The Impact of Artificial Intelligence in Health Care: A Literature Review
Shuhan Jia
Dr. Miscavige

In 1999 the cost of medical errors was estimated to be between $17 billion and $29 billion. Seven years later that figure is exceeding $200 billion dollars which means the cost of medical errors is increasing making healthcare a more expensive necessity than some argue it should be. The reason this cost increases is because the “prevention of medical errors often lead to adverse events” (Hammerling, 2012). This high cost negatively affects the hospitals and health centers as well as the patients. There will always be a desire for better healthcare which is measured by efficient, low cost, and accurate processes. With that being said, artificial Intelligence (AI) is recognized as the fourth revolution of science and technology, improved the quality of education, enabled stronger market analysis and automotive advancements. Recent literature on AI provided insight regarding the ability to alleviate complications, assist with patient care, assess risk of disease and evaluating treatment options (Becker, 2019). Thus, making it a viable option to better the public healthcare system in the United States.

Improving the current healthcare system would require major changes including automating functions through technology such as AI, monitoring quality indicators and enhancing professional training (Hammerling, 2012). Although it is impossible to achieve a diagnostic confidence rate of 100%, merging AI and medical professionals such as nurses and physicians enhances medical results such as accuracy (Miller, 2017). This literature review will discuss how current research interprets AI’s impact on healthcare. Implementation includes electronic heath record organization, comparative analysis, imaging, and diagnosing illnesses. More specifically, this review will focus on the impact AI will have on the cost of healthcare by looking at projected costs. Furthermore, this review will analyze the impact AI may have on accuracy in order to answer the question, how will artificial intelligence impact the cost and accuracy of healthcare?

II.Artificial Intelligence
Artificial Intelligence is the field of science where computer systems are being developed to preform tasks that normally require human intelligence such as analyzing images, speech recognition, diagnosing and detecting. In healthcare these tasks can range from organizing electronic health records to reading radiology images and making treatment plans accordingly. There are two types of medical AI including Artificial Narrow Intelligence (ANI) and Artificial General Intelligence (AGI). ANI includes simple functions such as voice to text translators and chat boxes whereas AGI is more complex and is still being developed. AGI describes the systems that will be able to reason and act accordingly across a variety of activities as opposed to a narrow set of tasks that ANIs are built to do. Hence the requirement of NLP and ML branches of AI to enable reasoning.

Natural Language Processing
AGI uses Natural Language Processing (NLP) to convert human language into a structured format for a computer to understand and analyze. NLP’s importance in health AI is high because it can be applied to a broad amount of tasks such as a clinical setting where information is transcribed into usable data that represents the patient’s problems, medications, allergies, past medical and surgical history, family history, an other factors that are typically recorded into an electronic health record (Abrahms, 2019). The practice of using NLPs to analyze and organize information and medical images can “improve decision-making efficiency and workflow” (Miller, 2017). Using NLPs, the AI will be able to contextualize medical records, which in turn, connects and compiles the useful data to generate more accurate results.

Machine Learning
Machine learning is another branch of AI. Machine learning learns from new data and improves the AI’s ability to make a prediction of determination. This enables the AI to stay up to date and increase accuracy as the data gets compiled. Machine learning has been implemented in other fields such as autonomous vehicles, web searching as well as tasks such as fraud detection. Machine learning is being developed to be “used for a variety of health care goals including better drug discovery and manufacturing, clinical trials and research, improved accuracy of radiology and radiotherapy diagnoses and treatments” (Abrahms, 2019).

There is an increase in demand for healthcare due to the population over the age of 60 having tripled in the last 50 years. With increased demand and a rising cost of labor cost a lower cost alternative is in demand. As researchers and organizations begin to explore the application of AI in healthcare the impact on cost becomes more relevant as it will be a deciding factor on whether it is implemented or not. The costs of medical attention continue to grow without commensurate improvements in outcomes in accuracy (Insel, 2019). Furthermore, some researchers believe it is unsure whether the cost would be ‘of more value’ to the patients however the cost would benefit the medical institution offering the service as the cost for healthcare has been rising and machines would help lower costs for these medical institutions. There is debate surrounding the implementation of machines and the long-term impact on traditional healthcare. Referencing Dr. Drew Simshaw, the lower cost AI will be able to provide more accurate results allowing the strain on the service providers to be relieved (Simshaw et al, 2016). Additionally, Thomas Insel argues that the implementation of AI will likely lower the long-term cost of medicine while improving the quality by ‘bringing empathy back to healthcare’ (Insel, 2019).

On the contrary other research suggests that the cost will be lower after the implementation of AI. Through using AI for medical imaging alone in an experimental study conducted through the analysis of 160,000 medical images it was noted that the AI method as opposed to a traditional radiologist was more cost-efficient, while being less time consuming and more accurate (Xiang et al, 2020). Further supporting this finding, Deloitte professionals state that through the use of NLP and ML there will be a more flexible array of options in where healthcare can be given in terms of location. The ability to offer remote healthcare ad diagnostics will lower cost, increase availability and allow physicians to leverage their time for the ‘top-of-license skill set” (Abrahms, 2019). While increasing availability this would also be more financially convenient for the patient as well as the providing institution because resources will be preserved while increasing availability.

Artificial Intelligence is promising to increase the ability to recognize meaningful patterns, advance the ability to prevent illness and in doing so Dr. Topol believes AI will help improve the accuracy of healthcare in diagnostic practices. This is supported by epidemiologists who believe AI can be used to map data with more precision in order to “monitor, control and prevent the spread of diseases” (Wiljer, 2019). This is done by using powerful analytical AI to recognize patterns in controlled situations and providing suggestions as to how to minimize the spread of a given disease. In addition to lessening the spread of disease through pattern recognition, research shows that diagnostic accuracy will be improved through implementing AI in fields such as histopathology. In reference to a study conducted in 2017, machine learning algorithms and AI combined were used to analyze 1,417 skin images to detect basal cell carcinoma and differentiate the malignant (harmful) from benign (not harmful) abrasions. The diagnostic accuracy of the automated analysis was >90%, which outperformed the prior results acquired by experts studying histopathology (Miller et al, 2017).

This research suggests that the accuracy of AI is greater than the accuracy of the conventional physician. In addition to increasing accuracy the research displayed an 85 percent decrease in human error (Miller et al, 2017) for analytical activities indicating that AI has a better chance of accurately analyzing images and diagnosing patients. These systems compare and store millions of images displaying tumor-positive and tumor-negative histological patches, the machines then use these data maps to compare the image at question and diagnose accordingly. According to Dr. Miller, the increased availability of data to train these AIs will result in higher diagnostic accuracy.

In addition to accurately analyzing medical images, research is proving that the capability of AI can also be used for mental health diagnostics. According to the National Journal of Medicine, depression affects approximately 6.8%-8.7% of the adult US population which results in over 8 million ambulatory emergencies (Miller et al, 2017). Furthermore, primary care institutions are not currently equipped to manage chronic depressive illnesses. After developing and applying deep learning to experimental studies it was apparent that through image mapping white matter and heat patterns, AI’s “recognition was 74% more accurate for predicting major depressive disorder” (Miller et al, 2017) which is significantly better than the traditional physician method being implemented. In addition to the diagnostic accuracy of AI, with the combination of AI and physicians’ studies showed a 2.5% rate of error from a sample of 1,600 mental health patients (Hernandez et al, 2017).

Provided this research it can be concluded that the accuracy of AI in diagnostic practices as well as analytical practices will impact healthcare positively by diminishing the rate of error. Although research has indicated the inability to achieve an 100% accuracy rate the marriage of physician care and the use of AI will “enhances system performance” (Miller et al, 2017) hence depicting an ample amount of evidence supporting AI’s impact on accuracy in healthcare being positive.

With the implementation of AI in healthcare confidentiality risks such as liability, security, and privacy factors arise. In addition to these factors, issues with data ownership, permission, and control also become relevant. Policy makers will also need to create rules in writing to enforce these issues (Feldman et al, 2019). Organizations require large data sets of electronic health records (EHRs). The adaptation of new policies will require the evolution of privacy policies for health care providers to ensure the confidentiality of patient data (Wiljer, 2019). In addition to the confidentiality ethical concerns also become relevant according to the American Journal of Medicine. The anticipated problems with this technology’s employment include lack of humanity in medicine, increased feeling of objectification, and loss of personal liberty as the machine will require patient data to build and improve AI technologies (Becker, 2019). Moreover, providers and distributors of data will need to consider “industry-level” security implementation to protect data from unauthorized access to patient data as well as the preservation of the integrity (Kamalnath, 2018).

There is a rapid increase in the demand for healthcare as population has been consistently increasing. As a result of the increasing demand, the cost of medical also increases. This demonstrates the necessity of a solution to decrease cost and the rate of medical error. With this in mind, AI uses NLP and ML to increase the abilities of the clinicians to lower costs and increase accuracy. The research displays a confidence in AI’s ability to increase diagnostic accuracy. Furthermore, the research finds that the cost of healthcare can also be lowered.

Abrams, K. (2019, January 1). The future of artificial intelligence in health care. Retrieved April 10, 2020, from
Becker, A. (2019). Artificial intelligence in medicine: What is it doing for us today? Health Policy and Technology, 8(2), 198–205.
Feldman, R. C., Aldana, E., & Stein, K. (2019). Artificial Intelligence in the Health Care Space: How We Can Trust What We Cannot Know. Stanford Law & Policy Review, 30(2), 399–419.
Hammerling, J. (2012). A Review of Medical Errors in Laboratory Diagnostics and Where We Are Today: Table 1. Laboratory Medicine, 43(2), 41–44.
Hernandez, I., & Zhang, Y. (2017). Using predictive analytics and big data to optimize pharmaceutical outcomes. American Journal of Health-System Pharmacy, 74(18), 1494–1500.
Insel, T. (2019, March 11). How Algorithms Could Bring Empathy Back to Medicine. Retrieved April 9, 2020, from
Kamalnath, A. (2018). Rethinking Liability and Licensing for Doctors in the Era of AI: Insights from Company Law. Asia Pacific Journal of Health Law & Ethics, 11(2), 33–50.
Miller, D., & Brown, E. (2017, November 7). Artificial Intelligence in Medical Practice: The Question to the Answer? Retrieved April 9, 2020, from
Simshaw, D., Terry, N., Hauser, K., & Cummings, M. (2016). Regulating Healthcare Robots in the Hospital and the Home: Considerations for Maximizing Opportunities and Minimizing Risks
Wiljer, D., & Hakim, Z. (2019). Developing an Artificial Intelligence–Enabled Health Care Practice: Rewiring Health Care Professions for Better Care. Journal of Medical Imagine and Radiation Sciences, 50(4). doi:
Xiang, Y. (2019, December 18). Implementation of Artificial Intelligence in Medicine: Status Analysis and Development Suggestions. Retrieved April 9, 2020, from