Introduction: In observational longitudinal studies, visit irregularity can lead to biased conclusions about the outcome and should therefore be accounted for in regression analyses. We propose metrics for quantifying the extent of irregularity. Setting: The time-period is split into bins to calculate the proportions of subjects with 0, 1 and > 1 visits per bin. We plot the mean proportions of 0 vs. > 1 visits per bin as bin width is varied and use the AUC to judge the extent of irregularity. Our metrics are applied to two studies. We perform simulations to assess properties of the AUC. Results: Simulations indicated the mean AUC increased as the variance of gap-times between visits increased, and as the percentage of missing values within each scheduled measurement occasion increased. Conclusion: Ignoring irregularity can lead to bias when the observed history is predictive of future visit intensity. Our metrics can assist in selecting the appropriate outcome approach.
Session
Date and Time
-
Language of Oral Presentation
English
Language of Visual Aids
English