What We Do
- Train and validate predictive and diagnostic algorithms using real-world data
- Quantitative and mixed-methods exploration to identify optimal use cases and implementation strategies
- Implementation and prospective evaluation of algorithm-enabled interventions
- Scaling algorithms for clinical impact
Palliative Care and Hospice Perspectives: Utilizing Mixed Methods
HACLab incorporates mixed-methods to better understand beliefs, experiences, and attitudes towards algorithm-based and digital health care delivery solutions. We integrate qualitative research via semi-structured interviews followed by detailed coding and analysis of transcripts to inquire about clincians' perspectives.
In the first of two white papers, HACLab researchers interviewed 23 clinicians from multiple specialties including hematology/oncology and family medicine to capture perspectives to design an effective virtual palliative care program. Key findings indicated that respondents felt that community-based palliative services (i.e. physical therapy, mental health counseling, and symptom management) would benefit their patients, but they had mixed feelings about automated referrals. Respondents were supportive of virtual palliative care for those having trouble accessing physician offices but felt that such care should only be provided after an initial in-person consultation. Respondents universally believed that more patients should have access to palliative care.
In the second white paper, we interviewed administrators and providers to learn how they envision palliative care and the relationship between hospice and palliative care. Respondents expressed that upstream palliative care is meaningful and can act as the “bridge” to hospice. Respondents thought that payers can facilitate uptake by broadening coverage of services providing more guidance in identifying eligible patients for palliative care; however, they were cautious about third party agencies acting as direct service provider. Automatic referrals triggered by objective measures can potentially increase access to upstream palliative care.
In the first entry of this series, we examined how algorithms can be biased on a variety of factors, including their underlying data, biased individuals creating models, or from statistical mischaracterizations from models themselves. In this entry, Caleb Hearn et al. explore how bias in algorithms is identified and quantified according to a few different statistical concepts and metrics.
Authors: Caleb Hearn, Sae-Hwan Park, Ravi Parikh
Updates from the HACLab for October 2023
Author: Will Ferrell
Algorithms leverage existing data to predict an outcome, using inputs that are associated with the outcome. One problem with the forthcoming tide of machine learning algorithms is that such algorithms can be biased. In this blog, we start exploring the concept of "Algorithmic Bias"
Author: Caleb Hearn
Algorithms are routinely used in the clinic to make decisions on patient care. Over time these algorithms may deteriorate in performance. Here, we start the explore the concept of “performance drift”.
Author: Likhitha Kolla
Updates from the HACLab for August 2023
Author: Will Ferrell
AI's role in our daily lives continues to become more omnipresent and healthcare is no exception. As AI becomes more routinely incorporated in healthcare, a central question becomes – how do we regulate it?
Author: Ravi Parikh
Caleb M. Hearn, MPH, CAPM will have a podium talk “Hospice Provider Perspectives on Providing Earlier Palliative Care for Patients with Serious Illness”
Jenna Steckel, MSW will have a poster presentation “Clinician Perspectives on Virtual Palliative Care for Patients with Advanced Illness”
Founded at the Perelman School of Medicine and the Abramson Cancer Center at the University of Pennsylvania, we are a laboratory focusing on the development, validation, implementation, and scaling of advanced algorithms in clinical care and health policy.
Author: Ravi Parikh
Real World Data?
Patient Generated Health Data?
A branch of AI and computer science which focuses on the use of data and algorithms to imitate the way that humans learn, gradually improving its accuracy (source:IBM)
The health and well-being of a person with cancer from the time of diagnosis until the end of life. This includes issues related to follow-up care, late effects of treatment, cancer recurrence, second cancers, and quality of life (source:cancer.gov)
Systematic and repeatable errors in a computer system that create unfair outcomes, such as privileging one arbitrary group of users over others. It also, occurs when an algorithm produces results that are systemically prejudiced due to erroneous assumptions in the machine learning process (source:FSU)
The data relating to patient health status and/or the delivery of health care routinely collected from a variety of sources- electronic health records (EHRs), claims and billing activities, pghd, product and disease registries (source:FDA)