Sir,

Response to the letter by Mr Depak Gupta,

I would like to thank Mr Gupta, Dr Hadinapola and Mr Eke for their interest in our paper and the contribution of their unit in providing data for the study.1 They have highlighted in a small survey of some of the doctors in their own department that not all carefully record all risk factors. They have not measured the accuracy of the responses, and therefore no firm conclusions can be drawn as to the magnitude of bias caused. Many of the risk factors identified in our study (1) are independent of the diligence of data recording by the surgeon (patient age, male gender, axial length >26 mm, use of alpha blockers) as they are recorded by nurses during pre-assessment and biometry or are highly likely to be recorded accurately (grade of surgeon). This leaves the following factors potentially susceptible to bias in their, and probably other units: (presence of glaucoma, diabetic retinopathy, brunescent/white cataract, no fundal view/vitreous opacities, pseudoexfoliation and pupil size).

Accuracy of data is an issue in any study. As all data were anonymised before extraction it was not possible for us to evaluate the accuracy of the data. The Medisoft electronic patient record (EPR) system attempts to balance speed of data entry with completeness of data capture by using default settings for some fields, but by forcing data entry in important fields such as co-pathology and complications. Using EPR completeness of data capture can be guaranteed, but not accuracy. I am constantly looking for ways to design out potential sources of inaccurate data entry; for example, we may implement automatic selection of co-pathology options based on previously entered diagnoses, so that over time the accuracy of data entry in the EPR may improve. As EPR systems begin to entirely replace paper records, rather than just being used to audit specific care pathways, there will be increasing opportunities for cross-checking the accuracy of data entry.

Despite the potential for bias introduced by inaccurate data entry for some of the risk factors, it is gratifying that the authors recognise that the findings are ‘broadly in line with our clinical experience’.