In the global race for superiority in artificial intelligence (AI), countries such as the U.S. and China are increasing their investment. The healthcare industry is in the midst of the AI revolution and stands to benefit greatly from so-called disruptive innovation. Currently, health is the biggest household expenditure in the U.S., and costs are expected to reach about 19.9% of the GDP by 2025.

AI offers the promise of reducing healthcare costs while increasing the quality of care. As the U.S. is shifting from its old fee-for-service model to one with payments tied to care quality, healthcare providers are turning to AI for solutions.

Researchers from the American Medical Association have identified multiple categories of healthcare spending that can be cut while increasing healthcare outcomes, including:

  • Reducing healthcare waste

  • Predicting epidemics

  • Avoiding preventable death

  • Improving efficiency of care

  • Developing new drugs and treatments

  • Improving quality of life

AI is the Key to Innovation in Healthcare

A class of solutions for addressing healthcare spending and quality of care focuses on integrating AI into the healthcare system. A survey of healthcare decision-makers revealed that 91% of respondents believe AI will provide predictive analytics for early intervention, 88% say it will improve care, and 83% agree it would improve the accuracy of medical diagnoses.

The recent increase in adopting electronic health records (EHRs) has allowed for AI and machine learning (ML) techniques in data extraction from free text, diagnostic and predictive algorithms, and clinical decision support systems to be applied in healthcare settings.

AI is already being used to automate daily tasks and reduce workloads, and aid in patient record linkage.  This is only the tip of the iceberg when it comes to the power of AI and ML when it comes to healthcare innovation.

Reduce Waste

Adopting powerful AI algorithms into the healthcare system can remedy much of the waste currently experienced.

Care coordination encompasses anything that bridges the gaps along the care pathway for a given patient and can go a long way in terms of reducing waste. Here, companies such as Google Deepmind Health are already leveraging AI technologies to mine medical records and provide better and faster services.

Healthcare spending is also wasted by overtreatment and unnecessary medical care. A study survey found that unnecessary medical care in the form of prescription medication, tests, and procedures were administered due to issues like difficulty accessing medical records. Currently, companies like Oncora Medical provide integrated digital databases that collect and organize patients’ EHRs to facilitate access.

Administrative complexity not only contributes to waste, but also creates a large work burden for clinicians.

AI can be used by insurance providers for identifying and compiling claims. Health insurers also use AI for fraud detection by identifying hotspots such as overpayments and what their causes may be. The previous manual method would take weeks to discover these cases, while use of AI can detect these in minutes.

Predict Epidemics

AI can drastically help epidemiologists by reviewing vast sets of real-time medical data to predict at risk populations. These predictive analytics tools process real-time data to identify and measure patterns. Epidemiologists can then gain the insight needed for earlier detection and analysis of the effectiveness of current programs. One such application already employed is the hc1 Opioid Dashboard used for tracking opioid usage patterns.

There is also a class of AI that uses relevant information including environmental features, to predict areas where outbreaks of diseases such as Dengue fever, are more likely.

Projects using AI to mine search term records, social media postings, and cell phone records have been deployed to predict trends in epidemics. AI clustering algorithms using NLP can make accurate future predictions by encoding similar terms and location data into features and running them through models trained on learned patterns.

Avoid Preventable Death

Causes of death can be classified in terms of preventable risk factors such as smoking, an unhealthy diet, or medical error. AI is an efficacious tool for classifying factors into features for models by processing huge quantities of data. Therefore, identifying at risk population segments can be effectively handled with these AI algorithms.

As an example, a 2018 study used an NLP system to extract relevant EHR data to identify whether the correct medications were prescribed to heart patients during discharge. Use of a system such as this can help physicians ensure they have not made mistakes in inpatient care and are adhering to guidelines.

In addition to identifying preventable risks, early detection significantly improves the chances for successful treatment. AI algorithms can cross-reference EHRs and patients’ genetic and lifestyle data to detect individuals that may be more at risk for developing a disease. For instance, Human Longevity combines large sets of genomic and phenotypic data, and then uses machine learning to transform this data into useful insight that can spot cancer or vascular disease at an early stage. This provides a proactive rather than reactive approach to healthcare.

Improve Efficiency of Care

Increasing efficiency in the healthcare industry is strongly tied to reducing waste. However, it has the added benefit of supplying consumers with faster and higher quality services.

Using text-to-voice algorithms, an AI voice enabled system can triage patients’ symptoms and direct them to lower-cost retail or urgent care settings when an ER visit is not required. This would cut down ER wait times and save patients from having unneeded emergency expenditures.

Additionally, AI nursing assistants can save RNs 20% of their time handling unnecessary visits. Nursing assistant apps can remotely assess a patient’s symptoms to deliver alerts to clinicians when care is needed. This means consumers will have access to health services faster and gain a reduction in time that wasted seeking unnecessary intervention.

Develop New Drugs and Treatments

Many companies are focusing AI efforts in the pharmaceutical industry. It is not feasible for a doctor to accurately predict a patient’s response to drug treatments by inferring all possible relationships among factors including a person’s metabolism and the distribution of the drugs compounds in the person’s body. However, machine learning can do this

AI algorithms can predict which potential medicines will work, and many companies are presently putting this to use to find novel uses for known drugs. Other applications are using supercomputers to search for therapies from databases of molecular structures.

Improve Quality of Life

Quality of life encompasses functional status, access to resources and opportunities, and a sense of well-being and is most often associated with managing chronic, disabling illnesses.

As wearables become more popular, AI is being used on harvested data from patients’ biosensors to detect subtle signs that may signal a problem will occur. By incorporating data from EHRs with this patient-generated health data, customized, real-time recommendations can be delivered to the clinician or patient

This can be particularly helpful for an individual struggling with a chronic illness. AI tools have the potential to transform the way clinicians are treating chronic diseases by shifting toward preventive medicine which, as previously noted, can decrease preventable death. Similarly, preventive medicine can also help patients manage illnesses before symptoms become too debilitating, thereby drastically improving their quality of life.

Linking It All Together

Record linkage is the glue that unites the vast amounts of fragmented, heterogenous data to give a complete, single source of a patient’s demographics and medical history.  Health data comes from various origins that collect electronic information in uncoordinated ways. During a routine visit to the doctor information on a person’s vitals will be collected. Biosensors worn by this same patient will collect similar data, but the data will not be stored in the same manner. To resolve duplicative or missing information, records from these various constituents must be linked.

For example, diseases such as cancer can last months or years, and patients often see clinicians at different organizations. Researchers require a full picture of the length of care and treatments given to cancer patients in order to form evidence based recommendations and treatments. Record linkage using AI can be used to gain the most insight from harvested data.

Generally, records are matched using a unique identifier key. However, in real life, a common key is rarely available. Linkage instead relies on matching partially identifying attributes. Complex linking algorithms using ML techniques such as probabilistic linkage or unsupervised linkage, must be deployed to handle the breadth and depth of this data.

AI record linkage has led to substantial improvement in the quality of records. The application of these techniques is increasing rapidly in the healthcare sector. In fact, the U.S. Department of Health and Human Services has instituted requirements for “meaningful use” of EHRs as a condition for providers to receive financial incentives. Basically, if providers want to meet these requirements, they need to be able to exchange data between various healthcare settings while maintaining accurate patient lists.

Final Thoughts

We live in an era where results are expected to happen instantaneously and the adoption of AI in our everyday lives has become quintessential. Healthcare organizations are using AI to help with automating tasks and are now looking to AI methods that can increase effectivity and efficiency in other areas.

AI research has been shown to deliver data insights that the human mind cannot, resulting in predictions that are more accurate than any made by clinicians alone. Tracking and “meaningful use” of EHRs is the key to leveraging the most out of the ever-growing sources of health data available. So, the adoption of AI into the healthcare system is imminent.

The next steps for companies and healthcare organizations is to prepare for incorporating AI into their business models. Shifting over to the use of AI will require experts to set the systems in place and educate healthcare workers on how to use them. Those that do not begin the adoption process will fall behind the curve. However, many software resources are available to help simplify the process.