Purpose: To compare the results of manifest refraction obtained with two different hand-held autorefractors (Sure-Sight, Welch Allyn Co. and Retinomax 1, Nikon Inc.) and with the Topcon RMA 6000 on-table autorefractor in order to estimate any potential bias between these refractometers and to compare the diagnostic performances of these two hand-held autorefractors as screening devices.
Methods: Ninety-eight children were refracted under manifest conditions with the three above-mentioned refractometers and under cycloplegic conditions with the Topcon on-table autorefractor, or by means of retinoscopy. The agreement between the manifest measurements obtained with the three different autorefractors was studied using the method of Bland and Altman. The validity of several thresholds of manifest refractive anomalies as measured with the Sure-Sight and with the Retinomax was estimated by receiver operating characteristic (ROC) curves using cycloplegic measures as reference. results There is a spherical positive bias of 1 D between the Sure-Sight and the Retinomax and better agreement between the Topcon and the Retinomax. The surface area indexes of the ROC curves and the diagnostic performances in term of sensitivity and specificity are better with the Retinomax in cases of hyperopia, astigmatism and anisometropia. For myopia, the Sure-Sight has better performance.
Conclusions: The results suggest that either device may be sufficient for assessing refractive errors in children in a screening setting. However, because of a bias between both refractometers in measuring the sphere, distinctive referral criteria must be chosen for the detection of hyperopia and myopia. The diagnostic performance is slightly in favor of the Retinomax.