There is a grand challenge in the healthcare system of utilizing information technology, namely artificial intelligence, to enhance outcomes and decrease costs. The United States continues to spend money to address the COVID-19 pandemic and other challenges, but the outcomes do not follow suit.
The US spends 16.9 percent of its gross domestic product on healthcare, yet it has the lowest life expectancy and highest suicide rates in comparison to 10 other countries with similar economic development.1 This was corroborated by a 2016 study whereby the US was found to spend more than twice as much as other countries yet had worse healthcare outcomes.2
Included in these healthcare costs are the administrative costs associated with running the operations. Here, too, the costs are higher in the US than in other countries. While the US spends 8 percent of healthcare money on administrative costs, other countries spend 1-3 percent.3 The National Health Expenditure Accounts’ research in 2018 found almost an exact figure at 8.5 percent.4 Overall costs, administrative costs, and outcomes need to be addressed to develop a more efficient healthcare delivery system.
AI to Reduce Costs
Information technology platforms, such as decision support systems and electronic health records (EHRs), have played a role in cost containment. EHRs that contained clinical decision support alerts were used to prevent errors and adverse medication reactions, which, in turn, saved money.5 A study by Lewkowicz et al. found that clinical decision systems reduced waste, which could be equated to cost savings.6 Studies on EHRs did not address administrative costs and thus falls short of combatting our initial grand challenge.
Artificial intelligence (AI) may hold the answer. AI is relatively new to healthcare, but it has multifaceted uses that impact costs, delivery, and quality. AI uses predictive analytics and, thus, it can have an effect from the very onset of the patient encounter.
AI can have a significant impact from the moment a patient makes the decision to encounter a provider (and even prior to their decision). The patient can register at a kiosk using fingerprint technology. Their data are brought forth both on their reason for the visit and their method of payment. Having the healthcare and insurance information linked will generate whether the patient has a copayment and, if so, how much. Once they have checked in, the provider has their medical information ready at hand—all of this without encountering a person.
We have learned in the pandemic about contactless care and restructuring the workforce such that jobs are not periled. Self-registration kiosks have been shown to increase patient satisfaction and efficiency and decrease costs.7 Providence St. Joseph Health found that for each visit scheduled with automation, it resulted in a cost savings of $3 to $4 for the appointment.8 Rather than have employees registering patients, which a kiosk can do, the employees can answer questions on insurance or convert to a clinical position.
Putting the patient in charge also occurs with wearable device technology. The device tracks the patient health information and alerts the patient and provider of any data, such as early symptoms, that raises a concern and necessitates a visit or a change in their plan. With this model of predictive analytics, intervening with early onset symptoms saves money and lives.
AI Data-Protection Challenges
With artificial intelligence capturing loads of data, concerns have arisen about the protection of the data. The Health Insurance Portability and Accountability Act (HIPAA) became law in 1996. Over the years, it has expanded to cover multitudes of situations. In 2003, the Privacy Rule was implemented and, six months later, the Transactions and Code Sets standards were implemented. This was followed in 2004 by the Standard Unique Identifier for Employers Rule. The next year, 2005, saw the Security Rule implemented, followed in five short months by the Claims Attachment Standards Rule. Finally, in 2006, the Enforcement Rule was implemented to establish how HIPAA would be managed and what the accompanying penalties by the federal government would be for noncompliance.
In 2007, the Standard Unique Healthcare Provider Identifier Rule, also known as the National Provider Identifier (NPI) number, was introduced. Then, in 2009, the Health Information Technology for Economic and Clinical Health Act (HITECH) was passed, which increased the fines for HIPAA violations. Along with HITECH in 2009 was the Breach Notification Rule, which covered what needed to be done if data was mishandled. Finally, in 2013, the Omnibus Rule was passed, which, 10 years later, strengthened the Privacy Rule and established the Genetic Information Nondiscrimination Act (GINA).
The challenge, though, becomes that, since 2013, HIPAA has had minor changes—just as the healthcare field is changing dramatically—all the more bolstered by artificial intelligence. Given the constant emergence of artificial intelligence applications in healthcare, HIPAA must incorporate changes that take into account all that is occurring. For example, are the self-registration kiosks that contain patient data or the AI computer applications that analyze patient data really business associates under HIPAA? Presently, the law does not deem them as such, but given that HIPAA was developed in the 1990s, AI uses in healthcare were very different and minimal as compared to the present.
In addition, wearable devices such as Fitbits and smart watches that collect patient health information from the wearer (i.e., the patient) are not covered under HIPAA, as the wearable device is the property of the wearer. Yet the data are transmitted and stored in large databases. Who sees that data? How are the data protected in accordance with the privacy and security standards? Here again, more challenges, from a regulatory standpoint, to conquer as AI continues to reinvent healthcare.
AI for the Investigation of Pathology Records and Tissue Repositories
With the increased accumulation of existing biomedical/registry data and tissue archives, it is becoming clear that exploration of these repositories is necessary to inform future research and guide biomedical discoveries. Healthcare facilities around the world are accumulating these data and rarely have capabilities to analyze them. These repositories are especially valuable in analyzing rare conditions, such as gynecologic malignancies, as few centers have a large number of these cases. Gynecologic cancers pose significant morbidity and mortality burden for US women. Endometrial cancer is the most common gynecologic malignancy in the US, while ovarian cancer is one of the deadliest. At the UPMC Hillman Cancer Center and University of Pittsburgh, investigators explored the feasibility of linking this well-curated, structured cancer registry data with unstructured text (i.e., pathology and radiology reports) using the Text Information Extraction System (TIES).
The TIES platform has been used to integrate breast cancer cases from the UPMC Network Cancer Registry system and then combine these data with other EHR data as a pilot use case that can be replicated for other cancers.9 Investigations of this nature hold a great promise to jump-start AI investigations on identifying which treatment options are most optimal for the vulnerable groups of patients with rare diseases. Also, the development of natural language processing (NLP) tools to facilitate automatic/unsupervised/minimally supervised extraction of specific discrete cancer-related data from various types of unstructured electronic medical records has a great potential to deliver high impact projects at a relatively low cost.
AI for Clinical Decision-Making
Electronic systems that support clinical decision-making and clinical processes often use AI methods to represent information, retrieve data, and develop functional and reliable decision support tools. With the evolution of digital and communication technologies plus innovative software methods, the ability to offer high-quality support to clinicians has resulted in impressive new capabilities and several commercial products.10 Examples of the existing use of AI in clinical practice include AI assisted robotic surgeries, use of AI algorithms to detect a disease (such as skin cancer), precision medicine and its implications to detecting disease risk, as well as drug discovery and drug compatibility issues. Recent review summarized the latest developments of applications of AI in biomedicine, including disease diagnostics, living assistance, biomedical information processing, and biomedical research.11
AI for Pandemic Surveillance
As demonstrated with the COVID-19 pandemic, infectious disease remains a pressing concern to public health. Timely, reliable, and coordinated outbreak response activities hold a great potential to improving population health. As of today, most infectious disease response programs, including COVID-19, are plagued by incomplete data, disconnected datasets, poor data communication efforts, misunderstanding of the problem, etc. A decade ago, influenza season forecasts were generated using web-based tools,12 initiating a quest for AI tools for epidemic modeling tools. Some AI-based algorithms have been showing efficacy in the fight against COVID-19, but wider implementation of these tools requires prospective validation studies of their performance.13 To help facilitate the use of AI throughout the pandemics, policymakers should encourage the sharing of medical, molecular, and scientific datasets and models on collaborative platforms to help AI researchers build effective tools for the medical community, and should ensure that researchers have access to the necessary computing capacity.14
Clearly, the grand challenge can be met with the use of artificial intelligence. Both pre-pandemic and post-pandemic, the utilization of artificial intelligence will effectuate the delivery of healthcare.
1. Commonwealth Fund. U.S. Health Care from a Global Perspective, 2019: Higher Spending, Worse Outcomes? Issue Briefs January 30, 2020 https://www.commonwealthfund.org/publications/issue-briefs/2020/jan/us-health-care-global-perspective-2019
2. Feldscher, Karen. What’s Behind High U.S. Healthcare Costs, The Harvard Gazette, March 13, 2018. https://news.harvard.edu/gazette/story/2018/03/u-s-pays-more-for-health-care-with-worse-population-health-outcomes
4. Tollen, L, Keating, E, Weil, A, How Administrative Spending Contributes To Excess US Health Spending. Health Affairs Blog, February 20, 2020. https://www.healthaffairs.org/do/10.1377/hblog20200218.375060/full/
5. Jason, C., Clinical Decision Support in the EHR Cuts Healthcare Costs. EHR Intelligence Use and Optimization News, September 21, 2020. https://ehrintelligence.com/news/clinical-decision-support-in-the-ehr-cuts-healthcare-costs
6. Lewkowicz, D., Wohlbrandt, A., and Boettinger, E. Economic impact of clinical decision support interventions based on electronic health records. BMC Health Services Research, 20:871, 2020. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7491136/
7. The Benefits of Self Registration Kiosks, June 11, 2020. https://stablerise.com/blog/the-benefits-of-self-registration-kiosks/
8. Arndt, R. Health systems save money using digital tools for scheduling appointments, administrative work. Modern Healthcare, July 7, 2018. https://www.modernhealthcare.com/article/20180707/TRANSFORMATION02/180709982/health-systems-save-money-using-digital-tools-for-scheduling-appointments-administrative-work
9. Linkov F, Silverstein JC, Davis M, Crocker B, Hao D, Schneider A, Schwenk M, Winters S, Zelnis J, Lee AV, Becich MJ. Integration of Cancer Registry Data into the Text Information Extraction System: Leveraging the Structured Data Import Tool. J Pathol Inform. 2018 Dec 24;9:47. doi: 10.4103/jpi.jpi_38_18. PMID: 30662793; PMCID: PMC6319041. https://pubmed.ncbi.nlm.nih.gov/30662793/
10. Shortliffe EH, Sepúlveda MJ. Clinical Decision Support in the Era of Artificial Intelligence. JAMA.2018;320(21):2199–2200. doi:10.1001/jama.2018.17163
11. Guoguang Rong, Arnaldo Mendez, Elie Bou Assi, Bo Zhao, Mohamad Sawan. “Artificial Intelligence in Healthcare: Review and Prediction Case Studies,” Engineering, Volume 6, Issue 3, 2020, Pages 291-301, ISSN 2095-8099, https://doi.org/10.1016/j.eng.2019.08.015
12. Jeffrey Shaman, Alicia Karspeck. “Forecasting seasonal outbreaks of influenza,” Proceedings of the National Academy of Sciences, 2022, https://doi.org/10.1073/pnas.1208772109
13. Danai Khemasuwan, Henri G Colt, “Applications and challenges of AIbased algorithms in the COVID-19 pandemic,” BMJ Innov, 2021, https://innovations.bmj.com/content/bmjinnov/7/2/387.full.pdf
14. OECD Policy Responses to Coronavirus (COVID-19), “Using artificial intelligence to help combat COVID-19,” 2020, https://www.oecd.org/coronavirus/policy-responses/using-artificial-intelligence-to-help-combat-covid-19-ae4c5c21/
Joan M. Kiel is a professor in the Department of Health Administration & Public Health and the chairperson of University HIPAA Compliance at Duquesne University.
Faina Linkov is an associate professor and chairperson in the Department of Health Administration & Public Health at Duquesne University.
Read a white paper on best practices for clinical documentation integrity and how advances in artificial intelligence technology are transforming CDI and the role of clinical documentation specialists (CDSs).
Take the CE Quiz