Democratizing Machine Learning Algorithms for Integrated Data-Sharing
On Artificial Intelligence and Ethics
The Tango of AI and Big Data
Transforming the Art Museum in the 21st Century.
Machine Learning: Enabling New Capabilities in Health and Beyond
Shiva Amiri, PhD, Director Research Infrastructure, 23andMe
Embedding AI in everything. Why we can't do without it now
Vishwa Kolla, AVP, Head of Advanced Analytics, John Hancock Insurance
Thank you for Subscribing to CIO Applications Weekly Brief
How AI and Machine Learning are Reinventing Healthcare Sector
By Julius Bogdan, Director of Analytics and Data Innovation, SCL Health
It seems that artificial intelligence (AI) and to a lesser extent machine learning (ML) is everywhere these days. It is permeating every industry, including retail (customer intelligence), finance (fraud detection), and manufacturing (predictive maintenance), leaving no industry untouched. It has even infiltrated healthcare with big promises of being a cure all for all of healthcare’s ills. There is definitely a lot of hype around AI but what is it and how can healthcare benefit from it?
Artificial Intelligence vs Machine Learning
Artificial intelligence is an umbrella term used to refer to machines that can perform “intelligent” tasks. There are several areas under the AI umbrella with one of the biggest being machine learning. At its essence, machine learning is a way of achieving AI by learning dynamically from data. It is actually the brains powering most predictive programs including recommendation engines, fraud detectors, and virtual assistants among many other applications. It is also the one that is most applicable to healthcare and the focus of the rest of this article.
Think of machine learning as a complimentary technology that provides physicians and administrators with new capabilities to improve patient care, reduce costs, and improve patient outcomes. With ML clinicians have more relevant data at their disposal to make more informed decisions about patient care. Physicians can use ML to risk stratify their patients focusing more of their attention on their high risk patients. Administrators can use ML to monitor and manage operations proactively giving them levers for improvement they historically did not have.
Machine Learning Use Cases in Healthcare
These are the basic use cases that healthcare systems have begun to implement.
• Prevent hospital acquired infections and conditions (HAIs/ HACs) – Healthcare systems can reduce CAUTI and CLABSI rates by predicting which patients will develop these conditions and alert clinicians so they can intervene and reduce that risk by focusing on individual patient risk factors.
• Reduce readmissions– Machine learning can provide patient centered guidance to clinicians on which patients would most likely be readmitted and how they might be able to reduce that risk, improving patient care.
Machine learning is improving by leaps and bounds and is becoming established in healthcare as a critical complimentary technology with its ability to empower clinicians, administrators, and patients
• Reduce length of stay (LOS) – Healthcare systems can lower the risk of HACs, improve mortality rates, and improve their bottom line by identifying patients that are at risk for an increased LOS and building best practices to address those.
• Distributed medical scheduling–Healthcare systems can improve demand forecasting and management utilizing machine learning to optimize scheduling improving patient satisfaction and the bottom line.
These are more advanced use cases focused on improving patient care that are on the horizon for healthcare systems.
• Early sepsis detection – ML can provide actionable predictive indicators for sepsis and insight into the relevance of different clinical traits reducing sepsis mortality rates and rates for sepsis in general.
• Predict no-shows – Healthcare systems can better predict which patients have a higher propensity for not showing up for an appointment allowing for a follow up to improve a patient’s chances of making their appointment.
• Predict chronic disease – Machine learning can help identify patients with undiagnosed or misdiagnosed chronic disease, predict the probability that they will develop chronic disease, and suggest treatment plans tailored to the individual to mitigate those risks.
• Reduce 1 year mortality – Machine learning can predict patients with a high 1 year mortality rate so healthcare systems can devise a continuum of care plan to better take care of the patient and undertake interventions to impact longevity and quality of life.
Barriers to Broader Adoption
There are many challenges to the broader adoption of machine learning in healthcare. The biggest challenge is that healthcare data is complex and murky at best. Healthcare data suffers from a lot of data quality issues including lack of standardization, missing elements due to workflow configurations, lack of data integrity in the data capture, and loss of fidelity in the many translations required to make sense of the data. In order to be better positioned to take advantage of machine learning we need to think more holistically about the data and what uses we will have for the data not just what our systems provide.
The skills required to work with data, including data engineers and data scientists are not only scarce but they are in high demand across all industries, making it a challenge for attaining them in healthcare. There are some technologies on the horizon, such as AutoML, that give us the ability to augment our roles with data science capabilities without hiring a whole team of data scientists. This can alleviate some of the constraints on resources but the underlying foundation has to be well positioned to take advantage of these new technologies.
The biggest barrier to broader adoption may be the perception of AI and machine learning. Every day there seems to be an article on how AI is going to replace physicians or how you are going to interact with an app as your provider. Why would physicians embrace something that is being touted as their replacement? According to a recent survey by consulting firm Accenture, a quarter of respondents who reported reluctance about using AI in healthcare said they were uncomfortable due to lack of understanding. This lack of meaningful education on both sides of care equation will slow the adoption of AI and ML in healthcare.
Machine learning is improving by leaps and bounds and is becoming established in healthcare as a critical complimentary technology with its ability to empower clinicians, administrators, and patients. Machine learning’s enormous predictive and pattern detection capabilities is helping unlock the value in our medical data. Every stakeholder along the continuum of care needs to recognize this and help shape its future or risk being left behind.