While not easily talked about, the ability of ACOs to aggregate patient-identifiable hospital, community and payment data in one place will be central to this work. The data needs to be identifiable to track and manage patients across care settings. Analysis will also be much easier when the data is in one place. Some ACOs, particularly those made up of community physicians, may outsource data analysis. In general, though, hospitals and health systems will be best equipped to implement the kind of data warehouse and analytic infrastructure necessary for supporting broad community data stores. An issue is whether the ACO savings-sharing discussions can overcome community provider distrust of local hospitals enough to allow them to fully access community provider data. At the same time, it will be critical for hospitals to be allowed to actually benefit from prevented admissions. Regardless of these political questions, ACOs that aggregate the most and best-structured clinical and claims data will be best positioned to analyze and benefit.

That ACO Sweet Spot

Analytics are usually done in non-transactional data stores, sometimes as data marts built off of the transactional store (like an EHR) and sometimes as stand-alone data warehouses. ACOs will have to merge large amounts of clinical care, process and claims data and be able to perform canned and dynamic queries. Some ACOs will likely build up existing hospital data warehouses and business intelligence software packages they may have already deployed. Generally speaking, health data in these systems are loaded, as they are available with only a limited degree of work done to clean them. A substantial effort is then involved in developing customized queries for that organization.

There are, however, a number of newer tools angling for the ACO sweet spot that focus on using structured data inputs like the HL7 Continuity of Care Document standard and lab messages and use proprietary storage schema that allow for more pre-developed query and data analytic capabilities. As a result, some of these products can also support outgoing services for alerting and reporting back into EHRs. At times this would mean supporting somewhat redundant transactional, data warehouse and dedicated analytic stores. But with data storage costs continuing to drop and with the challenges of working with large data sets on multi-use systems, this redundancy may be expedient and not extravagant.
   
Others may try to arrive at an ACO analytic sweet spot via health information exchange or even EHR software. While one can't rule out either of these approaches, for an ACO of any significant size that has multiple EHR software systems and other claims and process data needs, a separate infrastructure will probably make the most sense.

In our next issue, part 2: The HIT of ACOs – Data to Improve Quality Management.

John Loonsk, MD, FACMI, is chief medical officer for CGI Federal. From 2006-09, he was director of interoperability and standards in the Office of the National Coordinator for Health Information Technology.