Large language models, a form of artificial intelligence, are generating a lot of hype in healthcare circles, mainly due to their potential to transform and improve various aspects of healthcare delivery and management. The buzz is also driven by rapid advances in AI and machine learning.
But while there is significant potential, challenges and ethical considerations remain, including concerns about data privacy and security, persistent bias, regulatory issues, data precision and more.
In short, AI is ready to do big things – but can it also be used for doctors?
Medicomp Systems CEO David Lareau believes this is possible – if the industry adopts complementary technologies that take advantage of the power of AI.
Healthcare IT news sat down with Lareau to talk AI, LLMs, and the future of healthcare.
Q. You propose to task artificial intelligence with identifying clinical quality measures and coding hierarchical condition categories for risk adjustment. How can AI help doctors here? What can it do?
A. Artificial intelligence and large language models have powerful capabilities for generating textual content, such as composing encounter notes and identifying multiple words and phrases that have similar meanings.
An example of this is using ambient listening technology with LLMs to capture and present concept notes from a clinical encounter by taking what is spoken during the patient encounter and converting it into text notes.
AI and LLMs allow a system to hear the patient say, “I sometimes wake up at night and have trouble catching my breath,” and associate that with specific clinical concepts such as “shortness of breath,” “difficulty breathing,” “recumbent bicycles”. shortness of breath” and conditions or symptoms.
These concepts can have different diagnostic implications for a doctor, but by being able to associate what a patient says with specific symptoms or conditions that are clinically relevant to potential problems or diagnoses, the combination of AI/LLMs can help a doctor focus on conditions that are eligible for risk adjustment, such as in this case sleep apnea, heart failure, COPD or other conditions.
This powerful first step in identifying the potential applicability of clinical quality measures is crucial. However, additional tools are needed to evaluate complex and nuanced patient inclusion and exclusion criteria. These criteria must be clinically accurate and include additional content and diagnostic filtering of other information from a patient’s medical record.
Q: Regarding AI and CQM/HCC, you say that even with advanced AI tools, the challenges of data quality and bias are significant, as is the inherent complexity of medical language. Explain some of the challenges.
A. In clinical settings, factors such as gender, race and socio-economic background play a crucial role. However, LLMs often struggle to integrate these aspects when analyzing individual medical records. Typically, LLMs draw from a wide range of sources, but these sources tend to reflect the most common clinical presentations of the majority population.
This can lead to biases in the AI’s responses, potentially overlooking unique characteristics of minority groups or individuals with specific conditions. It is important that these AI systems take into account the diverse backgrounds of patients to ensure accurate and unbiased healthcare support. Data quality poses a significant challenge in effectively using AI for the management and documentation of chronic conditions.
This issue is especially relevant to the thousands of diagnoses eligible for HCC risk adjustment and CQMs. Several standard healthcare codes, including ICD, CPT, LOINC, SNOMED, RxNorm, and others, have unique formats and do not integrate seamlessly, making it difficult for AI and natural language processing to filter and present relevant patient information for specific diagnoses.
Furthermore, interpreting medical language for coding is complex. For example, the term ‘cold’ can be related to colds, sensitivity to lower temperatures or cold sores. Furthermore, AI systems such as LLMs struggle with negative concepts, which are crucial for distinguishing between diagnoses, because most current code sets do not process such data effectively.
This limitation hinders the ability of LLMs to accurately interpret subtle but significant differences in medical wording and patient presentations.
Q. To address these challenges and ensure compliance with risk-based reimbursement programs, you propose CQM/HCC technology that has the ability to analyze information from patient records. What does this technology look like and how does it work?
A. CQMs serve as proxies to determine whether quality care is being provided to a patient, given the existence of a set of data points indicating that a specific quality measure applies. Participation in a risk-adjusted reimbursement program such as Medicare Advantage requires that providers use the Management, Evaluation, Assessment and Treatment (MEAT) protocol for diagnoses included in HCC categories, and that documentation supports the MEAT protocol.
Since there are hundreds of CQMs and thousands of diagnoses included in the HCC categories, there is a clinical relevance engine that can process a patient record, filter it for information and data relevant to each condition, and normalize the presentation so that a clinical user can assess and take action will be a requirement for effective care and compliance.
Of Through the adoption of FHIR, the creation of the first QHINs, and the opening of systems to SMART-on-FHIR apps, companies have new options to keep their current systems in place while adding new capabilities to enable CQMs, HCCs, and address clinical data interoperability .
This requires the use of clinical data relevance engines that can convert text and disparate clinical terminologies and code sets into an integrated, computable data infrastructure.
Q. Natural language processing is part of your vision here. What role does this form of AI play in the future of AI in healthcare?
A. On request, LLMs can produce clinical text, which NLP can convert into codes and terminologies. This capability will reduce the burden of creating documentation for a patient encounter.
Once that documentation is established, other challenges remain because not only do the words have clinical meaning, but so do the relationships between them and the clinician’s ability to quickly find and act on relevant information.
These actions obviously include CQM and HCC requirements, but the bigger challenge is to enable the clinical user to mentally process the LLM/NLP outputs using a trusted ‘source of truth’ for clinical validation of the output of the AI system.
Our focus is on using AI, LLMs and NLP to generate and analyze content, and then process it using an expert system that can normalize results, filter by diagnosis or problem, and deliver actionable and clinically relevant information to the doctor can present.
Follow Bill’s HIT coverage on LinkedIn: Bill Siwicki
Email him: bsiwicki@himss.org
Healthcare IT News is a HIMSS Media publication.