Artificial Intelligence and Acupuncture: A Data-Driven Synergy

Artificial Intelligence Transforms Medicine

In the upcoming years, the integration of artificial intelligence (AI) in the field of medicine will bring about a profound transformation. As AI technologies continue to advance, they hold the potential to revolutionize how medical professionals approach patient care, from early detection and proactive preventive measures, to more accurate and personalized diagnoses and treatments. The incorporation of AI in health care workflows is expected to streamline and optimize processes, enhancing the overall efficiency of medical practices. This transformative shift also brings with it a re-evaluation of the roles and responsibilities within the health care system, as professionals adapt to and collaborate with AI technologies to provide more effective and patient-centric care.

Data are the lifeblood of AI systems, serving as the foundational material from which they learn and make informed recommendations. The quality, diversity, and quantity of data significantly influence the performance and accuracy of AI algorithms, directly and foundationally impacting the effectiveness of AI applications. Acupuncturists or their patients may hesitate to document data for AI systems due to concerns about privacy and data security, as well as uncertainties about how the information will be utilized and whether it aligns with ethical standards.

In addition, the potential for increased workload, coupled with the need for additional training in data collection methods, can contribute to the reluctance among acupuncturists to actively participate in the generation of data for AI applications. Depending on the country and the configuration of the respective health care system, acupuncturists in private practice might not have been invited to provide data. The chance that data from patients undergoing acupuncture will be missing in the databases utilized to train AI algorithms for diagnosing and treating symptoms is high. Therefore, acupuncture may not be incorporated into AI-generated recommendations. To avoid this, acupuncturists might even have to take an active approach to get their data integrated in larger health care databases.

With these concerns in mind, acupuncture researchers are taking a proactive approach. A systematic scoping review on AI in complementary medicine conducted in 2022 found that AI was already being employed in research, particularly in the selection of acupuncture points. ¹ In 2023, work began on the Topological Atlas and Repository for Acupoint research (TARA) project, which will include an electronic database containing the information for each acupuncture point, along with a search interface to find previously published data by related point, anatomical region, or by disease or condition, with each data set to be accompanied by a rich set of metadata. ²

Living up to the saying “with great power comes great responsibility,” caution should be exercised in the research and academic arena, where AI such as large language models such as ChatGPT can have a strong influence. In a recent editorial of this journal, Cramer highlighted the misconduct that can happen with AI, especially in the scientific environment: “As with any new technology, it is important to have appropriate safeguards in place to prevent and detect such misconduct.” ³

Recognizing the profound shifts in health care within the increasingly digitalized society, the integration of acupuncture treatment data into health care databases is crucial for inclusion in treatment plans and reimbursement processes. These data not only facilitate pertinent research but also play a pivotal role in enhancing the overall integration of acupuncture within the health care system.

Interoperability and Core Data Standards

Practitioners and researchers in complementary medicine often find themselves dealing with a substantial amount of data that include acupuncture treatment as part of the intervention. The collected data, however, are heterogeneous, perhaps even more than those collected in conventional medicine. The heterogeneity stems from numerous factors, notably the unique position of acupuncture in health care practices. When acupuncture is employed as a complementary therapy alongside conventional treatments, it is documented using existing medical data frameworks, where it is integrated as an adjuvant therapy. Conversely, in clinics where acupuncture serves as the primary treatment method, data are gathered using tailored structures designed to capture the intricate details of this specific therapeutic approach.

The purpose behind data collection also influences its structure. Data recorded for health insurance reimbursement purposes are often narrowly focused around the requirements of the reimbursement process, leading to restrictions in the language and coding used. Meanwhile, data collected for the sole purpose of documenting medical encounters may be unstructured and in free-text format for convenience, further contributing to the diversity. This divergence in documentation methods leads to a profound diversity in data structure and content, presenting a challenge in achieving data interoperability.

Data interoperability refers to the ability of different types and systems of data to be merged and aggregated. ⁴ This concept encompasses the processes, standards, and technologies that allow for the exchange, understanding, and use of information between various systems, applications, or organizations. For data to be interoperable, there must be a consistency in how phenomena are measured and formatted. For example, if one data set records birthdates while another captures age, the two are not immediately interoperable without significant data cleaning. Similarly, if one data set records acupuncture points in World Health Organization (WHO) standard acupuncture point locations ⁵ and the other using Chinese names, they are not interoperable.

Variation in acupuncture styles and needle stimulation techniques adds complexity to data interoperability. Although reporting standards for acupuncture details have been recommended or established for research, ^6,7 labeling the data can still exhibit variability, further posing challenges in aggregating and analyzing data across different practices.

Addressing the specific issue of semantic interoperability within acupuncture data requires a shared vocabulary and data structures across various entities. This would enable the exchange of consistent and comprehensible data. Although standardized classifications offer a uniform approach to recording patient conditions and treatments, their application to acupuncture is limited by their current inability to fully encapsulate the scope of acupuncture practice. Some of the international standardized classifications to consider are SNOMED Clinical Terms, ⁸ International Codes for Health Interventions (ICHI), ⁹ and International Classification of Diseases 11th revision (ICD-11). ¹⁰

Interoperability relies heavily on metadata and data documentation, as these elements inform researchers which data sets and variables are comparable. Data standards, which are community-agreed-upon methods for data collection and organization, often facilitate interoperability and establish minimum criteria for describing patient encounters, leading to systematic organization and standardization of acupuncture data. Without these standards, the rich data potential remains underutilized, hindered by interoperability and analytical challenges.

Descriptions of acupuncture points, needle usage, duration of treatment, electroacupuncture specifics, methods, adjunct treatments, and any supplementary activities prescribed to patients are some of the details that need to be discussed to build a core data standard. A core data standard with these details in consensus would serve to homogenize the descriptors used in data to identify acupuncture procedure in specific details and patient diagnosis for which acupuncture treatment was carried out. The standards would also standardize follow-up records, ensuring reliable measurements and interoperability, enabling an establishment of uniform data sets for seamless preprocessing and analysis and enhancing the use of acupuncture data in research.

In 2023, work began on the TARA project, which will include an electronic database containing the information for each acupuncture point, along with a search interface to find previously published data by related point, anatomical region, or by disease or condition, with each data set to be accompanied by a rich set of metadata. ² Furthermore, leveraging reporting standards ⁶ and other established recommendations in the field of acupuncture research, ⁷ as well as in the broader medical records context, ¹¹ is essential for standardizing acupuncture data. These guidelines are crucial for consistently documenting treatment regimens in both academic research and clinical settings, thereby producing data that are interoperable and universally accessible.

Society for Acupuncture Research Special Interest Group

To facilitate a responsible approach to AI for acupuncture research and clinical practice, the Society for Acupuncture Research (SAR) initiated in autumn 2023 the AI and Digital Health Special Interest Group (SIG). This SIG derived from engaging presentations and discussions on the topic at the SAR conference in New York in May 2023, and will provide a platform for collaboration, knowledge exchange, and joint research. Furthermore, SAR members in this SIG will lend their voices and expertise to develop recommendations and encourage an interactive look into the future alongside world-renowned colleagues in other SAR SIGs. ¹² As next steps, the AI and Digital Health SIG will develop strategic recommendations as well as a set of core standards for acupuncture data.