Matching data (structured records) such as patient, customer, account, company or product has existed for several decades. The two prominent technologies that have addressed this issue are rules-based (aka deterministic) and probabilistic. Neither of these approaches, however, have achieved the level of accuracy required in many industries like healthcare. This is underscored by the fact that the match solutions available today use these technologies and create "tasks" otherwise known as potential duplicates. This is their inability to automatically solve 10-20% of matches and instead create tasks for human review teams to address manually. These older technologies don't posses the ability to adapt to the nuance required for these harder match problems.
Machine learning does posses the ability to deal with this nuance and it does it very well. In fact it's technology is exactly nuanced based.
Here's an article that highlights the use of machine learning in healthcare and how EntityWise helped St Joseph with our machine learning solution.