Advantages and Risks of AI Usage in Healthcare

The healthcare sector can significantly benefit from the potential influence of generative artificial intelligence and machine learning applications. Patients, medical providers, and insurance companies can all reap the rewards of the widespread use of these technologies, including improved operational efficiencies and better treatment outcomes. However, all parties must consider the potential risks of deploying these cutting-edge tools in the medical field.

Providing healthcare services

The adoption of artificial intelligence (AI) in healthcare is on the rise, demonstrating effectiveness in diagnosing chronic ailments, enhancing staff efficiency, and improving the overall quality of care. AI tools are being leveraged for patient diagnosis, drug discovery and development, enhancement of physician-patient communication, and medical transcription tasks. Particularly in areas where large sets of data, including images, exist, AI has shown proficiency in diagnosing conditions that require visual comparisons. For instance, AI systems developed by Google and Stanford University have been successful in diagnosing and grading diabetic retinopathy and detecting 14 pathologies via X-ray review, respectively.

In enhancing patient interaction experiences, AI assistants and chatbots help patients locate physicians, schedule visits, and answer common queries. Healthcare providers also increasingly rely on AI to identify treatment protocols and the appropriate medication, thereby promoting efficiency. These AI-enabled assistant tools also help document patient interactions in almost real-time, aiding in the betterment of the patient documentation process and reducing the burden associated with the time-intensive documentation task. They are also being employed like insurance verification tools to secure prior authorization for procedures, limiting the occurrence of unpaid claims.

Despite the accuracy and efficiency provided by AI in healthcare, public sentiment needs to be more accepted. A survey by the Pew Research Center reveals that a significant proportion of Americans would be uncomfortable if their healthcare providers relied on AI for diagnosis or treatment suggestions. As per the survey data, 60% of respondents were uneasy about such a possibility, 57% believed that AI could negatively alter the patient-provider relationship, and just 38% conveyed optimism that AI could lead to improved health outcomes.

Racial and Gender Bias

Concerns have been raised about the potential of racial and gender bias in the algorithms of artificial intelligence (AI) used in healthcare. Studies have shown discrepancies based on race in these algorithms and gaps due to insufficient healthcare data for women and minority populations. In a May 2022 report, Deloitte emphasized the need to reevaluate pre-existing clinical algorithms to ensure equitable patient care. This would involve scrutinizing how race is factored into the algorithms and whether such inclusion is justified.

There are also issues with the collection and usage of data concerning race and ethnicity in healthcare, as noted by the Deloitte report. Owing to the absence of standardized methodology and existing misconceptions, substantial data about race and ethnicity needs to be included in healthcare. This was highlighted by the Centers for Disease Control and Prevention, who found that nearly 40% of the people testing positive for COVID-19 or receiving a vaccine did not have available race and ethnicity data. The American Medical Association (AMA) upholds the importance of using population-representative data in AI development and healthcare applications and advocates for using augmented AI rather than fully automated alternatives.

Regulators have also taken cognizance of the bias possible in healthcare AI. California Attorney General Rob Bonta issued letters to 30 hospital CEOs nationwide, seeking information about measures to identify and rectify racial and ethnic disparities in commercial decision-making tools. This is part of investigating the potential discriminatory implications of commercial healthcare algorithms based on race and ethnicity. However, the Pew Research Center poll suggests that a majority of Americans who acknowledge racial and ethnic bias in healthcare also believe that AI usage can help mitigate bias and unfair treatment.

Privacy of Health Data

Health data privacy in the context of AI applications in healthcare poses serious concerns. The development and operation of AI algorithms necessitate access to extensive health data, potentially exposing this sensitive information due to the algorithm's data retention capabilities or vulnerability to third-party data breaches. Partnerships between academic research centers and private-sector companies for AI research commercialization have sometimes resulted in inadequate data privacy protections. Such collaborations have led to instances where patients' control over their data usage was undermined or needed to be more adequately informed about privacy implications.

Furthermore, AI tools have demonstrated the capacity to re-identify individuals from anonymized health data repositories, even predicting personal non-health-related data in specific scenarios. This re-identification ability further accentuates privacy concerns. The healthcare sector and associated third-party vendors are susceptible to data breaches and ransomware attacks. The healthcare industry reported the most expensive data breaches, with an average cost of $10.93 million in 2023, per IBM Security's Cost of a Data Breach Report.

Despite these prevalent privacy concerns, advancements are being made to safeguard individual privacy in the growing field of healthcare AI. Ten U.S. states have instituted AI-related regulations within their more extensive consumer privacy laws, specifically focusing on data privacy and AI applications in healthcare. As healthcare continues integrating AI technology, all involved entities must recognize potential risks of bias or privacy loss and take proactive measures to mitigate them, thus ensuring the vast potential benefits for patients and providers.