Abstract
In the past year, several papers have been published that aggregate large numbers of historical treatments, addressing clinically important questions that are not well answered by controlled clinical trials. These questions include:
Does acupuncture improve prognosis among individuals with dementia? Among 4844 individuals with dementia, those who were treated with acupuncture were less likely to become disabled within the 1 to 9-year follow-up period compared with those who did not.
1
Is acupuncture safe for individuals with thrombocytopenia? Among 51 patients with thrombocytopenia (defined as having a platelet count of <20,000 per microliter) who received 815 acupuncture sessions, there were no bruising events reported, suggesting that acupuncture is safe even among these vulnerable individuals.
2
How common is hepatotoxic injury from Chinese herbal medicine—in which populations, and for which herbs? Comparing 1791 injured individuals, researchers demonstrated a three-factor causal model incorporating the herbal material, user characteristics, and other medications involved.
3
Does acupuncture reduce cardiac mortality in individuals with diabetes (either type 1 or 2)? Among 6700 individuals with diabetes who were treated with at least three acupuncture treatments, those treated with acupuncture had lower cardiac and all-cause mortality compared with those not treated.
4
The studies described above are from Taiwan, 1 Japan, 2 China, 3 and Korea 4 respectively. In those countries, a large volume of acupuncture treatments and relatively standard use of electronic health records (EHR) currently allow for conduct of large-scale retrospective analyses such as these. These so-called “Big Data” approaches both explore and begin to answer important research questions, addressing compelling issues such as safety and preventive health care. However, “the danger of being precisely inaccurate” 5 has long been identified as a concern regarding large-scale analyses of observational data. As artificial intelligence (AI) and machine learning enter the U.S. health care domains, they will be applied to such acupuncture data as is extant in health care systems.
The purpose of this article is twofold: first, to introduce Western readers to several projects that have taken a proactive approach to shaping what data are collected from routine acupuncture assessments and treatments. These projects all in different ways lay groundwork for large-scale practice-based research to advance clinical science in a way that is meaningful for acupuncturists, as well as individuals who receive treatment, hospital administrators, insurance providers, and policy-makers. Therefore, this article’s second purpose is to urgently invite acupuncture practitioners to join one or more of the conversations below regarding how acupuncture data is collected, aggregated, analyzed, and interpreted, regardless of previous personal interest in research. In “data democracy” as in geopolitics, not voting does not mean opting out of consequences, but simply not being represented.
No Safe Sidelines
As argued in a previous Turning Points article, 6 AI approaches will inevitably be used to mine large data sets, in countries that have them—and it cannot be assumed that the impact on acupuncture practice will always be positive. Recently, Korea has reported multiple initiatives to standardize collection of traditional diagnostic information so as to harmonize it more reliably with biomedical health records 7 and study the construct of constitution 8 more precisely, providing a more personalized total package of care. In Taiwan, machine learning techniques such as association rule mining are being used to develop clinical algorithms for point selection and diagnosis-guided protocols. 9
For better or worse, it can be anticipated that U.S. large health care institutions will use AI to determine (or at least make recommendations) for whether a given patient should receive acupuncture, 6 and perhaps which points should or should not be needled. These recommendations would in turn greatly impact acupuncture education, reimbursement, availability for patients, and job security for acupuncturists. The appropriateness of any recommendations made based on AI outcomes will rest, in part, on the quality of the data: is it complete and accurate? And, more importantly, who determines what data are collected and used as input to any algorithm? Acupuncturists are not currently in a position to determine how health care institutions use data to allocate resources. However, acupuncturists do have considerable power at this time to collectively organize their data output, so that it consistently reflects what they consider important about the treatment encounter and outcomes.
Five Projects That Open Conversations about Structuring Collection of Acupuncture Practice Data
In this context, several U.S. acupuncture practitioner–researchers have embarked on proactive courses to improve the means of acupuncture data production. These projects are presented as examples of an approach: they range greatly in their scope, aim, and maturity but they all share common goals of harmonizing acupuncture research and practice by tightening the feedback loop between them. Each project demonstrates that data can be collected straightforwardly and naturalistically from ordinary clinical activities in a manner that has “face validity” and value for participants. Each aims to expand possibilities for quantitatively and qualitatively understanding acupuncture’s whole-person impact on the individuals being treated in the real world. The Society for Acupuncture Research (SAR) is committed to moving this work forward through initiatives such as the Topological Atlas and Repository for Acupuncture Research (TARA) 10 and participation in international symposia for knowledge exchange such as the recent 2024 joint meeting of the Society for Acupuncture Research and the Research Center for Chinese Medicine Innovation at the PolyU Academy for Interdisciplinary Research in Hong Kong (SAR/RCMI 2024). Future Turning Points columns will discuss the extraordinary work seen there.
This selection of five projects below illustrates possibilities for a structured approach to collecting practice data in the United States. These projects represent only a small fraction of the work on practice-based research that is being done in the current global context and will be done in future.
From Case to Evidence: A Secondary Analysis Approach to Acupuncture Reports
Presented at SAR/RCMI 2024, this poster outlines a novel, mixed-method model for integrating qualitative and quantitative analyses from real-world data sets. Ten case reports adhering to the CARE guidelines (for CAse REports)
11
and data from ∼130 clinical registry cases will be analyzed to look at the distribution and common characteristics of patient demographics, health complaints, and treatment details within the dataset, as well as the change in pain before and after intervention quantified. Qualitatively, thematic analysis will examine changes in diagnostic reasoning and patient-reported self-care behaviors over time, utilizing rich time-series information from EHR. This part of the analysis focuses on detecting complex patterns, such as non-linear symptom changes, consistent with the traditional Chinese medicine (TCM) model. This comprehensive analytical framework aims to explore not just the efficacy of acupuncture treatments but also the underlying reasoning and patient behaviors associated with successful outcomes. It may serve as a model for larger dataset analyses of holistic practice outcomes. Co-Principal Investigators (PIs): Kathleen Lumiere and Lisa Conboy. For more information visit www.convergentpoints.com/fromcasetoevidence or contact
Acuaware
This project is currently assessing feasibility of an EHR that both provides a consistent template for entry of clinical data and also electronically pushes out standard outcome measures to patients for routine collection of reliable quantitative data on patient progress. Standard outcome measures for all patients include the PROMIS (Patient-Reported Outcomes Measurement Information System) 10-question “global health” questionnaire developed by the National Institutes of Health (NIH) to address well-being and health-related functional impairment. 12 Also included is the MAIA-2 (Multidimensional Assessment of Interoceptive Awareness). 13 Interoceptive awareness is the internal perception of physical and emotional sensations; 14 –16 it is disordered among those with chronic pain 17 and may be modulated by acupuncture. 18 ACU-Track’s EHR comprises adaptive templates for the TCM 10 questions, pulse, and tongue diagnostic assessments for all patients, with additional special purpose modules for assessing pain, mental health, and other patient concerns. University of Utah, PI: Lisa Taylor-Swanson. Find the ACUAWARE study at www.ACU-Track.org/acuaware-study.
Finding harmony in data sets: TCM practice-based common outcomes in the west and the east
In a workshop hosted at the SAR 2023 meeting in New York, 38 acupuncturists from 11 countries provided practitioner perspectives on the relative clinical importance of treatment outcomes used in clinical trials of acupuncture for women’s health across the lifespan, which the team had extracted by systematic review of the literature. Workshop participants’ feedback highlighted important gaps in the development of common outcome measures in TCM and showed the need to continue the conversation across continents. From the feedback provided, five surveys were developed to conduct a Delphi study as the next stage of the research. These surveys will be distributed in 2024 to consolidate outcome measure data to inform the development of common outcome sets across women’s lifespans. At the recent SAR/RCMI conference, we had the opportunity to further identify potential measures that accurately reflect the complexity of the TCM approach to whole-person care across conditions and cultures. The overall purpose was to support a translational research strategy in which practice-based data collection informs clinical trials and basic research, integrating a systems biology framework with an in-depth exploration of TCM factors. Contact:
Inpatient acupuncture quorum sensing
This project aims to gather responses to basic questions from inpatient acupuncturists throughout the U.S. These questions include what areas/units they provide inpatient acupuncture, how their program is funded (if they know), how many acupuncturists are in their program, and other similar basic information. The survey will use the snowball sampling method, which is designed to capture other sites through the participants sharing the survey with colleagues in other inpatient acupuncture programs. The survey will proceed iteratively in at least three rounds. The project will first informally engage 12 participants personally known to the authors and knowledgeable regarding inpatient acupuncture, who will give feedback on survey questions, as well as appropriate centers to be included in the primary round. In the snowball approach, each round is sent to a larger group while also incorporating feedback from previous rounds. After all those who have received the survey have had time to respond, the responses will be gathered, and any identifiable information that has been gathered or put into the responses will be removed. The goal of this survey is to gather preliminary information about current inpatient acupuncture programs and disseminate the information as an article published to help improve the quality of inpatient acupuncture programs nationwide. University of California Irvine (UCI), Project Lead: Scott Phelps. To participate in this survey, contact
Acupuncture quality improvement research initiative
This demonstration project will transparently explore the use of data analysis and AI techniques with a large acupuncture patient data set for two purposes: (1) to show in practice how data collection and analysis can inform and improve the quality of practitioner understanding and patient care, and (2) to provide an initial, practitioner-led investigation into best practices for acupuncture data collection, processing, analysis, interpretation, and reporting. The practice of AI, particularly machine learning, involves several steps, each of which entail a number of discrete decisions on the part of the researcher: designing collection systems; cleaning and preparing data sets; choosing, testing, evaluating, and deploying algorithms; and interpreting results. Each of these decisions affect the accuracy and appropriateness of outcomes. For example, are unusual results incorporated into data algorithms or regarded as outliers and removed? Do the analyses emphasize accuracy of prediction or interpretability of results? The principles for addressing these choices in biomedicine may not be appropriate for all acupuncture practitioners and acupuncture patients. It is hoped that other practitioners will be inspired to explore principles and approaches to what data is collected, through structured quality improvement projects like these, which can be conducted before IRB approval. Aligned Modern Health, Project Lead: Eric Hirsch. Do you have a large EHR data set you would like to work with in this way? Contact
EnergyPointsTM—Moving from Feasibility to Phase II Testing
From the patient’s point of view, EnergyPointsTM is a free smartphone app that educates and guides users through evidence-based acupressure protocols. The app gently gamifies the practice 19 by prompting 30–120 sec of stimulation of each acupoint within each “ritual” (protocol) selected and tracks self-reports of which points experienced the characteristic, achy sensation of de qi. EnergyPointsTM integrates with users’ fitness tracker data to correlate acupressure use with sleep and activity data. Users can download and print or email a PDF report for themselves, their acupuncturist and other health care providers; they can also opt in to “donate their data to science,” subject to deidentification and the HIPAA (U.S.), CPRA (California), and GDPR (European Union) privacy standards.
From a research perspective, the app can be a powerful data collection tool when used in an IRB-approved randomized controlled study. EnergyPointsTM app outcome measures include: (1) a daily self-report (four questions about quality of day overall, sleep, fatigue, and use of medication for sleep) and (2) baseline and weekly PROMIS measures including wellbeing, fatigue, and sleep. Two items also measure self-efficacy: confidence to self-administer acupressure, and to manage symptoms. After a successful feasibility trial conducted with participants reporting cancer-related fatigue, the EnergyPointsTM team is currently working towards a proposed Phase II decentralized clinical trial with an adult cancer survivor population. Contact: Dr. Susan Beck,
Where Does the Conversation Go Now?
The collection of practice data affects U.S. acupuncture practitioners, schools, and state and national associations. The issues that need discussion cannot be fully characterized or satisfactorily addressed in this opinion column. However, the authors make an urgent plea for these conversations to continue, contributing to the building of essential research infrastructure for resilience in a changing economic and social climate. Just as contemporary research on neuroplasticity leads to insights that we physically shape our brains through activities of daily living, acupuncturists in the U.S. are beginning to be aware that the data they collect and aggregate (or do not) will shape the societal container in which use of acupuncture grows (or shrinks).
In the interest of meeting AI’s inevitable growth with conscious and purposeful human intelligence, SAR will be initiating conversations about these and associated issues in the coming year. The conversations will largely be collaborations between several of SAR’s Special Interest Groups (SIGs): Dissemination and implementation, hospital-based acupuncture, and AI and digital health, all of which may be joined at www.acupunctureresearch.org, as well as the practitioner-facing forums of www.evidencebasedacupuncture.org and the Research Study Group led by Lisa Conboy at the free online platform www.whitepinecommunity.org.