Delving into AI Basics, Dispelling Myths, and Ensuring ROI in Healthcare Integration
Artificial Intelligence (AI) presents an exciting opportunity for revolutionizing healthcare, but for many healthcare professionals, it remains a perplexing and elusive technology. To understand the full measure of its potential, we must demystify its intricacies, gain insights into how healthcare institutions can formulate effective AI strategies, evaluate AI's compatibility with existing workflows and technology platforms, and, most importantly, quantify the genuine Return on Investment (ROI) that AI can deliver to the healthcare sector.
Deciphering the AI Puzzle: Grasping the Fundamentals
At its core, AI refers to computer programs capable of executing tasks that typically necessitate advanced cognitive processes, including perceptual learning, memory organization, and critical reasoning. In 1955, John McCarthy, who coined the term “artificial intelligence,” defined AI as “the science of making machines do things that would require intelligence.”
However, AI is not a monolithic concept; it encompasses various approaches and types, each tailored to address specific challenges and scenarios (and certain approaches can fall into multiple categories):
- Rules-based (commonly not thought of as AI): This approach relies on predefined rules and logic. It is commonly seen in Robotic Process Automation (RPA) and is brittle when it comes to gray areas.
- Statistical Inference: Statistical models and data analysis generate predictions based on historical data patterns.
- Generative AI: Generative models create novel data or content that emulates human creativity, such as generating art, music or text.
- Natural Language Processing (NLP) and Natural Language Understanding (NLU): Focus on enabling machines to comprehend, interpret, and respond to phrases rather than just individual words. For example, differentiating “sepsis” vs. “unlikely to be sepsis”.
- Machine Learning (ML) and Deep Learning (DL): These are subsets of AI that deploy advanced statistical models and artificial neural networks to empower machines to learn from data and enhance their performance over time.
Dispelling AI Misconceptions: A Multifaceted Reality
One of the prevalent misconceptions surrounding AI is the belief in a one-size-fits-all approach. In reality, the effectiveness of an AI approach—or a combination of approaches—depends on the specific problem it aims to solve. Healthcare institutions must carefully evaluate the nature of the issue at hand and tailor their AI strategy accordingly.
Moreover, terms like "NLP" or "AI" can be misleading. Not all companies claiming to offer NLP or AI solutions employ the same technology or methodologies. It is crucial to understand that NLP encompasses various methodologies, each yielding distinct results. Consequently, healthcare leaders must exercise due diligence in comprehending the technology behind the label.
Additionally, developing robust AI systems demands substantial expertise and experience, and not all companies marketing AI possess the necessary capabilities. Unlike regulated food labels governed by the USDA, AI technology lacks standardized nomenclature, allowing companies to label their products as they see fit. This absence of regulation necessitates thorough scrutiny and evaluation of AI providers to ensure their technology aligns with the healthcare institution's objectives and requirements.
Moving Past Hype and Hastiness in AI Implementation
While AI holds immense potential in healthcare, its adoption should not be rushed or driven solely by industry hype. One critical area where AI can significantly contribute is in mitigating physician burnout—a pervasive issue in healthcare. AI can streamline documentation processes, enhance efficiency, and reduce the administrative burden on healthcare professionals. However, history has shown that hastily implementing large-scale AI initiatives can lead to regrets and challenges in garnering buy-in or measuring outcomes.
Thus, for hospitals with limited operating margins, embarking on AI implementation through broad strategies with vague outcome measurements is not advisable. Instead, a more prudent approach involves implementing focused solutions that impact a select group of users and processes, enabling a clear and meaningful definition of ROI. Healthcare institutions can further mitigate financial risks by structuring contracts with vendors on a contingency basis, departing from traditional SaaS contracts.
One example of a narrow but effective process is using AI for prebill review, which involves scanning charts for additional documentation and coding opportunities after final coding but before the bill is dropped. This is effective because the endpoints are clear (revenue and quality from added diagnoses that are net new relative to what was captured); the users and processes are narrowly defined (a Clinical Documentation Improvement Specialist or Medical Coder that works a QA process); and the operational risks are low (no rip-and-replace of existing solutions). Hospitals that have implemented AI in this stage have realized millions of dollars in revenue lift with minimal operational or financial risk.
Embracing AI with Clarity and Purpose
AI in healthcare possesses the potential to revolutionize patient care and alleviate the burdens on healthcare professionals. However, realizing this potential requires a clear understanding of the technology, strategic planning, and the ability to effectively measure ROI. Healthcare leaders must actively engage in the discourse on AI regulation and best practices, ensuring that AI becomes a valuable and trustworthy ally in the continually evolving healthcare landscape. By doing so, healthcare institutions can unlock AI's genuine potential and deliver enhanced care to patients while optimizing their operations.
Michael Gao, MD, is CEO and co-founder of SmarterDx.
###