Skip to main content
Offcanvas
Some text as placeholder. In real life you can have the elements you have chosen. Like, text, images, lists, etc.
What are you looking for?

Promoting Health Equity Through AI Integration: Navigating Challenges and Embracing Regulation

Blog

US biopharmaceutical companies spent about $1 billion to bring each of their new drugs to market between 2009 and 2018[1]. Considering this substantial cost, it is essential to utilize technology to minimize disruptions and control data validity for more effective trials. The integration of artificial intelligence (AI) in healthcare holds immense potential in making the drug development process more effective and to ensure successful outcomes for patients.

There has been an explosion in the collective psyche understanding of AI, but industry is still working to fully understand how it may help within a healthcare setting. Warren Whyte, vice president and head of ERACE at ConcertAI; and Nick Kenny, chief scientific officer at Syneos Health®, met at the American Society of Clinical Oncology (ASCO) Annual Meeting and had a discussion to clarify if– and how–AI can enhance decision-making and ensure inclusivity in healthcare processes. To take full advantage of these powerful tools, there is a need for effective regulation, data validation and the mitigation of bias to ensure responsible and equitable use of AI.

Using Technology for Informed Decision-Making

Leveraging AI and data-driven approaches to inform decision-making in all stages of the drug development lifecycle is imperative, and offers some promise to enhance inclusivity of patient recruitment. The lack of diversity in clinical trials is a long-standing issue, impacting the generalizability of research findings and exacerbating health disparities. By synthesizing different datasets such as social determinants information, claims data and patient registry data, AI models offer an opportunity to provide insight into patient demographics and optimize selection of clinical trial locations.

This approach holds the potential to drive:

  • More inclusive trial participation, particularly among underrepresented communities
  • Insight for providers to better aid in patient recruitment for clinical trials
  • Greater awareness on the overall sentiment in the community about participating in trials
  • Tailored communication with patients so that they may better understand their condition, treatment options and the benefits of clinical trial participation.

AI offers the prospect to healthcare providers to be able to scan a large range of resources to quickly assimilate thousands of articles and publications for quicker, in-depth insights to inform treatment decisions. The ability to search social determinant, claims, patient registry information and community assessment reports helps inform providers to make predictions about an individual’s care, lifestyle and circumstances that affect their overall health. In addition to improved patient recruitment for clinical trials, AI holds promise to augment, support and expedite clinical decision making, for faster diagnoses, better patient-specific treatment options, and a more informed understanding of outcomes associated with treatments.

Data Validation and Data Biases

While the potential of AI in healthcare seems infinite, it is important to acknowledge the presence of biases and limitations associated with data informed decision making. To ensure the accuracy and trustworthiness of insights provided by AI, rigorous data validation and identification of source origination is required. Mitigating biases, during the development and training of AI models will require proper regulation to guide development and deployment, including improved inclusion of more diverse data sets and the removal of human created medical algorithm biases.

AI systems rely on incoming data sets to digest information and provide insights. It is important to understand source data–what it is, what it isn’t–to be able to train and retrain an AI tool. Misleading or inaccurate clinical information, provided deliberately or not, creates inherent bias within any analysis that AI systems produce, which will potentially lead to less inclusive clinical trials. Types of data biases that need to be considered in drug development, include:

  • Sampling bias – the population of interest in that data set is not representative or reflective of the population the study is examining
  • Selection bias – protocols are written in such a way that the criteria exclude certain groups from being able to participate
  • Algorithmic bias – the system that is generating a result shows an implicit value that was based off of the creator
  • Human bias – the source data provided to the AI included inherent bias, resulting in that bias informing all analysis

For accurate and useful trial results, it is important to understand an AI system’s data source, be able to validate that source, recycle the information and understand the training algorithms that have been used.

The Fourth Industrial Data Revolution

According to FDA Commissioner, Robert Califf, we are entering a “fourth industrial revolution” with diverse sources of data made available in real time, including large language models, based on massive databases and parameters[2]. However, he warns that the widespread medical misinformation resulting from data biases, is the “leading cause of preventable death,” and will need to be regulated strictly to see the benefits of AI[3].

Such regulation needs to consider the intended use of data and AI systems, distinguishing between those that provide insights and those that automate clinical decisions. Striking a balance between providing guidelines and avoiding excessive restrictions is a key challenge. Clear data standards can help ensure responsible and equitable application of AI by ensuring source data is complete and accurate. This can be accomplished by:

  • Establishing guidelines and standardizations of sources to ensure it meets a minimum threshold in regard to data fidelity
  • Using data imputation to retain the majority of the dataset's data and information by substituting missing data with a different value.
  • Combing data sets with benchmarking data
  • Leveraging other data sources outside of the study organization to ensure that information captured is similar in terms of the current benchmarks.

If used wisely, data can inform the entire lifecycle of the drug development process–from patient recruitment processes to the benefits and risks of utilizing medical interventions for both HCPs and patients. “If we’re not nimble in the use and regulation of large language models, we’ll be swept up quickly by something that we hardly understand,” Califf said. “The regulation of large language models is critical to our future”[4].

Toward a Bright Future of AI

The potential of AI in healthcare is undeniable, and includes opportunity around health equity, enhancing clinical trial diversity and improving decision-making. To realize these benefits, it is imperative to address biases, validate data sources and implement effective regulation. By combining AI's power with rigorous data validation and control, intentional planning and community outreach, healthcare stakeholders can foster inclusivity and ensure equitable patient care. The journey towards harnessing AI's potential must be guided by a comprehensive regulatory framework that ensures patient safety, minimizes biases and promotes the responsible use of this transformative technology.


Sources

  1. Gardner J. New estimate puts cost to develop a new drug at $1b, adding to long-running debate. BioPharma Dive. March 3, 2020. https://www.biopharmadive.com/news/new-drug-cost-research-development-market-jama-study/573381/#:~:text=from%20your%20inbox.-,New%20estimate%20puts%20cost%20to%20develop%20a%20new%20drug%20at,adding%20to%20long%2Drunning%20debate&text=U.S.%20biopharmaceutical%20companies%20spent%20about,published%20in%20JAMA%20on%20Tuesday
  2. Lawrence L. “Hard to catch up”: FDA commissioner on Regulating New Digital Health Tools. STAT. May 9, 2023. https://www.statnews.com/2023/05/09/fda-ai-digital-health-care/
  3. McNamara D. “Exciting time”: FDA commissioner Talks Ai and misinformation. WebMD. May 31, 2023. Accessed July 18, 2023. https://www.webmd.com/a-to-z-guides/news/20230530/fda-commissioner-talks-ai-and-misinformation
  4. Payerchin R. Ai could revolutionize health care, but regulation is needed, FDA commissioner says. MedicalEconomics. May 15, 2023. Accessed July 18, 2023. https://www.medicaleconomics.com/view/ai-could-revolutionize-health-care-but-regulation-is-needed-fda-commissioner-says

Contributors

Nick Kenny, Chief Scientific Officer, Syneos Health

Interested in Syneos Health?