Predictive value of screening tests for visually significant eye disease.

PURPOSE: To determine the predictive value of ophthalmic screening tests with visually significant eye disease in a cohort of American Indian/Alaskan Natives from the Pacific Northwest.

DESIGN

Validity assessment of a possible screening protocol.

METHODS

Ophthalmic technicians performed a screening examination including medical and ocular history, best-corrected visual acuity, limbal anterior chamber depth assessment, frequency-doubling technology perimetry (FDT, C-20-5), confocal scanning laser ophthalmoscopy, nonmydriatic digital photography, and tonometry on 429 participants. An ophthalmologist performed a comprehensive eye examination on subjects with 1 or more abnormal screening tests and a random selection of those with normal screening tests. We used univariate and multivariate logistic regression to determine the association between abnormal screening test results and visually significant eye disease. We also determined the predictive value of screening tests with ocular disease.

RESULTS

Univariate analysis identified history of eye disease or diabetes mellitus (P < .001), visual acuity <20/40 (P < .001), abnormal/poor-quality confocal scanning laser ophthalmoscopy (P < .001), abnormal FDT (P < .001), and abnormal/poor-quality nonmydriatic imaging (P < .001) as associated with visually significant eye disease. A multivariate analysis found visually significant eye disease to be associated (P < .001; receiver operating characteristic curve area = 0.827, negative predictive value = 84%) with 4 screening tests: visual acuity <20/40, abnormal/poor-quality nonmydriatic imaging, abnormal FDT, and abnormal/poor-quality confocal scanning laser ophthalmoscopy.

CONCLUSIONS

Ophthalmic technicians performing a subset of screening tests may provide an accurate and efficient means of screening for eye disease in an American Indian/Alaskan Native population. Confirmation of these results in other populations, particularly those with a different profile of disease prevalence, is needed.