Q: What are the challenges in conducting data analysis studies, and where did these modes of analysis succeed?
Especially if you want to make a meaningful impact in one of the most complex sectors, the healthcare sector. Various stakeholders are involved in the healthcare sector. Especially in the United States, where healthcare is highly decentralized, it is highly regulated in areas such as data collection and data use. Analytics-based solutions that help one part of this sector can harm others, making it very difficult to find the best global solutions for this sector. Finding a data-driven approach that can impact society is therefore no easy task.
And the implementation presents various challenges. In my lab, we can design advanced machine learning and AI algorithms with superior performance. However, if they are not implemented in practice, or if the recommendations they provide are not followed, there will be no concrete impact.
For example, in some recent experiments, we found that the algorithm we designed outperformed specialists in one of the major hospitals in the United States. Interestingly, when we provided the algorithm-based recommendations to doctors, doctors gave less weight to the advice they got from the algorithm, and ignored it when treating their patients, although they could have outperformed them. I knew it was hot.
Next, we considered how to remove this obstacle. Combining human expertise with recommendations provided by algorithms not only makes doctors more likely to value algorithmic advice, but also integrates better than both the best algorithms and human experts. I found that I also get recommendations for
Similar challenges were also seen at the policy level. For example, we developed advanced algorithms trained on massive amounts of data to help the Centers for Disease Control and Prevention improve opioid-related policies. The opioid epidemic killed more than 556,000 people in the United States between 2000 and 2020, but officials are fully aware of what they can do to effectively control this deadly epidemic. not. Our algorithm made recommendations that we believe are better than the CDC. But, again, the key challenge is getting the CDC and other authorities to listen to these good recommendations.
We don’t want to imply that policy makers and other authorities are always against these algorithm-driven solutions, but their usefulness is consistently underestimated and, in fact, disregarded. I think there are many.
Q: What are your thoughts on the role of oversight and regulation in this area of new technology and data analytics models?
It is important to impose appropriate regulations. However, it is a piece of paper. New tools and advancements must be protected from misuse, but regulation must not prevent these tools from reaching their full potential.
As an example, in our published paper, National Academy of Medicine By 2021, the use of mobile health (mHealth) interventions (primarily enabled by advanced algorithms and smart devices) will enable healthcare providers, industry, and governments to develop more efficient ways to deliver healthcare. I explained that it is increasing rapidly all over the world because we are asking for it. Despite technological advances, increasingly widespread adoption, and support by leading voices from the medical, government, financial, and technology sectors, these technologies have not reached their full potential.
Part of the reason is that there are scientific challenges that need to be addressed. For example, as we describe in our paper, mHealth technology uses more sophisticated algorithms and statistical experiments in determining how best to adapt treatment content and delivery timing to the user’s current context. design should be used.
However, various regulatory challenges remain, including how best to protect user data. In his 2019 statement, the Food and Drug Administration encouraged “the development of mobile medical apps (MMAs) that improve healthcare,” but “oversees the safety and efficacy of medical devices, including mobile medical apps. He also stressed public health responsibility. Striking a balance between encouraging new development and ensuring that such development follows the well-known principle of “do no harm” is not an easy regulatory task.
Ultimately, two things are needed: (a) advances in the underlying science, and (b) well-balanced regulation. Once these are met, the possibilities for using advanced analytical science methods to solve persistent social problems are endless.
—
Banner art by Gremlin/Getty Images
