Quantifying Clinical Subjectivity: Text-Derived Agitation Indices to Assess Emergency Psychiatric Interventions

• In Event: Open Refereed Roundtables
  In Refereed Roundtable (90 minute): Table 13: Stress and Mental Health

Abstract

The objective of this study is to evaluate the appropriateness of involuntary intramuscular (IM) medications for psychiatric inpatients in an agitated state by bridging the gap between clinical text and patient behavior. Race sits at the heart of many studies on emergency interventions in psychiatric inpatient environments, from physical restraints to IM therapies. While previous research explores these disparities using standard sociodemographic and diagnostic variables, it rarely controls for the severity of the behavioral disturbance that triggered the psychiatric chemical intervention orders.

Currently, there are several systematic evaluation tools designed to objectively measure agitation in real time. However, when psychiatric facilities lack a standardized system to consistently record these measurements, unstructured clinical notes become the primary record of patient behavior. To address this gap, we created composite agitation indices derived from these notes to systematically evaluate time-sensitive, on-the-ground patient assessments. Using a five-year retrospective dataset of adult psychiatric inpatients, this study employs four concurrent text-mining techniques (dictionary matching, sentiment analysis, TF-IDF, and Latent Dirichlet Allocation) to build a reliable behavioral control variable. Furthermore, over five hundred unique psychotropic medications were categorized into four behavioral tiers (maximum restraint, emergency escalation, negotiated high-arousal, and minimal intervention) anchored in three clinical standards (The Maudsley Prescribing Guidelines, NICE NG10, and ACEP policies).

Anchoring the analysis in Labeling Theory and Focal Concerns Theory, we view maximum restraint medications as potential institutional mechanisms of social control deployed to manage perceived danger. We hypothesize that after controlling for the text-derived agitation index, racial and ethnic disparities in medication administration will become statistically insignificant. Additionally, using separate multivariable models, we hypothesize that the duration of treatment effectiveness will correlate with the assigned potency of the psychotropic medication. By quantifying clinical subjectivity, this research isolates whether demographic disparities reflect objective patient behavior or unequal clinical perception and documentation practices.

Authors

• Yalun Zhang, Harverd University, VA Boston Health Care

• Kuniko Chijiwa, Riverside University Health System

• John David Reitzel, California State University San Bernardino