RT Journal Article SR Electronic T1 Estimating patient demographic profiles from practice location JF Canadian Family Physician JO Can Fam Physician FD The College of Family Physicians of Canada SP 414 OP 419 VO 58 IS 4 A1 Michael Shortt A1 William Hogg A1 Rose Anne Devlin A1 Grant Russell A1 Laura Muldoon YR 2012 UL http://www.cfp.ca/content/58/4/414.abstract AB Objective To test the accuracy of imputing a practice population’s average socioeconomic characteristics (such as average education levels and average income) using census data centred on the location of the practice. Design Comparison of census data with survey data collected in primary care offices. Setting Ontario. Participants A cross-sectional sample of patients from 116 urban practices. Main outcome measures Patient data were compared with census data at different levels of aggregation using mean absolute relative error (ARE), median ARE, and Spearman rank correlations. Results A total of 4413 patient surveys were collected. Differences between patient profiles and census data were large. Most mean AREs were clustered between 0.70 and 0.80, and median AREs were as high as 1.67. Correlations were low (ρ = 0.02) to moderate (ρ = 0.48). These results held across both levels of aggregation. Conclusion The use of imputation techniques based on practice location is inadvisable, given the large differences that were observed.