Research: Your Walking Pattern Predicts Depressive Episode Risk Years in Advance

Wrist-worn sensors can predict incident depressive episodes up to 9 years before diagnosis by analyzing daily walking patterns. This groundbreaking UK Biobank study of 72,359 participants reveals that reduced activity, slower gait, and irregular walking patterns are powerful early indicators of mental health risk—opening new possibilities for early intervention through passive monitoring (Chan et al., 2023).

The Core Finding

Researchers from Neuroscience Research Australia and UNSW tracked 72,359 middle-aged and older adults (ages 40-73) for an average of 7.4 years after measuring their gait patterns using wrist-worn accelerometers (Axivity AX3). The median sensor wear time was 6 days (interquartile range: 6 to 7 days). During follow-up, 1,332 participants (1.8%) experienced incident depressive episodes.

The analysis revealed striking associations between walking patterns and future risk of depressive episodes. The strongest univariable predictors were:

Longer walk durations (51% decreased risk per standard deviation increase in median walk duration)
More daily steps (24% decreased risk per SD increase)
More variable gait (23% increased risk per SD increase in step-time variability)
Longer continuous walks (22% decreased risk per SD increase in longest walk duration)
More running activity (20% decreased risk per SD increase in log running duration)
Faster walking speed (both usual and maximal speed were protective)

In other words, people were significantly more likely to experience depressive episodes if they were less active, walked slower, had more variable gait, completed shorter uninterrupted walks, and had fewer arm movement patterns while walking.

After adjusting for all established risk factors—including age, lifestyle, sleep patterns, and medical conditions—three digital gait biomarkers remained independent predictors: no running activity (10% increased risk per SD), fewer steps per day (12% increased risk per SD), and irregular walking patterns (10% increased risk per SD).

As the study authors conclude: “We found people were more likely to develop depressive episodes if they were less active, walked slower, had more variable gait, completed shorter uninterrupted walks, and had fewer arm movement patterns while walking.”

Why This Matters

This research demonstrates that simple, passive monitoring through wrist-worn devices—technology that millions already wear—can identify risk of depressive episodes years before clinical symptoms emerge. The implications are profound:

Early Intervention Window: Detecting risk 7-9 years before diagnosis creates unprecedented opportunities for preventive interventions when they’re most effective.

Objective Assessment: Unlike self-reported symptoms or clinical questionnaires, digital gait biomarkers provide continuous, objective measurement of behavioral patterns that signal underlying mental health changes.

Accessible Screening: In this study, the method required only a wrist sensor worn for a median of 6 days—no blood tests, brain scans, questionnaires, or clinical visits. This makes population-scale mental health screening feasible for the first time.

Real-World Behavior: By capturing how people actually move in daily life rather than in clinical settings, wrist sensors reveal functional capacity more accurately than laboratory tests.

The Mechanisms: Why Walking Predicts Depressive Episodes

The study authors propose three pathways linking reduced mobility with future depressive episodes:

Biochemical Changes: Decreased physical activity leads to lower levels of vitamin D, brain-derived neurotrophic factor (BDNF), and key neurotransmitters (dopamine, endorphins, serotonin, noradrenaline). These biochemical changes may increase risk of depressive episodes years before symptoms appear.

Functional Loss: Mobility impairment can cause socio-psychological distress through loss of independence and functional status. This distress accumulates over time, eventually manifesting as clinical depressive episodes.

Shared Pathophysiology: Depressive episodes and mobility impairments may result from similar underlying processes. Gait disturbances can indicate cerebrovascular disease, and lesions in brain regions affecting gait (prefrontal cortex and basal ganglia) may also increase risk of depressive episodes. Chronic inflammation may lead to both mobility decline and incident depressive episodes.

What This Means for Digital Health

While this research demonstrates the predictive power of just 6 days of sensor data, Sahha goes beyond one-time assessments to provide continuous, live tracking and trend analysis over 90 days. This transforms static risk prediction into dynamic health monitoring.

This research validates what Sahha has built: behavioral intelligence systems that transform raw sensor data into actionable health insights. Our approach recognizes that mental health monitoring requires analyzing a broad range of biomarkers.

At Sahha, we provide:

Live tracking and trend analysis across 90 days, not just one-time snapshots
Personalized insights that detect changes from individual baselines, not population averages
Comprehensive behavioral analysis integrating movement, sleep, and circadian patterns
Multi-dimensional assessment revealing emerging risks through pattern changes over time

Our Mental Wellbeing Score incorporates movement patterns alongside other behavioral markers to provide validated mental health insights. When combined with our Sleep, Activity, and Readiness scores, developers can build applications that detect concerning patterns as they emerge—not years later.

The difference: This study showed what’s possible with 6 days of data. Imagine the insights from continuous monitoring that tracks trends, detects deviations from personal baselines, and provides actionable intelligence in real-time.

The Future of Mental Health Prediction

This study represents one of the largest investigations of digital biomarkers and mental health, with rigorous methodology including:

72,359 participants tracked for up to 9 years
Depression diagnoses based on ICD-10 clinical criteria (not self-report)
Comprehensive adjustment for sociodemographic, lifestyle, and health factors
Validated gait biomarkers showing strong agreement with clinical gold standards
Sensitivity analysis confirming findings weren’t due to reverse causality

The research demonstrates that passive, continuous monitoring of real-world behavior can predict mental health outcomes years in advance. As wearable technology becomes ubiquitous, we have unprecedented opportunities for early intervention—detecting risk when preventive measures can make the greatest difference.

For Researchers and Developers

If you’re building health applications or conducting research on behavioral health, Sahha provides the intelligence layer that makes this kind of prediction practical:

Live tracking and 90-day trend analysis for continuous monitoring
Personalized insights for end users based on individual patterns
Validated biomarkers from smartphones and wearables
Behavioral patterns that signal health changes
Easy API integration for rapid development
Research-backed scores developed with university partners

The gap between raw sensor data and actionable health insights is exactly where we operate. Rather than building data processing infrastructure from scratch, focus on what matters: using behavioral intelligence to improve health outcomes.

Learn more about Sahha’s platform | Explore our API documentation | Contact our research team

References

Chan, L.L.Y., Brodie, M.A., & Lord, S.R. (2023). Prediction of Incident Depression in Middle-Aged and Older Adults using Digital Gait Biomarkers Extracted from Large-Scale Wrist Sensor Data. Journal of the American Medical Directors Association, 24(7), 1106-1113. https://doi.org/10.1016/j.jamda.2023.04.008

Study Objective: “To determine if digital gait biomarkers captured by a wrist-worn device can predict the incidence of depressive episodes in middle-age and older people.”

Key Finding from Abstract: “The study findings indicate digital gait quality and quantity biomarkers derived from wrist-worn sensors are important predictors of incident depression in middle-aged and older people. In univariable models, we found people were more likely to become depressed if they were less active, walked slower, had more variable gait, completed shorter uninterrupted walks, and had fewer arm movement patterns while walking.”

Study Details:

Participants: 72,359 adults ages 40-73 from UK Biobank
Follow-up: Mean 7.4 ± 1.1 years (range 0.1-9 years)
Incident Depressive Episodes: 1,332 participants (1.8%)
Sensor: Axivity AX3 wrist-worn accelerometer
Wear Time: Median 6 days (interquartile range: 6 to 7 days)
Analysis: Cox proportional-hazard regression models
Outcome Definition: ICD-10 Code F32 (depressive episode)

Read the full paper