A plus sign can make a lot of difference. A doctor found this out when he hired a data scientist to create an algorithm—one that could analyze his notes and determine which patients with breast cancer had a high risk of recurrence after treatment.
The algorithm, adapted from one originally used by Facebook, disappointed the doctor with mediocre results. Then he noticed the program removed punctuation before analyzing his notes, a major problem because he used plus signs to indicate specific types of breast cancer. After modifying the algorithm to keep punctuation, its accuracy skyrocketed.
“You wouldn’t expect a data scientist to understand the importance of punctuation in clinical notes,” says Kendall Kiser, a fourth-year student at McGovern Medical School at UTHealth, who likes to tell the story to help others understand the importance of medical experts knowledge about health informatics. “This shows how you can’t just reuse an algorithm that Facebook created that was not built for health care.”
Kendall would not have understood algorithms at all before the fall of 2017, when during his third year of medical school he happened to attend a lecture by Jiajie Zhang, PhD. Zhang enthralled Kendall with the applications of artificial intelligence to medicine—specifically how an algorithm could distinguish between melanoma and harmless skin lesions with the accuracy of a board-certified dermatologist.
“He framed it by showing how these technologies will empower less specialized health care personnel to perform more specialized tasks,” says Kendall.
After doing his own research, Kendall decided to postpone his medical education to earn a master’s degree in biomedical informatics from UTHealth School of Biomedical Informatics. He worked with curriculum directors to condense the two-year program into one, graduating in May 2019.
“Now in addition to the technical skillsets I gained, I can speak the languages of medicine and information technology together,” he says. “That positions me as an interpreter between data scientists and the practitioners who will use their software.”
After he completes medical school, Kendall plans to specialize in radiation oncology—a technology-intensive field that uses advanced systems to target radiation beams at tumors while sparing healthy tissue. He anticipates his informatics background will help him communicate with software developers and perhaps create his own software based on clinical needs.
“Radiation oncologists draw lines to designate treatment areas,” he says. “Artificial intelligence algorithms are approaching the point of being able to do this, which would free physicians to focus on other aspects of care.”
In the medical field as a whole, Kendall sees tremendous potential for a doctor-technology bridge in electronic health records. In the last 10 years, these digital patient records have gained widespread use among physicians, who rely on them for nearly every aspect of documentation. However, some doctors find the systems ineffective and cumbersome, opening an opportunity to improve electronic health records with additional input from clinicians.
In a way, it illustrates the broader reason he earned a degree in informatics: to be a connection between technology—with all its potential—and the real people who use it every day.
“We need the voices of physicians, nurses, and every clinical provider who uses these systems,” he says. “Clinicians have an incredible opportunity to help shape the way we implement health information technologies in medicine.”
“Informatics holds vast potential in terms of what it can do for medicine,” says Zhang. “Having clinicians who also understand data science will really help bring that potential to life.”
