An international study led by RISE-Health demonstrated that simple statistical models can be more effective than advanced algorithms in cardiovascular risk prediction. The research compared 11 different methods of machine learning and regularized regression.
The researchers analyzed different approaches to evaluate which one could predict patient cardiovascular risk with greater accuracy. The results indicate that regularized regression techniques achieved performance superior or equivalent to the more complex machine learning methods.
This study questions the widely held belief that more sophisticated algorithms are necessarily more accurate in predictive medicine. The research suggests that the simplicity of traditional statistical models may be sufficient, and even advantageous, in certain clinical contexts.
The results have important implications for medical practice, since simpler methods tend to be more interpretable, easier to implement, and less costly than more complex artificial intelligence-based solutions.
