Activity in AI and aMCC of Clinicians Predicts Pain Assessment and Treatment via Stereotypes

2344 

Abstract Submission 

Theoni Varoudaki¹, Nikta Khalilkhani², Morgan Gianola³, ELIZABETH LOSIN²

¹Penn State Univeristy, State College, PA, ²Penn State University, State College, PA, ³University of Miami, Coral Gables, FL

Theoni Varoudaki  
Penn State Univeristy
State College, PA

Nikta Khalilkhani  
Penn State University
State College, PA

Morgan Gianola  
University of Miami
Coral Gables, FL

ELIZABETH LOSIN, Ph.D.  
Penn State University
State College, PA

Νon-Hispanic White populations are regularly over-prescribed opioids and other analgesics, which leads them to born a greater burden of the opioid epidemic, while minoritized groups and women are typically under-prescribed pain medication relative to clinical guidelines and experience poor pain management outcomes. One potential mechanism underlying this phenomenon is clinicians adjusting their prescription according to inaccurate stereotypes about the pain sensitivity and opioid abuse rates of different demographic groups or the idea that clinicians experience less shared pain when examining patients of a different demographics than their own, leading to lack of motivation to provide pain relief and an associated decrease in analgesic prescription. To understand the psychological and brain mechanisms underlying these health disparities, we conducted an fMRI study where clinicians watched vignettes of mock pain patients of different demographics.

65 medical trainees (age: 22-32 years old, 34 female, 31 male) attended a single 3-hour visit. During the visit, they had an fMRI scan where they saw profiles and short videos of participants experiencing evoked pain, meant to represent the clinical exam of 36 musculoskeletal pain patients (12 non-Hispanic African American, 12 non-Hispanic White and 12 Hispanic White). For each patient, clinicians made pain assessment and treatment decisions.

In order to understand how the pain expressions of patients and vicarious pain of clinicians contributed to the clinicians' decisions the following analyses took place: Pain facial expressions of the patients were analyzed using to draw conclusions about their effect on the clinicians' treatment decisions. A single-trial analysis focusing on two a priori regions (AI and aMCC) of interest from a meta-analysis on pain empathy was conducted. Z-scores of mean activation were extracted and added to a Structural Equation Model (SEM) to test the contributions of pain stereotypes, pain expression, neural indicators of vicarious pain (AI and aMCC activity), pain assessment, and treatment decisions in a single SEM model. The clinician-patient dyad was the unit of analysis with a total of 2,340 dyadic interactions.

We found that clinicians assessed the pain of their patients based on the pain sensitivity stereotypes that they held about the patient's demographic. Activation in the AI and aMCC predicted both the clinician's treatment decisions via the clinicians' sensitivity stereotypes and pain assessment. However, the two areas had opposite effects on assessment and treatment when controlling for all other variables, with higher anterior insula activation predicting lower pain assessment and higher pain treatment, and higher anterior medial cingulate activation predicting higher pain assessment and lower pain treatment, indicating that they may play separate roles in the clinicians' decisions. The clinicians' gender was also an important predictor of pain assessment, sensitivity stereotypes and activation of the anterior insula, with women assessing all patients as more sensitive to pain and their pain as higher, but having a lower activation of their anterior insula when observing patients in pain than men. This finding could suggest that compared to male clinicians, female clinicians' pain treatment decisions may be driven more by affective (previously associated with the aMCC) than cognitive empathy (previously associated with the AI). Finally, clinicians with more years of medical education were less likely to prescribe any analgesics to patients but assessed patients' pain as higher and exhibited lower activation in the medial cingulate cortex when observing patients in pain during the mock clinical exam.

The finding that AI activity predicted treatment decisions whereas aMCC predicted pain assessment, could indicate that clinicians may share the pain of their patients but make their treatment decisions based on their objective impression of it.

Emotional Perception

Social Interaction ²

Activation (eg. BOLD task-fMRI)

BOLD fMRI ¹

Perception: Pain and Visceral

Cognition

Data analysis

FUNCTIONAL MRI

MRI

Pain

Social Interactions

Statistical Methods

STRUCTURAL MRI