A recent study by the National Institutes of Health (NIH) looked at how well an AI system called GPT-4V works in clinical diagnoses. GPT-4V answered 207 medical quiz questions from the New England Journal of Medicine’s Image Challenge. It showed a high rate of correct diagnoses by using clinical images and text summaries.
When doctors had to rely only on memory without any reference materials, GPT-4V did better at choosing the right diagnosis than the doctors. But when doctors were allowed to use reference materials, they performed better than the AI, especially on harder questions. This shows both the strengths and limits of AI in medical settings.
Even though GPT-4V often picked the right answers, it struggled to explain why. It also made mistakes when describing clinical images, especially when lesions or conditions looked similar from different angles. This means that the AI does not fully understand the context, and human experience still matters a lot.
Dr. Stephen Sherry, Acting Director at the National Library of Medicine, said, “AI can help medical professionals diagnose patients faster and start treatment sooner. But it cannot replace the detailed knowledge and skill of human doctors yet.”
For healthcare managers, the message is clear: AI can help speed up diagnosis, but people must still check its work to make sure it is safe and accurate.
AI can improve clinical workflows and patient care in several ways:
Despite the benefits, the NIH study and other research show some important limits and risks of AI in healthcare:
Healthcare leaders must know that using AI in clinics and hospitals requires following changing rules and ethical standards. Recent studies point to the need for strong governance to make sure AI is used safely and fairly.
Key rules focus on protecting patient data, making AI decisions clear, and having people oversee AI results. Groups like the National Library of Medicine (NLM) do research to help guide safe AI use.
Health authorities require licensing, certification, and ongoing checks of AI systems to reduce risks from bias, mistakes, or unintended effects. Healthcare administrators should work with legal and IT experts to make sure AI use follows these laws and rules.
AI also helps automate many clinical and administrative tasks in healthcare. This is important for healthcare managers who want to improve how their offices and hospitals run.
AI that uses Natural Language Processing (NLP) can handle routine patient calls and messages. This includes appointment reminders, symptom checks, and answers to common questions. Using AI to manage phone systems reduces the work for office staff because AI can handle many calls quickly and correctly.
In the U.S., where many phone calls come for appointments, prescription refills, or billing, AI-powered answering systems work 24/7. These systems help improve patient satisfaction by giving faster service.
Robotic Process Automation (RPA), a type of AI, automates repetitive tasks like billing, managing claims, and setting appointments. This leads to fewer errors and less manual work. Faster billing helps keep money flowing and lowers claim rejections.
RPA also frees up staff time to focus on patient care tasks that need human attention.
AI uses predictive analytics to warn healthcare managers about expected patient visits. This helps with planning staff schedules and managing resources in advance. This is especially useful for big clinics and hospitals where patient numbers change a lot and affect costs and care quality.
AI can connect with EHR systems to give doctors real-time suggestions and patient history information. This makes workflows smoother because doctors do not have to check many separate systems while treating patients.
Using AI in clinical work needs to follow strict data security and privacy rules. Healthcare data is sensitive and must be protected. Programs like the HITRUST AI Assurance Program set guidelines for managing risks, transparency, and compliance with security standards.
HITRUST works with cloud companies like AWS, Microsoft, and Google to make sure AI systems meet security rules. Healthcare leaders should choose AI partners and systems that meet these standards to keep patient data safe and build trust.
Healthcare administrators, owners, and IT managers should take a careful approach when bringing in AI:
AI use in healthcare and clinical decisions is still growing. Evidence from NIH and other studies shows AI can speed up diagnoses and help with administrative jobs that take a lot of staff time.
But AI also has current weaknesses, like trouble explaining its answers and ethical risks. This means healthcare leaders should bring in AI carefully, using facts and tests to guide decisions. They should keep checking and improving AI as rules and clinical knowledge change.
The U.S. healthcare system, with many patients and complex setups, could gain a lot from using AI to help make decisions and handle workflows. Still, human skill, good management, and following laws and ethics are key to making AI useful for safe and good patient care.
By using AI thoughtfully, healthcare providers can improve care, run operations better, and serve patients well in today’s busy medical world.
The NIH study found that the AI model GPT-4V performed well in diagnosing medical images but struggled with explaining its reasoning, highlighting both its potential and limitations in clinical settings.
The AI selected correct diagnoses more frequently than physicians in closed-book settings, while physicians using open-book resources performed better, particularly on difficult questions.
The AI often misinterpreted medical images and failed to correlate conditions despite accurate diagnoses, demonstrating gaps in its interpretative capabilities.
It’s crucial to assess AI’s strengths and weaknesses to understand its role in improving clinical decision-making and ensure effective integration into healthcare.
The study was led by researchers from NIH’s National Library of Medicine (NLM) in collaboration with several prestigious medical institutions including Weill Cornell Medicine.
The tested model was GPT-4V, a multimodal AI capable of processing both text and image data, relevant to diagnosing medical conditions.
NLM supports biomedical informatics and data science research, aiming to improve the processing, storage, and communication of health information.
Despite AI’s capabilities, human experience is essential for accurately diagnosing patients, as AI may lack contextual understanding necessary for correct interpretations.
Further research is required to compare AI capabilities with those of human physicians to fully understand its potential in clinical settings.
The findings suggest that while AI can enhance diagnosis speed, its current limitations necessitate careful evaluation before widespread implementation in healthcare.
Simbo is using AI agents to automate all phone related workflows. Connect now to know more.
Schedule AI Agents Demo Now© Since 2019, Simbo AI.
