In many cases of medical research, the problem is not one of gathering sufficient data. Existing medical records of patients can provide a rich source of information that can be mined for useful insights. Fortunately, artificial intelligence (AI) tools are becoming more sophisticated and powerful, capable of revealing valuable correlations that might have been missed in past analyses.
For example, the Look AHEAD (Action for Health in Diabetes) program studied 5,000 overweight and obese patients with Type 2 diabetes, tracking them for up to 13 years. It sought to find out if moderate weight loss could lead to reduce mortality rates and serious health events such as heart attacks and strokes. The program was ended early by the National Institutes of Health (NIH) because there was no evidence of reduced cardiovascular events.
Then the data science researchers at The Arnhold Institute for Global Health of the Icahn School of Medicine at Mount Sinai took a second look at the data. They used machine learning technology to analyze the study results. They found that there was indeed a clear benefit from weight loss for 85% of the study group. The problem with the original analysis was that for the remaining 15%, weight loss actually led to worse health outcomes. Furthermore, the new AI analysis was able to identify characteristics of this sub-group with the negative results. They were patients whose diabetes was mild or well-controlled, tended to have a negative view of their health, poor compliance with exercise intervention, and lower improvement in other areas such as mental health and blood pressure.
Not only does this new analysis indicate that weight loss is indeed a good idea for most patients with Type 2 diabetes, it also makes it possible to identify the one out of seven patients who are not likely to be helped by weight loss. This is an important step to advance the concept of personalized medicine, which will help physicians and other healthcare professionals choose the most effective course of treatment based on individual characteristics. The result will be better outcomes for patients overall.
