A Model for Setting Optimal Data-Acquisition Policy and its Application with Clinical Data

Manual data acquisition is often subject to incompleteness – data attributes that are missing due to time and data-availability constraints, which might damage data usability for analyses and decision making. This study introduces a novel optimization model for setting mandatory versus voluntary attributes in a dataset. This model may direct the decision of whether or not to enforce the acquisition of certain attributes, given certain constraints and dependencies. The feasibility and the potential contribution of the proposed model were evaluated with a clinical dataset that reflects Colonoscopy procedures performed in a large hospital over a 4-year period. The evaluation demonstrated that the model can be reasonably estimated within the given context, and that its implementation may contribute important insight toward improving data quality. The current data-acquisition setup was shown to be suboptimal, and some further evaluation identified factors that influence incompleteness and may require revisions to current data acquisition policies.