AIMS: There is a need for evaluation of screening and grading services for diabetic retinopathy (DR) in compliance with quality-assurance (QA) standards. We describe the screening/grading QA programme set up for OPHDIAT over the 2005-2006 period. METHODS: Screening and grading objectives, evaluation criteria and minimum acceptable QA standards were set. To ensure the quality of DR photos, the proportion of nongradable photos in at least one eye had to be less than 10%. To ensure grading accuracy, intergrading agreement had to be greater than 90%. Grader-generated reports had to be available in less than 48 h for more than 80% photos. Readers had to grade 500 to 3000 photos per year. RESULTS: Sixteen screening centres were opened between June 2004 and December 2006, and 14,769 patients were screened. Percentages of nongradable photos were consistently below the QA requirement (less than 10%). Overall, 800 photos were graded a second time by a reader blinded to original grading; agreement between graders ranged from 92 to 99%. More than 90% of grader-generated reports were produced within 48 h. The number of readings by each grader nearly achieved the QA standard. CONCLUSION: QA for DR telescreening should be a continuous process to provide performance feedback, thus guaranteeing a high standard for delivered results. Almost all of the predetermined QA standards in OPHDIAT for screening and grading were met. Besides the quality/sensitivity of the screening/grading modalities, it is important to evaluate at-risk patients so that they can be treated efficiently; this should be addressed in a global QA programme.
AIMS: There is a need for evaluation of screening and grading services for diabetic retinopathy (DR) in compliance with quality-assurance (QA) standards. We describe the screening/grading QA programme set up for OPHDIAT over the 2005-2006 period. METHODS: Screening and grading objectives, evaluation criteria and minimum acceptable QA standards were set. To ensure the quality of DR photos, the proportion of nongradable photos in at least one eye had to be less than 10%. To ensure grading accuracy, intergrading agreement had to be greater than 90%. Grader-generated reports had to be available in less than 48 h for more than 80% photos. Readers had to grade 500 to 3000 photos per year. RESULTS: Sixteen screening centres were opened between June 2004 and December 2006, and 14,769 patients were screened. Percentages of nongradable photos were consistently below the QA requirement (less than 10%). Overall, 800 photos were graded a second time by a reader blinded to original grading; agreement between graders ranged from 92 to 99%. More than 90% of grader-generated reports were produced within 48 h. The number of readings by each grader nearly achieved the QA standard. CONCLUSION:QA for DR telescreening should be a continuous process to provide performance feedback, thus guaranteeing a high standard for delivered results. Almost all of the predetermined QA standards in OPHDIAT for screening and grading were met. Besides the quality/sensitivity of the screening/grading modalities, it is important to evaluate at-risk patients so that they can be treated efficiently; this should be addressed in a global QA programme.