Developing a Healthcare Equity Tool: Combining Technology and Stakeholder Engagement to Uncover Hidden Bias in Patient-Provider Interactions

Academic Health 2021


Research Objective: Hidden biases in patient-provider interactions can result in poor communication in clinical interactions and healthcare disparities. The UnBIASED Research Project is combining stakeholder engagement with computational sensing to automatically detect hidden biases and provide automated feedback on communication quality during clinical encounters. Our aim is to develop technology that will be used in provider training to improve patient-centered communication and quality of clinical interactions. We focus on interactions between primary care providers and people who are Black, Indigenous, People of Color (BIPOC ) and/or LGBTQ+ people. Input of key stakeholders to inform technology design is critical.
Study Design: We created a three-phase approach to develop technology that makes hidden bias visible in clinical interactions. First, we engaged patients, providers, and other key informants in virtual interviews to report experiences of poor clinical interactions due to bias to help inform the design of automated feedback. Second, we are auto-assessing verbal and non-verbal social signals that are associated with implicit bias in patient- provider communication using previously recorded real-world clinical interactions. We use social signal processing (SSP), a computational approach that analyzes and synthesizes social signals observed in interpersonal interactions. Social signals refer to verbal and nonverbal communication behaviors, such as what is said and how it is said through body movement and placement, facial expression, or voice (e.g. talk time, interruptions, tone). SSP will assess behavior cues from multimodal audio, video, and body tracking data collected from patients and providers interacting during clinical encounters. Finally, we will combine our design and SSP efforts to develop and test in the lab and field a healthcare equity tool that makes hidden bias visible in clinical interactions and gives feedback to providers and patients.
Population Studied: Our study sample consists of primary care providers in California and Washington State. We are interviewing key informant patients who are BIPOC and LGBTQ+, primary care physicians and other providers like Cultural Case Mediators (CCMs) to better understand elements of poor patient-provider communication, and experiences of unsatisfactory clinical interactions.
Principal Findings: Early insights from stakeholder engagement highlight the importance of nonverbal communication as social signals of hidden bias. We interviewed CCMs as key informants regarding their observation of critical incidents of hidden bias that contributed to poor clinical interactions. CCMs reported that providers’ body language and actions can negatively impact clinical interactions. Examples include; the provider moving back from a patient conveyed a fear of aggression from the patient; a seated patient experienced a provider’s hand on their shoulder as a condescending gesture; time pressure negatively impacted communication; lack of greetings from the provider inhibit trust building; and lack of awareness of patients’ social circumstances is a barrier to communication.
Conclusions: Identifying cues associated with hidden bias and unsatisfactory patient- provider clinical interactions is an important first step to develop a healthcare equity tool that will detect and provide feedback on hidden biases in clinical interactions.
Implications for Policy and Practice: An automated tool to detect and give feedback on hidden biases in clinical interactions has the potential to improve patient-provider communication and quality of healthcare.

2021 Academy Health Virtual Annual Research Meeting
Nadir Weibel
Nadir Weibel
Associate Professor of Computer Science and Engineering
Steven Rick
Steven Rick
Ph.D. Candidate