Digital Phenotyping Data to Predict Symptom Improvement and Mental Health App Personalization in College Students: Prospective Validation of a Predictive Model.

BACKGROUND: Mental health apps offer a transformative means to increase access to scalable evidence-based care for college students. Yet low rates of engagement currently preclude the effectiveness of these apps. One promising solution is to make these apps more responsive and personalized through digital phenotyping methods able to predict symptoms and offer tailored interventions. OBJECTIVE: Following our protocol and using the exact model shared in that paper, our primary aim in this study is to assess the prospective validity of mental health symptom prediction using the mindLAMP app through a replication study. We also explored secondary aims around app intervention personalization and correlations of engagement with the Technology Acceptance Model (TAM) and Digital Working Alliance Inventory scale in the context of automating the study. METHODS: The study was 28 days in duration and followed the published protocol, with participants collecting digital phenotyping data and being offered optional scheduled and algorithm-recommended app interventions. Study compensation was tied to the completion of weekly surveys and was not otherwise tied to engagement or use of the app. RESULTS: The data from 67 participants were used in this analysis. The area under the curve values for the symptom prediction model ranged from 0.58 for the UCLA Loneliness Scale to 0.71 for the Patient Health Questionnaire-9. Engagement with the scheduled app interventions was high, with a study mean of 73%, but few participants engaged with the optional recommended interventions. The perceived utility of the app in the TAM was higher (P=.01) among those completing at least one recommended intervention. CONCLUSIONS: Our results suggest how digital phenotyping methods can be used to create generalizable models that may help create more personalized and engaging mental health apps. Automating studies is feasible, and our results suggest targets to increase engagement in future studies. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): RR2-10.2196/37954.

Deep learning reconstruction for cardiac magnetic resonance fingerprinting T1 and T2 mapping.

PURPOSE: To develop a deep learning method for rapidly reconstructing T1 and T2 maps from undersampled electrocardiogram (ECG) triggered cardiac magnetic resonance fingerprinting (cMRF) images. METHODS: A neural network was developed that outputs T1 and T2 values when given a measured cMRF signal time course and cardiac RR interval times recorded by an ECG. Over 8 million cMRF signals, corresponding to 4000 random cardiac rhythms, were simulated for training. The training signals were corrupted by simulated k-space undersampling artifacts and random phase shifts to promote robust learning. The deep learning reconstruction was evaluated in Monte Carlo simulations for a variety of cardiac rhythms and compared with dictionary-based pattern matching in 58 healthy subjects at 1.5T. RESULTS: In simulations, the normalized root-mean-square error (nRMSE) for T1 was below 1% in myocardium, blood, and liver for all tested heart rates. For T2 , the nRMSE was below 4% for myocardium and liver and below 6% for blood for all heart rates. The difference in the mean myocardial T1 or T2 observed in vivo between dictionary matching and deep learning was 3.6 ms for T1 and -0.2 ms for T2 . Whereas dictionary generation and pattern matching required more than 4 min per slice, the deep learning reconstruction only required 336 ms. CONCLUSION: A neural network is introduced for reconstructing cMRF T1 and T2 maps directly from undersampled spiral images in under 400 ms and is robust to arbitrary cardiac rhythms, which paves the way for rapid online display of cMRF maps.

Increasing the value of digital phenotyping through reducing missingness: a retrospective review and analysis of prior studies.

BACKGROUND: Digital phenotyping methods present a scalable tool to realise the potential of personalised medicine. But underlying this potential is the need for digital phenotyping data to represent accurate and precise health measurements. OBJECTIVE: To assess the impact of population, clinical, research and technological factors on the digital phenotyping data quality as measured by rates of missing digital phenotyping data. METHODS: This study analyses retrospective cohorts of mindLAMP smartphone application digital phenotyping studies run at Beth Israel Deaconess Medical Center between May 2019 and March 2022 involving 1178 participants (studies of college students, people with schizophrenia and people with depression/anxiety). With this large combined data set, we report on the impact of sampling frequency, active engagement with the application, phone type (Android vs Apple), gender and study protocol features on missingness/data quality. FINDINGS: Missingness from sensors in digital phenotyping is related to active user engagement with the application. After 3 days of no engagement, there was a 19% decrease in average data coverage for both Global Positioning System and accelerometer. Data sets with high degrees of missingness can generate incorrect behavioural features that may lead to faulty clinical interpretations. CONCLUSIONS: Digital phenotyping data quality requires ongoing technical and protocol efforts to minimise missingness. Adding run-in periods, education with hands-on support and tools to easily monitor data coverage are all productive strategies studies can use today. CLINICAL IMPLICATIONS: While it is feasible to capture digital phenotyping data from diverse populations, clinicians should consider the degree of missingness in the data before using them for clinical decision-making.

Digital Phenotyping Models of Symptom Improvement in College Mental Health: Generalizability Across Two Cohorts.

Smartphones can be used to gain insight into mental health conditions through the collection of survey and sensor data. However, the external validity of this digital phenotyping data is still being explored, and there is a need to assess if predictive models derived from this data are generalizable. The first dataset (V1) of 632 college students was collected between December 2020 and May 2021. The second dataset (V2) was collected using the same app between November and December 2021 and included 66 students. Students in V1 could enroll in V2. The main difference between the V1 and V2 studies was that we focused on protocol methods in V2 to ensure digital phenotyping data had a lower degree of missing data than in the V1 dataset. We compared survey response counts and sensor data coverage across the two datasets. Additionally, we explored whether models trained to predict symptom survey improvement could generalize across datasets. Design changes in V2, such as a run-in period and data quality checks, resulted in significantly higher engagement and sensor data coverage. The best-performing model was able to predict a 50% change in mood with 28 days of data, and models were able to generalize across datasets. The similarities between the features in V1 and V2 suggest that our features are valid across time. In addition, models must be able to generalize to new populations to be used in practice, so our experiments provide an encouraging result toward the potential of personalized digital mental health care.

The impact of guided versus supportive coaching on mental health app engagement and clinical outcomes.

Although mobile mental health apps have the unique potential to increase access to care, evidence reveals engagement is low unless coupled with coaching. However, most coaching protocols are limited in their scalability. This study assesses how human support and guidance from a Digital Navigator (DN), a scalable coach, can impact mental health app engagement and effectiveness on anxiety and depressive symptoms. This study aims to detach components of coaching, specifically personalized recommendations versus general support, to inform scalability of coaching models for mental health apps. 156 participants were split into the DN Guide versus DN Support groups for the 6-week study. Both groups utilized the mindLAMP app for the duration of the study and had equal time with the DN, but the Guide group received personalized app recommendations. The Guide group completed significantly more activities than the Support group. 34% (49/139) of all participants saw a 25% decrease in PHQ-9 scores and 38% (53/141) saw a 25% decrease in GAD-7 scores. These findings show mental health apps, especially when supported by DNs, can reduce depression and anxiety symptoms when coupled with coaching, suggesting a feasible path for large-scale deployment.

EPViz: A flexible and lightweight visualizer to facilitate predictive modeling for multi-channel EEG.

Scalp Electroencephalography (EEG) is one of the most popular noninvasive modalities for studying real-time neural phenomena. While traditional EEG studies have focused on identifying group-level statistical effects, the rise of machine learning has prompted a shift in computational neuroscience towards spatio-temporal predictive analyses. We introduce a novel open-source viewer, the EEG Prediction Visualizer (EPViz), to aid researchers in developing, validating, and reporting their predictive modeling outputs. EPViz is a lightweight and standalone software package developed in Python. Beyond viewing and manipulating the EEG data, EPViz allows researchers to load a PyTorch deep learning model, apply it to EEG features, and overlay the output channel-wise or subject-level temporal predictions on top of the original time series. These results can be saved as high-resolution images for use in manuscripts and presentations. EPViz also provides valuable tools for clinician-scientists, including spectrum visualization, computation of basic data statistics, and annotation editing. Finally, we have included a built-in EDF anonymization module to facilitate sharing of clinical data. Taken together, EPViz fills a much needed gap in EEG visualization. Our user-friendly interface and rich collection of features may also help to promote collaboration between engineers and clinicians.

Utility of Smartphone-Based Digital Phenotyping Biomarkers in Assessing Treatment Response to Transcranial Magnetic Stimulation in Depression: Proof-of-Concept Study.

BACKGROUND: Identifying biomarkers of response to transcranial magnetic stimulation (TMS) in treatment-resistant depression is a priority for personalizing care. Clinical and neurobiological determinants of treatment response to TMS, while promising, have limited scalability. Therefore, evaluating novel, technologically driven, and potentially scalable biomarkers, such as digital phenotyping, is necessary. OBJECTIVE: This study aimed to examine the potential of smartphone-based digital phenotyping and its feasibility as a predictive biomarker of treatment response to TMS in depression. METHODS: We assessed the feasibility of digital phenotyping by examining the adherence and retention rates. We used smartphone data from passive sensors as well as active symptom surveys to determine treatment response in a naturalistic course of TMS treatment for treatment-resistant depression. We applied a scikit-learn logistic regression model (l1 ratio=0.5; 2-fold cross-validation) using both active and passive data. We analyzed related variance metrics throughout the entire treatment duration and on a weekly basis to predict responders and nonresponders to TMS, defined as ≥50% reduction in clinician-rated symptom severity from baseline. RESULTS: The adherence rate was 89.47%, and the retention rate was 73%. The area under the curve for correct classification of TMS response ranged from 0.59 (passive data alone) to 0.911 (both passive and active data) for data collected throughout the treatment course. Importantly, a model using the average of all features (passive and active) for the first week had an area under the curve of 0.7375 in predicting responder status at the end of the treatment. CONCLUSIONS: The results of our study suggest that it is feasible to use digital phenotyping data to assess response to TMS in depression. Early changes in digital phenotyping biomarkers, such as predicting response from the first week of data, as shown in our results, may also help guide the treatment course.

Digital phenotyping correlations in larger mental health samples: analysis and replication.

BACKGROUND: Smartphones can facilitate patients completing surveys and collecting sensor data to gain insight into their mental health conditions. However, the utility of sensor data is still being explored. Prior studies have reported a wide range of correlations between passive data and survey scores. AIMS: To explore correlations in a large data-set collected with the mindLAMP app. Additionally, we explored whether passive data features could be used in models to predict survey results. METHOD: Participants were asked to complete daily and weekly mental health surveys. After screening for data quality, our sample included 147 college student participants and 270 weeks of data. We examined correlations between six weekly surveys and 13 metrics derived from passive data features. Finally, we trained logistic regression models to predict survey scores from passive data with and without daily surveys. RESULTS: Similar to other large studies, our correlations were lower than prior reports from smaller studies. We found that the most useful features came from GPS, call, and sleep duration data. Logistic regression models performed poorly with only passive data, but when daily survey scores were included, performance greatly increased. CONCLUSIONS: Although passive data alone may not provide enough information to predict survey scores, augmenting this data with short daily surveys can improve performance. Therefore, it may be that passive data can be used to refine survey score predictions and clinical utility may be derived from the combination of active and passive data.

Digital Phenotyping Data to Predict Symptom Improvement and App Personalization: Protocol for a Prospective Study.

BACKGROUND: Smartphone apps that capture surveys and sensors are increasingly being leveraged to collect data on clinical conditions. In mental health, this data could be used to personalize psychiatric support offered by apps so that they are more effective and engaging. Yet today, few mental health apps offer this type of support, often because of challenges associated with accurately predicting users' actual future mental health. OBJECTIVE: In this protocol, we present a study design to explore engagement with mental health apps in college students, using the Technology Acceptance Model as a theoretical framework, and assess the accuracy of predicting mental health changes using digital phenotyping data. METHODS: There are two main goals of this study. First, we present a logistic regression model fit on data from a prior study on college students and prospectively test this model on a new student cohort to assess its accuracy. Second, we will provide users with data-driven activity suggestions every 4 days to determine whether this type of personalization will increase engagement or attitudes toward the app compared to those receiving no personalized recommendations. RESULTS: The study was completed in the spring of 2022, and the manuscript is currently in review at JMIR Publications. CONCLUSIONS: This is one of the first digital phenotyping algorithms to be prospectively validated. Overall, our results will inform the potential of digital phenotyping data to serve as tailoring data in adaptive interventions and to increase rates of engagement. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): PRR1-10.2196/37954.

Smartphone digital phenotyping, surveys, and cognitive assessments for global mental health: Initial data and clinical correlations from an international first episode psychosis study.

Objective: To examine feasibility and acceptability of smartphone mental health app use for symptom, cognitive, and digital phenotyping monitoring among people with schizophrenia in India and the United States. Methods: Participants in Boston, USA and Bhopal and Bangalore, India used a smartphone app to monitor symptoms, play cognitive games, access relaxation and psychoeducation resources and for one month, with an initial clinical and cognitive assessment and a one-month follow-up clinical assessment. Engagement with the app was compared between study sites, by clinical symptom severity and by cognitive functioning. Digital phenotyping data collection was also compared between three sites. Results: By Kruskal-Wallis rank-sum test, we found no difference between app activities completed or digital phenotyping data collected across the three study sites. App use also did not correlate to clinical or cognitive assessment scores. When using the app for symptom monitoring, preliminary findings suggest app-based assessment correlate with standard cognitive and clinical assessments. Conclusions: Smartphone app for symptom monitoring and digital phenotyping for individuals with schizophrenia appears feasible and acceptable in a global context. Clinical utility of this app for real-time assessments is promising, but further research is necessary to determine the long-term efficacy and generalizability for serious mental illness.