Hodges started off his presentation with a light-hearted joke about how his mother does not deem his research work on how to apply artificial intelligence in the healthcare industry, as equally important to when he is mending a broken leg, or conducting a surgery.
However, Hodges explained to the audience that the work they are doing with developing AI is actually just as important, as it will be very beneficial for the future of healthcare. “AI solutions enable average doctors to perform like the best doctors”, Hodges said. Working as ‘just a doctor’ often includes having a private practice, rising costs and the data gathered often has apparent value, but there is a lack of tools and strategy to make the most of this data.
This is where AI comes to play, Hodges explained, looking at three different use areas. The first is the Simple Rules Engine, that provides the doctor with the ability to identify, prioritise and gain prompt patient follow-ups. It also provides the staff of the medical practice more time to deliver care, rather than doing mundane time-consuming tasks.
The second is Analytic Products, the “next generation enterprise data warehouse”, which enables increased analysis which again becomes increased data challenges. The third is Real-time Processing that is very beneficial as data availability and IT requirements are exploding. Real-time processing allows the data infrastructure and data sharing to improve.
Hodges mentions that they can receive up to 10,000 requests per day, which is the equivalent of up to 100 fax pages of unstructured data per request. This amount of data takes time to analyse, hence why AI is very beneficial to this industry.Magellan wants to close the gap between the average and best, applying machine learning to allow the machine start, but the human to finish the research, and create a rising tide in your labour force.
Magellan’s partner is Hindsait, artificial intelligence for better healthcare with a team of rocket scientists, medical professionals, statistical analysts and engineers that can use their expertise to transform healthcare.Magellan reviews hundreds of thousands of requests, and the reviews are conducted by trained doctors and nurses. AI and historical data is leveraged to “score” patient records on likelihood of medical necessity.
Scored records are routed to the most appropriate reviewer for each case, and top clinicians are focused on the toughest cases. What Hindsait does, Hodges told us, is that it works in three different stages, applying natural language processing, machine learning and predictive analysis.
The natural language processing works with analysing “free texts” from Doctor’s notes, extracts clinical concept and quantities, negation and detection and it is integrated with medical ontologies and lexicons. Machine learning learns from historical outcomes data, ingests clinical appropriateness criteria/guidelines/business rules and provides continuous learning from newer datasets.
The predictive analytics scores PA requests in near real time, where the predictive score ranges from 0-1, and predicts PA requests approvability based on appropriateness criteria.
The results are performance change and cost savings, and to summarise it provides consistency across physicians, more efficient use of time and skills and discovery of outliers. This means that physicians make different decisions than their peers given the same fact pattern.
“The metrics can be tied to immediate reduction in spend, down stream cumulative savings and benefits from operational efficiencies”, Hodges said. Impact estimates from applying AI is 25-100% improvement in prevention of unnecessary health service, $1.4 million saved from 25,000 PA monthly requests, and the average time to process have gone from 3.35 days, to 1.03 day.