Why Healthcare Machine Learning Adoption is Skyrocketing

The industry is adopting machine learning to address numerous use cases, tap its automation potential, and derive greater value from data.

The healthcare industry is embracing machine learning. Chandra Kalle, Senior Director, Engineering for LeanTaaS, says factors driving artificial intelligence (AI) and, specifically, machine learning adoption include ever-increasing healthcare costs (over $9 trillion globally), an aging, longer-living population, and a perennial provider shortage. “Thus, there is an ever-greater need for using AI to do more and, at the same time, more mature AI technology available to do it. The latter includes advances in machine learning, such as deep learning, as well as better AI tools and computational infrastructure developed by technology leaders like Google and Amazon,” Kalle explains.

In the following Q&A, Kalle shares additional insights on machine learning in the healthcare space:

What are the primary types of use cases for machine learning?

Kalle: This is a really vast space. Machine learning can be used for:

  • Diagnostic imaging, for example, identifying disease by processing images interpreting radiology images or skin images looking for cancers.
  • Telehealth that uses chatbots for clinical assessment, monitoring and diagnosis
  • Clinical decision support, e.g., what cancer treatment to prescribe next given how things have gone so far
  • Drug discovery using lab data to predict the impact of a potential drug on the whole human even before engaging in lengthy and costly clinical trials

Another huge area is personalized medicine, where treatment is informed by patient phenotype, for example, to choose the most effective medication and gauge the right dose (pharmacogenomics).

The list goes on and on.

How does machine learning automate routine or manual processes? Is this widely used or something that we’ll see more of in the future?

Kalle: We’ll absolutely continue seeing more and more of this. In the clinical space, a human’s capacity to remember relevant information and make decisions is increasingly becoming a limiting factor as we are able to gather greater amounts of data on patients and disease progression and make increasingly subtle treatments available.

In the operations space, we have been seeing time and time again how schedule optimization improves access to hospitals and other providers, allowing patients to receive care they need faster and from more specialized providers at some of the best hospitals in the country.

How is machine learning helping healthcare providers deal with the large volumes of data that they need to collect and use?

Kalle: Data volume is an enormous challenge in any industry but particularly in healthcare where data tends to sit idle in databases managed by dated EHR systems. Many companies build their business on getting large volumes of data from these systems to make them available and actionable as they power predictive analytics, decision support, imaging, operation optimization, and other applications. Other organizations make extensive use of insurance claims data that have recently become available through state governments. However, guidelines for gaining access for commercial purposes are nascent utilizing complex processes, so successful applications have been few and far between.

Is machine learning helping to diagnose conditions? Can you provide examples of how it’s used now and how it may be used in the future?

Kalle: Yes. Machine learning is helping radiologists interpret imaging results that may have been previously missed by a clinician. For example, a study at Stanford created an algorithm that could detect pneumonia, at this particular site, in those patients involved, with a better average than the radiologist in that trial.

Additionally, the use of AI to describe and evaluate the outcome of maxillofacial surgery or cleft palate therapy has had recent advancements. On average, the human dermatologist accurately detected 86.6 percent of skin cancers from the images, compared to 95 percent for the CNN machine.

Are there opportunities for software developers to help healthcare providers leverage machine learning? What are the greatest opportunities going into 2020?

Kalle: We think many of the big opportunities being explored this year such as leveraging machine learning to diagnose disease, early intervention and disease management, physician productivity, telemedicine, and remote assistance, are quite promising and will continue to gain momentum next year.

One area that hasn’t gotten a lot of attention in the past and is starting to is leveraging machine learning to increase operational efficiency. Healthcare organizations are constantly under pressure to “do more with less,” and helping them improve operational efficiency, which in turn leads to better patient experience and staff satisfaction, is incredibly valuable. In the past four years, we’ve built several products that leverage data science and machine learning to mathematically balance supply and demand and improve operations at specific departments such as cancer centers, operating rooms, primary care clinics, and inpatient beds and have seen phenomenal success. More than 200 hospitals are relying on our products to lower patient wait times and improve staff satisfaction and overall financial performance.

In our experience working with hundreds of healthcare organizations, we’ve learned three things that developers should keep in mind when working with healthcare organizations on machine learning ideas:

  1. Start with data that is available and accessible. Data interoperability and security continues to be a challenge. New standards like FHIR are making it easier but are far from widespread adoption. Security and compliance is a big factor as well. Start with the data that is accessible and available and come up with bottom-up ideas that can provide value.
  1. Aim for clear, significant, repeatable ROI. Healthcare organizations are resource-constrained and have a lot going on. With the move toward value-based care and newer government regulations, they are under a lot of pressure to do more with less. And they are sick of vendors with fancy machine learning ideas that overpromise and under-deliver. So, it’s important to have an ROI that is clear, significant and repeatable.
  1. Make it easy to implement and adopt. Healthcare IT departments are overwhelmed with projects and security risks. Every new solution adds a lot of work for them. So your solution not only needs to have a significant ROI, it should also have a simple implementation, fit well into their workflow, not require a lot of user training, and not pose any security risks. In short, it should be “IT friendly.” 

 

Chandra Kalle, senior director, engineering, brings more than a decade of experience in engineering and product management to LeanTaaS. Prior to LeanTaaS, Chandra was VP of product at Collegefeed (acquired by AfterCollege), and at Symantec, he was a lead engineer where he developed malware detection technology that protects more than 200 million users worldwide. Chandra studied computer science at North Dakota State University and holds several patents in computer security and mobile design. For more information on LeanTaaS, a Silicon Valley software innovator that helps hospitals and health systems increase access and lower cost through predictive and prescriptive analytics, please visit https://leantaas.com/, and follow the company on Twitter @LeanTaaS, Facebook at www.facebook.com/LeanTaaS and LinkedIn at www.linkedin.com/company/leantaas


SHARE

The former owner of a software development company and having more than a decade of experience writing for B2B IT solution providers, Mike is co-founder of DevPro Journal.