To the Editor:

RE: The reliability of visual acuity measurements from inpatient referrals to ophthalmology

I read with interest, Hazelwood and Nderitu’s March article [1] concerning the use of portable Snellen charts and smartphone apps for inpatient assessment of visual acuity (VA), and applaud their proper use of intraclass correlation coefficients (ICCs) and Bland–Altman analysis, often neglected in VA reliability investigations. However, the reported statistical performance in this context is worryingly poor. ICCs of 0.309 and 0.224 [1] represent very poor agreement [2] between inpatient and clinic VA, and this is borne out in Bland–Altman plots exhibiting 95% limits of agreement (LOA) at ±0.62–0.88 logMAR [1], variability that could span 6/6–6/38 Snellen; a patient could have been assessed as sight impaired in one setting, and with perfect vision in another. This compares to LOA of around ±0.15 logMAR reported in the clinical setting [3].

These results highlight the lack of reliability of inpatient VA assessment, suggesting that (a) training in VA assessment is insufficient, and/or (b) reliable tools for non-specialists have not been developed or tested. Given that ophthalmological clinics exhibit far better reliability, I would suggest that the clinician-directed element of VA testing may be the underlying factor driving the variability of inpatient readings. To fix this problem, either training must be improved, or tools must be implemented that require less clinician involvement. Validated applications exist that require little to no clinician input [4, 5], but none have yet been trialled in the inpatient setting. Further work is indicated to improve the quality of referrals from inpatient wards, and while training of doctors in other specialties in conventional methods would be ideal, adoption of alternative tools may be a more feasible solution.