Data Quality Assessment of Medical Insurance Claims
Data Quality Assessment of Medical Insurance Claims

Summary/Abstract: High quality data lay a solid foundation for downstream research. Data quality assessment (DQA) is a necessary yet challenging step in data analytics projects involving coded administrative health data. Eight rules were established based on the presence of age and gender incompatible diagnosis. Using a rule-based method, we assessed data quality of a large medical insurance claim database, the Truven Health Analytics MarketScan Database. DQA results were reported based on nine years of inpatient claim data (Y2006-2014) and three years of outpatient data (Y2006, 2010, and 2014). In both inpatient and outpatient data, the most prevalent data issues are related to newborn conditions. Among the over 2.8 million inpatient records with newborn specific conditions, more than 22% belong to patients who are not newborn. The corresponding error rate in outpatient data is around 2%. Among the over 4 million inpatient maternal records, around 6% contain newborn specific codes. It was observed that the error rates are lower in outpatient data across the board and different patterns of data quality issues are associated with inpatient vs. outpatient records. To the best of our knowledge, this is the first work that systematically assesses the data quality of MarketScan databases.

• Details
• Contents