A new peer‑reviewed study in JAMA Network Open reports that everyday signals from smartphone sensors — from how far we move to when we sleep and how often we charge our phones — can reveal broad and specific mental health risks. The research, conducted by teams at the University of Michigan, University of Minnesota, and University of Pittsburgh, followed 557 adults for 15 days and found shared behavioral patterns linked to a general risk dimension across mental illnesses, as well as distinct signatures tied to particular domains like social detachment or impulsivity. With more than nine in ten people in Thailand now online and mobile phones ubiquitous, the findings raise timely questions about how the kingdom could adopt “digital phenotyping” to spot trouble earlier while safeguarding privacy under the Personal Data Protection Act.
The study, summarized by University of Michigan communications and reported by outlets such as SciTechDaily, shows that major forms of psychopathology — mood, anxiety, and related conditions — are detectable from passive smartphone data. The authors demonstrated that people with higher general mental health risk tended to move less, stay home more, go to bed later, and keep their phone at a lower battery charge. “These findings suggest that major forms of mental illness are detectable from smartphone sensors, indicating that this technology could potentially be used for symptom monitoring and research on wide-ranging psychiatric problems,” said the senior author, a University of Michigan psychologist, in university coverage of the paper and subsequent reporting (JAMA Network Open; MedicalXpress; SciTechDaily).
For Thai readers, the significance lies in timing and fit. Thailand’s digital adoption is among Southeast Asia’s highest: in early 2025 there were 65.4 million internet users (91.2% penetration) and 99.5 million active mobile connections, equivalent to 139% of the population, with messaging app LINE alone reporting about 56 million monthly active users in-country (DataReportal, Digital 2025: Thailand). Meanwhile, mental health needs remain pressing. Recent coverage citing the Department of Mental Health’s suicide prevention center reported 5,217 suicide deaths in 2024, or roughly 15 a day, underscoring the urgency of earlier detection and support (The Nation Thailand). The Ministry of Public Health and National Health Security Office have integrated the 1323 mental health hotline into the Universal Coverage Scheme to widen access to help, building on a whole‑of‑society prevention approach highlighted by the World Health Organization (NHSO; WHO feature). The new research points to a complementary digital channel: ethically using the phones already in nearly every Thai hand to nudge support sooner — if privacy and consent are handled right.
The JAMA study’s premise is deceptively simple. Modern phones carry sensors that unobtrusively log activity, location, motion, phone call patterns, screen engagement, and battery status. Over two weeks in 2023, the team collected these “passive” signals every few seconds to minutes and summarized them into 27 daily behavioral features (for example, time spent at home, number and duration of calls made, total distance traveled, walking time, number and length of screen sessions, sleep window inferred from accelerometer data, and average battery level). Participants also completed a validated, dimensional mental health questionnaire that maps symptoms onto six broad domains: internalizing (e.g., depression and anxiety), detachment (social withdrawal), disinhibition (impulsivity), antagonism (hostility), thought disorder (psychosis‑related features), and somatoform (physical symptom distress). The researchers then used multilevel statistical models to connect the sensor‑derived behavior patterns to both the six domains and to a higher‑order “p‑factor,” a general dimension of psychopathology that researchers have debated for a decade (JAMA Network Open; background on the p‑factor: American Journal of Psychiatry 2014 via PMC).
What they found breaks new ground for digital psychiatry. First, the “p‑factor” — a common risk signal that cuts across multiple diagnoses — had clear, everyday correlates: lower physical mobility, more time spent at home, later sleep timing, and lower average phone charge. In plain terms, people at higher general risk tended to be more sedentary, socially less out‑and‑about, delayed in their sleep schedule, and less fastidious about charging — a constellation the authors argue could reflect broader emotional, motivational, or executive functioning challenges. Second, after accounting for this general factor, several domain‑specific markers emerged. Higher antagonism was associated with making fewer and shorter phone calls (a sign of reduced social initiative), detachment with less walking and more time at home (behavioral disengagement), disinhibition with lower average battery charge (a possible proxy for planning difficulties), internalizing with shorter, more frequent screen engagement bouts (potential worry‑checking or fragmentary interactions), and somatoform with less walking (consistent with perceived or real physical barriers to activity). Notably, thought disorder showed no unique markers once the general factor was considered, suggesting that passive movement and phone-use sensors may be less suited to cognitive‑perceptual disturbances, which might require language or voice analysis instead (JAMA Network Open).
These insights echo and advance prior work. The idea of “digital phenotyping” — quantifying human behavior in situ via personal devices — was first articulated in 2016 as a way to capture moment‑by‑moment patterns without burdening users, incorporating both “active” inputs (like surveys) and “passive” signals (like GPS and accelerometer data) (Wikipedia entry, summarizing Torous et al. 2016). Systematic reviews have found promise but also inconsistency in linking phone‑derived markers to specific diagnoses, in part because traditional diagnostic categories lump together varied symptom profiles that differ behaviorally and overlap across disorders. The new work’s dimensional approach helps reconcile that problem by separating a shared, general risk signal (p‑factor) from more nuanced domain signatures (JMIR 2023 systematic review). As the senior author put it in university and media coverage, progress had been “modest” partly because “most digital psychiatry work has not used what we know about how mental illness is organized within people when selecting targets to predict and monitor.” The ability, he added, “to use passive sensing to connect someone with help before things get really bad would have huge benefits, including better outcomes, reduced costs, and lower stigma” (MedicalXpress; SciTechDaily).
Thailand is unusually well‑positioned to explore this science because phones and chat apps are deeply woven into daily life. DataReportal’s 2025 update reports 65.4 million internet users (91.2% of the population), 51.0 million social media user identities (71.1%), and 99.5 million mobile connections — with LINE estimating about 56 million monthly active users nationally, a platform already used by hospitals, universities, and public agencies to reach citizens at scale (DataReportal, Digital 2025: Thailand). In practical terms, that means passive sensing could be architected to run largely on‑device for pattern detection, with opt‑in prompts connecting people to Thai services they already know — for instance, the 1323 hotline, community mental health clinics, or trusted hospital networks — without broadcasting raw data to the cloud. Such a design would align with core safeguards of the Personal Data Protection Act (PDPA), which specifically treats health information as “sensitive data” requiring explicit, informed consent, purpose limitation, and strong security, with rights for data subjects to access, correct, and withdraw consent (DLA Piper: Data protection laws in Thailand). Any pilot in Thailand would need to meet PDPA standards and likely go further — for example, by favoring on‑device analytics, minimizing data collection, using open‑source auditability for the detection algorithm, and creating clear red‑flag thresholds co‑designed with Thai clinicians and ethicists.
Still, the science is early, and the authors are explicit about limitations. The study analyzed 15 days of data, a relatively short window; the sample skewed female; some sensor streams (like detailed app usage or text content) were not captured; operating system differences were present; and passive signals can be influenced by context unrelated to mental health (for example, carrying a phone in a handbag vs a pocket). Importantly, missing data in some sensors were handled statistically but remain a real‑world challenge for any rollout. And, critically, the study does not “diagnose” — it maps associations between behaviors and symptom dimensions at the group level. That means any attempt to apply it must avoid overreach: smartphones may be useful as early “smoke alarms,” not definitive tests (JAMA Network Open).
Even with caveats, the public‑health logic is compelling in a Thai context. The WHO has urged a whole‑of‑society approach to suicide prevention in Thailand, integrating health services with law enforcement and community organizations. The Department of Mental Health and NHSO recognized the value of easy access when they folded the 1323 hotline into universal coverage, reducing barriers to care (WHO feature; NHSO). Passive smartphone sensing — if voluntary, transparent, and privacy‑protective — could help reach people earlier, especially younger Thais and working‑age adults who are mobile‑first and time‑poor. In 2021, UNICEF flagged high levels of stress, anxiety, and depression among Thai adolescents during the pandemic, reinforcing the need for low‑friction, youth‑friendly pathways into support. While conditions have evolved, the lesson remains: digital touchpoints matter for early engagement (UNICEF Thailand press release).
To understand what a Thai pilot might look like, imagine a voluntary feature inside a hospital’s existing LINE account. With explicit consent, the app would locally analyze a handful of privacy‑preserving indicators proven in the JAMA study — weekly mobility trends, sleep timing estimated from phone accelerometer, and average battery level — and, only if a sustained high‑risk pattern persists (for example, two or more weeks of low mobility plus progressively later sleep), gently prompt the user: “Would you like to chat with a counselor via 1323?” No raw GPS coordinates leave the device, and users can opt out anytime. Aggregate, anonymized statistics could help the Department of Mental Health track population signals without identifying individuals, aiding planning while respecting PDPA constraints (JAMA Network Open; DLA Piper: PDPA; DataReportal: Thailand).
For readers wary of false positives and stigma, the authors’ nuanced framing is helpful. The “general risk” pattern — moving less, staying home more, later bedtimes, lower average charge — is not a label; it’s a nudge to check in with yourself or someone you care about. Why might battery level matter? The researchers suggest it could be a proxy for planning and executive function — the sort of small life‑management friction that accumulates when people are struggling. Similarly, later sleep can be a cross‑cutting marker tied to mood, stress, and circadian rhythms, which many Thais already address with mindfulness, evening routines, or family support. Recognizing these signals early — before a crisis — is the point (JAMA Network Open).
Global researchers have long debated whether digital phenotyping can deliver dependable, equitable tools outside of labs. Reviews note mixed results when predicting specific diagnoses, underscoring the need for large, diverse samples and transparent validation. The new dimensional approach is a promising course correction, but it must be tested in Thai populations, across urban and rural settings, and in multiple languages and device types. On the privacy side, Thailand’s PDPA provides a robust baseline framework (consent, purpose limitation, data minimization, security, and user rights), yet mental health data demands even stricter practices — including on‑device processing, end‑to‑end encryption for any alerts, third‑party code audits, and independent ethics oversight (JMIR 2023 review; DLA Piper: PDPA).
Cultural fit also matters. Thai society’s strong family networks and community ties can be protective, but stigma still deters many from seeking formal help. Digital prompts delivered through familiar platforms like LINE may feel less threatening, especially if they lead to anonymous chats or text‑based counseling before a phone call. But such design choices should be co‑developed with Thai users, clinicians, Buddhist clergy engaged in community health, educators, and youth groups to ensure they respect norms and reduce, not amplify, stigma. The WHO’s “whole‑of‑society” guidance for Thailand emphasizes exactly this kind of inclusive design and partnership (WHO feature).
For health systems, the biggest opportunity may be not diagnosis, but triage and continuity. Many Thai patients struggle to sustain follow‑up after a clinic visit. Passive, consensual sensing could serve as a quiet safety net between appointments, flagging when support should step up, then stepping back when patterns normalize. It could also amplify community health volunteers’ impact by pointing them (with consent) to households that may need a check‑in. For universities and employers, opt‑in wellness programs could use similar signals to offer resources without accessing raw data, preserving trust.
Looking ahead, three developments are likely. First, technical: expect movement from raw sensor feeds toward composite, explainable metrics co‑validated with clinicians — for example, “weekly mobility variability” rather than latitude/longitude — processed on device to limit data exposure. Second, regulatory: Thai authorities may consider issuing PDPA‑aligned guidance specific to digital mental health, clarifying best practices for consent flows, risk communication, algorithm auditing, and the boundaries between wellness features and medical devices. Third, evidence: Thai universities and hospitals will need to run pragmatic trials — randomized or stepped‑wedge designs — to test whether these alerts actually increase help‑seeking, reduce symptom burden, or prevent crises in Thai settings. Publishing negative as well as positive results will be crucial to avoid hype.
What should Thai readers do now? A few practical steps:
If you’re curious about mental health apps or phone‑based wellness features, treat them as supportive tools, not diagnoses. Look for clear consent, minimal permissions, and the option to opt out anytime. Be cautious of apps that request precise location or contact lists without a compelling reason. Thailand’s PDPA gives you rights to access and withdraw your data — use them if needed (DLA Piper: PDPA).
Consider simple, evidence‑informed habits that align with the study’s signals: aim for regular sleep and wake times, gentle daily movement (even neighborhood walks), and small social check‑ins. If a loved one’s patterns shift — staying home, sleeping later, ignoring phone charge — try a compassionate conversation.
If you or someone you know is struggling, free help is available. The 1323 mental health hotline operates 24/7 and is now integrated into universal coverage. You can also reach out through hospital networks or local community health services (NHSO).
For clinicians and educators, watch this space. The JAMA study provides a rigorous, dimensional map of which passive behaviors matter and why. If your institution pilots digital supports, insist on on‑device analytics where possible, transparent algorithms, and clear opt‑in consent. Evaluate outcomes, not just engagement (JAMA Network Open).
For policymakers and developers, co‑design with Thai users and protect privacy by design. Favor on‑device processing; collect the least data needed; make algorithms auditable; communicate plainly to users; and establish independent oversight. Align designs with PDPA’s sensitive‑data rules — and go beyond the minimum where mental health is concerned (DLA Piper: PDPA).
This study does not end the debate over digital phenotyping; it re‑frames it. By showing that smartphones capture a general mental health “pulse” and some domain‑specific rhythms, it sets a realistic target for what phones can and cannot do: help us notice when patterns drift into risk and open a door to care. In Thailand, where smartphones are near‑universal and services like 1323 stand ready, the promise is not surveillance, but timely support — if we build with consent, context, and compassion.
Sources and further reading: The original research is open access in JAMA Network Open (July 3, 2025): “Passive Smartphone Sensors for Detecting Psychopathology,” DOI: 10.1001/jamanetworkopen.2025.19047 (JAMA Network Open; PMC version). University summaries and science news coverage include SciTechDaily and MedicalXpress, with press notices via EurekAlert!. For background on the p‑factor, see American Journal of Psychiatry 2014 (open via PMC). For digital phenotyping concepts and reviews, see the Wikipedia overview and a 2023 systematic review in JMIR (JMIR). For Thai digital adoption statistics, see DataReportal: Digital 2025 Thailand. For national suicide prevention context and services, see the WHO feature on Thailand’s approach (WHO) and NHSO’s notice integrating the 1323 hotline into universal coverage (NHSO). For PDPA guidance, see DLA Piper’s Thailand data protection overview.