A version of this article was originally published by Becker’s Hospital Review on April 25, 2017.
As the healthcare ecosystem continues to transition to value-based care models, we will see significant investments and related advancements in applications of machine learning – which is the underpinning of artificial intelligence (AI) of computers and processes used in healthcare.
From machine learning, we are seeing the explosion of another common buzz phrase natural language processing (NLP), which is an application of the computer learning “on its own” through the use of neural networks to build predictive algorithms based on a feedback loop of learned understandings – which continues to get smarter with the more experiential data ingested and analyzed.
Historically, the healthcare system in the United States has been plagued with the obstacle of rising healthcare costs and inefficiencies attributed to the fragmented nature of care delivery. As the frameworks for interoperability and interconnectivity evolve to bridge the gaps between healthcare silos, so too do the opportunities to more quickly aggregate data and leverage advanced parallel processing to employ cognitive computing and AI-like technologies. Using systems and platforms that integrate and aggregate disparate real-time data from historically fragmented sources and making those data available to the healthcare delivery system provides for the basis for AI to change how payors, providers and other healthcare organizations engage with patients and drive outcomes in a growing value-based, outcomes-based environment.
Whether it’s a hospital, health plan, medical group, pharmaceutical company or medical device manufacturer, the focus over the next decade and beyond will be on patient-specific innovative technologies and data-driven solutions that improve the quality of care while simultaneously reducing costs. Machine-learning technologies will play an ever-important role in this, ensuring that the healthcare system as a whole is continually “learning” and applying those insights to real-time decision-making – for individual patients – at the point of care.
Today’s machine learning platforms expand upon the classic regression techniques which are found in traditional predictive analytic engines in order to derive a more dynamic, patient-specific understanding of health care systems and most importantly, individuals. For data scientists, however, the Holy Grail is doing all this in real-time in a broad-spectrum manner leveraging disparate, big data – another buzz phrase of recent. The ability for a computer system to analyze thousands of disparate and evolving data points to create patient-level predictions driving more precise treatments is a game-change for many clinical conditions.
We’re already seeing real-world applications of machine learning that promises to drive greater efficiencies in healthcare. With the application of NLP, managed care organizations and health plans can benefit by conceivably no longer needing to exhaust often scarce and costly resources to manually read and analyze millions of patient medical records for data-submission audits by CMS and other regulatory authorities. Applications of NLP – powered by machine learning – can now be used to pre-screen medical records with a greater efficiency and completeness over the traditional approach of human clinical review of often lengthy medical records delivering bottom-line savings in the process.
The future of AI is supported by scalable cloud-based platforms and these integrated diverse data sets will impact clinical organizations through a shift in the role of the provider from one of diagnostician informed by training and evidence-based practices, to decision makers informed by both that training and practices, as well as informed by real-time patient specific analytics to guide the clinician at the point of care. These processes will help to change the healthcare system through improvements and efficiencies in areas such as having appropriate clinical and risk indicators at the point of care, reduction in medical errors, improved diagnosis within the mental health domain, detection of cancer and various others.
You see this application starting to emerge across the healthcare ecosystem in areas such as telemedicine and specialty pharmacy delivery where clinicians may soon be armed with a longitudinal patient record during their virtual visit that is tailored to the specific genotypic, phenotypic, social economic circumstances of that patient, as of that moment in time.