❮ Back to Blog

Should The Use of Machine Learning in Healthcare Be Embraced or Met With Skepticism?

by Nolan Rhem on September 11, 2017 at 6:26 PM

blog-machine-learning-healthcare.png

Medical science is always evolving, experiencing both gradual advances over extended periods of time as well as dramatic, even seminal, innovations that propel physicians’ ability to improve patient outcomes at unimagined rates.

One of those formative innovations was the “Antiseptic Principle of the Practice of Surgery,” published by Joseph Lister in 1867. In the first four years of his tenure as Professor of Surgery at the University of Glasgow, Lister reported that 50 percent of his amputation cases died from sepsis. At that time, the prevailing clinical belief was that patients succumbed to sepsis from infections to wounds exposed to miasma, or bad air. Surgeons were not required to wash their hands, bed sheets and surgical clothing were reused and medical equipment was not cleaned between surgeries. Rather, hospital wards were merely aired out at midday to reduce infection from miasma.

A New Way of Doing Things: Using Antiseptic Barriers

To expand on the preventative measures, Lister theorized that an antiseptic barrier could be created to inhibit infection. He experimented, with great success, in applying carbolic acid to equipment, surgeons’ hands and the wounds themselves. Lister extended this concept of sterile surgery to regularly washing surgical equipment, hospital linens and physicians’ hands.  In the subsequent four years of his tenure at Glasgow, Lister noted that surgical mortality fell to 15 percent.

Despite his individual success, Lister was initially unable to generate excitement about sterile surgery in the broader medical community. Reaction throughout Europe and America ranged from skepticism to disbelief. Lister authored paper after paper and gave speech after speech defending his work, but it wasn’t until his retirement in 1893, 26 years after publication of his original paper, that the sterile surgery methodology was universally accepted.[1]

Is There a Future For Machine Leaning in Healthcare?

Today, there is a development in healthcare that is generating similar levels of both skepticism and excitement. As medical science continues to gradually progress, the application of machine learning provides an opportunity to dramatically increase improvements in patient outcomes and reduce costs.

Because we live in the era of big data, it is possible to process ever-increasing quantities of data in ever-decreasing amounts of time. Applying machine learning tools in a big data environment accelerates the medical community’s ability to draw clinical conclusions and identify treatment options.

Machine learning leverages historical data to deliver unobserved insights, highly correlated patient classifications and even predictive models. For example, machine learning assists with:

  • Natural language processing (NLP) of unstructured free-text content, such as that found in electronic health record (EHR) clinical notes, used to diagnose disease. NLP of observational data may be more insightful than review of medical diagnoses codes like those found in ICD-9 and CPT.
  • Machine learning algorithms to identify patients eligible for clinical trials through review of historical EHR data. This could improve patient suitability for and outcome from the clinical trial.
  • Predicting mortality rates after major surgery by learning from historical data. This could inform post-surgery treatment and medication plans.
  • Using models to predict staffing levels, procedure durations and necessary provider training and skills. This would allow provider systems to optimize facility operations, delivering higher quality care and reducing costs.
  • Using data provided by the patient upon admission to the hospital and prior to the collection of vital signs, laboratory results or patient history to classify that patient’s risk for a particular disease state. These classifications can be used to expedite examinations, diagnoses and treatment plans.

Consider Lister’s sterile surgery development in the context of the potential for machine learning in healthcare. Both developments are interesting for the skepticism with which they are initially received and the profound impact they have had on the practice of medicine. Lister’s methodology reduced sepsis rates from 50 to 15 percent in four years. Could machine learning be leveraged to further reduce sepsis rates? What would be the impact from applying machine learning to patient records upon arrival at the hospital, as described in the final bullet above?

Machine learning has positioned healthcare at the threshold of another pivotal moment in its evolution. As in the case of the sluggish recognition of the benefits of sterile surgery, skepticism about machine learning approaches are a disservice to the industry and place patients at increased risk. However, if such skepticism is pushed aside in favor of open-minded consideration and a swift adoption of machine learning approaches, the door will be opened to rapid improvements in both healthcare delivery and clinical outcomes.

New Call-to-action
author avatar

This post was written by Nolan Rhem

Nolan Rhem is the Vice President of Operations at Intermedix. In his position, Nolan is responsible for project management, product implementation and client services within the data science division. Nolan has 25 years of experience in technology and healthcare solutions delivery with an overall emphasis in project management and operational efficiency. Prior to joining Intermedix, Nolan served as Chief Operations Officer for WPC Healthcare, which was acquired by Intermedix in 2017, and Executive Vice President of Quinnian Health Inc. Nolan obtained his bachelor’s degree in international relations from University of Virginia and his master’s degree in Business Administration from the University of Memphis.

Connect with Nolan