The healthcare sector has been working on implementing artificial intelligence (AI) to improve data mining. Over the years, AI has proved increasingly more valuable to data mining. Yet it still hasn’t reached what many thoughts was its full potential. Less-than-stellar results are now causing some to openly wonder if there is a way to combine AI with human data abstraction to improve results.
The question is worth exploring when you consider our tendency to think of data abstraction and AI as mutually exclusive. We have a tendency of thinking it must be one of the other. We either abandon AI and find ways to improve human data abstraction, or we abandon human input and go all-in on AI.
As with most things related to technology, the best results for healthcare data mining probably reside in a hybrid system. AI algorithms perform mundane tasks that do not require a ton of human intervention. Meanwhile, human data abstractors handle those complex problems that computer algorithms cannot deal with.
More About Data Abstractors
Understanding why AI has produced less-than-stellar results in healthcare data mining requires understanding what a human data abstractor does. Data abstractors go through endless volumes of medical data to find specific information. They are often tasked with reconciling conflicting information or correcting errors. They also input information as well.
A data abstractor in a hospital might scour patient records looking for reasons explaining billing errors. An abstractor working for a private practice may spend some time entering data from paper medical forms into the practice’s computer system. Data abstractors can perform any number of tasks depending on the work environment.
Abstraction Requires Logical Thought
On the surface, it might seem like most data abstraction tasks could be performed by computers and software. But digging a little deeper reveals that the job requires logical thought. Consider the fact that there are tens of thousands of codes utilized for medical billing. The same procedure could be coded multiple ways depending on the circumstances surrounding it.
One of AI’s shortcomings in this regard is its inability to correctly sort out coding issues. Computer algorithms cannot account for the nuances that influence how medical providers code certain procedures. It takes a human brain to figure those nuances out.
AI Can Do a Lot
None of this is to say that AI doesn’t have a place in healthcare data mining. It actually has a very big place. AI can do a lot with mountains of data that humans would require too much time to mine. For example, consider the physician database offered by iMedical Data. The database provides professional and contact information for thousands of physicians across the country.
This is qualified, first-party data voluntarily offered by the physicians themselves. It is data that recruiters can use in their search for doctors. Healthcare marketers can use the database to market to private practices, group practices, etc. How can AI help?
An AI-enabled system can scour that physician’s database in search of candidates that meet specific criteria for recruiting. What it would take a recruiter days to do can be accomplished by AI in just a few minutes. The AI system mines and analyzes data to produce a list of the most qualified candidates for the recruiter.
AI is a great tool for analyzing research data and predicting healthcare trends. It is a great tool for gathering and reporting. But there are some things it cannot do. For that reason, the best model moving forward is a hybrid model that combines both human data abstraction and AI capabilities.