In 2017, I worked as a summer researcher with the Department of Biomedical Informatics at UCSD. Although I had essentially no knowledge of medicine, healthcare, or even biology; I wanted to work there so that I could explore ways to use my Computer Science skills in order to help others.

My project for the summer was to work on a system that could help identify hidden stages for chronic diseases based on patient records. For example, certain chronic illnesses such as kidney disease and diabetes are not fully understood in the medical field. Many people with these diseases progress through a series of illnesses and symptoms that may highlight hidden stages that are currently unknown to medical professionals. People with chronic disease spend more time and money on healthcare than most other people do. According to the American Diabetes Association, the cost of diabetes was $327 billion in 2017 (http://www.diabetes.org/advocacy/news-events/cost-of-diabetes.html).

Naturally people with chronic diseases are affected by this, but they are not alone. Their families also face a huge emotional burden and often help to take on the financial costs. Additionally, a lot of the financial support comes from taxpayers. Doctors and other medical professionals are also interested in better understanding these diseases so they can prescribe better treatment to keep their patients from visiting hospitals and clinics so often.

Originally I thought that I was well suited to address the problem. I had technical skills, the data at hand, and prior experience working with patient records. I felt that through machine learning I could finding something that would make a huge impact. After working on the project, however, I feel that this is not a problem that I can solve alone. Patient records contain a vast amount of data, much of which was unknown to me. The implementation of algorithms to view and manipulate the data is just one small part of the puzzle. Without a true understanding of medicine, I had a hard time determining what to examine and how to interpret the results of what I found. Even medical professionals could not say with certainty which factors are the most important. I feel that this problem (like most) cannot be solved by any one individual, but must instead be solved by a team of people who have the necessary combined knowledge to understand the data and determine what might be most helpful to examine. In particular, it seems that people who have chronic diseases must also be part of the solution. Without fully understanding the factors that bring them in for extra medical care, and those that don’t, it is difficult to find the right solution.

Our proposal to use machine learning in conjunction with patient records introduces the risk of data leaks. Anytime patient records are stored and acted on, there is a risk that private information may be exposed. Perhaps just as dangerous, though, is the potential consequence of adding to a system where people are often the afterthought, and not the forefront, of a solution. Trying to articulate the problem, understand the data, and devise a solution without any knowledge of the domain suggests that simply having technical skills gives people the privilege to build systems they don’t fully understand. I know now that domain knowledge is much more valuable than a skill, and hope to demonstrate this in my future work.