Active Research in the Computationally-Enhanced Health Care Lab

Our Lab is composed of a multidisciplinary team whose research efforts focus on the collaborative challenge of reimagining ambulatory care using AI and computer technology.  Our research merges engineering with clinical medicine and communication science.

Artificial Intelligence (AI) in Medicine

  • REDUCE - How can translate outpatient clinic visit interactions into data and action to support patient care?

  • Automating Messaging -  How can large language models (LLMs) transform patient-provider communication?

  • The Observer Project - How can we create a FAIR data repository for sociotechnical ethnography and engineering science?

Our lab also is exploring research to improve Message Reach and Impact and to enhance EHR Cognitive Support.

AI in Medicine

This project is rooted in the idea that a deeper understanding of the interactions between healthcare providers and patients can significantly improve patient health outcomes and the efficiency of healthcare services. Our goal is to explore how we can automate the actions based on an encounter. This project is called REDUCE

In addition, we plan to use the data we collect for REDUCE to create a comprehensive digital archive of clinical visits called the Observer repository.

The Potential of AI with Multimodal Data

AI can revolutionize healthcare analytics and patient care when fed rich, multimodal data from a comprehensive repository. AI algorithms can analyze complex datasets encompassing video, audio, and textual information to uncover subtle yet crucial patterns in patient-provider interactions. This analysis can lead to developing advanced diagnostic tools, predictive models for patient outcomes, and personalized treatment plans. AI can also assist in identifying effective communication strategies for healthcare providers, optimize clinical workflows, and even detect early signs of conditions that might be missed in traditional analyses. AI can train healthcare providers by offering insights into effective patient communication and care strategies. Integrating AI with a comprehensive repository of healthcare interactions opens a new frontier in healthcare innovation, promising more accurate diagnoses, tailored treatments, and improved patient care and provider efficiency.


Reduce Logo

Goal: Develop a system that can qualitatively, quantitatively, and equitably summarize clinical encounters, freeing healthcare providers from manual documentation's burdensome and time-consuming tasks. REDUCE (Reimagining Documentation Using Computation from Clinical Encounters) aims to detect and analyze spoken and unspoken elements of patient care, identify gaps in symptoms and treatment, and generate AI-enabled visit documentation. This approach is expected to enhance cognitive support for providers, improve the efficiency and satisfaction of clinical visits, and ensure more equitable care by identifying and addressing previously unnoticed patient needs in real time.

Talks: Visit REDUCE talks to learn more through pre-recorded presentations.



Reduce Timeline

Funding: Supported by the National Institutes of Health.

Goal: Investigate the potential of cutting-edge large language models, such as GPT, to generate high-quality responses in complex and diverse patient-provider communications. Through a targeted survey of Penn primary care providers, we aim to validate this potential and refine our machine-learning model for improved accuracy in distinguishing between AI-generated and genuine patient messages. The ultimate objective is to utilize AI to enhance the efficiency and quality of patient-provider interactions and significantly reduce healthcare providers' workload, thereby enhancing the overall healthcare experience for both providers and patients.

Goal: Address the critical challenge of enhancing diversity in clinical research, recognizing the need for medical advancements to be assessed using a population that reflects demographic diversity. Given the historical underrepresentation of groups such as black and brown-skinned individuals in clinical studies, our collaboration with esteemed collaborators  Andy Tan, PhD, MPH, MBA, MBBSJessica Fishman, PhD / the Message Effects Lab, and Susan Furth, MD, PhD focuses on empirically investigating whether personalized appeals can effectively increase their willingness to participate in clinical trials. By overcoming trust barriers in the recruitment process, this initiative aims not only to improve current methods but also to lay the groundwork for leveraging data science and AI in automating and optimizing recruitment messages in the future, facilitating a more inclusive and equitable research landscape.

Funding:  Supported in part by the Penn Medical Communication Research Institute of the University of Pennsylvania.

Goal: Address the persistent challenge of integrating computer-based patient records effectively into healthcare settings to enhance provider decision-making without exacerbating burnout, a prevalent issue associated with the use of electronic technology in patient care. Despite numerous technological and informational advancements across academia and industry, the healthcare sector requires assistance to fully realize these innovations' potential benefits. In collaboration with CMS and following the publication of a viewpoint in JAMA, our lab is committed to developing "SMARTER" guides. These guides aim to assess and quantify how these advancements can provide cognitive support to healthcare providers, reducing cognitive load rather than adding to it during care episodes. Through partnerships with private and public entities, we aim to bridge the gap between potential technological benefits and practical application in healthcare, ultimately enhancing patient care and improving provider experience.

Improving Cognitive Support without Adding Cognitive Load

Observer Logo

Goal: Establish a centralized repository for securely storing and sharing multimodal visit-related data, such as video, audio, transcripts, surveys, computer logs, and sensor information collected from clinical encounters. By aggregating and helping make unique datasets available, the repository intends to support the discovery of novel insights into patient-provider interactions, predictive models for patient outcomes, and enhance the understanding of communication dynamics within healthcare settings. Ultimately, the goal is to leverage these insights to improve patient care, inform healthcare policies, and drive medical research and practice innovations.

Dashboard: In development...

Observer aims to utilize outcomes from the REDUCE project to aid other researchers in anonymizing and adding their data to the repository, making it accessible for further research.