Ecological Momentary Assessments and Passive Sensing in the Prediction of Short-Term Suicidal Ideation in Young Adults

Abstract

ImportanceAdvancements in technology, including mobile-based ecological momentary assessments (EMAs) and passive sensing, have immense potential to identify short-term suicide risk. However, the extent to which EMA and passive data, particularly in combination, have utility in detecting short-term risk in everyday life remains poorly understood.ObjectiveTo examine whether and what combinations of self-reported EMA and sensor-based assessments identify next-day suicidal ideation.Design, Setting, and ParticipantsIn this intensive longitudinal prognostic study, participants completed EMAs 4 times daily and wore a sensor wristband (Fitbit Charge 3) for 8 weeks. Multilevel machine learning methods, including penalized generalized estimating equations and classification and regression trees (CARTs) with repeated 5-fold cross-validation, were used to optimize prediction of next-day suicidal ideation based on time-varying features from EMAs (affective, cognitive, behavioral risk factors) and sensor data (sleep, activity, heart rate). Young adult patients who visited an emergency department with recent suicidal ideation and/or suicide attempt were recruited. Identified via electronic health record screening, eligible individuals were contacted remotely to complete enrollment procedures. Participants (aged 18 to 25 years) completed 14 708 EMA observations (64.4% adherence) and wore a sensor wristband approximately half the time (55.6% adherence). Data were collected between June 2020 and July 2021. Statistical analysis was performed from January to March 2023.Main Outcomes and MeasuresThe outcome was presence of next-day suicidal ideation.ResultsAmong 102 enrolled participants, 83 (81.4%) were female; 6 (5.9%) were Asian, 5 (4.9%) were Black or African American, 9 (8.8%) were more than 1 race, and 76 (74.5%) were White; mean (SD) age was 20.9 (2.1) years. The best-performing model incorporated features from EMAs and showed good predictive accuracy (mean [SE] cross-validated area under the receiver operating characteristic curve [AUC], 0.84 [0.02]), whereas the model that incorporated features from sensor data alone showed poor prediction (mean [SE] cross-validated AUC, 0.56 [0.02]). Sensor-based features did not improve prediction when combined with EMAs. Suicidal ideation-related features were the strongest predictors of next-day ideation. When suicidal ideation features were excluded, an alternative EMA model had acceptable predictive accuracy (mean [SE] cross-validated AUC, 0.76 [0.02]). Both EMA models included features at different timescales reflecting within-day, end-of-day, and time-varying cumulative effects.Conclusions and RelevanceIn this prognostic study, self-reported risk factors showed utility in identifying near-term suicidal thoughts. Best-performing models required self-reported information, derived from EMAs, whereas sensor-based data had negligible predictive accuracy. These results may have implications for developing decision algorithms identifying near-term suicidal thoughts to guide risk monitoring and intervention delivery in everyday life.

Machine Learning for Multimodal Mental Health Detection: A Systematic Review of Passive Sensing Approaches

“…Wearable devices have further enabled the collection of physical activity, movement, sleep, and physiological signals like heart rate (HR), electrodermal activity (EDA), skin temperature (ST), and galvanic skin response (GSR). Some examples of wearables are Empatica E4 wristbands [ 149 , 172 , 178 ], Microsoft Band 2 [ 150 ], Fitbit Charge or Flex trackers [ 151 , 155 , 164 , 180 , 181 , 182 , 191 ], and the Galaxy S3 smartwatch [ 169 ]. Data gathered through these devices were transmitted directly to an internet-connected server [ 215 ] or transferred via Bluetooth [ 210 ] to dedicated mobile applications that handle the transmission as described above.…”

“…Depression AV [43, SM [20,25,98, SS [99,100,104,105, WS [149][150][151][155][156][157][158]164,169,[171][172][173][178][179][180][181][182] Suicidal intent AV [100,183,184] SM [185][186][187][188][189] SS [100,147,190] WS [181,182,191] Bipolar disorder AV [101][102][103][192][193][194][195][196][197][198][199][200] SM…”

Subgrouping suicidal ideations: An ecological momentary assessment study in psychiatric inpatients

“…With the finegrained assessment facilitated by EMA (22), instances of SI can be captured that would have gone unnoticed by retrospective self-report measures (23). Contrary to prior assumptions that in addition to the more subjective assessment with EMA, an objective digital marker (e.g., data from passive sensors) might be a crucial contribution to better prediction models, recent evidence suggests that EMA data alone was a good predictor for next-day SI and passive sensing data did not improve prediction (24).…”

Identifying factors impacting missingness within smartphone-based research: Implications for intensive longitudinal studies of adolescent suicidal thoughts and behaviors.

“…Intensive longitudinal designs and smartphone-based research offer opportunities to frequently and continuously measure real-world adolescent behavior and, thus, provide a unique window into short-term risk factors (Allen et al, 2019). Both self-report-based assessment (i.e., experience sampling) and passive smartphone or wearable sensors (e.g., geolocation, accelerometer, keyboard inputs, and app usage) have shown promise in characterizing digital phenotypes and identifying proximal factors associated with clinical acuity and STB (Auerbach et al, 2023; Czyz, King, et al, 2023; Glenn, Kleiman, Kandlur, et al, 2022; Moreno-Muñoz et al, 2020). Smartphone and wearable measures can also continuously collect data that track affective and behavioral dynamics (Ren et al, 2023; Russell & Gajos, 2020) without requiring in-person lab visits, thus lowering barriers to participation among disadvantaged and hard-to-reach populations (Sugie, 2018).…”