Abstract
Purpose: :
This study analyzed the accuracy of a morphology-based grading protocol of wide-field digital imaging (WFDI) studies to identify patients with referral-warranted retinopathy of prematurity (RW-ROP) as determined using a "gold standard" indirect ophthalmoscopy (IO) examination by an expert in ROP.
Methods: :
This is an independent, retrospective, masked, cross-sectional validation study. An existing database of simultaneously acquired IO exams and WDFI studies of premature infants at risk for ROP was used. Masked grading of the WFDI studies were performed by the examining physician three years after the original IO exams. All grading was done using a structured, morphology-based, ten-point questionnaire. Outcomes measured are accuracy (sensitivity, specificity, predictive values) of the graded WFDI studies resulting in a referral recommendation relative to the "gold standard" of RW-ROP as determined by IO examination.
Results: :
All patients with both IO and WFDI studies initially performed by 30 weeks PMA were included in this per patient analysis (N=466 studies on 41 patients) for each evaluation. Results are of a "per eye" analysis comparing image grader score to clinical RW- ROP from indirct ophthalmoscopy. The prevalence of referral-warranted ROP (RW-ROP) was 32%. The sensitivity of grader evaluation in detecting referral warranted ROP was 95.1% (95% CI) and a specificity of 89.7%. The positive predictive value was 65% and the negative predictive value was 98.9% for this sample.
Conclusions: :
When compared with the "gold standard" the proposed grading of WFDI studies of infants at risk for ROP provides highly accurate ROP referral recommendations. The inherent advantages of digital imaging systems with structured evaluation provide high quality referral decisions and can be reliably used for evaluating ROP.
Keywords: retinopathy of prematurity • clinical (human) or epidemiologic studies: health care delivery/economics/manpower