Scientist and clinician Andrea Sikora, PharmD, MSCR, FCCP, FCCM, has joined the Department of Biomedical Informatics (DBMI) at the University of Colorado School of Medicine as an associate professor of biomedical informatics, a position she says she is excited about because it is going to open new opportunities to translate academic findings to the bedside.
"The ability to connect diverse disciplines from critical care to machine learning to pharmacy with the goal to improve patient outcomes is an exciting opportunity," says Sikora. The work has a multi-faceted approach to improving medication use, including prediction modeling for intensive care unit (ICU) complications, integration of medications into artificial intelligence (AI) methodology, and ICU workload prediction.
Sikora shared that her vision centers on "data-driven comprehensive medication management (CMM) for all patients." She explained, "CMM is the data-driven component that involves leveraging AI and clinical decision support systems to improve our ability to make individualized decisions at the bedside." She added, "Currently, CMM is not uniformly accessible to all patients in the U.S., even within large academic medical centers, and information technology tools to support this type of cognitive service are lacking."
"Dr. Sikora brings a research program focused on providing each critical care patient with the right medications at the right times with the right dosages for them, based on a multitude of complex data points. She also has thought and written deeply about mentoring, which will enhance the training landscape of our department," adds Casey Greene, PhD, chair of DBMI.
In 2022, Sikora was awarded a $1.86 million R01 Grant from the Agency for Healthcare Research and Quality (AHRQ). With this funding, she is working to provide data-driven, optimal pharmacotherapeutic care for every critically ill patient. Sikora hopes to develop a clinician-friendly health IT tool that will visualize AI-informed predictions from the perspective of the critical care pharmacist, who will ultimately prevent those adverse drug events.
Leveraging data and AI to improve clinical decision making
Clinicians need detailed information to accurately interpret medications—like dosage, formulation, frequency, and patient-specific data. Sikora shared, "High-dimensional data often poses difficulties for traditional modeling techniques." She explained that when extracting medication data from electronic health records (EHR), the information can be "messy and unstructured." She pointed out that she is particularly interested in making medications machine-readable, allowing for better integration into AI systems. The integration of common data models and ontologies designed for inpatient medication use are important next steps.
Sikora is investigating how to incorporate drug data into predictive modeling to enhance clinical decision-making in the ICU. For instance, if a clinician considers administering fluids to a patient, knowing the patient's risk for fluid overload could significantly inform that decision. She explained, "If you have a septic, critically ill patient and you are considering giving them more fluids vs. starting vasopressors, it would be nice to provide the team with data-driven risk predictions, 'The decision to give fluids is associated with a 60% chance of fluid overload but reduces acute kidney injury risk by 80%.' So many clinical decisions are made heuristically, but combining objective risk predictions with clinician judgment could improve outcomes."
Sikora shared that she is excited to be at the University of Colorado Anschutz Medical Campus (CU Anschutz) because of the combination of the progressive academic medical center and advanced informatics tools, including sepsis alerts at UCHealth. She shared, "These are incredibly neat tools that can predict sepsis before its onset and provide early intervention. The next step, though, is thinking about how to include data in the treatment process for antibiotic and fluid selection, when to start pressors, and all of the other decisions we make at the bedside."
Data-driven workload prediction and optimization
An important challenge to ICU care is thinking about clinician workload. How does the ICU manage patient surges (like bringing up backup attendings and teams) or even something more routine like a pharmacist taking time off? Overload can lead to "suboptimal care for patients." Sikora is part of the OPTIM study, which has enrolled over 36,000 ICU patients, making it one of the largest workload studies for healthcare teams ever. She and the team of researchers recently completed data collection and preprocessing and expect to publish findings soon. They hope to use AI methods to develop predictive modeling to optimize workloads.
Future plans for the Sikora Lab
Sikora looks forward to many things as she transitions to CU Anschutz, "I'm excited about the potential for collaboration across various departments and specialties, computer scientists, biostatisticians, intensivists, critical care pharmacists, and more. I recently had a wonderful experience meeting with members of the pulmonary division, where we discussed topics like antibiotic stewardship and ICU practices, and the Virtual Health Center at UCHealth gave me a tour. There is a lot of innovation happening here."
Collaborating with the software engineering team in DBMI is also at the top of Sikora's list. She said, "One of my goals is to build software tools that enhance medication management in the ICU, including the MRC-ICU tool." She added, "I am eager to blend clinical insights with technical capabilities to advance our efforts."
Having written books on mentorship and the power of healthcare teams with the UGAC3, she is passionate about team science and creating an innovative and caring lab: she looks forward to integrating diverse specialties to improve medication use and patient outcomes.
