i202 Fall 2009 School of Information, UC Berkeley

A recent report by the department of Health and Human Services (HHS) found that the sole central database for Medicaid claims and eligibility information is riddled with problems. The database, operated by the Centers for Medicare and Medicaid Services (CMS), is incomplete, out of date, and less accurate than claimed.

Coverage from modernhealthcare.com mentions that the database is missing roughly half the elements needed to effectively detect and combat fraud; important basic data like the identifier of the care provider for a treatment is currently not being captured. Given that these types of information were flagged as important by CMS, their absence speaks to major failings in the database schema.

Coverage at Mass Device details further problems. The HHS report looked at Medicaid data collected for the period between 2004 and 2006, which was supposed to be released to the public. Fully 46% of data files are still not complete, and the rest took roughly a year and a half on average to release. A large part of this latency is explained by aggregation issues - data has to first be submitted by states, which are often late. In fact, 10% of all files for the 2004 to 2006 period still haven't been sent in by states.

Once the data is submitted, it takes fully four months for the CMS to validate it, presumably due to issues with the incoming data; the report also indicated that error tolerance for validation had to be increased to account for formatting inconsistancies. The increased error tolerance had not previously been publicly disclosed. The errors at the root of the problem speak to the difficulty of aggregating massive amounts of data from different sources, which may use different formats than the official central database.