Having selected the states, NORC needed to identify carriers from those states operating in the small group and individual/conversion market. This began with obtaining the most current NAIC Supplement Health Exhibit from 2011, which lists the carriers operating in each state and market, as well as total premium dollars and enrollment. 27 We used this data to identify all health insurance carriers in the individual and small group markets that had a market share of greater than one percent. A floor of five carriers from each state and market was also set, regardless of market share, allowing additional sampling from states with highly-concentrated markets. When applying these two standards (at least one percent market share or top five carriers), NORC used the one that allowed the inclusion of more carriers. The NAIC exhibit also provided data on member enrollment which was essential for constructing weights (as was done for the Trends study).
Research staff began by searching through each state’s website. Because of the design of these sites and the organization of the rate filing data at the state level, we found there were multiple ways to search for and access filings for specific carriers and applicable markets. Furthermore, we found that in some cases different approaches to searching lead to unexplained differences in the filings retrieved. Therefore, we built a list of all possible query options (search terms and paths) for locating filings submitted by the carriers identified in each state. NovaRest, an actuarial consulting firm and NORC’s subcontractor on the Trends study, reviewed this list of potential query options, to determine which queries from each state’s total list would be used in the data extraction phase. This was necessary because states do not use a universal system for classifying insurance products, and so NovaRest’s familiarity with various state markets was needed to identify labels associated with major medical products. Once the list of queries had been winnowed to fit the study’s parameters, we employed this constricted list to create a query guidance document. We used this document to develop a list of queries for each state to obtain all health insurance filings for major medical policies in the individual and small group markets.
After validating the query guidance document and creating the query list, research analysts employed these documents to conduct data extraction by downloading filings from each state’s website. Data extraction began when an analyst entered the first unique query (for most states, a permutation of attribute-based filters) from the state’s query list in each state’s web portal. The output of this search was captured as part of a relational database created in Microsoft Access, to associate all filings produced with a unique source query, and the process continued until all unique query permutations were searched. For each filing we used in the study, we have captured the exact parameters entered into a State’s search portal and saved a screenshot associated with that query to record all filings that resulted from the query at that point in time. We also documented the number of results returned by each unique query permutation.
The individual filings were then entered into a Data Extraction form. We captured all relevant tracking variables including:
- Market type
- Carrier type
- NAIC Carrier Code
- Three digit record code
- SERFF number
- State ID number
- Extraction date
- Tracking information (Name of extracted file, URL, and extraction date)
This process continued until every filing document connected to the unique query from the permutation list had been extracted. Each Data Extraction form corresponded with a downloaded PDF of the filing found on the state’s web portal. For quality assurance purposes NORC used pre-populated drop-down menus in the Access user interface wherever possible to reduce transcription error.
Some states included in this study use the SERFF system, which is a standardized, automated electronic system that allows insurance companies to submit insurance rate filings and allows states to review, comment, and respond (with an approval or rejection). Querying of state portals that were hosted by the SERFF system could be conducted using standardized query terms. Some states, in addition to using a SERFF system, also used a standardized SERFF template for rate filings, and these templates presented information in a consistent format allowing for easier review of important data elements. However, all filings, regardless of whether they used the SERFF template or not, often had missing information or had information captured only in the correspondence attached to the filing form.
27 Comparable data from the California Department of Managed Health Care (DMHC), which is not a member of NAIC, was used to supplement information from the California Department of Insurance, which is. This and all subsequent references to NAIC data also include this additional DMHC data.