PHOENIX—Early adopters of artificial intelligence solutions are beginning to see success in clinical areas such as predicting readmissions and avoidable emergency department visits, according to a joint report from KLAS Research and the College of Healthcare Information Management Executives (CHIME).
KLAS and CHIME polled early adopter healthcare organizations using AI software, specifically machine learning and natural language processing, to evaluate the gains they’ve achieved in clinical, financial and operational areas.
"The most exciting insight from our research is that artificial intelligence (machine learning and natural language processing) has truly begun to make a difference in healthcare. It’s not all just smoke," Ryan Pretnik, director of research and strategy at KLAS and co-author of the study, said via email. "Artificial intelligence is driving outcomes, saving patient lives, and driving operational and financial efficiencies for providers and payers."
Pretnick added, "We also validated performance among artificial intelligence vendors, which is a big step forward for the market.”
The research is based on interviews with IT leaders—including chief information officers, chief medical information officers and data scientists—at 57 healthcare organizations using purpose-built AI vendors and analytics platforms with AI infrastructure. Electronic health record (EHR) vendors and health IT application vendors with AI capabilities were not included in KLAS’ evaluation.
Based on the providers profiled, the KLAS-CHIME report identified 37 distinct use cases where AI software is being applied across clinical, financial and operational areas, from predicting hospital bed availability to detecting claims fraud and predicting hospital-acquired pressure injury.
Population health and clinical decision support are two of the most common use cases where AI software is being applied.
Healthcare providers profiled in the report cited early success with clinical outcomes such as reducing readmissions and detecting sepsis risk.
In the report, one healthcare senior vice president and CMIO said using AI software from technology company Jvion has helped the organization reduce readmissions through the identification of patients more likely to be admitted with an accuracy of 80% to 90%.
"The system is giving us a better indication of who to go after with our resources, and it is providing us with some intelligence on what to consider doing going forward," the executive said.
A healthcare provider CIO said the organization is using predictive modeling for things like the likelihood of a frivolous lawsuit. "Someone can file a lawsuit without legal backing; people can say that they fell when they really didn’t or that they were mistreated when they really weren’t. Those lawsuits are the kinds of things that we would want to predict before they happened," the CIO said in the report.
Which vendor ranks best?
In a separate version of the healthcare AI report, KLAS rated customer satisfaction for six leading AI healthcare vendors: Jvion, DataRobot, KenSci, Clinithink, IBM Watson Health and Health Catalyst. The evaluation specifically looked at solutions that provide machine learning and natural language processing capabilities.
While work in AI is still in an early stage, vendor performance in the AI space is high compared to other software market segments, KLAS reported.
Researchers noted that because healthcare AI is such a new market, only one of those six—Jvion—has enough evaluations (at least 15) to be considered fully rated. Findings on all other vendors are based on limited data (6–15 evaluations).
DataRobot, which offers an automated machine learning platform for readmissions prediction and prevention, received a high customer satisfaction score, 94, from customers. KenSci also got a high satisfaction score, 92.8, for its length-of-stay prediction tool. Both companies collaborate closely with healthcare organizations that have helped drive outcomes, customers said, according to the KLAS report.
IBM Watson Health has turned the corner in performance but still has room to improve in providing value, healthcare customers said.
Historically, development partners felt the IBM Watson Health solution took a long time to learn and did not achieve desired outcomes. Today, most validated customers use Watson’s machine learning and natural language processing technology with a clinical focus and fairly narrow scope, according to the KLAS report.
Watson AI has the most comprehensive store of indexed medical publications, and this is used for highly valuable use cases such as medical research, genomics analysis and education for cancer patients.
“A couple of customers question whether the product, while helpful, is worth the price tag long term. Two of the seven interviewed IBM customers report nickel-and-diming, the highest percentage among measured vendors,” KLAS said.
Jvion has the largest client base in the healthcare AI market, according to KLAS. The vendor’s AI software targets emergency department and readmission prediction. Clients report mediocre satisfaction with the software and implementation challenges, according to the report, and the company received a customer satisfaction score of 84.4.
Health Catalyst customers gave the company a performance score of 85.5, citing long implementation times. The company has the strongest healthcare expertise, according to KLAS, but customers report that outcomes are either early or will take a while to be realized.