Subtle differences in cadence, step acceleration, and daily movement emerged outside the laboratory, offering a glimpse of how mental health symptoms may be reflected in everyday motion.
Study: Passively sensing anxiety and depression symptoms in daily life among middle-aged adults. Image Credit: africa_pink / Shutterstock
In a recent ‘Article in Press’ in the journal Scientific Reports, researchers investigated how anxiety and depression symptoms relate to motor patterns.
Anxiety and depression each affect about one-fifth of adults each year. Existing treatments are limited by high relapse and non-response rates. Many affected individuals exhibit gait and posture changes, with altered sensorimotor predictions and movement. Yet current treatments do not account for the altered sensorimotor control. Studies have found a slowing of gait speed with increasing depression severity.
In addition, anxiety has been associated with gait and posture changes. Previous studies have also linked anxiety to increased lumbar movement and postural sway. But research on the links between these conditions and changes in movement patterns has primarily involved laboratory-based tasks. For this reason, understanding whether these associations hold outside the laboratory setting is crucial for assessing the generalizability and ecological validity of these biomarkers.
About the study
In the present study, researchers investigated how depression and anxiety symptoms relate to motor patterns. Adults aged 40–64 years whose questionnaire scores met criteria for moderate-to-severe depression and/or anxiety on the Patient Health Questionnaire (PHQ)-9 and Generalized Anxiety Disorder Questionnaire-IV were recruited in the United States. Individuals with changes in treatment within the past 30 days, acute psychosis, bipolar disorder history, suicidal thoughts, or cancer history were excluded.
Of 100 adults who consented and enrolled, 92 entered the 90-day study and downloaded a mobile application that collected passive data, including phone acceleration, illuminance, angular velocity, and location. Accelerometer data were collected at random periods. Participants completed ecological momentary assessments (EMAs) once daily using the application. The EMA comprised validated symptom questionnaires: PHQ-8 and three anxiety questions adapted from the PHQ-Somatic, Anxiety, and Depressive Scales.
A subset of 25 participants wore hip and wrist actigraphs for seven consecutive days, with usable matched actigraphy and EMA data available for 17. The low actigraph sampling rate prevented calculation of cadence and peak height, so the analysis used summary activity measures. Gait metrics, such as gait cycle time, peak height, and cadence, were derived from the higher-frequency phone accelerometer data. The team fit a hidden Markov model (HMM) to peak height and cadence.
A transition probability matrix was derived from the HMM. The model estimated the likelihood that participants would remain in their current gait state or shift to another at the next time step. The relationship between gait state and mood was evaluated using omnibus likelihood ratio tests. Linear mixed effects models were used with hidden states as predictors of anxiety and depression.
Findings
Among the analytical samples, 57 participants contributed usable phone accelerometer and EMA data, while 17 contributed usable hip and wrist actigraphy data matched to EMAs. Most participants were female (89%) and White/Caucasian (89%). Four gait states were identified using HMM: nominal, elevated, high-amplitude, and low-amplitude. Peak height and cadence were moderate in the nominal gait state. In the elevated gait state, both were higher than in the nominal gait state.
In the high- and low-amplitude gait states, cadence was normal, but the peak height was higher and lower, respectively. Elevated, nominal, low-amplitude, and high-amplitude gait states were 98.9%, 83.9%, 72.2%, and 31.7% stable, respectively. The elevated state had the highest probability of remaining in the same state at the next modeled transition. Most frequent transitions were to the nominal or low-amplitude gait state.
Gait states defined by peak height and cadence were linked to anxiety and depression scores in the overall tests. In exploratory pairwise analyses, the elevated gait state was associated with lower depression scores than the nominal and high-amplitude states, whereas its difference from the low-amplitude state was not statistically significant. The pairwise analyses were not corrected for multiple testing.
The overall anxiety test was significant, but none of the individual pairwise comparisons reached statistical significance; elevated gait showed a trend toward lower anxiety scores. In the exploratory wrist analysis of 17 participants, higher anxiety scores were associated with lower movement counts and vector magnitudes.
Conclusions
Taken together, the study explored associations between data from varying sensors, such as phone, hip, and wrist accelerometers, and anxiety and depression symptoms in middle-aged adults. Phone-derived gait states varied with daily anxiety and depression scores, with the elevated state generally corresponding to lower symptom scores. Wrist accelerometer measures showed a relationship with anxiety scores in the small actigraphy subsample.
The study’s limitations include its small sample size and the inability to account for potential confounders (e.g., age, sex). The authors also cited the small actigraphy subsample, skewed sex and phone operating-system distributions, limited within-person data, and exploratory uncorrected pairwise comparisons. The study found that differences in anxiety and depression symptom severity corresponded to measurable variation in everyday movement patterns. These results support further investigation into passively sensed gait biomarkers as potential tools for clinical monitoring of anxiety and depression.
