When data are manually entered, there are possibilities for errors to be made. For dense longitudinal data, these errors are easy to identify when the data are plotted because they appear as huge spikes away from the longitudinal trend. However, in large datasets it is too tedious to manually examine every subject. Hence, we employ b-splines to detect data incorrectly entered since they are commonly used to create smooth trends. Briefly, we design an automated procedure to fit multiple b-splines, select the better fitting spline with AIC, then perform a residual analysis leveraging the Bonferroni correction. In this talk, we show the utility of this method by applying it to epidemiological data from an electronic medical record.