September 2021

Ethics Corner

John C. Quinn, MD
Ethics & Professionalism Committee Member

Artificial intelligence is in use today across a vast array of industries, including health care, banking, retail, and manufacturing. Machine learning algorithms have shown tremendous potential as tools to improve clinical decision making and patient care however, it must be recognized that use of these technologies present unique ethical challenges. A major challenge with implementation of machine learning tools in healthcare is ensuring that these algorithms are trained with high quality data that is sufficiently diverse and that societal biases such as those due to race, ethnicity and socioeconomic status aren’t reflected in the algorithms. Failure to do so may contribute to further propagation of socioeconomic disparities in healthcare if not recognized and accounted for in the design and development of these tools.

A particular challenge in the development of machine learning algorithms for ASD is the disparity in utilization of ASD surgery in racial and ethnic minority populations. While the population of the US has been growing more diverse there exists a dramatic racial disparity in the utilization of surgery for ASD. A recent analysis of the NIS database showed that of the 28,921 surgeries performed for adult spinal deformity between 2004 and 2014 25,953 (89.7%) were performed on Caucasians and 2968 (10.3 %) African Americans.4 This study also pointed out that while overall complication rates were similar that there was a significant decrease in complication rates over time among Caucasians patients that was not observed in African Americans patients. There also exists a paucity of data regarding the impact race/ ethnicity or socioeconomic status has on outcomes following surgery for ASD. While several studies suggest no such differences exist others provide some evidence that differences in ethnic/ cultural expectations may impact outcomes and patient satisfaction following surgery for ASD.5,6 These studies highlight potential challenges to the development of highly accurate algorithms that can be used across diverse patient populations.

Implementation of Artificial intelligence has the potential to revolutionize the way health-care is delivered. It is imperative however, that we recognize not just its strengths but also the limitations of this technology as we move forward with widely adapting this technology. In a recent article “Potential Biases in Machine Learning Algorithms Using Electronic Health Record Data” the authors describe some steps to be taken to minimize the impact health care disparities may have in the development of machine learning clinical tools.3 They suggest that in order to improve the accuracy and reliability of these machine learning algorithms across a diverse patient population that these algorithms be built and tested in socioeconomically diverse healthcare system. They also stress that key variables such as race, ethnicity, language and social determinants of health are being captured and included in algorithms and that these algorithms be tested routinely for discriminatory behavior.

Use of advanced computational tools and machine learning has shown significant promise in the advancement of treatment of patients with adult spinal deformity through the ability to identify risk at the individual level which represents a significant step forward in treatment. While there exists significant promise for use of these techniques it is important to recognize current limitations that exist and to further develop these tools for use in increasingly diverse patient population.


1. Joshi RS, Lau D, Scheer JK, et al. State-of-the-art reviews predictive modeling in adult spinal deformity: applications of advanced analytics. Spine Deform. 2021;9(5):1223-1239.
2. Ames CP, Smith JS, Pellisé F, et al. Development of Deployable Predictive Models for Minimal Clinically Important Difference Achievement Across the Commonly Used Health-related Quality of Life Instruments in Adult Spinal Deformity Surgery. Spine (Phila Pa 1976). 2019;44(16):1144-1153.
3. Gianfrancesco MA, Tamang S, Yazdany J, Schmajuk G. Potential Biases in Machine Learning Algorithms Using Electronic Health Record Data. JAMA Intern Med. 2018;178(11):1544-1547.
4. Wang KY, Puvanesarajah V, Xu A, et al. Growing Racial Disparities in the Utilization of Adult Spinal Deformity Surgery: An Analysis of Trends from 2004 to 2014 [published online ahead of print, 2021 Aug 18]. Spine (Phila Pa 1976).

5. Sanford Z, Taylor H, Fiorentino A, et al. Racial Disparities in Surgical Outcomes After Spine Surgery: An ACS-NSQIP Analysis. Global Spine J. 2019;9(6):583-590.
6. Yagi M, Ames CP, Hosogane N, et al. Lower Satisfaction After Adult Spinal Deformity Surgery in Japan Than in the United States Despite Similar SRS-22 Pain and Function Scores: A Propensity-Score Matched Analysis. Spine (Phila Pa 1976). 2020;45(17):E1097-E1104.


Chair: Steven D. Glassman Committee: B. Stephens Richards III, Past Chair; Jacob M. Buchowski; David A. Hanscom; Joseph F. Baker (C); Thomas M. Gavin (C); Frank T. Gerow (C); John F. Lovejoy III (C); Robert F. Murphy (C); John C. Quinn (C); Sherif M. El Ghamry; Hee-Kit Wong; S. Samuel Bederman; John P. Lubicky; Dale Blasier; Daniel J. Hedquist