Real-World Interobserver Reproducibility Of Autokeratometry: Does Operator Attention Matter?

To evaluate whether operator attention influences autokeratometry measurements in a busy academic cataract surgery center.

Vajira Hospital, Navamindradhiraj University, Bangkok, Thailand.

In this prospective study, 48 patients (98 eyes) scheduled for cataract surgery underwent autokeratometry performed by two groups of operators: routine medical personnel conducting measurements as part of standard clinical workflow, and ophthalmic staff who were aware of the study and optimized patient positioning, encouraged stable fixation and instructed blinking prior to measurement. Interobserver agreement was assessed using Bland–Altman analysis.

For flat keratometry, the mean bias was –0.099 D with (95% limits of agreement (LoA)–0.777, 0.578 D). Steep keratometry showed a mean bias of 0.005 D with (95% LoA  –0.907, 0.917 D). For mean keratometry, the mean bias was –0.041 D, with (95% LoA –0.747, 0.665 D). For astigmatic parameters, the magnitude of deviation demonstrated a mean bias of –0.105 D with (95% LoA  –1.245, 1.036 D). Astigmatic axis showed a mean bias of 2.143°, with (95% LoA  –71.225°,  75.510°). No proportional bias was observed.

Autokeratometry demonstrated strong interobserver agreement despite differences in operator attention under real-world, busy routine clinical conditions. Mean interobserver biases were minimal across all keratometric parameters, although wider limits of agreement were observed for astigmatic axis measurements. Overall, these findings support autokeratometry as a reliable and operator-friendly tool. It may serve as an additional modality for cross-checking corneal power measurements alongside optical biometers, thereby increasing confidence in intraocular lens power calculation.