Sundar Pichai, CEO of Google, once said, “AI is one of the most profound things we're working on as humanity. It's more profound than fire or electricity.” While some may argue that this claim is ambitious, it is clear that many industries, including healthcare, will undergo significant transformation in the coming years.
One of the pioneering developments in this space involves large language models and generative AI like ChatGPT. Some early demonstrations have included using these models to generate medical notes, to help summarize information for patients, and to reduce administrative tasks such as prior authorization.
However, like any technology, AI offers both promises and perils. As these language models become more integrated into healthcare, it's vital to acknowledge their limitations and adopt processes that mitigate potential drawbacks. A concern that stands out is 'hallucination', where the AI creates unrelated or incorrect information. Moreover, with the impending tsunami of AI-generated notes, there's no guarantee that physicians will meticulously check each sentence for accuracy (in fact, the authors are willing to take a bet with anyone who thinks they will!).
Michael Gao, MD, the CEO of SmarterDx and an Assistant Professor of Medicine at Weill Cornell, reflects on his residency, “Residents were often tempted to copy and paste information from other physicians’ notes. The attendings would often ask, ‘Where’s the primary data?’ and ‘What is the original study supporting that diagnosis?’ I quickly learned that directly reviewing underlying data was necessary to truly understand the patient and that patient notes would often have errors–either false information or omission.”
In the era of ChatGPT, it is highly likely that the role of Clinical Documentation Improvement (CDI) professionals will amplify, not diminish. Monica Watson, a Senior Director at the University of Arkansas for Medical Sciences (UAMS), underscores this: "Our role in ensuring that clinical indicators specific to the patient’s episode of care, such as appropriate monitoring, evaluation, assessment, and treatment are meticulously documented becomes even more vital in the era of generative AI and hallucinations."
SmarterDx helps hospital CDI and Coding specialists identify clear and consistent criteria by using its proprietary AI technology to scrutinize the primary data – labs, medications, vitals, orders, radiology, pathology, and notes from nursing and ancillary staff – and to ‘true up’ that data with physician notes and ICD-10-CM/PCS codes. SmarterDx complements concurrent CDI and Coding tools by acting as a Prebill safety net that scans 100% of charts to catch any opportunities.
Watson notes, "We use SmarterDx in conjunction with our concurrent CDI and Coding computer assisted coding software. It scans all raw data, regardless of formatting or templates, and uncovers additional clinical documentation, quality, and coding opportunities before bill submission. We were pleasantly surprised at how effective it was because we were already at the 90th percentile in MedPAR prior to using SmarterDx. But it’s not only helped us capture additional opportunities but has provided insights for further program improvement.”
Hospitals like the University of Arkansas for Medical Sciences are capitalizing on the AI revolution to identify, on average, $800K per 100 beds in net new revenue, all fully supported by clinical indicators.
However, it is crucial to remember that the success of these technologies ultimately hinges on the symbiosis between AI systems and human oversight. As AI continues to evolve and shape medical documentation, professionals like Monica Watson will play an increasingly important role in maintaining the accuracy and integrity of patient records.
Dr. Gao concludes, "This is just the beginning of AI's potential in healthcare. Our task is to adapt and leverage the promises of these technologies without succumbing to their perils. A new chapter in medicine is unfolding, and we are all the authors of this exciting journey."