Health insurance exchanges (HIEs) are a key part of the Affordable Care Act (ACA). Exchanges and insurance plans will be bucketed into four tiers based on their quality ratings. Insurers that aren’t rated cannot participate.
The National Committee for Quality Assurance (NCQA) rates health insurers based on Healthcare Effectiveness Data and Information Set (HEDIS) measures. To calculate an insurer’s HEDIS rating, the NCQA requires:
- Member data that contains member profiles and demographics;
- Enrollment data that tells the HEDIS rating engine if a particular member is enrolled in medical, pharmacy and other insurance lines;
- Claims: all types, along with multiple diagnosis and procedure codes;
- Lab results, such as blood sugar levels for diabetics;
- Children’s immunization records; and
- Provider data, including the name, address, zip code and phone number of each office.
Member data is typically stored and maintained in different systems. For example, preferred provider organization (PPO) and HMO member data are often kept in different systems with different requirements.
The only place these data are brought together is in an operational data store (ODS). Hence, the ODS is the best system for most business intelligence (BI) projects, including providing HEDIS data.
At many insurers, the ODS, built on older systems, hasn’t been properly tested for data-quality issues. It doesn’t provide a good way to identify individuals as they move from PPO plans to HMO plans. As a result, a single person often appears to be two different people, and sometimes two people with the same name become a single person. Additionally, there are character limitations for first names, last names are missing in up to 60 percent of patients, and there is often no phone number on file.
Similar problems exist for provider data – hospitals, doctors, pharmacies, dentists, etc. Each different type of provider data is managed by a different department and exists in different databases with different shapes and forms. It is challenging to bring them together under one structure. There are different internal identifiers to identify different providers. These databases often have overlapping records of providers; for example, doctors’ data might also include dentists.
Claims data is equally challenging. Different types of claims – such as professional, facility, drug and vision – go through different systems and follow a different processing cycle. By the time claims reach the ODS, they’ve been de-normalized of member and provider data such as names and stripped of internal identifiers. This makes it hard to link them back to the provider and member data.
An incorrectly spelled name from the transaction system will cause the de-normalized claim to have the wrong name, and the name can no longer be matched with the original member. That makes the claim unusable for HEDIS.
Older claims systems don’t quickly adapt to the changes in the industry. For example, an older claims system cannot typically capture more than one diagnosis per claim.
Insurers get claims for lab work but don’t get results. Fortunately, some companies can gather lab data from various labs and provide it to insurers. The challenge is to match the member data with the patient information.
Similarly, child immunization data is also not readily available to carriers via claims data. However, several states maintain an immunization repository, which makes it easy for insurance companies to get data. The challenge is to match member data with the state’s patient data.
Carriers can act to ensure they are well positioned for the NCQA and other data. Establishing a “single source of truth” would be ideal, but there are operational and financial barriers. Insurers can implement other, less-intrusive solutions to improve data quality.
One problem with member data is master data management (MDM). Person-matching software that takes in several attributes can eliminate the problem of identifying the members as they jump from one insurance product to another or change jobs. It’s easier to implement this before the data gets into the ODS. Another solution is improving extract transform load (ETL).
For provider data, there are many products that manage different types of providers in one system, thus allowing the downstream process to get the data from one place.
For claims, the best solution is to move to one of the latest claims systems. Other problems with claims could also be addressed by improving ETL processes.
About the author
Yunus Burhani is a senior architect with X by 2. For more on X by 2, click here.
