A sick child comes to the hospital for the third time in a year with an asthma crisis. His coughing and difficulty breathing improve with the inhaled medication the doctor administers, only to get worse again 2 hours later in the small room in the ED.

While patients are overflowing from the waiting room, his doctor tries to decide whether another nebulized treatment will make him well enough to go home, or whether he needs to be admitted to the hospital overnight.

Another doctor, after seeing 30 patients that day, struggles to find time to write notes in the electronic medical records before going home. At 6 p.m. she sighs and puts her laptop in her bag, thinking she will do it after she gives her kids dinner and helps them with their homework. Her day never seems to end.

A nurse struggles in the ICU. She has three patients assigned to her, but one of them has had a severe head injury and needs almost 1:1 care.

These are some of the scenarios where AI can be helpful in healthcare that I discussed at Drexel University’s Executive Leadership in Academic Medicine program in late January.

Helping the patient and clinician

Healthcare is at a watershed moment. With the impactful forces of clinician burnout, the rise of chronic disease, and overarching health inequities, innovation and out-of-the-box solutions are urgently needed.

Almost 40% of healthcare providers are looking to leave medicine in the next few years, and over 60% of primary care providers state they are overwhelmed and fatigued. For two important stakeholders, the clinician and the patient, AI may be a solution.

Some areas where AI could help the patient include earlier diagnosis for rare conditions using large sets of data and better outcome prediction for chronic diseases. Also, large language models can guide preventative care and behavioral change outside of the clinic and hospital.

For clinicians, AI-enhanced technologies could reduce the administrative burden of writing patient notes, billing and even prior authorizations. In addition, predictive analytics have demonstrated the ability to improve staffing efficiency in many hospital systems.

Safety of AI

Is AI safe to use? It is essential to understand the guardrails around AI to ensure safety, absence of bias, and privacy.

Large diverse data sets are mandatory to avoid bias in any algorithm. Additionally, it is key to understand that data sets may not be similar in different areas of the country — such as in urban vs. rural settings — as this can bias an AI tool and limit its capabilities.

Finally, there must be constant maintenance and checks and balances to ensure that the AI “fits” as health trends change. For example, an AI tool built prior to the COVID-19 pandemic may not have had the right capabilities during the pandemic due to changed health patterns.

Part of daily life

AI is not new and continues to be implemented by many industries at an exponential rate. From search engines to natural language processing, it is part of our daily lives.

In healthcare, AI can be used to accurately read radiology images, reduce diagnosis times and enhance technologies, and AI startups are finding solutions for EMR transcription. These are just some examples.

As we head into this brave new world, we must understand the principles of AI to improve our capabilities and ensure that what we do is safe and best for everyone in healthcare.