gms | German Medical Science

Introduction: It is a current trend in healthcare for medical practitioners to affiliate in cooperative medical supply centers to increase their power to compete [1]. In such an integration scenario [2], data quality is an important concern: On the one hand, integration can benefit from data quality, on the other hand, many data quality problems only become evident when comparing several sources of data [3]. Moreover, for certain use cases in economic controlling certain kinds of data quality are more important than others: some data quality problems translate directly to decreased income for the center. Our goal is to establish useful quality considerations, as well as ways to assess them.

Methods: We conducted interviews with practice managers, asking about their notions of data quality and their most pressing needs for improvement. Data quality is traditionally measured along several dimensions, many of which are well established [4], although the exact definitions vary. Constructing metrics for these is commonly regarded as highly non-trivial [5].

Results: Our interviews exposed currency, correctness and completeness [4] as the dimensions practice managers are most interested in. Completeness is separated into several sub-dimensions [3]. One of these is population completeness [6], which denotes the percentage of real-world entities that have a distinct corresponding entry in the database. Measurements along this dimension take an exceptional position in a clinical context since, according to practice managers, they can directly influence a supply center’s revenue. As a guiding example, consider tracking the benefits a medical unit provides to its patients. As long as the sum of the proceeds of these benefits stays below a certain budget, the Association of SHI Physicians reimburses the full sum. If the proceeds go over budget, however, only a fractional amount is remitted. For this reason, the dimension of completeness is tied closely to the remaining budget capacity. As long as a medical unit stays below its budget, population completeness of the creditable benefits is highly important for the practice manager. According to our interviews, an accurate value enables the manager to countersteer if necessary. Unfortunately, we cannot establish a reference relation [7] due to the distributed and independent nature of our data sources. A comprehensive source with guarantees of completeness does not inevitably exist in a medical supply center. That is why it is necessary to estimate rather than measure the population completeness. We developed, and are currently evaluating, a completeness metric to support the centers’ economic planning. The metric is based on the evaluation of past account data and the prediction of future occurrences of deductible benefits.

Conclusion: Population completeness is an important consideration for economic planning in medical supply centers. In this domain, completeness of data may be impossible to measure. Therefore, we need to develop methods to estimate population completeness, a non-trivial challenge. The metric we developed ties in with our general approach to data quality [8] inspired by Wang’s TDQM [9].

Acknowledgements: The project is supported by the German Federal Ministry of Education and Research (BMBF), project grant No. 01EX1013D.