New York City-based Mount Sinai School of Medicine researchers developed an artificial intelligence-powered algorithm that can learn to read patient data from EHRs as accurately as the traditional, labor-intensive "gold-standard" method, the organization said Sept. 2.
The gold-standard method requires much more manual labor to develop and perform; currently, scientists use a set of established computer programs to mine medical records for new information. Storage and development of these algorithms is managed by a system called the Phenotype Knowledgebase. While the system is highly effective at correctly pinpointing patient diagnoses, it is time consuming.
The Mount Sinai researchers' new method, called Phe2vec, was built based on existing research and involves the computer learning on its own to spot disease phenotypes.
To build Phe2vec, a computer was programmed to scour through millions of EHRs and learn how to find connections between data and diseases. This programming relied on "embedding" algorithms that had been previously developed by other researchers to study word networks in various languages. Then the computer was programmed to use what it learned to identify the diagnoses of nearly 2 million patients whose data was stored in Mount Sinai Health System's records.
The research team compared the effectiveness between the new Phe2vec system and the old systems and found that Phe2vec was as effective as, or performed slightly better, than the gold standard phenotyping process at correctly identifying nine out of 10 diseases tested and recorded in the EHRs. A few of the diseases tested were dementia, multiple sclerosis and sickle cell anemia.