Scroll to top

Association of Selective Serotonin Reuptake Inhibitor Use with Abnormal Physical Movement Patterns as Detected Using a Piezoelectric Accelerometer and Deep Learning in a Nationally Representative Sample of Noninstitutionalized Persons in the U.S.


Heinz MV, Price GD, Ruan F, et al.


To evaluate differences in physical movement among individuals treated with selective serotonin reuptake inhibitors (SSRIs) compared to a control group and to characterize features of movement-related adverse effects associated with SSRI treatment.


This cross-sectional study of a population-based sample of non-institutionalized persons in the U.S. (N=7,162) evaluated the intensity of body movement recorded by an accelerometer worn on the hip for one week. Use of SSRI medicationswasself-reported; adherenceanddosage was not available. Existing techniques were used to smooth and reshape the movement data. Modeling started with a simple logistic regression model test followed by the construction of a deep learning model capable of encoding time series data to compare performance. The deep learning model included multiple convolutional-long short- term memory layers (Conv-LSTM). To account for possible confounding, a parallel deep learning model incorporated participant depression scores and movement data.

  • The sample had a mean age of 33.7 years, 51.7% were female, and 39.9% identified as White, 27.0% as Black, 25.5% Mexican American, 4.7% other or multi-racial, and 2.9% other Hispanic. 3.7% of the sample were taking a SSRI.
  • Overall, fair model performance was demonstrated, with all models showing moderate sensitivity and specificity, high negative predictive value, low positive predictive value, and high stability.
  • On average, the results showed significantly less movement in the SSRI treated group compared to the control group, even when controlling for depression.
  • The SSRI group also showed slower morning increase in movement and a slower evening decrease in movement suggesting less well-defined sleep-wake boundaries.
  • This is the first large-scale study of SSRI use and movement collected in an ecologically valid and naturalistic way, demonstrating that passively collected data can be useful for characterizing and exploring adverse effects of medications using time series deep learning models.
  • Results show associations between SSRI use and movement, as well as the existence of a movement pattern characteristic of SSRI use.
Read More