Hospitals started using AI to reduce the number of patients who suffer from dangerous falls during their hospitalization
Falls that occur in hospitalized patients are a widespread and serious threat to patient safety. El Camino Hospital in Silicon Valley set to reduce the number of patient suffering from dangerous falls during their hospitalization.
A software that predicts which patients are most likely to fall by deriving from Electronic Health Records (EHR) who are the patients with risk factors to fall and improving the predictive model with real-time tracking of patients.
Six months after the hospital implemented artificial intelligence technology, the rate at which patients suffered dangerous falls dropped 39 %.
Source: Beckers Hospital Review
The management of Lucile Packard Children's Hospital in Palo Alto were interested to predict the likelihood of delays to surgeries. Through this distribution of delay-likelihoods, decision makers could be prompted with options to adjust the schedule.
Implemented real-time prediction platform that allows real-time information about the provided treatments. Due to the real-time information regarding a surgery delay or a predicted run over decision maker can now act proactively versus reactively. Used data included the type of procedure, the surgeon's historical record and the patient's medical history.
10% reduction in case overruns and 15% decrease in case under-runs within four months of implementing the software. The percentage of cases delayed by more than 10 minutes decreased by 11 %, cumulatively resulting in 520 hours of delays prevented since the launch.
As banks go digital they face new levels of fraud, or fraud threats that have to be checked at very high speed. Often the major cost is turning away good business because fraud detection is too restrictive and turns down lucrative deals. Denmark’s Danske Bank was automatically identifying 1,200 potential frauds per day in its transaction monitoring whereas 99.5 % of them were falsely identified
after reviewing some of the anti-fraud software packages, the bank decided to build its own predictive model to better customise for their data landscape. The prediction model hereby leveraged the numerous data sources within the bank.
With a machine learning solution, the bank was able to reduce falsely identified frauds by 35 % and improve detection of actual frauds at roughly the same percent. When the solution was extended to base on deep learning, the numbers almost doubled to a 60 % reduction in mistakenly identified frauds and about 50% improvement in detecting actual fraud
Thanks to AI technology JPMorgan Chase saved 360 thousand work hours of lawyers with an automated system able to accomplish the same task at higher quality within seconds.
The bank had to spend 360,000 hours of work each year by lawyers and loan officers tackling a slew of rather mundane tasks, such as interpreting commercial-loan agreements for filing purposes.
An AI software was developed to parse financial deals that once kept legal teams busy for thousands of hours. It was trained on the results of real lawyers on many past loan agreements and now interprets them quicker than human lawyers.
The software has helped reduce loan-servicing mistakes that were often attributable to human error in interpreting 12,000 new contracts per year.
Approximately 7-10% of AXA’s customers cause a car accident every year. While most of these are small accidents involving insurance payments in the hundreds or thousands of dollars, about 1% are so-called large-loss cases, that require pay-outs over $10,000. It was important for AXA to understand which clients are at higher risk for such cases in order to optimize the pricing of its policies.
AXA’s R&D team in Japan initially developed a machine learning model to predict if a driver may cause a large-loss case during the insurance period. First, the team implemented a traditional machine-learning technique, called Random Forest, which led to a prediction accuracy comparable to random chance. The team then improved the accuracy by developing an experimental deep learning (neural-network) model and achieved 78% accuracy in its predictions.
The model achieved 78% accuracy in its predictions. This improvement could give AXA a significant advantage for optimizing insurance cost and pricing, in addition to the possibility of creating new insurance services such as real-time pricing at point of sale.
Increasing the precision of fraud detection systems.
Investigating health providers suspected of billing for fraudulent procedures can be a costly and time-consuming process, since there are relatively few bad actors: a typical needle-in-a-haystack problem.
A prediction model that scores and ranks cases by risk to direct investigators to those cases with the highest potential for fraud
Increased the fraud detection rate from 5% to 48% for the top 50 riskiest providers identified by the model. Hereby increased efficiency of investigative resources.