As part of the Affordable Care Act, taxpayers who obtain insurance through healthcare.gov receive a 1095a tax form which explains their coverage and any premium tax credit they have received. It is then used to reconcile any coverage detail discrepancies reported by the consumer after receiving the form at tax time. A case must be opened to resolve the matter and have a corrected 1095A issued to the consumer, when applicable.
The reconciliation process was largely manual, often taking analysts 30 minutes to research and resolve a single case. To determine the correct resolution, the type of discrepancy (case type) must first be identified. Case types include but are not limited to, address, coverage date, tax credit, and multiple issues. To determine the correct information and resolve the complaint, the consumer’s claim must be compared with government, issuer, and consumer data existing across multiple large databases.
Our team leveraged artificial intelligence to identify the case type. Once the case type is known, a series of case-type-specific rules can be applied to resolve the case. We used Spark and Python, available in Microsoft Azure HDInsight, to extract the necessary information from the large amounts of data to resolve the case.
High volumes of cases can now be resolved at the rate of one case per second with our automated basic data comparisons. We have estimated that between 60 and 80 percent of cases can now be solved automatically.