Deep Learning in Palliative care (DeepLeaP) Collaborative
Communication that clarifies patient care preferences is widely endorsed as a quality indicator, especially for frail and seriously ill patients. However, routine assessment of these conversations is nearly impossible because the information is documented within the free-text of clinical notes. We have developed AI methods to identify documentation of patient care preferences, including goals of care discussions and limitations on life sustaining treatments. In this project, we will implement these methods at two tertiary academic hospitals and three community hospitals in a regional healthcare system. We will analyze the results at the hospital, unit, and clinician levels and then design a template for producing a quality-measure report.
“My long-term career goal is to promote patient-centered care by combining the power of novel methods from computer science with clinical palliative care. Since joining the faculty at the Dana-Farber Cancer Institute in 2016, my research has focused on building computational methods to assess process and quality measures relevant to serious illness. I have built a cross-disciplinary lab with team members from Dana-Farber, Partners HealthCare, Harvard School of Public Health, and the Massachusetts Institutes of Technology. My lab has developed several methods based on natural language processing and deep learning to identify patients with serious illness and to assess established quality indicators, e.g., the documentation of goals of care discussions regarding critically ill older adults. Moving forward, my goal is to scale our methods to work on data from multiple electronic health record systems so that palliative care measures can be assessed within and across healthcare systems. The Sojourns Scholar Leadership Program is a unique opportunity that would afford me the time and resources to develop the skills necessary to successfully lead these efforts.”