Simple, Transparent, and Flexible Automated Quality Assessment Procedures for Ambulatory Electrodermal Activity Data.

Academic Article


  • OBJECTIVE: Electrodermal activity (EDA) is a noninvasive measure of sympathetic activation often used to study emotions, decision making, and health. The use of "ambulatory" EDA in everyday life presents novel challenges-frequent artifacts and long recordings-with inconsistent methods available for efficiently and accurately assessing data quality. We developed and validated a simple, transparent, flexible, and automated quality assessment procedure for ambulatory EDA data. METHODS: A total of 20 individuals with autism (5 females, 5-13 years) provided a combined 181 h of EDA data in their home using the Affectiva Q Sensor across 8 weeks. Our procedure identified invalid data using four rules: First, EDA out of range; second, EDA changes too quickly; third, temperature suggests the sensor is not being worn; and fourth, transitional data surrounding segments identified as invalid via the preceding rules. We identified invalid portions of a pseudorandom subset of our data (32.8 h, 18%) using our automated procedure and independent visual inspection by five EDA experts. RESULTS: Our automated procedure identified 420 min (21%) of invalid data. The five experts agreed strongly with each other (agreement: 98%, Cohen's κ: 0.87) and, thus, were averaged into a "consensus" rating. Our procedure exhibited excellent agreement with the consensus rating (sensitivity: 91%, specificity: 99%, accuracy: 92%, κ: 0.739 [95% CI = 0.738, 0.740]). CONCLUSION: We developed a simple, transparent, flexible, and automated quality assessment procedure for ambulatory EDA data. SIGNIFICANCE: Our procedure can be used beyond this study to enhance efficiency, transparency, and reproducibility of EDA analyses, with free software available at
  • Authors

  • Kleckner, Ian R
  • Jones, Rebecca M
  • Wilder-Smith, Oliver
  • Wormwood, Jolie
  • Akcakaya, Murat
  • Quigley, Karen S
  • Lord, Catherine
  • Goodwin, Matthew S
  • Status

    Publication Date

  • July 2018
  • Keywords

  • Adolescent
  • Algorithms
  • Autistic Disorder
  • Child
  • Data Accuracy
  • Female
  • Galvanic Skin Response
  • Humans
  • Monitoring, Ambulatory
  • Quality Control
  • Reproducibility of Results
  • Signal Processing, Computer-Assisted
  • Digital Object Identifier (doi)

    Start Page

  • 1460
  • End Page

  • 1467
  • Volume

  • 65
  • Issue

  • 7