Research Day
MIXED METHODS ANALYSIS OF OPEN ENDED RESPONSES ON A PEDIATRIC BEHAVIORAL RATING SCALE
Document Type
Abstract
Date
2021
Abstract
Objective: Child behavior rating scales often contain both forced-response and open-ended prompts, however there is little data on the utility of analyzing the open-ended responses. This preliminary study explores mixed-methods analysis of the sentiment of open-ended responses (concerns about behaviors, strengths of child) and correlation with the composite scores on the Behavior Assessment System for Children - Third Edition (BASC-3).
Methods: This retrospective cross-sectional study included 174 subjects from an eligible pool of 185 caregivers and teachers of children/adolescents under 18 years old. The participants were referred to an academic behavioral health clinic of an academic health system and completed BASC-3 assessments over a period of 18 months. Two human raters rated the sentiment of open-ended responses. T-scores were calculated for the Behavioral Symptoms Index, Internalizing Problems, Externalizing Problems, and Adaptive Skills Composite, and subsequently classified as "clinically significant", "at risk", or "average". Spearman correlation was used to correlate sentiment with T-scores via and simple linear regression was used to evaluate the linear relationship between sentiment and each T-score of interest.
Results: The authors found a positive correlation between the Adaptive Skills and Strengths comment sentiment. Additionally, the Concerns comment sentiment demonstrated greater correlation with Externalizing Problems than with Internalizing problems. Word count did not significantly impact the correlation.
Conclusion: Analysis of the sentiment of open-ended responses can complement the quantitative data of behavior rating scales and contribute to clinical decision making.