Practice Tips

Online library

Artificial Intelligence in Healthcare

Artificial Intelligence (AI) a technology that simulates the performance of complex tasks (reasoning, solving problems, or making decisions) that normally require a human to perform, is rapidly transforming healthcare delivery. This technology is increasingly integrated into various healthcare applications from patient records management to diagnostic procedures. Some organizations have welcomed the technology while others are taking a slower approach. It may surprise clinicians that they are probably already using AI indirectly through vendors who have incorporated various uses of AI into the products they offer. As AI becomes more commonplace, healthcare professionals and organizational leaders need to educate themselves and develop strong risk mitigation strategies to keep their patients and organizations safe from liability.

Types of AI in Healthcare

There are several AI categories classified by system type and level of intelligence. Most AI tools currently used in healthcare fall into the “limited memory” machine-learning categories and can be seen within several fields such as clinical documentation, imaging diagnosis, patient monitoring, and even assisting surgeons in procedures (AI-powered robotics).

Generative AI

Generative AI uses algorithms to create new content, which can be in the form of language, audio, images, and videos. Generative AI applications like ChatGPT and ambient clinical documentation can allow providers to spend more time with patients and less time on paperwork. During patient encounters, AI tools can listen and generate draft notes, significantly easing the documentation burden for clinicians.

Risk Management Consideration:
AI-generated notes can contain the same types of documentation errors as person-generated and dictated/transcribed documentation. The clinician is responsible for carefully reviewing the content to ensure accuracy prior to authentication. Use of a templated statement within your note indicating there may be typos and inaccuracies related to the use of AI documentation is strongly discouraged. This type of well-meaning statement will erode trust in the accuracy of your documentation, leaving you open to liability issues.

Predictive AI

Predictive AI, also known as predictive analytics, analyzes data to identify trends and patterns, make future predictions, anticipate events and behaviors, and diagnose conditions. In healthcare, organizations are using this type of AI to improve outcomes with falls, sepsis, clinical deterioration, and diagnostic imaging.

Risk Management Consideration:
As a rule, clinical decision support tools such as predictive AI models should not be a replacement for medical decision making. They are meant as a guide/prompt but do not relieve the provider from responsibility for the patient's outcome.

Implementing AI Safely

As AI transforms healthcare delivery, it is important to develop a comprehensive multidisciplinary approach to ensure this innovation does not have unanticipated consequences for patients, clinicians, and organizations.

Risk Assessment:
Before deploying AI technologies, conduct a thorough risk analysis focusing on potential impacts on patient populations, possibilities for bias, and error rates. This assessment should also extend to vendor-supplied AI tools, which may affect your organization directly.
Policy & Procedure:
Develop comprehensive policies to govern AI usage that comply with legal standards and regulatory standards and ensure equitable care. New AI initiatives should begin with a detailed risk analysis to prepare for and mitigate potential adverse effects. The policy should include that staff can only use technologies fully vetted and adopted by the organization.
Quality Improvement:
Integrate AI applications into a comprehensive quality improvement program including the broader organizational Quality Assurance and Performance Improvement (QAPI) programs. This integration ensures that AI-driven innovations align with overall quality improvement goals and are monitored for effectiveness and safety.
Machine-Learning Data:
As healthcare organizations adopt predictive analytics AI, it is crucial to ensure that the data used for machine learning is current, unbiased, and relevant. For instance, if a predictive analytics AI tool is being used to diagnose clinical deterioration in pediatric patients, the dataset for AI training must not be limited to adult data. It is essential to continuously update the data to maintain accuracy and reliability.
Regulatory Compliance:
Maintain vigilance for emerging regulatory requirements that are expected to have continued focus from the federal government. With AI's rapid advancement, regulatory landscapes are continually evolving.
HIPAA Compliance:
With the proliferation of AI in clinical documentation, maintaining HIPAA compliance is crucial. Identify and address any potential risks to patient data privacy and security associated with AI tools.

As AI continues to advance, staying informed and proactive is essential for healthcare organizations and professionals. Keeping abreast of the latest research literature and case studies will help the healthcare community harness AI's potential responsibly and innovatively. By embracing these technologies thoughtfully, healthcare providers can enhance patient care, improve operational efficiencies, and navigate the complexities of modern healthcare environments.


HealthTech. (2023 July). What Types of AI Are Being Used in Healthcare?

The White House. (2023 October 30). Executive Order on the Safe Secure and Trustworthy Development and Use of Artificial Intelligence. (2024 February). Generative AI vs Predictive AI: A comparative guide.