Table A3.
Studies of smartphones and wearables for monitoring neuropsychiatric illness
| Reference | Key aim | Population | Sensors | Design |
|---|---|---|---|---|
| Abdullah et al. 2016 | Estimate social rhythms (assessed via SRM questionnaires) using smartphone data | Seven subjects with BD | Smartphones recorded GPS data, accelerometry, microphone audio, and social communication | Offline retrospective |
| Aguilera et al. 2015 | Assess relationship between daily / weekly mood scores and PHQ-9 scores | 33 subjects | Smartphone administered PHQ-9 surveys | Offline retrospective |
| Albert et al. 2017 | Distinguish subjects with PD from controls using accelerometry of hand tremor | Eight subjects with PD and 18 controls | Smartphones recorded accelerometry of hand tremor during motor tasks | Offline retrospective |
| AlHanai et al. 2017 | Classify subject mood while reading happy or sad stories using wearable data | Ten healthy subjects | Audio was recorded using Apple iPhones. Samsung Simband smartwatches recorded PPG, ECG, accelerometry, skin impedance, galvanic skin response, and skin temperature | Online real-time |
| Apiquian et al. 2017 | Assess motor activity and sleep time before and after antipsychotic treatment | 20 subjects with schizophrenia and 20 controls | Wrist-worn devices recorded accelerometry | Offline retrospective |
| Barnett et al. 2018 | Predict clinical relapse from behavioral anomalies in two-week window prior to event | 17 subjects with schizophrenia | Smartphones recorded mobility, social activity, and questionnaires | Offline prospective |
| Beiwinkel et al. 2016 | Depressive and manic symptoms (assessed via HAMD and YMRS questionnaires administered every three weeks) were classified using smartphone data | 13 subjects with BD | Smartphones recorded GPS, accelerometery, and cell tower data; mood states were assessed via a self-reported two-item questionnaire | Offline retrospective |
| Ben-Zeev et al. 2015 | Correlate smartphone features with daily stress ratings, PHQ-9, PSS, and Revised UCLA Loneliness Scale scores | 47 healthy subjects | Smartphones recorded GPS, accelerometry, sleep duration, and time proximal to human speech | Offline retrospective |
| Berle et al. 2017 | Assess motor activity and rest-activity characteristics | 46 subjects with schizophrenia and 32 controls | Wrist-worn devices recorded actigraphy | Offline retrospective |
| Bullock et al. 2014 | Assess rest-activity metrics in BD patients with low and high trait vulnerability (assessed via the GBI questionnaire) | 72 subjects with BD | Wrist-worn devices recorded accelerometry | Offline retrospective |
| Burns et al. 2011 | Correlate EMA survey scores with smartphone features | Eight subjects with MDD | Smartphones recorded GPS, accelerometry, ambient light, and recent calls | Offline retrospective |
| Canzian et al. 2015 | Correlate and predict PHQ score deviations with smartphone features | 28 healthy subjects | Smartphones recorded GPS and accelerometry | Offline prospective |
| Capecci et al. 2016 | Identify freezing of gait events using accelerometry | 20 subjects with PD | Smartphones recorded accelerometry while subjects walked and were video recorded | Offline retrospective |
| Cella et al. 2017 | Assess autonomic dysfunction in schizophrenia using wearable device data | 30 subjects with schizophrenia and 25 controls | Empatica E4 devices recorded skin conductance, HRV, and accelerometry | Offline retrospective |
| Ellis et al. 2015 | Compare outcome measures of gait and gait variability in subjects with PD versus controls | 12 subjects with PD and 12 controls | Steps were captured via a smartphone, heel-mounted sensors, and a sensor mat | Offline retrospective |
| Kamdar et al. 2016 | Estimate variance of emotional state from wearable data via random forest | 13 healthy subjects | Samsung Gear S smartwatches recorded accelerometry, ambient light, heart rate; web app administered mood surveys | Offline retrospective |
| Moore et al. 2012 | Forecast mood time series using previous week’s self-rated mood data via exponential smoothing and Gaussian process regression | 100 subjects with BD | Mood surveys recorded via SMS | Offline prospective |
| Faedda et al. 2016 | Distinguish BD from ADHD using wearables data | 48 subjects with BD, 65 subjects with ADHD, and 42 controls | Belt-worn devices recorded accelerometry for five minutes | Offline retrospective |
| Faurholt-Jepsen et al. 2015 | Correlate smartphone data with depressive and manic symptoms via HDRS-17 and YMRS scores assessed monthly | 61 subjects with BD | Smartphones recorded speech duration, social activity, and accelerometry | Offline retrospective |
| Maria et al. 2016 | Classify depressive and manic states (via HDRS-17 and YMRS scores) using smartphone data and voice features | 28 subjects with BD | Smartphones recorded voice features (pitch, duration, etc.), speech duration, social activity, and accelerometry | Offline retrospective |
| Fasmer et al. 2015 | Fit resting and active periods to power law distributions and assess differences in MDD | 47 subjects with MDD and 29 controls | Wrist-worn devices recorded accelerometry | Offline retrospective |
| Griffiths et al. 2012 | Assess features of dyskinesia and akinesia from wearable data, and identify improvements in UPDRS scores after medication | 34 subjects with PD and 10 controls | Wrist-worn devices recorded accelerometry | Offline retrospective |
| Grünerbl et al. 2015 | Depressive and manic symptoms (assessed via HAMD and YMRS questionnaires administered every three weeks) were classified using smartphone data | Ten subjects with BD | Smartphones recorded GPS, accelerometry, number and length of phone calls, and speech and voice features | Offline retrospective |
| Hauge et al. 2011 | Assess motor activity and rest-activity characteristics | 24 subjects with schizophrenia, 25 subjects with depression, and 32 controls | Wrist-worn devices recorded actigraphy | Offline retrospective |
| Kassavetis et al. 2016 | Correlate UPDRS scores with smartphone data | 14 subjects with PD | Smartphones recorded accelerometry while subjects performed motor tasks | Offline retrospective |
| Kheirkhahan et al. 2016 | Correlate impaired mobility from wearable data | 1,135 subjects | Hip-worn devices recorded accelometry | Offline retrospective |
| Kim et al. 2015 | Classify freezing episodes from normal walking using accelerometry | 15 subjects with PD | Smartphones recorded accelerometry while subjects walked and were video recorded | Offline retrospective |
| Kostikis et al. 2014 | Correlate accelerometry features with UPDRS hand tremor scores | 23 subjects with PD | Smartphones recorded accelerometry of hand tremor during motor tasks | Offline retrospective |
| Kostikis et al. 2015 | Distinguish subjects with PD from controls using accelerometry of hand tremor | 25 subjects and 20 controls | Smartphones recorded accelerometry of hand tremor during motor tasks | Offline retrospective |
| Krane-gartiser et al. 2014 | Assess mean activity, variance, symbolic dynamics, and power spectral features | 18 subjects with mania and 12 subjects with BD | Wrist-worn devices recorded accelerometry | Offline retrospective |
| Kuhlmei et al. 2013 | Associate activity with apathy and depression (assessed via AES and BDI questionnaires) | 32 subjects with dementia, 21 subjects with MCI, and 23 controls | Wrist-worn devices recorded accelerometry during motor tasks | Offline retrospective |
| Lee et al. 2015 | Compare RR peak detection, HRV measures, and stress detection from wearable versus Holter monitor | 17 subjects | Custom ECG patch was developed to record cardiac activity | Offline retrospective |
| Lee et al. 2016 | Correlate UPDRS scores with smartphone data | 103 subjects with PD | Smartphones recorded hand dexterity via timed tapping test, rapid alternating movements, tremor tracker via tracing between two parallel lines, and a cognitive interference test | Offline retrospective |
| Martin et al. 2006 | Assess time in bed, sleep consistency, daytime sleeping, and circadian rhythm regularity | 28 subjects with schizophrenia and 28 controls | Wrist-worn devices recorded accelerometry and light exposure | Offline retrospective |
| Nakamura et al. 2007 | Fit resting and active periods to power law distributions and assess differences in MDD | 14 subjects with MDD and 11 controls | Wrist-worn devices recorded accelerometry | Offline retrospective |
| Nero et al. 2015 | Define accelerometer cut points for different walking speeds in adults with PD | 30 subjects with PD | Waist-worn devices recorded accelerometry | Offline retrospective |
| Niwa et al. 2011 | Assess if medication status, MMSE scores, activity, and HRV features differed by disease severity (assessed via UPDRS scores) or disease duration | 27 subjects with PD and 30 controls | Wrist-worn devices recorded accelerometry and Holter monitors recorded ambulatory ECG | Offline retrospective |
| O’Brien et al. 2016 | Assess relationship between quality of life, ADLs, learning, and depression (assessed via SF-36 and IADLS questionnaires) and smartphone data | 29 subjects with MDD and 30 controls | Wrist-worn devices recorded accelerometry. Quality of life, ADLs, learning, and depression were assessed via SF-36 and IADLS questionnaires | Offline retrospective |
| Osipov et al. 2015 | Classify schizophrenic subjects from controls using rest-activity characteristics and HRV features | 16 subjects with schizophrenia and 19 controls | Adhesive patches recorded locomotor activity and ECG | Offline retrospective |
| Palmius et al. 2017 | Estimate depressive symptoms (assessed via QIDS-SR16 questionnaires administered weekly) and detect depression using smartphone data | 22 subjects with BD and 14 controls | Smartphones recorded GPS data | Offline retrospective |
| Pan et al. 2015 | Correlate accelerometry features with UPDRS scores, and use features to detect hand resting tremor and gait difficulty | 40 subjects with PD | Smartphones recorded accelerometry of hand tremor and gait during motor and walking tasks | Offline retrospective |
| Patel et al. 2009 | Estimate UPDRS scores using wearable data | 12 subjects with PD | Arm and leg-worn devices recorded accelerometry | Offline retrospective |
| Place et al. 2017 | Estimate depression and PTSD symptoms (assessed via SCID questionnaires) using smartphone data | 73 subjects with at least one symptom of PTSD or depression | Smartphones recorded GPS, accelerometry, calls and SMS activity, device use, and voice audio | Offline retrospective |
| Reinertsen et al. 2017a | Classify patients with PTSD using time-domain, frequency-domain, and complexity features from RR interval time series | 23 subjects with PTSD and 25 controls | A Holter monitor recorded RR intervals for 24 hours | Offline retrospective |
| Reinertsen et al. 2017b | Classify schizophrenic subjects from controls using rest-activity characteristics and HRV features, and evaluate relationship between number of days of data and classifier accuracy | 16 subjects with schizophrenia and 19 controls | Adhesive patches recorded locomotor activity and ECG | Offline retrospective |
| Roh et al. 2014 | Compare RR peak detection, signal-to-noise, and HRV measures from wearable versus Holter monitor | 12-41 subjects (varied by test) | Custom ECG patch was developed to record cardiac activity | Offline retrospective |
| Roy et al. 2011 | Classify tremor and dyskinesia from wearable data | 11 subjects with PD | Arm and leg-worn devices recorded accelerometry | Offline retrospective |
| Saeb et al. 2015 | Classify low from high PHQ-9 scores using smartphone features | 28 healthy subjects | Smartphones recorded GPS and phone usage | Offline retrospective |
| Saeb et al. 2016a | Correlate PHQ-9 scores with smartphone features from weekend vs. weekday data | 48 healthy subjects | Smartphones recorded GPS and phone usage | Offline retrospective |
| Sano et al. 2012 | Fit resting and active periods to power law distributions and assess differences in schizophrenia | 19 subjects with schizophrenia and 11 controls | Wrist-worn devices recorded accelerometry | Offline retrospective |
| Sano et al. 2013 | Distinguish stressed from non-stressed states using wearable data | 18 subjects | Wrist-worn devices recorded accelerometry and skin conductance. Smartphones recorded call and SMS activity. Surveys assessed stress, mood, sleep, tiredness, general health, alcohol or caffeine intake, and electronics usage. | Offline retrospective |
| Sano et al. 2015 | Estimate PSQI, PSS, and MCS questionnaire scores from wearable data | 66 subjects | Wrist-worn devices recorded accelerometry and skin conductance. Smartphones recorded call and SMS activity. Sleep, stress, and mental health were assessed via PSQI, PSS, and MCS questionnaires respectively | Offline retrospective |
| Shin et al. 2016 | Correlate symptom severity (assessed via the PANSS questionnaire) with activity levels | 61 subjects with schizophrenia | Wrist-worn devices recorded accelerometry | Offline retrospective |
| Stamatakis et al. 2013 | Classify UPDRS score categories from wearable data | 36 subjects with PD and 10 controls | Finger-worn sensors recorded accelerometry during a tapping test | Offline retrospective |
| Tung et al. 2014 | Compare area, perimeter, and mean distance from home in subjects with AD versus controls using smartphone data | 19 subjects with AD and 33 controls | Smartphones recorded GPS | Offline retrospective |
| Walther et al. 2009b | Assess if motor symptoms (assessed via PANSS questionnaires) correlate with wearables data | 55 subjects with schizophrenia | Wrist-worn devices recorded actigraphy | Offline retrospective |
| Walther et al. 2009a | Assess if activity differs by schizophrenia subtype | 60 subjects with schizophrenia | Wrist-worn devices recorded actigraphy | Offline retrospective |
| Wang et al. 2014 | Correlate smartphone data with PHQ-9, PSS, flourishing scale, and UCLA loneliness scale scores | 48 healthy subjects | Smartphones recorded accelerometry, conversations, sleep, and location | Offline retrospective |
| Wang et al. 2016 | Determine associations between EMA survey scores and smartphone data via generalized estimating equations | 21 subjects with schizophrenia | Smartphones recorded accelerometry, voice audio, light sensor readings, GPS data, and application usage | Offline retrospective |
| Weenk et al. 2017 | Evaluate association between changes in HRV measures and stress in surgeons | 20 subjects | Adhesive patch measured single-lead ECG, respiratory rate, skin temperature, body posture, activity, and steps | Offline retrospective |
| Wichniak et al. 2011 | Measure association between activity levels and mental status (measured via PANSS and CDSS questionnaires) | 73 subjects with schizophrenia and 36 controls | Wrist-worn devices recorded accelerometry | Offline retrospective |
| Winkler et al. 2005 | Assess if light therapy can improve sleep efficiency and stability in people with seasonal affective disorder (SAD) | 17 subjects with SAD and 17 controls | Wrist actigraphy was recorded from which sleep-wake amplitude, phase, and sleep efficiency was estimated | Offline retrospective |
| Woods et al. 2014 | Distinguish PD from essential tremor using accelerometry | 14 subjects with PD and 18 subjects with essential tremor | Smartphones recorded accelerometry of hand tremor during motor tasks | Offline retrospective |
| Vallance et al. 2011 | Assess relationship between depression (assessed via PHQ-9 questionnaires) and activity | 2,862 subjects | Wrist-worn devices recorded accelerometry | Offline retrospective |