Introduction
All Payer Claims Databases (APCDs) currently exist in ten states, with another twelve in the development stage. In addition, there are a number of non-state government claims databases that exist throughout the country. As APCDs are being created, the next step in the progression of obtaining more comprehensive, timely, and accurate healthcare data is to link the APCDs with other databases. The three databases most commonly discussed include two historic databases (hospital administrative and vital statistics) and data from the newly created health information exchanges (HIEs), which will be the focus of this discussion.
HIEs have the potential to enhance existing APCDs with clinical information for quality and outcomes reporting. Because of their content and purpose, HIEs and APCDs will be distinctly separate initiatives as they are developed. If both are integrated, data will exist for comparative effective research, population health applications, and to improve risk adjustment, clinical studies, and outcomes research.
With The Health Information Technology for Economic and Clinical Health Act’s (HITECH) roll out of electronic health records and the establishment of Federal grants to expand health information technology (e.g. – American Recovery and Reinvestment Act State Grants to Promote Health Information Technology), which greatly augmented existing private initiatives, HIEs now can be found in some form in every state in the US. Simultaneously, a number of Federal initiatives associated with, or driven by, the Patient Protection and Affordable Care Act (e.g. – creation of accountable care organizations, patient-centered medical homes, and CMS funded Multi-Payer Advanced Primary Care Practice Demonstration pilots and Performance Measurement pilots) have elevated the discussion regarding the integration of administrative data derived from APCDs with clinical data generated by HIEs to a new level.
However, while the linkage of claims data with clinical data has taken on a greater significance, overlapping technical and legal impediments exist which may prevent the successful merger of these two robust data sources.
Technical Issues
The critical points of linkage between APCDs and HIEs occur with member/patient identification and rendering provider identification.
Patient data derived from HIEs usually contain name, date of birth, gender and patient account/control numbers. Social security number and place of residence of the patient may or may not exist and are not always accurate. Member data from APCDs usually contain name, date of birth, gender, social security number, plan specific contract number, and patient account/control numbers. However, most state and non-state APCDs collect the member names, social security number, plan specific contract number from the payers in a hashed/encrypted form. In order to accurately perform the linkage, the equivalent HIE patient data must be processed through the same hashing/encrypting algorithm.
This problem can be rectified if APCDs collect identifiable data that are not hashed or encrypted or if a master patient/member index is established in a state, which would provide all individuals receiving health care with a unique numeric identifier that is used by both health care payers and providers. The establishment of such a system has its own technical problems (who assigns the number and how does it move throughout the system) as well as potential political ramifications with individuals expressing concerns about privacy.
Linking the rendering providers in the two databases is less problematic because provider name(s), NPIs, and the location of the service(s) provided appear in both the APCD and HIE systems. However, historically, three issues have emerged with APCDs that can impact the rendering provider linkage. First, due to contractual issues and other factors, payers have submitted the billing provider in lieu of the rendering provider. Second, physician assistants and nurse practitioners may have their own NPIs but also provide services under the direction of a supervising physician, which can result in a situation where the NPI does not associate with the name of the rendering provider. This creates confusion with the rendering provider attribution process and with the linkage. Third, due to the nature of prescription transactions, the accurate identification of prescribing physicians within APCDs has been incomplete.
Although none of the identified issues are insurmountable, it means that the linkage process cannot rely solely on matching the NPIs.
Two more serious technical issues must be understood when attempting to link APCD data with HIE data.
First, APCDs do not contain data on all individuals receiving medical care in a state or regional geographic area. Some APCDs have no Medicare data and/or claims data from TPAs. Most do not have claims for Federal employees. APCDs do not have data on uninsured patients and, to date, Tricare data. Although the acquisition of these data may occur in the future, the current omissions will impact certain analytic uses of the merged data.
Second, one large impediment can exist to make the APCD/HIE data linkage extremely difficult, if not impossible. HIEs can be designed in two basic models: distributed and consolidated. In a distributed model, the data resides with each provider and is pulled across the system only when needed. In a consolidated model the data moves from the individual providers to a central data repository (which can easily be converted into a database) and then to the users. In order for APCD and HIE data to be linked in a distributed system, the linkage would need to occur with every individual provider, which would make it technically cumbersome and cost-prohibitive.
Legal Issues
As mentioned previously, in order to most accurately link the APCD data to the HIE data, a master patient/member index would be ideal. If an index does not exist, the actual names of patients and members in both databases would provide accurate linkages, followed by the actual names in the HIE and encrypted names in the APCD. However, at present, a number of states have enacted legislation prohibiting the collection of, and/or the release of, APCD data that directly identifies individuals, which will certainly make the linkage of the APCD and HIE data much more difficult.
The current statutory status relative to the collection of member/patient identifiers for states with existing or planned APCDs is as follows:
State |
Status |
Colorado |
All data transmitted from the carriers, including patient identifiers will be encrypted during transmission and while stored within the APCD. Data will be decrypted briefly as received from the carriers so that a unique identifier can be attached to each patient, and then re-encrypted. All data will be released without direct patient identifiers. |
Kansas |
Not currently allowed for commercial data, but Kansas currently collects identifiable information for state employees and Medicaid. |
Maine |
Allowed by law, but prohibited by law from being disclosed; not currently collected. |
Maryland |
Allowed by law. Currently collecting unencrypted patient identifiers. |
Massachusetts |
Allowed by law. Currently collecting unencrypted patient identifiers. |
Minnesota |
Not currently allowed. |
New Hampshire |
Not currently allowed. |
New York |
Allowed by law. System not implemented yet. |
Oregon |
Currently collecting a subset of unencrypted patient identifiers. |
Rhode Island |
Not currently allowed. |
Tennessee |
Not currently allowed. |
Utah |
Allowed by law. Currently collecting unencrypted patient identifiers. |
Vermont |
Allowed by law. Currently collecting encrypted patient identifiers. |
West Virginia |
Allowed by law to be collected, but not disclosed. |
Source: All-Payer Claims Databases 2.0: The Next Evolution, July, 2011; APCD Council
Legal issues also exist with the data generated by HIEs. At present, all of the HIEs in the US are private organizations. Most are established to electronically transfer clinical data between providers in a particular health care system. These data are owned by the providers participating in the HIE and are shared with patients. If proposals are developed to link and share APCD and HIE data, it will be necessary for the governing bodies of the HIEs to approve the transfer of their private data to the state agencies, where it would be used to produce various public reports. It is unclear at this time how much HIE data will be allowed to migrate to the public domain.
The issue of data ownership is further complicated by the current structural state of HIEs in the US. There are few states which have operational statewide HIEs covering a large portion of the state. According to a February, 2011 publication (State HIE Strategic and Operational Plan Emerging Models) of the Office of the National Coordinator for Health Information Technology, there are only ten states that have centralized exchanges. Consequently, approval for data transfer would need to come from a large number of organizations rather than one, which will greatly complicate the process of data transfer.
Conclusion
Without question, the merger of APCD and HIE data will be tremendously important to the future of health care analytics. However, the process will be technically, legally, and politically complex. Without modifications to most of the states’ APCD legal/regulatory framework governing the collection and release of data, the linkage of the two databases will be very difficult and will hinder (or prevent) many of the anticipated uses. Furthermore, until the APCD and HIE data are actually linked and used for meaningful analytics that are available to the public, state governments will be leery to make legal/regulatory changes, and so we may find ourselves in a “catch 22” situation for the foreseeable future.