Implementing Psychophysiology in Clinical Assessments of Adolescent Social Anxiety: Use of Rater Judgments Based on Graphical Representations of Psychophysiology

Abstract

Social stressor tasks induce adolescents' social distress as indexed by low-cost psychophysiological methods. Unknown is how to incorporate these methods within clinical assessments. Having assessors judge graphical depictions of psychophysiological data may facilitate detections of data patterns that may be difficult to identify using judgments about numerical depictions of psychophysiological data. Specifically, the Chernoff Face method involves graphically representing data using features on the human face (eyes, nose, mouth, and face shape). This method capitalizes on humans' abilities to discern subtle variations in facial features. Using adolescent heart rate norms and Chernoff Faces, we illustrated a method for implementing psychophysiology within clinical assessments of adolescent social anxiety. Twenty-two clinic-referred adolescents completed a social anxiety self-report and provided psychophysiological data using wireless heart rate monitors during a social stressor task. We graphically represented participants' psychophysiological data and normative adolescent heart rates. For each participant, two undergraduate coders made comparative judgments between the dimensions (eyes, nose, mouth, and face shape) of two Chernoff Faces. One Chernoff Face represented a participant's heart rate within a context (baseline, speech preparation, or speech-giving). The second Chernoff Face represented normative heart rate data matched to the participant's age. Using Chernoff Faces, coders reliably and accurately identified contextual variation in participants' heart rate responses to social stress. Further, adolescents' self-reported social anxiety symptoms predicted Chernoff Face judgments, and judgments could be differentiated by social stress context. Our findings have important implications for implementing psychophysiology within clinical assessments of adolescent social anxiety.

ACKNOWLEDGMENTS AND FUNDING

This work was supported, in part, by an internal grant from the University of Maryland at College Park (College of Behavioral and Social Sciences Dean's Research Initiative) awarded to Andres De Los Reyes. This work was also partially supported by a Predoctoral National Research Service Award to Sarah Thomas from the National Institute on Drug Abuse (F31-DA033913). We thank Anna Swan, Michael Van Wie, Ho-Man Yeung, and William Lechner for assisting with data collection. We also thank Martin Buschkuehl and Jeffrey S. Chrabaszcz for assistance with Chernoff Face and coding procedures.

Notes

¹The composition of family ethnicity/race exceeds 100% because there was overlap among the ethnic/racial categories (i.e., participants could select more than one category).

Note: All heart rates that we report in this table are in beats per minute. The highest mean heart rate values used to create the proportion values represent the largest mean heart rate value within the corresponding segment of the task (e.g., Baseline Minute 1) across all patients included in the sample.

Note: All heart rates that we report in this table are in beats per minute. The highest mean heart rate values used to create the proportion values represent the largest mean heart rate value within the corresponding segment of the task (e.g., Baseline Minute 1) across all patients included in the sample.

Note: These values represent mean proportion values for a single patient and the normative control values to which we compared patient values, based on 1-min segments of each context (i.e., baseline, speech preparation, and speech-giving). See Figures 1, 2, and 3 for illustrations of how these values map onto actual Chernoff Face comparisons.

²Originally, coders judged features on patients' Chernoff Faces as wider, narrower, or equal, relative to normative control Chernoff Faces. This resulted in at least one coder (i.e., master or reliability coder; or both coders) making 32 “equal” judgments. To have complete data on all 264 Chernoff Face judgments, we had coders return and make “wider” or “narrower” judgments on the 32 comparisons originally judged as equal. Thus, our data are based on the 264 wider/narrower judgments. It is important to note that the accuracy rate for the master coder's 232 judgments (i.e., all judgments originally judged wider/narrower) was only slightly higher (i.e., 89%) than the final rate of 86% (i.e., after coders recoded judgments originally coded as equal).

³For comparative purposes, we calculated the mean heart rates (in beats per minute) for the three assessment contexts, using the mean heart rates across the 4 min used to create patient Chernoff Faces. These mean heart rates were 73.94 for baseline, 77.06 for speech preparation, and 89.33 for speech-giving. These rates approximated the ranges of heart rates for assessments taken during baseline and stressful assessment contexts in prior work (e.g., Allen et al., 2007; Gonzalez et al., 2011; Miers et al., 2011; Weems et al., 2005).

Note: The reference code for the dependent variable (i.e., dichotomous Chernoff Face codes for adolescent heart rates at baseline, speech preparation, and speech-giving contexts) was 0. Context factor contrasts based on comparisons of factors in descending order, with “Baseline” serving as the reference. The Context factor (coded in ascending order) was coded Baseline, Speech Preparation, and then Speech-Giving. The adolescent gender factor (coded in ascending order) was coded Male and then Female. OR = odds ratio; B = unstandardized beta; SE = standard error; 95% CI = 95% Wald confidence interval.

Note. bl = blank. Blank cells correspond to other Chernoff Face feature options in R. The template above assigns patient values to the four facial features used for the current study (i.e., width of face, width of mouth, width of eyes, width of nose). However, additional features are available.

Note. The labels and main title generated for the pdf output of the Chernoff Face images may be modified from the code above.