Perceptions about Artificial Intelligence Use in Vascular Surgery: Insights from Healthcare Providers and Industry Professionals
Poster #: 005
Session/Time: B
Author:
Ryan Mancoll, BS
Mentor:
David Dexter, MD
Research Type: Mixed-methods Survey/Interview
Abstract
INTRODUCTION:
Artificial intelligence (AI) and Machine learning have become widely used in healthcare for a multitude of tasks. Following this trend, AI research and use in vascular surgery have increased as the field navigates obstacles such as acceptance and access to training data. Current AI uses include note-writing assistance, imaging interpretation, and case planning in various disease states. This study seeks to evaluate vascular health care providers' and industry representatives' perceptions and concerns about this rapidly emerging technology.
METHODS:
We performed a prospective, mixed-methods, convergent parallel design cohort study via a quantitative survey and qualitative interview. Participants were recruited from the SVS member directory, as well as known contacts. Participants fell into two groups: a healthcare provider (HCP) in the vascular space (surgeon, trainee, APP), or vascular device industry professionals. The survey was primarily Likert-scale-based, comprising of sections on demographics, understanding, derivation/validation of AI, performance, and uses in vascular surgery. Surveying was completed using the secure online tool RedCap.
RESULTS:
The survey invitation was distributed to approximately 3,500 individuals, of whom 106 (~3.0%) completed the survey. Among them, 88 qualified for the health care provider group, 71 of those being surgeons, and 18 qualified for the industry group. Six participants agreed to be interviewed, and two followed up and completed the interview. Both groups had comparable knowledge of AI terminology, interest, use, and evidence requirements for implementation. While large proportions of participants from both groups were at least mildly concerned about AI misuse in the clinical setting (31% HCP vs. 44% Industry) HCPs rated themselves more strongly/very strongly concerned (44% HCP vs. 17% Industry). The majority of HCPs (52%) felt that AI would bring notable or radical change to vascular surgery in the next 5 years, in comparison to 34% of industry. More respondents from industry felt that AI should perform better than health care providers prior to being put into practice (32% HCP vs. 61% Industry). While more HCPs felt comparable performance between AI and healthcare providers(44% HCP vs. 28% Industry) is an acceptable threshold. The usefulness of already established or proposed uses of AI in vascular surgery was also assessed and outlined, showing generally positive attitudes across both groups towards all uses other than direct surgical assistance. As well, Industry rated these indirect and direct applications of AI as marginally more useful than HPCs in all categories other than tracking patient complications and outcomes.
CONCLUSION:
Despite the increasing prevalence of AI, evaluation of expectations and perceptions in vascular surgery remains largely absent. Our data shows overall homogeneity between vascular surgery health care providers and our industry partners in most aspects of AI evaluated, including knowledge, interest, use, and requirements. However, it also elucidates differences in aspects like levels of concern, perception of impact, performance requirements, and general application. Our findings suggest the necessity for discussion and collaboration between all groups in the field to support effective and accepted AI implementation in vascular surgery.
Artificial intelligence (AI) and Machine learning have become widely used in healthcare for a multitude of tasks. Following this trend, AI research and use in vascular surgery have increased as the field navigates obstacles such as acceptance and access to training data. Current AI uses include note-writing assistance, imaging interpretation, and case planning in various disease states. This study seeks to evaluate vascular health care providers' and industry representatives' perceptions and concerns about this rapidly emerging technology.
METHODS:
We performed a prospective, mixed-methods, convergent parallel design cohort study via a quantitative survey and qualitative interview. Participants were recruited from the SVS member directory, as well as known contacts. Participants fell into two groups: a healthcare provider (HCP) in the vascular space (surgeon, trainee, APP), or vascular device industry professionals. The survey was primarily Likert-scale-based, comprising of sections on demographics, understanding, derivation/validation of AI, performance, and uses in vascular surgery. Surveying was completed using the secure online tool RedCap.
RESULTS:
The survey invitation was distributed to approximately 3,500 individuals, of whom 106 (~3.0%) completed the survey. Among them, 88 qualified for the health care provider group, 71 of those being surgeons, and 18 qualified for the industry group. Six participants agreed to be interviewed, and two followed up and completed the interview. Both groups had comparable knowledge of AI terminology, interest, use, and evidence requirements for implementation. While large proportions of participants from both groups were at least mildly concerned about AI misuse in the clinical setting (31% HCP vs. 44% Industry) HCPs rated themselves more strongly/very strongly concerned (44% HCP vs. 17% Industry). The majority of HCPs (52%) felt that AI would bring notable or radical change to vascular surgery in the next 5 years, in comparison to 34% of industry. More respondents from industry felt that AI should perform better than health care providers prior to being put into practice (32% HCP vs. 61% Industry). While more HCPs felt comparable performance between AI and healthcare providers(44% HCP vs. 28% Industry) is an acceptable threshold. The usefulness of already established or proposed uses of AI in vascular surgery was also assessed and outlined, showing generally positive attitudes across both groups towards all uses other than direct surgical assistance. As well, Industry rated these indirect and direct applications of AI as marginally more useful than HPCs in all categories other than tracking patient complications and outcomes.
CONCLUSION:
Despite the increasing prevalence of AI, evaluation of expectations and perceptions in vascular surgery remains largely absent. Our data shows overall homogeneity between vascular surgery health care providers and our industry partners in most aspects of AI evaluated, including knowledge, interest, use, and requirements. However, it also elucidates differences in aspects like levels of concern, perception of impact, performance requirements, and general application. Our findings suggest the necessity for discussion and collaboration between all groups in the field to support effective and accepted AI implementation in vascular surgery.