Using machine learning in the healthcare industry to predict health outcomes

Machine learning (ML) is a powerful tool that uses data analysis to learn and automatically improve. After being trained on set algorithms, ML can function independently and predict outcomes at rates much faster than a human. This can be a tremendous benefit to the healthcare industry.

In late 2020, two studies were released that demonstrated ML’s contributions to the fields of psychology and cardiology.

Benefits of supplementing human predicting analysis with ML predictive analysis

As shared in the Journal of the American Medical Association (JAMA) in early December 2020, a study by Nikolaos Koutsouleris, MD, and others out of Munich, Germany sought to find if a diagnosis of psychosis could be predicted in patients “with clinical high-risk states or recent-onset depression by optimally integrating clinical, neurocognitive, neuroimaging, and genetic information with clinicians’ prognostic estimates.”

The study used machine learning models to combine clinical and biological data with clinicians’ estimates for 668 patients, half of which were patients and half of which were part of a control group. The ML model “correctly predicted disease transitions in 85.9% of cases across geographically distinct patient populations,” reported Koutsouleris in the JAMA article.

Though a noted small sample size of psychosis transitions increased the risk of “overly optimistic prediction results,” Koutsouleris and the other study leads feel that supplementing human predictive analysis with ML predictive analysis may result in improved assessments for risk of certain patient populations developing psychosis.

Challenges of finding proper algorithms for ML and predicting cardiovascular outcomes

An article for Scientific Reports published in September 2020 highlights how ML is being used at increasing rates for predicting cardiovascular disease. Study author Chayakrit Krittanawong and others worked “to assess and summarize the overall predictive ability of ML algorithms in cardiovascular diseases.”

They sought to determine if ML could accurately predict coronary artery disease, heart failure, stroke, and cardiac arrhythmias, and which ML algorithms were the most successful at predictions.

The study leads reviewed 344 existing studies where ML was used to predict cardiac-related outcomes. This group was reduced to 103 cohorts including over 3.3 million individuals who met the study criteria. A number of ML algorithms were identified as being used in predicting cardiac health outcomes, including boosting algorithms, custom build algorithms, support vector machine (SVM) algorithms, and convolutional neural network (CNN) algorithms.

The study found that ML algorithms used in predicting cardiac health outcomes were often successful, but there is a lot of work to be done before offering conclusive guidance on which ML algorithm is best.

“SVM and boosting algorithms are widely used in cardiovascular medicine with good results,” reported Krittanawong, “However, selecting the proper algorithms for the appropriate research questions, comparison to human experts, validation cohorts, and reporting of all possible evaluation matrices are needed for study interpretation in the correct clinical context.”

Room to improve, grow, and flourish

While both of these studies show the value of ML in predicting health outcomes, they also demonstrate the need for additional research and data. However, this is no doubt this area will continue to grow over time.

If a future career in Machine Learning interests you, take a deeper look at Capitol Tech. Our university offers bachelor’s, master’s, and doctoral degrees in both computer science and data science. Both areas of study include coursework in machine learning and offer courses in-person and online. For more information, email admissions@captech.edu.