By Sam Koslowsky, Senior Analytic Consultant, Harte Hanks
Of the many disciplines that are active users of machine learning and artificial intelligence, perhaps the most significant contributions and most meaningful impacts can be made in the health care industry. In addition to the potential improvement in overall human health, the vast availability of data provides for a rich reservoir to mine. Big pharma, with their associated research and development activities, has resulted in expansive data bases. Insurance companies have organized their patient files. Physicians are recording patient histories. And hospitals are finally maintaining a plethora of records, as well.
Like all industries, cost savings, revenue enhancements, and productivity gains are key drivers for developing health care applications. One out of every five dollars of our GNP emerges from health-care-a staggering sum.1
Machine learning and related technologies are becoming more prominent in business and society, and are now being harnessed in healthcare in a number of different domains. These new tools are indeed transformative for patient care, as well as for managerial practices within provider, payer and pharmaceutical organizations.
While there are numerous approaches for classifying machine learning health-care associated applications, I prefer using the following categories.
- Disease diagnosis and treatment
According to a report from the Association of American Medical Colleges, there will be a disturbing disparity between demand and supply of the health care workforce in the near future. The projections show that in a decade, there is will be a shortfall of between 46,900 and 121,900 physicians in the United States.2 As a result, early prediction of health care risks is critical to improve health care quality and reduce health care costs. Predictive analytics uses historical data and algorithms based on machine learning to develop predictive models that capture important trends. Predictive models developed using machine learning technologies are commonly applied for various health care problems such as disease diagnosis, treatment selection, and therapy personalization.3
Medicine has typically relied on the judgment of practitioners. For example, a physician would have to recommend appropriate treatments based on a patient’s symptoms and perhaps some diagnostic tests. However, this is not always accurate. Now, with the continual enhancements in technology and machine learning algorithms, particularly in deep learning, researchers have demonstrated extraordinary progress in disease diagnosis and treatments. Take, for example, image-recognition tasks. Historically, in radiology practice, trained physicians visually assessed medical images for the detection, prognosis and diagnosis of disease. Neural networks has become indispensable in image recognition. They’re most commonly used to analyze visual imagery. Image classification converts an input (like an XRAY) and outputs a class (like “healthy”) or a likelihood that the input is a particular class.4
Lancet, a highly regarded medical journal concluded, “Our review found the diagnostic performance of deep learning models to be equivalent to that of health-care professionals. However, a major finding of the review is that few studies presented externally validated results or compared the performance of deep learning models and health-care professionals using the same sample. Additionally, poor reporting is prevalent in deep learning studies, which limits reliable interpretation of the reported diagnostic accuracy. New reporting standards that address specific challenges of deep learning could improve future studies, enabling greater confidence in the results of future evaluations of this promising technology.”5
Pathologists are employing machine learning to provide more accurate diagnoses. Reduction in errors during the diagnostic stages of disease discovery allows for more time to treat patients at an earlier point in time resulting in the saving of additional lives.
Medical equipment has also benefited from AI. With AI improved microscopes, physicians can more easily identify harmful substances. And this allows for more accurate prediction of disease.6
While many medicines are available for current disease, it is sometimes overwhelming mapping these treatments to other ailments. For example, during the height of the covid epidemic, scientists were searching for existing medicines that could successfully address covid. AI is allowing the speedier match between promising existing treatments for certain diseases to be applied to different disorders.
Machine learning is also providing medical practitioners the ability to stay current on the latest developments. Providers often find it difficult to stay updated with the latest medical advances. Biomedical data collected can be analyzed by machine learning technologies to provide timely, dependable answers to health care providers.7
In many cases, health data and medical records of patients are stored as health provider notes resulting in difficulty in deciphering and interpretation. Thus, requiring negotiating unstructured data. Machine learning tools can obtain, collect, store and analyze non-quantitative medical data providing fast, accurate, and targeted treatment plans for patients.8
In addition to the significant values discussed above, other machine learning administrative benefits are also being realized. Take chatbots. A chatbot is a computer program that uses artificial intelligence and natural language processing (NLP) to understand customer questions and automate responses to them. Complex neural networks paradigms are being developed using machine learning technologies, permitting patients to ask and comment concerning appointments, bill payments, and more. Chatbots also provide ‘advice’ to patients regarding their illness and symptoms, which in turn frees up valuable time for the health care provider.9
Surgeons are now employing robots. ‘Augmented’ robots, and associated machine learning tools, enhance the skills of the doctors. The robot provides an enlarged scene so that the surgeon can view minute details, that would otherwise be difficult to observe.10 Surgeries that are assisted by AI-implemented robots result in fewer difficulties, relatively less pain for the patients, and a speedier return to normal for the patient.
As the patient recovers, there is a need to discharge him when conditions warrant. Hospitals can utilize predictive analytics via machine learning to identify patients with high likelihood of imminent readmission. Additionally, with machine learning tools, patients most likely to miss an appointment can be isolated and appropriate services can be offered to assist them.11
In addition to the diagnosis and treatment applications, researchers are also devoting resources to identifying illicit health insurance-related conduct, more commonly referred to as fraud. The Internal Revenue Service recently announced they intend to hire thousands of new associates. While some of these new employees will provide customer service functions, many will be involved in identifying fraudulent tax-related activity. The IRS is now employing machine learning with AI, to more adeptly identify tax fraud and other questionable activity. Fraud in health care is no different.
While most health care providers and associated health care facilities behave in a legitimate manner, there does exist a subsample of entities that participate in fraudulent activities. The Department of Justice reports that they have recouped in excess of $2.5 billion coming from healthcare fraud.12 Monies emerging from healthcare fraud cases involve many types of activities including drug and medical device manufacturers, health care providers, hospitals, pharmacies, hospice organizations, laboratories, and physicians. Primary fraudulent activities include:
- billing for unnecessary and expensive services
- falsifying patient records Disguising non-covered services as covered services.
- employing incorrect diagnosis/procedure codes.
Estimates are that 3% of healthcare claims in the United States are fraudulent.13 This converts into a hundred billion dollars lost annually. Employing machine learning technologies, the healthcare industry can expose fallacious claims before they are paid, and accelerate approval for legitimate ones. Machine learning algorithms can be developed to determine whether malicious intent exists. This may help to reduce the funds lost, and dissuade the scammers from future attempts. Often, these scammers develop new schemes that require updating the machine learning algorithms.
“This past summer, the Department of Justice announced criminal charges against 36 defendants in 13 federal districts across the United States for more than $1.2 billion in alleged fraudulent telemedicine, cardiovascular and cancer genetic testing, and durable medical equipment (DME) schemes.”14
Yes-there are dishonest professionals that are costing the public huge sums of money. But machine learning and artificial intelligence technologies are certainly working to mitigate these losses. On the other hand, there are others engaging in machine learning to further improve drug development, disease diagnosis, and treatment therapies. It is these researchers that will discern exceptional insights into enhancing the lives of patients.
Sam Koslowsky serves as Senior Analytic Consultant for Harte Hanks. Sam’s responsibilities include developing quantitative and analytic solutions for a wide variety of firms. Sam is a frequent speaker at industry conferences, a contributor to many analytic related publications, and has taught at Columbia and New York Universities. He has an undergraduate degree in mathematics, an MBA in finance from New York University, and has completed post-graduate work in statistics and operations research.