A central task in realizing the health statistics Vision is to find ways, within the context of privacy protections, to link data that now exist in separate “silos.” Consistent and controlled access to data from various sources are necessary because at present, it is difficult to do any of the following important forms of analysis:
- link or share data from different sites of care (hospitals/nursing homes/ambulatory care)
- understand the relationships between health status, health determinants, and health services
- link data from surveys and administrative and encounter data
- link complementary data from private and public databases
- combine comparable data from local, state, and federal systems
Because most data systems were created to address specific needs, with no common framework or consensus on an overall “system,” it has been difficult to compare or link data between data systems. There are many reasons for this difficulty: constraints on the use of specific data sets to compensate for inadequate national privacy protections, differing practices and systems within branches and levels of government, the way data elements are defined and coded, and the technical complexity of combining data collected with different methodologies.
It should be understood that the goal is not to replace the data “silos” of today with some sort of consolidated mega-database¾the mythical “Central Database in the Sky.” Rather, the goal is to provide controlled mechanisms for accessing and combining data from different sources for defined public health purposes. (This conception is much like that for the National Health Information Infrastructure.) Such linkages could enhance understanding of the causes of ill health and what to do about them. They also could make our statistical system more efficient, less burdensome to data providers, and more responsive to emerging data needs.
With a concerted effort, it should be possible to remove unnecessary barriers to an integrated, efficient system. The barriers can be overcome by greater attention to standardization and comparability, improvements in technology, and creative approaches to making data available for analytic use in ways that do not jeopardize individual privacy.(16)
Similar effort will be required to overcome the philosophical and practical differences preventing greater consistency, sharing, and integration across public and private entities. While many of our most historically useful data systems have been conducted by public agencies for public purposes, an increasing array of health information initiatives originate in the private sector, where data are initially justified as being useful for proprietary or institutional purposes. While these data could be of great use if placed in the public domain, it is not always clear that private organizations have the necessary incentives or protections to make data widely available. Similarly, data derived from systems established for non-statistical purposes (e.g., payment systems) can be of broader use, but may not be readily converted to a useful form. A well functioning health statistics system must find ways to encourage such mutually supportive collaborations.