Despite the myth that artificial intelligence (AI) will replace clinicians and deskill eyecare professions, optometrists hold positive attitudes toward the future use of AI to support the diagnosis of retinal disease, SHARON HO and Dr ANGELICA LY have discovered in their new study.
In today’s technological age, artificial intelligence (AI) represents a contemporary and innovative approach to eyecare, sparking tremendous interest due to its potential to increase efficiency, productivity, affordability, quality and accessibility. AI has chiefly been applied to image analysis, where computerised algorithms are trained to analyse clinical images and provide diagnostic and/or management recommendations. Previous research has shown these AI clinical decision support systems (CDSS) can achieve high accuracy for the detection of a number of ocular diseases including diabetic retinopathy,1-3 age-related macular degeneration,4,5 glaucoma6 and retinopathy of prematurity.7
Despite their proven potential, AI CDSS are not widely implemented in routine clinical practice. This may be in part due to general misunderstanding and distrust toward new technology. Along with Associate Professor Gordon Doig, we investigated this issue in our study entiled: ‘Attitudes of Optometrists toward Artificial Intelligence for the Diagnosis of Retinal Disease: A cross-sectional mail-out survey’.8
We designed and conducted a survey of optometrists to identify their attitudes toward using AI in clinical practice to assist in diagnosing retinal disease. A total of 252 surveys were mailed to randomly selected practising optometristres across Australia, of which 133 were returned and included in the study.
On average, optometrists reported positive attitudes toward using AI as a tool to aid the diagnosis of retinal disease. They also agreed there will be an overall need for AI in primary eyecare and were excited by future increased use of AI. This is promising for the future implementation of AI CDSS into clinical practice as it suggests that optometrists’ attitudes will not be a major limiting factor.
Motivators and barriers to future use
Understanding both what encourages clinicians to use AI and what holds them back are key in ensuring the successful implementation of AI CDSS into clinical practice.
Our survey identified several key factors for consideration. Increased patient accessibility to healthcare and more reliable diagnoses were ranked as the two most important potential benefits of using AI. This suggests that optometrists may be more receptive to using AI CDSS in communities with relatively limited access to eyecare, such as in rural or regional communities. Meanwhile, multiple studies already exist demonstrating the reliability of diagnoses made by AI CDSS.1,3,9 Disseminating this information to optometrists may help reinforce positive attitudes toward the new technology.
In our survey, optometrists also responded positively to the potential for AI to improve diagnostic accuracy, save time, and be more cost-effective than current processes. Furthermore, they valued AI CDSS that fit well into their clinical workflow and that are easy to learn to use. These features are instrumental in driving the future use of AI and important qualities to developers for ensuring their products are well-suited to end-users.
While motivators for the adoption of AI mainly revolve around patient-centred benefits, potential barriers appear related to the rigour of validation and impact on professional autonomy from using AI CDSS.
Respondents in our survey agreed that AI should be validated through higher quality randomised controlled trials rather than lower quality retrospective studies. However, the conduct of such studies has so far been limited due to resource and practical constraints. Interestingly, government approval was perceived by optometrists as the least important requirement for validating AI, being associated with neutral attitudes, indicating that this will not pose a significant barrier to the future use of AI CDSS.
Perhaps one of the biggest myths relating to the use of AI is that it will replace clinicians. Other concerns have been voiced about the risk of overreliance on technology leading to automation bias, compromise of independent decision-making and deskilling of clinicians.10
Meanwhile in our survey, optometrists had mixed attitudes about whether they would limit their use of AI if it caused a neglect in their clinical skills. At the same time, they were apprehensive about “the potential [for AI] to bypass optometry and refer…based solely on the acquisition of a scan” and further asserted that “optometry is based on personable relationships…[AI] detracts from this”. This emphasises their concern for professional autonomy and value placed on the human aspect of clinical care. AI is oftentimes intended as a support system to enhance, not replace, human intelligence.11 Proactively educating optometrists on this may ease anxiety and increase awareness of the benefits of AI. Early efforts to help optometrists establish a healthy relationship with AI will ensure their potential is fully realised.
Using AI in different clinical scenarios
Clinicians will use CDSS only if sufficiently convenient to do so, and if they are used in a manner concordant with their needs and preferences. Of particular importance is the way in which the system is integrated into the clinical workflow.
For example, one clinical scenario could involve using AI during the consultation to provide a diagnostic recommendation at the point-of-care, while a different clinical scenario could be to use AI after the consultation to provide a second opinion on diagnosis. An advantage of the first clinical scenario is that the result is immediately available to the optometrist who can then make an informed decision on diagnosis and management while the patient is in the chair. However, using AI in this way entails an extra step in the clinical workflow whereby CDSS that are interruptive and time-consuming will detract from their original purpose. On the other hand, the second clinical scenario may solve the potential issue of disrupted workflow, though access to the AI-derived diagnostic recommendation is delayed.
Our survey compared these two specific clinical scenarios and found that optometrists held no significant preference for using AI during the consultation at the point-of-care or after the consultation. Future investigation is needed to better understand optometrists’ preferences for using AI to help maximise its clinical potential and ensure its successful translation into practice.
Factors influencing attitudes toward AI
Finally, we also found that optometrists who used a greater number of computerised systems in the workplace had more positive attitudes toward AI. This suggests that exposure to other computer-based technologies in the workplace may encourage optometrists to have a stronger positive belief in the future use of AI. It further implies at a relationship with general innovation readiness and digital confidence, which describe whether users are receptive to and capable of innovation as well as their digital literacy and self-efficacy.12 These measures help predict the success of healthcare innovation spread and appear to drive optometrists’ positive attitudes toward AI.
Other factors – optometrists’ gender, work experience, workplace location (rural vs. urban), number of patients seen daily, and level of accessibility to ophthalmology services – were not significantly associated with attitudes toward AI. Of note, there was no difference in attitudes toward AI between optometrists of different ages. This stands in contrast to the greater reluctance in accepting AI in those aged 55 years and over compared to younger members of the general public.13
Hype or hope? Friend or foe? AI CDSS are a state-of-the-art technology with the accuracy and reliability to dramatically enhance optometric practice. Despite the myth that AI will replace clinicians and other negative preconceptions surrounding deskilling, optometrists hold positive attitudes toward the future use of AI to support the diagnosis of retinal disease. The findings of our recent survey study – described in this article and covering motivators and barriers, possible clinical scenarios, and individual factors influencing attitudes toward the use of AI CDSS by optometrists – may be applied to assist the future implementation and uptake of AI CDSS for retinal disease in clinical practice.
The authors thank co-investigator, A/Prof Gordon Doig, and all optometrists who participated in the survey.
1 Gulshan V, Peng L, Coram M et al. Development and Validation of a Deep Learning Algorithm for Detection of Diabetic Retinopathy in Retinal Fundus Photographs. JAMA 2016; 316: 2402-2410.
2 Shahriari MH, Sabbaghi H, Asadi F et al. Artificial intelligence in screening, diagnosis, and classification of diabetic macular edema: A systematic review. Surv Ophthalmol 2023; 68: 42-53.
3 Abràmoff MD, Lavin PT, Birch M et al. Pivotal trial of an autonomous AI-based diagnostic system for detection of diabetic retinopathy in primary care offices. NPJ Digit Med 2018; 1: 39.
4 Lee CS, Baughman DM, Lee AY. Deep learning is effective for the classification of OCT images of normal versus Age-related Macular Degeneration. Ophthalmol Retina 2017; 1: 322-327.
5 Peng Y, Dharssi S, Chen Q et al. DeepSeeNet: A Deep Learning Model for Automated Classification of Patient-based Age-related Macular Degeneration Severity from Color Fundus Photographs. Ophthalmology 2019; 126: 565-575.
6 Ahn JM, Kim S, Ahn KS et al. A deep learning model for the detection of both advanced and early glaucoma using fundus photography. PLoS One 2018; 13: e0207982.
7 Travis KR, John Peter C, James MB et al. Evaluation of a deep learning image assessment system for detecting severe retinopathy of prematurity. British Journal of Ophthalmology 2019; 103: 580.
8 Ho S, Doig GS, Ly A. Attitudes of optometrists towards artificial intelligence for the diagnosis of retinal disease: A cross-sectional mail-out survey. Ophthalmic Physiol Opt 2022.
9 Gargeya R, Leng T. Automated Identification of Diabetic Retinopathy Using Deep Learning. Ophthalmology 2017; 124: 962-969.
10 Garg AX, Adhikari NK, McDonald H et al. Effects of computerized clinical decision support systems on practitioner performance and patient outcomes: a systematic review. Jama 2005; 293: 1223-1238.
11 OCT News.2018 Notal Vision Announces FDA Grants Breakthrough Device Designation for Pioneering Patient-Operated Home Optical Coherence Tomography (OCT) System. Available from: http://www.octnews.org/articles/8387113/notal-vision-announces-fda-grants-breakthrough-dev/.
12 Tim B. Digital innovation evaluation: user perceptions of innovation readiness, digital confidence, innovation adoption, user experience and behaviour change. BMJ Health &amp; Care Informatics 2019; 26: e000018.
13 Bristows.2018 Artificial Intelligence: Public Perception, Attitude and Trust.