The transition to value-based care has made it more important than ever for healthcare organizations to use their data to determine how best to allocate limited resources to achieve high quality care. In response, an onslaught of business intelligence vendors have descended on healthcare with data warehouses, reporting tools, and dashboard analytics that have led to tremendous efficiency in other industries.
However, healthcare is not like other industries. Decades of “fee for service” care have left us with up to 50% of clinically relevant information stored as unstructured free text. In addition, structured data as fundamental as what diseases people suffer from (ICD codes) can be up to 80% inaccurate. This isn’t likely to improve as we increase the number of codes 5-fold with the transition to ICD-10.
As value-based care organizations are discovering, these multi-million dollar investments in fee for service analytics are useful for understanding what happened – how many beds were filled, drugs prescribed, surgeries performed – but are limited in their ability to answer the fundamental questions of value-based care: what should happen, to whom, and when.
Beneath a rather thick veneer of buzzwords from cognitive computing, big data, artificial intelligence, and a dozen other synonyms is a set of techniques derived from mathematics and computing that has proven to be well-suited for analyzing healthcare data. Rather than rely on rules written from a few data points, ‘machine learning’ discovers patterns of relevance from millions of data points in fractions of seconds.
Hundreds of studies over the past 20 years have demonstrated machine learning’s superior predictive performance across cardiothoracic surgery, urology, orthopedic surgery, transplant, trauma, neurosurgery, cancer prediction and prognosis, and intensive care unit morbidity. Combined with natural language processing technology, these approaches can learn from the 50% of clinically relevant information currently locked away as free text.
However, these advantages will not improve patient care when they’re only accessible to PhDs and data scientists. To matter, they must be simple enough to be useful, yet robust enough to contribute across all of healthcare. This was the impetus behind the research and now, the company that is Cyft. Our goal was to create software that could be used by non-technical end users to build highly accurate predictive models for healthcare.
For years we attempted to transfer this capability to healthcare via open source software. We gave talks, created videos, and setup demo data with instruction for healthcare-specific use cases. While it was satisfying to see researchers at institutions across the globe put this technology to use as a teaching and research tool, it became clear that widespread adoption by healthcare organizations would require scaled infrastructure, sales, and implementation support – not the type of work typically supported by research funding nor typical of the skills possessed by researchers.
And so we formed Cyft. The team has grown but the goal remain the same: to give healthcare a single solution capable of turning organizations’ own data – from EMR to claims to call center transcriptions – into highly accurate predictions for the value-based problems that matter most to them quickly, and a fraction of the cost of traditional analytics. And while we used to measure results in terms of accuracy and speed, the results in terms of return on investment have been encouraging. Early adopters of Cyft are seeing results in days and realizing returns in the tens of millions of dollars while improving care by creating models for their most pressing revenue recognition, cost savings, and quality improvement applications.
On behalf of the clinicians, researchers, mentors, and members of our current team that have worked hard to make this idea a reality, I thank you sincerely for your interest in making prediction accessible to all of healthcare. I look forward to learning how we can work together to make value-based care a success for both patients and the organizations that serve them.
Leonard D’Avolio, PhD
CEO and Co-Founder