv) with rich annotation (especially support strategy) in a help-seeker and supporter mode. To ensure a corpus of high-quality conversations that provide examples of effective emotional support, we take extensive effort to design training tutorials for supporters and several mechanisms for quality control during data collection. Finally, we evaluate state-of-the-art dialog models with respect to the ability to provide emotional support. Our results show the importance of support strategies in providing effective emotional support and the utility of ES-Conv in training more emotional support systems 1 .