BMI PhD students Erica DePasquale, PJ Van Camp, and Lei Liu
PhD students Erica DePasquale, PJ Van Camp, and Lei Liu recently presented their research from the Biomedical Informatics Practicum course.

The Biomedical Informatics Practicum is a project-oriented course that uses state-of-the-art techniques and current data sets to improve healthcare. For Biomedical Informatics PhD students, it serves as a cornerstone experience in the transition from coursework to hands-on research.

Scientists begin their careers by learning theoretical concepts and basic skills and transition to performing their own research. The Practicum course is designed to help students do exactly that, with experienced faculty guiding them through the process of applying the knowledge they have gained to solve pressing health problems. 

The Concept

“Transition.” This word stuck with Graduate Program Director Jarek Meller, PhD, as he considered the journey of his PhD students. How could they gain more hands-on experience, stepping into the role of medical informaticists?

In fall 2016, Meller, Michael Wagner, PhD and Brett Harnett, MS-IS developed the idea for the Biomedical Informatics Practicum.

“The idea is to give the students experience with electronic health record (EHR) data, which can be noisy and dirty and have all sorts of problems,” says Wagner, a course co-director. “In theory, everyone is using these systems—and producing lots of data. They tend to think, ‘Oh, it’ll be easy to analyze this.’ But in reality, problems and questions arise. That’s where medical informaticists come in. We help physicians get a realistic picture of what can and cannot be answered with EHR data.”

The Course

This year’s practicum course focused on EHR data from patients with a diagnosis of asthma, chronic obstructive pulmonary disease (COPD) and related diseases.

The data comes from UC Health’s clinical repository, Epic Clarity, where students do not have full access. However, the UC Center for Health Informatics (CHI) has created a local repository that is limited in size, but contains the tables and fields needed for students to execute queries, create subsets, and analyze appropriately. Over 30,000 patients with a diagnosis of asthma are represented in the dataset—an extremely large and robust cohort.

In the course, students first review literature from previous studies that used similar data. They attend lectures by local experts on asthma and COPD, gaining background on treatment and established EMR data.

This research culminates in the next step: Developing a hypothesis of their own. Students build on previous studies to add new analytical ideas, formulating research questions to be explored through the EMR data.

“We chose asthma because it is so prevalent,” says Wagner. “Due to HIPPA regulations, we had to use a limited dataset, so students couldn’t see notes or names. This made it important for the dataset to be large—if you can’t go into the details, it becomes more difficult to resolve research questions. Due to the robust cohort, however, we were able to overcome these challenges.”

The asthma dataset was also supplemented with structured data from a variety of other sources, giving students another dimension of insight to formulate their research questions. When studying asthma and COPD, many factors come into play: Where does the patient live? How is the air quality there? Does greenery dominate their surroundings, or do highways?

Publically available data from the census bureau, Environmental Protection Agency, and Geographic Information System allowed students to identify environmental exposure values.

After the data was extracted and presented in a user-friendly format, it was time to ask questions.

The Projects

Each student worked on their own project, though instructors encouraged collaboration and discussion of ideas. In the end, evaluation was based on research creativity, practical problem solving and strength of research conclusions.

For his project, PJ Van Camp was inspired by Dr. Cole Brokamp’s lecture on an application he developed to link patients to air pollution exposure. Van Camp decided to use the dataset to find a link between air pollution and asthma exacerbations. He was particularly interested in studying how daily fluctuation of air pollution levels affect people differently, based on the air quality of their environment or their smoking habits.

Lei Lui’s research interest is machine learning, so she was intrigued by Dr. Muhammad Zafar’s lecture on COPD superusers (patients who have the highest frequency of re-admittance after initial medical care for COPD.) Lui decided to develop a machine learning model to predict COPD superusers from EMR data using key patient characteristics. Through early identification of these superusers, preventive interventions can be made to address care failures and reduce medical costs.

After hearing Dr. Jonathan Bernstein’s lecture on markers for asthma and COPD subtypes, Erica DePasquale became interested in the role eosinophils (disease-fighting white blood cells) play in the progression and severity of disease. Her project highlighted significant differences in demographics and comorbidities of asthma, COPD, and ACO (asthma-COPD overlap syndrome) patients with differing levels of eosinophils. Understanding the role of eosinophils in these diseases can generate new hypotheses and improve patient outcomes.

The Takeaways 

Although the practicum course has ended, the resulting research is still ongoing. Van Camp is partnering with Dr. Anh Dao to analyze complex EMR data for allergen research. Lui and Dr. Zafar are planning to expand the cohort definition and try more data processing techniques in future analysis. DePasquale is following up with Dr. Bernstein to finalize her data and deliver the results in a format that could be used for future publications.

These ongoing partnerships between students and researchers are a testament to the value of hands-on mentorships begun in the course.

“We were really fortunate to have a group of six guest lecturers who all generously agreed to help with mentoring students,” says Wagner. “The students were able to share their ideas, and mentors could help them decide whether or not they were actionable. They worked together to form research questions that were practical and realistic. Students learned to speak the language of clinical researchers.”

Even the instructors learned something new—Harnett says he was intrigued by new insights on how air pollution and proximity to major roadways can drive health outcomes. “It was great to see that the students actually learned and applied many of the principles from my intro class,” he says. “The clinical mentors were critical. They helped educate students on core issues and coach them through the research.”

Next year, course organizers would like to expand the dataset, including pediatric patients or other diseases. They also hope that more clinicians will get involved, partnering with students to act as mentors. Once again, questions will arise, insights will emerge, and another group of students will become clinical researchers.

Tagged on: