The first automated hospital benchmarking engine was launched earlier this week by OnlyBoth, a new company co-founded by Raul Valdes-Perez—who also co-founded Vivisimo, which was purchased by IBM in 2012 and is currently the search engine used to help power the Watson platform. The new benchmarking engine is available on the company’s website "as a public servic," said Valdes-Perez, and includes public data on more than 4,800 U.S. hospitals. Although it contains almost a hundred attributes, it was just updated with seven new ones, including outpatient procedure data from 2012. Valdes-Perez said he has already been approached to expand the attributes further for hospitals.
Benchmarking was introduced by Xerox in the 1980s as a way to reduce production costs and later developed into a method for continuous quality improvement. But in the healthcare sector, it dates back to the 17th century with mortality comparisons between hospitals. Its initial use in the U.S. was for comparing hospital outcomes to rationalize funding, according to a Healthcare Policy article. Today, with the bar for better performance set high due to the Affordable Care Act and the pressing need to reduce costs, benchmarking is playing a more important role for hospitals.
What sets OnlyBoth apart from others? First, says Valdes-Perez, is that the software is automated and combines novel algorithms for knowledge discovery and uses artificial intelligence to provide unbiased comparisons. Federal data from the Medicare Hospital Compare Quality of Care site and several published U.S. News and World Report hospital rankings are utilized to identify differentiators by qualifiers such as hospital size, staff and patient experience, along with hundreds of other characteristics.
The software makes huge numbers of comparisons where hospitals differ within a peer group, in ways that may be good, bad, neutral or ambivalent. The resulting insights are communicated in read-to-publish English, instead of graphs or charts. "When we consult with executives, they tell us they don't want to see another dashboard," said Valdes-Perez. "What they want is information on how to improve." Results of the benchmarking will help hospitals decide what is working and what isn't.
"As is true of benchmarking, in general, hospital comparisons have been based on subjective choices of narrow peer groups and few metrics. The novel power of benchmarking engines, powered by artificial intelligence, enables comparing and contrasting every hospital to every peer along every attribute, uncovering those hidden nuggets and writing them up, all automatically," said Valdes-Perez in a statement. "By eliminating lengthy manual benchmarking cycles, we enable organizations to concentrate on evaluating the findings and determining the next steps."
Benchmarking needs software automation, explained Valdes-Perez. It has several flaws that are holding back its potential, including data that is not readily available, expensive and complex, limited in scope (potentially leading to bias); and quantitative measurements that lead to simple comparisons, debatable assumptions about the range of outstanding or even acceptable performance, and the economic value is limited to one organization versus 1,000 peer organizations. The benefit of software automation is that it “looks at all the combinations of peer groups and quantitative/qualitative data attributes, discovers insights, ranks them, writes them up, and reports them and does all these well. Properly programmed artificial intelligence can handle all this. But humans just can’t cope with the numerous combinations. To illustrate: there are 161,700 ways to form three-attribute peer groups out of 100 data attributes," according to the white paper.
The data provided by the benchmarking engine also promotes self-improvement, Valdes-Perez explained, by revealing exceptional outcomes in the context of significant peer groups. Also, when applied to public data, as with the hospitals, it encourages transparency in the area of public interest. "I think all businesses can improveùa universal betterment through automated benchmarking."