Journal Logo


Technology in the Assessment, Treatment, and Management of Depression

Bader, Caroline S. MD; Skurla, Miranda BS; Vahia, Ipsit V. MD

Section Editor(s): Pizzagalli, Diego A. PhD; Editror

Author Information
doi: 10.1097/HRP.0000000000000235
  • Free

The diagnostic standard for major depressive disorder (MDD) continues to rely on the latest, fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5), with its nine criterion domains of depression: (1) depressed mood, (2) loss of interest or pleasure, (3) appetite/weight disturbance, (4) sleep disturbance, (5) psychomotor changes, (6) loss of energy, (7) feelings of worthlessness/guilt, (8) decreased concentration, and (9) suicidal ideation. This structure creates a heterogeneous diagnostic group, however, and does not account for the widespread recognition that there may be several individual phenotypes of depression that bear scarce resemblance to each other.1 For example, older adults are much more likely to experience loneliness, helplessness, and grief as part of a depressive syndrome, but traditional measures of depression do not capture such entities. The current diagnostic approach also fails to capture subthreshold depressive symptoms, which may be up to three times more common than MDD, cause significant declines in quality of life and functioning, and may be diagnosed by the presence of just one symptom of depression.2 Further, these categories, while on the whole being useful to diagnose MDD, also can lack specificity. Sleep disturbance may include insomnia or hypersomnia. Similarly, loss of appetite and weight loss are grouped into the same category as increase in appetite and weight gain. The Patient Health Questionnaire (PHQ)–9, a screening tool based on these DSM-5 criteria, is increasingly being used in primary care settings as the sole method of screening for and diagnosing depression. However, with the rapid increase in availability of new technologies to facilitate the mapping of behaviors,3 it may now be feasible to measure each domain of depression distinctly, paving the way for more accurate and individualized data about a given person’s mood. Digital phenotyping using technologies can also serve as a window into the patient's broader psychosocial environment and provide a higher quality of collateral information than present measures.4 In this commentary, we review how specific technologies may facilitate the monitoring of each individual domain of the depression syndrome. We also discuss how such technologies can facilitate treatments for individual aspects of the depression syndrome and move the field toward more precise care of specific elements of depression.


Overall, the scope of technologies that may be applied to the measurement of depression is broad, ranging from active approaches to data collection, such as ecological momentary assessment (EMA), to highly passive approaches that require minimal engagement by the patient. Researchers have extracted signals of mood from a variety of sources, including frequent self-report (i.e., EMA),5 motion (via accelerometry or other passive sensing methods),6 language (via natural language processing [NLP] or measuring keystrokes),7 and spatial location (via global positioning system [GPS] technology).8 Frequent self-report based on EMA through surveys applied via text messaging9 or mobile application10 is one of the best-validated approaches to real-time data collection, and although it is not a new technology, the growing prevalence of personal mobile devices makes EMA an increasingly attractive and feasible approach. Motion detection has been considerably simplified by the ubiquitous availability of accelerometers in mobile devices and wearables. The use of NLP is based on the recognition that vocal changes are known to be present in depressed mood, including slowed rate, reduced pitch, and errors in articulation. Features such as the number of conversations, speaking length, pitch, and volume have been used to assess mood states, energy, and social rhythm, though their clinical utility has not been fully established.11 The ability to generate temporally dense data on motion using various technologies, including mobile devices, wearables, and passive environmental sensors,6 can help map an array of behaviors and also sleep. GPS technology, especially when accessed via mobile devices, can help quantify an individual’s movements in small spaces such as a house (smartphone GPS can accurately locate an individual within under five meters)12 and across much larger geographical fields such as entire cities. Each of these technologies may be deployed to capture specific domains of depression (Table 1). In the section below we illustrate how these technologies are being used by researchers to capture depression-related parameters.

Table 1
Table 1:
Technologies Used to Assess, Manage, or Treat Different Domains of Depression


Depressed Mood

Since depressed mood is a subjective state, it has been challenging to quantify it. EMA, however, which collects real-time, contextual in vivo information about a person’s mood state and avoids recall bias, can help track variations in depressed mood. EMA has been widely validated and is reliable, sensitive to changes in medication management, and feasible for use in patients with MDD.13–16 EMA has also been studied for use specifically in children and adolescents, though many studies still use an EMA protocol that involves phone calls to participants, rather than smartphone apps or digital reminders that link to surveys on web pages. In addition, EMA has also been demonstrated as a feasible approach in older adults, indicating that this approach can have utility across the lifespan.17,18 Changes in language use are also seen in depressed mood. NLP has been studied as a method for extracting information on depressed mood from several sources, including electronic health records19 and social media.20 For example, in one study, researchers demonstrated how the analysis of publicly posted content on Facebook could predict subsequent diagnosis of depression in their medical records up to three months in advance.21 Similarly, another group created a classifier to distinguish between depressed and non-depressed Twitter accounts and to detect risk of developing MDD based on characteristic ways of when and how users tweeted;7 they found depressed users were more likely to tweet at night, less likely to respond to others’ tweets, and less likely to use the first and third person in their tweets (i.e., tweet about themselves or others).


Detection of depressed mood and anhedonia often overlap, and technologies may detect related signals. NLP, for example, has the potential to detect a person’s mood state and enjoyment in an activity. It has been shown to be possible to extract PHQ-9 scores using NLP in clinical notes.22 Similarly, another group created a classifier to distinguish between depressed and non-depressed Twitter accounts and to detect risk of developing MDD based on characteristic ways of when and how users tweeted;7 they found depressed users were more likely to tweet at night, less likely to respond to others’ tweets, and less likely to use the first and third person in their tweets (i.e., tweet about themselves or others). Methods used to predict or identify depressed mood, as cited above, may also potentially be used in relation to anhedonia and engagement with activities. Future directions include combining NLP and machine learning to create a real-time intervention (ecological momentary intervention [EMI]) tailored to the individual based on data collected about the person’s mood state at that specific time. This intervention could then be delivered via handheld devices instantaneously. While this approach remains in its infancy and is still speculative, it has the potential to shift the paradigm of psychiatric care from the current model of “diagnosis and treatment” to a more continuous, real-time assessment model with immediate delivery of tailored care.5

GPS data may also serve as an approximation for anhedonia by showing a person’s inclination to leave the home, socialize with others, and participate in activities.23 These measures also overlap with other depressive symptoms, including low energy and motivation. A study of bipolar patients in a community sample was able to successfully detect depressive episodes using geographic location recordings from smartphone data.24 One pilot study combined data from GPS and Bluetooth interactions to map clusters of activity (e.g., home, workplace, gym, cafeteria), demonstrating the feasibility of mapping pleasurable activity and changes in routine.25 It has been acknowledged that no single one of these technologies may approximate anhedonia or depressed mood and that a multimodal approach may be most suitable; in one study, for example, Twitter users’ accounts were analyzed with NLP, GPS monitoring, and analysis of the social network, to approximate depression.7

Changes in Sleep

Disruption in sleep can be both a predictor and a symptom of depressed mood.26 The current gold standard for objective measurement of sleep quality is polysomnography (PSG), which involves an overnight stay in a sleep laboratory. The widespread use of actigraphy technology and innovations in electroencephalography (EEG) have allowed sophisticated, objective measures of sleep to became portable and accessible. Actigraphy has been used in combination with self-reported measures like the Pittsburgh Sleep Quality Index and sleep diaries to understand the relation between sleep disturbances and the risk of depression in adults. Objectively poor sleep quality as measured by actigraphy has been associated with greater odds of subjectively worse depression symptoms in both men and women.27,28 Additional benefits of actigraphy include the ability to analyze sleep quality by specific metrics, such as sleep efficiency, sleep latency, time awake after sleep onset, total sleep time, and long-wake episodes, all of which provide an individualized picture of sleep behavior.28

In addition to actigraphy, portable, in-ear, and frontopolar electroencephalograms have been evaluated by testing their level of agreement with PSG in categorizing sleep stages. Overall, the newer technologies had highest agreement with PSG when identifying REM stage sleep and transitions into and out of REM stage sleep.29,30 The newer EEGs had lower percentages of agreement with PSG when differentiating between light stages of sleep, such as N1 and N2, and wakefulness.31 An advantage of some in-home EEG devices is the ability to autoscore the raw data, which eliminates the need for an expert reader, as is usually done in PSG and EEG. While these devices have not been studied extensively for sleep in depression, they hold great potential.

Changes in Appetite

Wrist-worn actigraphy has been employed to objectively measure appetite by counting and tracking bites of food taken during a meal. Number of bites taken during a meal has been correlated with the number of kilocalories consumed, suggesting that bites can be used as a proxy measure for food intake.32 In one study, participants had to initiate and terminate the “bite-count” mode on the wrist-worn device by pressing a button at the beginning and end of their meal.32 Video and infrared camera technology have also been studied as a way to automate the bite-counting process. One study combined a Microsoft Kinect motion-sensing device with a Samsung SUR40 tabletop touchscreen computer33 but noted that a wrist-worn sensor was needed to augment the infrared technology in order to obtain sufficient sensitivity and precision in bite detection.

A commercial technology that has not been academically studied but that merits mention is the “smart” Samsung Family Hub Refrigerator. The refrigerator contains three cameras that allow remote viewing of the food inside, and supports apps that can keep track of expiration dates and order groceries.34 Hypothetically, these features may help quantify frequency of ordering groceries or the amount of expired food disposed, both of which could indicate an altered appetite.

Changes in Energy Levels

Outside of self-report measures, it is difficult to objectively quantify a person’s energy level. While EMA can help improve the quality of self-report data, GPS sensors can enhance our ability to objectively analyze gross motor activity and movement between locations. A systematic review identified that patients with depression may show less daytime motor activity than healthy controls.35 Specific variables that have been mapped using GPS include circadian movement (regularity in 24-hour movement patterns), location variance (variance in movement, or overall mobility), and entropy (how time was distributed across different locations). These features, as well as the percentage of time spent at home and the number of location clusters that an individual visited, were found to be correlated with depressive symptom severity.36 Additionally, one study found that the correlation was greater on weekends than weekdays, presumably because individuals had fewer social constraints and therefore more control over their own actions.37 One study supplemented motor activity assessed by actigraphy with EMA to allow subjects to report their energy levels throughout the day and found increased motor activity was associated with increased subjective energy levels later in the day.38 The relationship was bidirectional, as greater subjective energy levels were associated with increased motor activity later in the day. The study also found a positive association between higher motor activity and improved future self-reported mood level.


Like depressed mood, guilt is a self-reported emotion39 and difficult to quantify objectively. Thus, EMA may again serve as useful in more frequent and detailed assessment of feelings such as guilt and feeling like a burden on others, as well as the often-related emotions of regret, shame, and self-recrimination. In a heterogeneous illness like MDD, these nuances are important for understanding an individual’s experience of depression and in treating the individual case. Interestingly, a recent study notes that a growing number of therapists are incorporating information from patients’ electronic media (e.g., text messages, email, social media feeds) into the process of psychotherapy.40 Such information can be conceptualized as digital collateral information. This suggests that digital communications may be a means to gain more quantified information on subjective mood states, and that NLP may have a future role in their detection.

Loss of Concentration/Cognitive Changes

The measurement of concentration (or loss of concentration) and other cognitive changes have traditionally relied on standardized neurocognitive testing. Technology is now facilitating novel ways of measuring cognitive function with the use of web-based testing batteries,41 adaptive web-based testing,42 and smartphone-based testing.43 While DSM-5 lists loss of concentration as a primary criterion for MDD, depression may be accompanied by a host of cognitive deficits, including working memory, attention, biased self-referential information processing, and executive function. As technology helps cognitive testing to move from laboratory or office-based testing toward more naturalistic quantification, this field is in evolution. While technologies have simplified the process of testing and have reduced costs and time burden, the eventual goal of developing a way to estimate cognitive function based on changed real-world behavior remains conceptual.

More promisingly, technology can facilitate cognitive training, which may improve mood.44 The observation that interventions targeting cognitive symptoms of depression may actually improve mood is leading to a reevaluation of the interplay of mood and cognition and is advancing our understanding of depression.

Changes in Psychomotor Activity

Psychomotor activity, whether appearing as an increase or decrease in speech and movement, is a common behavioral manifestation of depression, especially in older adults. By breaking down both speech and motion into component variables, actigraphy, gait-measurement technology, and NLP can add great precision to measurement.

Actigraphy can measure acceleration magnitude and entropy, thus quantifying the suddenness, predictability, and repetitiveness of movements. This technique was used in a study of older adults with late-life depression, who displayed slower fine-motor movement than healthy older controls.45 Technology to measure gait provides a full-body picture of psychomotor speed and activity. In specially built environments (incorporating cameras, sensors, and walkways), it is possible to detect a variety of measures, including pressure, force, inertia, foot-to-ankle angles, stride length, and muscle tension, all of which give biomechanical information about the character of a patient’s gait.46


The ability to accurately measure and predict suicide risk (and lack of accuracy thereof) has been a long-standing challenge in psychiatry.47 A major focus of research in the technology domain has been the development of ways to more accurately predict suicidality. Several approaches have been tried, based on the assumption that large data sets could better objectively identify an individual person’s risk at a given moment, as well as serve as a more specialized target for resources. One study of service members in the U.S. Army created a machine learning–based algorithm to predict soldiers’ post-hospitalization suicide risk; over 50% of the suicides occurred in the 5% of patients having the highest predicted risk.48 A large, recent systematic review on existing suicide-prediction models found, however, a predictive rate lower than 1%, indicating that the field is a long way from viability.47 Recently, Facebook introduced a novel service, based partly on NLP, where it contacts the local emergency services for users of its platform who may have expressed suicide-related content. While the potential for such an approach is clear, it is limited by several factors such as lack of transparency about the analytic algorithm and a high rate of false positives.49


A rapidly growing field in psychiatry is that of digital therapeutics. While a complete discussion of this topic is beyond the scope of this commentary, technologies are facilitating a host of novel approaches to treatment, primarily by providing globally scalable access to non-pharmacological interventions.50 Depression is a primary focus of digital therapeutics, though challenges such as low rates of engagement and sustained use remain. Across the literature, the criteria used to measure acceptability, usability, and user satisfaction of digital therapeutics such as mobile phone applications are highly variable—which impedes our ability to evaluate the real-world efficacy of these technologies.51 However, with rapid advances in technology, especially the maturing of virtual reality and the potential for artificial intelligence–based interactions to supplement human interactions, this field is poised to influence care significantly over the next decade.50 Approaches such as computerized cognitive training have been shown to improve both mood and cognition, suggesting that technology-based interventions may be able to improve multiple aspects of neuropsychiatric functioning.52


Whether psychiatry as a whole is able to move toward a more digitally enabled model of care may eventually be determined by whether the personal data collected through digital methods can be obtained with clear consent while protecting privacy and securing such data.53 There is no current global standard, and existing regulations such as the General Data Protection Regulation in Europe have not been consistently implemented.54 Of the technologies discussed here, privacy policies of mobile apps have been the best studied. While the available evidence suggests that the majority of apps designed for depression care have privacy policies,55 details about how data collected by those apps is transmitted to third parties is much less transparent. Similar trends have also been reported in apps assessing other disorders, such as dementia.56 An ongoing challenge for clinicians and researchers is that most data used to create predictive or diagnostic information around depression from such technologies are likely created by persons outside the realms of Health Insurance Portability and Accountability Act protection, and that such data may therefore be controlled by third parties.57 In the near term, while the complexities of regulation and issues of data ownership, protection, and security are worked out, clinicians and patients may be best served by discussing these risks and weighing them against the benefits of digital information. Decisions as to whether the use of the technology is justified can then be made by the clinician and patient on a case-by-case basis by weighing privacy risks against the perceived clinical benefit. For instance, when considering the use of a commercial wearable fitness monitor to guide the management of disrupted sleep or apathy, it may not be immediately clear how the device manufacturer may be handling data captured by the device. In such a case, the clinical immediacy of the symptoms may be a significant factor to be weighed in deciding whether to deploy the wearable.


While the conceptual framework guiding the diagnosis of major depressive disorder remains anchored in DSM-5’s syndrome-based model, the clinical and research implications of assessing individual domains of the depression syndrome have long been recognized by psychiatry. It is also recognized that the current diagnostic standards may not capture conditions like subsyndromal depression. Recognizing this deficit in the current nosology of psychiatric conditions, the National Institute for Mental Health has promoted the use of the Research Domain Criteria (RDoC) to study mental disorders, which groups psychological constructs based on function. For the RDoC to transition from a theoretical model to an operationalized approach to patient care, digital tools will play a crucial role since they can provide dense data on specific behavioral domains, as we outline above. The early evidence suggests some success in predicting acute changes in individual domains of mood—in particular, by combining large volumes of data with sophisticated computational approaches. For example, in the context of bipolar disorder, emergent suicidal ideation may be identified through EMA data.58 While this finding bears extensive replication, especially in the context of a broader literature that points at low success in predicting suicidality, these early studies may nevertheless represent at least some degree of predictive validity. The array of new technologies that can already facilitate these approaches is growing rapidly.3 While some domains such as guilt, hopelessness, and depressed mood represent subjective emotional states, others such as poor concentration, psychomotor retardation, and disorders of sleep may be more easily quantified objectively. The level of evidence to support technology-based measurement of individual domains varies greatly; while actigraphy-based sleep monitoring is an established standard of care, our ability to predict suicidality remains well below the point where it can be clinically meaningful. Overall, clinical acumen remains the gold standard, but technologies can provide highly sophisticated digital collateral information to refine diagnosis and simultaneously create avenues for real-time intervention, often aided by technology such as apps. The ability to translate the potential of technology into regular clinical practice will furthermore depend on developing transparency and trust surrounding issues of data storage and privacy.

Declaration of interest

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the article.


1. Zimmerman M, Ellison W, Young D, Chelminski I, Dalrymple K. How many different ways do patients meet the diagnostic criteria for major depressive disorder? Compr Psychiatry 2015;56:29–34.
2. Vahia IV, Meeks TW, Thompson WK, et al. Subthreshold depression and successful aging in older women. Am J Geriatr Psychiatry 2010;18:212–20.
3. Torous J, Onnela JP, Keshavan M. New dimensions and new tools to realize the potential of RDoC: digital phenotyping via smartphones and connected devices. Transl Psychiatry 2017;7:e1053.
4. Onnela JP, Rauch SL. Harnessing smartphone-based digital phenotyping to enhance behavioral and mental health. Neuropsychopharmacology 2016;41:1691–6.
5. Schueller SM, Aguilera A, Mohr DC. Ecological momentary interventions for depression and anxiety. Depress Anxiety 2017;34:540–5.
6. Collier S, Monette P, Hobbs K, Tabasky E, Forester BP, Vahia IV. Mapping movement: applying motion measurement technologies to the psychiatric care of older adults. Curr Psychiatry Rep 2018;20:64.
7. Calvo RA, Milne DN, Hussain MS, Christensen H. Natural language processing in mental health applications using non-clinical texts. Nat Lang Eng 2017;23:649–85.
8. Pratap A, Atkins DC, Renn BN, et al. The accuracy of passive phone sensors in predicting daily mood. Depress Anxiety 2019;36:72–81.
9. Asselbergs J, Ruwaard J, Ejdys M, Schrader N, Sijbrandij M, Riper H. Mobile phone-based unobtrusive ecological momentary assessment of day-to-day mood: An explorative study. J Med Internet Res 2016;18:e72.
10. Kenny R, Dooley B, Fitzgerald A. Ecological momentary assessment of adolescent problems, coping efficacy, and mood states using a mobile phone app: an exploratory study. JMIR Ment Health 2016;3:e51.
11. Or F, Torous J, Onnela JP. High potential but limited evidence: Using voice data from smartphones to monitor and diagnose mood disorders. Psychiatr Rehabil J 2017;40:320–4.
12. Van Diggelen F, Enge P. The world’s first GPS MOOC and worldwide laboratory using smartphones. In: Proceedings of the 28th international technical meeting of the Satellite Division of the Institute of Navigation (ION GNSS+ 2015). Tampa, FL: Institute of Navigation, 2015:361–9.
13. Moskowitz DS, Young SN. Ecological momentary assessment: what it is and why it is a method of the future in clinical psychopharmacology. J Psychiatry Neurosci 2006;31:13–20.
14. Burke LE, Shiffman S, Music E, et al. Ecological momentary assessment in behavioral research: addressing technological and human participant challenges. J Med Internet Res 2017;19:e77.
15. Hartmann JA, Wichers M, Menne-Lothmann C, et al. Experience sampling-based personalized feedback and positive affect: a randomized controlled trial in depressed patients. PLoS One 2015;10:e0128095.
16. van Winkel M, Wichers M, Collip D, et al. Unraveling the role of loneliness in depression: the relationship between daily life experience and behavior. Psychiatry 2017;80:104–17.
17. Baltasar-Tello I, Miguelez-Fernandez C, Penuelas-Calvo I, Carballo JJ. Ecological momentary assessment and mood disorders in children and adolescents: a systematic review. Curr Psychiatry Rep 2018;20:66.
18. Depp CA, Schkade DA, Thompson WK, Jeste DV. Age, affective experience, and television use. Am J Prev Med 2010;39:173–8.
19. Dreisbach C, Koleck TA, Bourne PE, Bakken S. A systematic review of natural language processing and text mining of symptoms from electronic patient-authored text data. Int J Med Inform 2019;125:37–46.
20. Yazdavar AH, Al-Olimat HS, Ebrahimi M, et al. Semi-supervised approach to monitoring clinical depressive symptoms in social media. Proc IEEE ACM Int Conf Adv Soc Netw Anal Min 2017;2017:1191–8.
21. Eichstaedt JC, Smith RJ, Merchant RM, et al. Facebook language predicts depression in medical records. Proc Natl Acad Sci U S A 2018;115:11203–8.
22. Adekkanattu P, Sholle ET, DeFerio J, Pathak J, Johnson SB, Campion TR Jr. Ascertaining depression severity by extracting Patient Health Questionnaire–9 (PHQ-9) scores from clinical notes. AMIA Annu Symp Proc 2018;2018:147–56.
23. Ben-Zeev D, Scherer EA, Wang R, Xie H, Campbell AT. Next-generation psychiatric assessment: using smartphone sensors to monitor behavior and mental health. Psychiatr Rehabil J 2015;38:218–26.
24. Palmius N, Tsanas A, Saunders KEA, et al. Detecting bipolar depression from geographic location data. IEEE Trans Biomed Eng 2017;64:1761–71.
25. Prociow PA, Crowe JA. Towards personalised ambient monitoring of mental health via mobile technologies. Technol Health Care 2010;18:275–84.
26. Zhai L, Zhang H, Zhang D. Sleep duration and depression among adults: a meta-analysis of prospective studies. Depress Anxiety 2015;32:664–70.
27. Maglione JE, Ancoli-Israel S, Peters KW, et al. Subjective and objective sleep disturbance and longitudinal risk of depression in a cohort of older women. Sleep 2014;37:1–9.
28. Paudel M, Taylor BC, Ancoli-Israel S, et al. Sleep disturbances and risk of depression in older men. Sleep 2013;36:1033–40.
29. Vacas S, McInrue E, Gropper MA, et al. The feasibility and utility of continuous sleep monitoring in critically ill patients using a portable electroencephalography monitor. Anesth Analg 2016;123:206–12.
30. Levendowski DJ, Ferini-Strambi L, Gamaldo C, Cetel M, Rosenberg R, Westbrook PR. The accuracy, night-to-night variability, and stability of frontopolar sleep electroencephalography biomarkers. J Clin Sleep Med 2017;13:791–803.
31. Looney D, Goverdovsky V, Rosenzweig I, Morrell MJ, Mandic DP. Wearable in-ear encephalography sensor for monitoring sleep. preliminary observations from nap studies. Ann Am Thorac Soc 2016;13:2229–33.
32. Scisco JL, Muth ER, Hoover AW. Examining the utility of a bite-count–based measure of eating activity in free-living human beings. J Acad Nutr Diet 2014;114:464–9.
33. Manton S, Magerowski G, Patriarca L, Alonso-Alonso M. The “smart dining table”: automatic behavioral tracking of a meal with a multi-touch-computer. Front Psychol 2016;7.
34. Eadicicco L. Behold Samsung’s new $5,800 smart refrigerator. Time 2016; May 4.
35. Burton C, McKinstry B, Szentagotai Tătar A, Serrano-Blanco A, Pagliari C, Wolters M. Activity monitoring in patients with depression: a systematic review. J Affect Disord 2013;145:21–8.
36. Saeb S, Zhang M, Karr CJ, et al. Mobile phone sensor correlates of depressive symptom severity in daily-life behavior: an exploratory study. J Med Internet Res 2015;17:e175.
37. Saeb S, Lattie EG, Schueller SM, Kording KP, Mohr DC. The relationship between mobile phone location sensor data and depressive symptom severity. Peer J 2016;4:e2537.
38. Merikangas KR, Swendsen J, Hickie IB, et al. Real-time mobile monitoring of the dynamic associations among motor activity, energy, mood, and sleep in adults with bipolar disorder. JAMA Psychiatry 2019;76:190.
39. Steinmetz SE, Gray MJ, Clapp JD. Development and evaluation of the Perpetration-Induced Distress Scale for measuring shame and guilt in civilian populations. J Trauma Stress 2019;32:437–47.
40. Hobbs K, Monette P, Owoyemi P, et al. Incorporating electronic and social media data into patient care: a clinician survey. J Med Internet Res 2019;21:e13218.
41. Dodell-Feder D, Ressler KJ, Germine LT. Social cognition or social class and culture? On the interpretation of differences in social cognitive performance. Psychol Med 2019:1–13.
42. Skandali N, Rowe JB, Voon V, et al. Dissociable effects of acute SSRI (escitalopram) on executive, learning and emotional functions in healthy humans. Neuropsychopharmacology 2018;43:2645–51.
43. Guimond S, Keshavan MS, Torous JB. Towards remote digital phenotyping of cognition in schizophrenia. Schizophr Res 2019;208:36–8.
44. Calkins AW, McMorran KE, Siegle GJ, Otto MW. The effects of computerized cognitive control training on community adults with depressed mood. Behav Cogn Psychother 2015;43:578–89.
45. O'Brien JT, Gallagher P, Stow D, et al. A study of wrist-worn activity measurement as a potential real-world biomarker for late-life depression. Psychol Med 2017;47:93–102.
46. Muro-de-la-Herran A, Garcia-Zapirain B, Mendez-Zorrilla A. Gait analysis methods: an overview of wearable and non-wearable systems, highlighting clinical applications. Sensors 2014;14:3362–94.
47. Belsher BE, Smolenski DJ, Pruitt LD, et al. Prediction models for suicide attempts and deaths: a systematic review and simulation. JAMA Psychiatry 2019.
48. Kessler RC, Warner CH, Ivany C, et al. Predicting suicides after psychiatric hospitalization in US Army soldiers: the Army Study To Assess Risk and Resilience in Servicemembers (Army STARRS). JAMA Psychiatry 2015;72:49–57.
49. Barnett I, Torous J. Ethics, transparency, and public health at the intersection of innovation and Facebook’s suicide prevention efforts. Ann Intern Med 2019;170:565–6.
50. Fairburn CG, Patel V. The impact of digital technology on psychological treatments and their dissemination. Behav Res Ther 2017;88:19–25.
51. Ng MM, Firth J, Minen M, Torous J. User engagement in mental health apps: a review of measurement, reporting, and validity. Psychiatr Serv 2019:70:538–44.
52. Motter JN, Pimontel MA, Rindskopf D, Devanand DP, Doraiswamy PM, Sneed JR. Computerized cognitive training and functional recovery in major depressive disorder: A meta-analysis. J Affect Disord 2016;189:184–91.
53. Torous J, Roberts LW. Needed innovation in digital health and smartphone applications for mental health: transparency and trust. JAMA Psychiatry 2017;74:437–8.
54. Donnelly M, McDonagh M. Health research, consent and the GDPR exemption. Eur J Health Law 2019;26:97–119.
55. Huckvale K, Torous J, Larsen ME. Assessment of the data sharing and privacy practices of smartphone apps for depression and smoking cessation. JAMA Netw Open 2019;2:e192542.
56. Rosenfeld L, Torous J, Vahia IV. Data security and privacy in apps for dementia: an analysis of existing privacy policies. Am J Geriatr Psychiatry 2017;25:873–7.
57. Glenn T, Monteith S. Privacy in the digital world: medical and health data outside of HIPAA protections. Curr Psychiatry Rep 2014;16:494.
58. Depp CA, Thompson WK, Frank E, Swartz HA. Prediction of near-term increases in suicidal ideation in recently depressed patients with bipolar II disorder using intensive longitudinal data. J Affect Disord 2017;208:363–8.

data; depression; psychiatry; smartphone; technology

© 2020 President and Fellows of Harvard College