MedPath

Implementation of an Integrated System of Artificial Intelligence and Referral Tracking for Real-time Diabetic Retinopathy Screening

Not Applicable
Conditions
Diabetic Retinopathy
Screening
Artificial Intelligence
Interventions
Diagnostic Test: Artificial Intelligence
Registration Number
NCT05166122
Lead Sponsor
Rajavithi Hospital
Brief Summary

This research study aims to bring an artificial intelligence system to screen for diabetic retinopathy (DR) along with referral tracking systems to the screening unit in Uthai Hospital in Phra Nakhon Sri Ayutthaya to assess the effectiveness of screening and follow-up of patients referred to Phra Nakhon Sri Ayutthaya Hospital. It will be compared with the existing screening system and follow up with regular referral by personnel

Detailed Description

Diabetic retinopathy is the most common ocular complication in people with diabetes. It is a leading cause of vision loss and blindness in people aged 20-64 years around the world because in the early stages of the disease there is no warning, causing the patients to be unaware. If the blood sugar content is allowed to increase, severe diabetic retinopathy can occur leading to blindness.

The incidence of diabetic retinopathy in diabetic patients tends to increase with the duration of diabetes. And according to the age of the patient, it was found that within 20 years, patients with diabetes type 1 with diabetic retinopathy is about 99% and diabetes type 2 with diabetic retinopathy is about 60%.

Screening for diabetic retinopathy is accepted and performed in health systems around the world. Evidence shows that screening can reduce blindness(1-3). Thailand uses the percentage of diabetic patients who have been eye tested. It is one of the indicators of service quality of the Eye Health District of the Ministry of Public Health. Screening for diabetic retinopathy using the retinal imaging method is cost-effective. It provides diabetic patients in distant places access to screening, such as bringing a mobile retina camera to take pictures in the community in conjunction with the use of teleophthalmology technology in screening(4-6). But according to a report by the Ministry of Public Health in the HDC system in 2015-2017, it was found that only 40% of the patients who were screened for diabetic retinopathy had not reached the 60% target.

In 2016, Rajavithi Hospital, in collaboration with researchers in Google Health, assessed the use of artificial intelligence to read retina images of diabetic patients in all 13 health districts of Thailand. It found that the artificial intelligence system was able to identify patients for referral to ophthalmologists (moderate non-proliferative diabetic retinopathy \[NPDR\]) with 95% sensitivity and 96% specificity, which is 73% higher than screening personnel specificity 98%.

From thereon, a prospective study with the introduction of artificial intelligence system was conducted to screen real patients in the project titled "Thailand-Google Prospective, Real-World Deployment of Artificial Intelligence for Diabetic Retinopathy Screening" (THAIGER) (NCT TCTR 20190902002) in 2018 to 2020 to assess the feasibility, including obstacles to implementing an intelligence-based screening process. The project integrated AI into the nation-wide screening system of the country. By conducting research in the primary care facilities, Rajavithi Hospital and 9 community hospitals in Pathum Thani Province and Chiang Mai, the diabetic patients in the THAIGER project received the results of reading images by artificial intelligence in real time. However, it was found that of the patients who were referred, very few actually went to see a doctor. There are also images that were unreadable (ungradable) by the artificial intelligence. And the artificial intelligence used in THAIGER has not yet been fully integrated into the screening system, including with a patient tracking system.

This research study aims to bring an artificial intelligence system to screen for diabetic retinopathy (DR) along with referral tracking systems to the screening unit in Uthai Hospital in Phra Nakhon Sri Ayutthaya to assess the effectiveness of screening and follow-up of patients referred to Phra Nakhon Sri Ayutthaya Hospital. It will be compared with the existing screening system and follow up with regular referral by personnel.

Recruitment & Eligibility

Status
UNKNOWN
Sex
All
Target Recruitment
1600
Inclusion Criteria
  • Patients aged 18 years and over.
  • Patients who have been screened for diabetic retinopathy at Uthai Hospital Phra Nakhon Sri Ayutthaya Province that can refer patients to Phra Nakhon Sri Ayutthaya Hospital to see an ophthalmologist
  • People with diabetes who are listed on the civil registry
  • Able to take pictures of the retina at least 1 eye.
Exclusion Criteria
  • Being a patient in a community hospital with an in-house ophthalmologist
  • Patients who were previously diagnosed for the following conditions / diseases: retinal edema, diabetic retinopathy (NPDR, PDR). The retina is affected by radiation (Radiation retinopathy) or retinal vein blockage (RVO).
  • Past history of laser retinal treatment or retinal surgery
  • Having other eye diseases (non-diabetic retinopathy) that requires referral to an ophthalmologist.
  • Inability to take pictures of the retina (for any reason)

Study & Design

Study Type
INTERVENTIONAL
Study Design
PARALLEL
Arm && Interventions
GroupInterventionDescription
AI workflowArtificial IntelligenceIn AI work flow, patients will be screened by taking normal retinal images and all images will be assessed for the severity of diabetic retinopathy by a computerized artificial intelligence system immediately after the photograph is taken via the Internet and retinal images will be sent to the retinal ophthalmologist for overreading.
Primary Outcome Measures
NameTimeMethod
Referral adherence6 months

Total number of patients who completed referral visit in each arm (ie, presented to tertiary eye care center)

Secondary Outcome Measures
NameTimeMethod
Screening throughputCompare time unit of 1 day for each arm

Assess the number of patients who successfully completed screening in a given day in the AI versus manual arm

Assess AI performance6 months

Confirm sensitivity and specificity of AI reading as demonstrated in previous prospective study (THAIGER, TCTR20190902002)

User trust and acceptability6 months

Assessment of staff satisfaction with workflows and patient experience in each arm

Trial Locations

Locations (1)

Rajavithi hospital

🇹🇭

Bangkok, Thailand

© Copyright 2025. All Rights Reserved by MedPath