Primitive AI tools, introduced almost 50 years ago in healthcare relied on human-defined rules to produce outputs. Since then, we have come a long way. The currently used cases of AI in healthcare are vast and varied. AI had impacted both clinical and administrative functions in healthcare with different specialties at different stages of adoption. Radiology leads the race with cardiology and neurology being the distant second and third. In the October 2023 list released by FDA on AI/ML medical devices(Artificial Intelligence and Machine Learning (AI/ML)-Enabled Medical Devices | FDA), around 77% are in Radiology, 10% in Cardiology and 3% are in neurology. Some of the very commonly used examples in clinical settings are summarizing medical notes, creating discharge summaries, predicting future adverse events, assistive diagnosis, personalized treatment plans etc. In administrative settings, the power of AI is being used for predicting patient volumes and associated staffing needs, automating pre-authorization, generating draft responses to patient portal messages, documenting billing codes, medical charts or visit notes, Triaging cases for prioritization support.
To strategically understand the areas where AI is impacting most and is expected to continue making biggest impact in a structured way, I have tried to fit in all the use cases in a framework with the acronym PEAR. P stands for Personalize treatments, E stands for enhancing diagnostics Accuracy, A is for accelerate advances in drug discovery and R defines the reduction in administrative burden. All the current and future AI use cases will mostly fall under this framework.
Personalization – In my view, Personalized treatment is the area where AI will make the biggest impact in healthcare delivery. For example, dermatologists can develop personalized treatment options and skin care routines based on patient’s skin condition, genetic structure and other factors which are very unique to every individual. Same way, Cardiologist can assess personalized risk factors for each patient. Ob/gyn can personalize birth and treatment plans based on individual medical histories. Pediatricians can analyze the voice, video samples of each child to predict and correct any development delays for each newborn. Oncologists can study the genetic structure of every individual and design suitable treatment options for any form of cancer. AI will help us achieve what sounded unthinkable till some time back – Each person having his own unique treatment plan designed by a well-trained AI model.
Enhance – If Personalization is the future of AI in healthcare, Enhancement is the present. Different AI models are already helping clinicians to detect and predict different diseases conditions tools faster, accelerate time to treatment, raise alerts and notifications if deviations occur from prescribed treatment plans, Automate triage of imaging exams and multiple other areas. For example, Pediatricians are monitoring vitals to detect deviations and predict onset of serious illnesses. Ophthalmologists are using retinal imaging to identify diabetic retinopathy cases. AI tools are analyzing EEG images to help Neurologists locate the places of seizures in brain and cerebrovascular abnormalities. Similarly, AI tools are analyzing ECG images to help cardiologists detect Ischemia, arrhythmias and other heart abnormalities. AI in many ways is “Augmenting” the whole process of healthcare delivery by providing clinicians the option to make use of these tools for better diagnosis, treatment and follow-up. That’s the reason why American Medical Association (AMA) calls AI as “Augmented Intelligence” rather than “Artificial Intelligence”.
Accelerate – The one area where AI has made the greatest impact in biomedical science is drug discovery and development. Recognizing the increased use of AI/ML throughout the drug development life cycle, FDA’s Center for Drug Evaluation and Research (CDER), in collaboration with Center for Biologics Evaluation and Research(CBER) has recently issued an initial discussion paper last month.
A typical drug discovery process starts with target identification with evidence of association to diseases, identifying potentially interacting molecules by screening of compound libraries, optimizing compounds for favorable drug properties, testing in pre-clinical and clinical settings and then finally the FDA approval. AI is impacting all of these stages, in a big way.
Reduction- While the first three areas of PEAR framework focus on clinical settings, AI is making an equally great impact in reducing the administrative burden on healthcare providers. Given the dearth of trained staff in almost all 50 states of USA, AI tools are already being commonly used for identifying appropriate billing and service codes based on medical notes, predict hospital staffing volumes and requisite staffing needs, predict likelihood of claim denials, provide answers to patients’ queries through the use of automated chatbots, analyze patient feedback and surveys, support prior authorization processes and multiple other areas.
All in all, AI is here to stay, and we need to keep looking for new applications of AI in healthcare delivery with the one and only aim of making healthcare more accessible, personalized and equitable. At the same time, it’s important to work towards some of the problems that use of AI models brings. The prominent ones are privacy risks, biases, risk of new liability claims and Hallucinations.
Disclaimer- I borrowed the title of this piece from my company’s theme for HIMSS 2024– Aiding the Future of Healthcare. The copyright of the title belongs to Happiest Minds Technologies. Happiest Minds Technologies is a leading consulting company that helps providers, payors, pharma and MedTech companies to implement many of the use cases that I have described here.
