The study collected over 4,000 potential risk factors and analyzed them using machine learning models. The models found that mental health issues like depression and anxiety were the most significant. Compared to models that only considered historical behavior, these models can more accurately predict future risks.