Waterloo Eye Study: Data Abstraction and Population Representation

    loading  Checking for direct PDF access through Ovid


Purpose.To determine data quality in the Waterloo Eye Study (WatES) and compare the WatES age/sex distribution to the general population.Methods.Six thousand three hundred ninety-seven clinic files were reviewed at the University of Waterloo, School of Optometry. Abstracted information included patient age, sex, presenting chief complaint, entering spectacle prescription, refraction, binocular vision, and disease data. Mean age and age distributions were determined for the entire study group and both sexes. These results were compared with Statistics Canada (2006) estimates and information on Canadian optometric practices. Inter- and intraabstractor reliability was determined through double entry of 425 and 50 files, respectively; the Cohen kappa statistic (K) was calculated for qualitative data and the intraclass correlation coefficient (ICC) for quantitative data. Availability of data within the files was determined through missing data rates.Results.The age of the patients in the WatES ranged from 0.2 to 93.9 years (mean age, 42.5 years), with all age groups younger than 85 years well represented. Females comprised 54.1% and males 45.9% of the study group. There were more older patients (>65 years) and younger patients (<10 years) than in the population at large. K values were highest for demographic information (e.g., sex, 0.96) and averaged slightly less for most clinical data requiring some abstractor interpretation (0.71 to 1.00). The two lowest interabstractor values, migraine (0.41) and smoking (0.26), had low reporting frequencies and definition ambiguity between abstractors. Intraclass correlation coefficient values were >0.90 for all but one continuous data type. Missing data rates were <2% for all but near phoria, which was 7.4%.Conclusions.The WatES database includes patients from all age groups and both sexes. It provides a fair representation of optometric patients in Canada. Its large sample size, good interabstractor repeatability, and low missing data rates demonstrates sufficient data quality for future analysis.

    loading  Loading Related Articles