Recent findings from a study published in the JAMA Ophthalmology have shed light on the underutilized potential of artificial intelligence (AI) in diagnosing diabetic retinopathy (DR), despite its proven effectiveness. The study meticulously examined the use of AI in comparison to traditional imaging methods across an extensive database of over 107 million patients, focusing specifically on those with diabetes.
Diabetic retinopathy, a severe vision-threatening complication of diabetes, requires early detection for effective management. However, the study identified that out of nearly 5 million diabetic patients assessed from January 2019 to December 2023, only a fraction (0.4%) underwent any ophthalmic imaging specifically for diabetic retinopathy detection. Among these, the use of AI-based imaging tools was exceedingly rare, accounting for just 0.09% of all screenings.
The adoption of AI imaging, coded under Current Procedural Terminology (CPT) code 92229 since January 2021, showed minimal usage. By 2023, the frequency of this AI imaging reached about 58.6 instances per 100,000 diabetic patients—an indication of slow uptake. In contrast, traditional imaging methods like Optical Coherence Tomography (OCT, CPT code 92134) and fundus photography (CPT code 92250) were more prevalent, with OCT used in 30.3% and fundus photography in 35.0% of cases.
Moreover, the data showed that AI-based screening methods, although few, resulted in a higher referral rate to OCT at 7.74% compared to traditional remote imaging techniques at 5.53%. This suggests a potential for better clinical outcomes when AI is employed, albeit the overall use of remote imaging methods saw a significant increase by 90.16% between 2021 and 2023, with AI-based screening lagging behind.
The geographical distribution of AI imaging usage displayed significant regional disparities. More than 80% of AI-based imaging cases were reported in the South, a region comprising only 40% of other imaging modalities. Additionally, the demographic analysis revealed that nearly half of the patients screened with AI systems were Black, contrasting with about a quarter in other imaging methods, highlighting a disparity in healthcare access and technology utilization.
The sluggish adoption of AI systems, approved by the FDA and epitomized by technologies like LumineticsCore and EyeArt, is hindered by factors such as cost, awareness, and integration challenges within existing healthcare workflows. Despite these obstacles, there are innovative programs like the Stanford Teleophthalmology Autonomous Testing and Universal Screening initiative, which aim to streamline processes and enhance collaboration between primary care providers and ophthalmologists.
The study advocates for enhanced support systems and patient-centered approaches to improve the uptake of AI technologies. By fostering better workflow integration and offering targeted strategies to healthcare providers, the potential of AI to improve early detection of diabetic retinopathy and patient outcomes could be fully realized.
Such integration is not only about adopting new technologies but also about adapting healthcare infrastructures to leverage these advancements effectively. Ensuring that AI tools are accessible and integrated into the daily routines of healthcare professionals can bridge the gap between traditional methods and cutting-edge technology, ultimately leading to better health outcomes for patients with diabetes.
In conclusion, while AI holds transformative potential for managing diabetic retinopathy, significant efforts are required to enhance its adoption. This includes education on AI capabilities, subsidy programs to offset costs, and policies to support technological integration. As AI continues to evolve, its role in transforming healthcare practices, especially in chronic disease management like diabetic retinopathy, becomes increasingly critical. By addressing these challenges, the healthcare system can better harness AI's capabilities to improve care delivery and patient outcomes.
Source：
Shah, S. A., Sokol, J. T., Wai, K. M., Rahimy, E., Myung, D., Mruthyunjaya, P., & Parikh, R. (2024). Use of Artificial Intelligence–Based Detection of Diabetic Retinopathy in the US. In JAMA Ophthalmology. American Medical Association.
https://doi.org/10.1001/jamaophthalmol.2024.4493
Post comments