A continuous Markov-Chain model of data quality transition: Application in insurance-claim handling

Abstract

Data quality (DQ) might degrade over time, due to changes in realworld entities or behaviors that are not reflected correctly in datasets that describe them. This study presents a continuous-time Markov-Chain model that reflects DQ as a dynamic process. The model may help assessing and predicting accuracy degradation over time. Taking into account cost-benefit tradeoffs, it can also be used to recommend an economically-optimal point in time at which data values should be evaluated and possibly reacquired. The model addresses data-acquisition scenarios that reflect real-world processes with a finite number of states, each described by certain data-attribute values. It takes into account state-transition probabilities, the distribution of time spent in each state, the damage associated with incorrect data that fails to reflect the real-world state, and the cost of data reacquisition. Given current state and the time passed since the last transition, the model estimates the expected damage of a data record and recommends whether or not to correct it, by comparing the potential benefits of correction (elimination of potential damage), versus reacquisition cost. Following common design science research guidelines, the applicability and the potential contribution of the model is demonstrated with a real-world dataset that reflects a process of handling insurance claims. Insurants' status must be kept up-to-date, to avoid potential monetary damages; however, contacting an insurant for status update is costly and time consuming. Currently the contact decision is guided by some heuristics that are based on employees' experience. The evaluation shows that applying the model has major cost-saving potential, compared to the current state.