Cost of Interconnecting Health Information Exchanges to Form a National Network

Abstract

Despite the demonstrated need for a national health information network (NHIN), there has been little progress in turning this need into reality beyond limited local demonstrations. One barrier is the lack of information evaluating the potential costs of connecting these local networks to form a national network. The Center for Information Technology Leadership (CITL), in conjunction with national experts, developed assumptions around the components needed to develop the NHIN. These assumptions were largely based on the architectural approach suggested by the Connecting for Health Common Framework for such a network. Using these assumptions, CITL collected cost data from three different markets engaging in healthcare information exchange (HIE). These costs were then extrapolated to the nation based on population density data from the U.S. Census Bureau. The CITL model projected an initial deployment cost of $97 million and an annual maintenance cost of $41 million for HIE across the NHIN.

Introduction

In 2004, the Office of the National Coordinator for Health Information Technology (ONCHIT) released a 10 year strategic framework calling for the creation of a national health information network (NHIN)¹ to enable the electronic exchange of health information nationwide. The building blocks of the proposed NHIN are Regional Health Information Organizations (RHIOs). RHIOs are multi-stakeholder organizations that enable the secure, electronic exchange and use of health information to improve healthcare quality, safety and efficiency².

Significant progress has been made to identify the steps necessary to develop the NHIN^3–6. However, the costs of connecting organizations to exchange secure health information are less well understood.

Several regions have been involved in healthcare information exchange efforts, such as the Indiana Health Information Exchange (IHIE) in Indianapolis, Massachusetts Simplifying Healthcare Among Regional Entities (MA-SHARE), and Mendocino Health Records Exchange (Mendocino-HRE) in California. All have evolved into RHIOs. Each of these three regional exchanges has its own history of development and investment, and can provide insight on the costs required for the development of HIE and, ultimately, the NHIN.

In contrast to a prior estimate of cost for the NHIN which included a broader assessment of all possible Healthcare Information Technology (HIT) and connectivity costs⁷ or CITL’s previous work on achieving national HIT interoperability⁸, this estimate focuses on the cost of connecting RHIOs to one another. Using the Markle Foundation’s Connecting for Health ‘Common Framework’⁹ as an approach to a NHIN architecture, as well as actual costs from these three regional exchanges, CITL developed a model to project the costs of connecting local HIE networks to form a national network.

Methodology

Overview

To estimate costs of the proposed NHIN architecture to connect RHIOs, CITL began with the architecture described by the Markle Foundation’s Connecting for Health¹⁰. In addition, CITL consulted an expert panel that was available throughout the project to vet assumptions and outcomes of the cost model (see acknowledgements section). While largely influenced by the architecture proposed by the Common Framework, this work can not be interpreted as a true cost model for the framework, but instead as a reasonable approximation of the costs of connecting RHIOs to create a NHIN.

The Common Framework

Connecting for Health is a public-private collaborative designed to address the barriers to the development of an interconnected health information infrastructure. In conjunction with experts in information technology, health privacy law, and policy, the Connecting for Health Common Framework was developed to help health information networks share information locally among their members and nationally while protecting privacy and allowing for local autonomy and innovation. The Common Framework describes reliable authorization, a common set of data exchange standards, and a minimum set of capabilities required to participate in the network, which became the framework of CITL’s cost components and assumptions. Although the Common Framework includes underlying policy costs associated with connecting RHIOs, CITL only included the technical costs incurred by the RHIOs, once each RHIO is already in place.

Cost Markets

CITL collected actual costs from the IHIE, MA-SHARE and Mendocino-HRE organizations. Each of these three communities has successfully evolved into RHIOs and each represents a distinct geographic and demographic setting in the United States. Additionally, recognizing that there are entities that are linked by factors other than geography, such as a payer organization or large integrated delivery system, CITL also considered non-geographic entities that would share health information. Examples include the Veteran’s Health Administration, Department of Defense, Kaiser Permanente, and the Indian Health Services. Therefore, instead of using the term “RHIO” to connote health information exchanges, CITL used Connecting for Health’s term “Sub-Network Organization (SNO)” to label these entities¹¹. A SNO is defined as a business structure comprised of entities that agree to share clinical information in accordance with a minimum set of technical and policy requirements and may be organized by either geography or other relationships not determined by location¹¹.

Cost Model Assumptions

The architecture framework for the NHIN would be a network of SNOs and each SNO would use a Record Locator Service (RLS) to locate the consumer’s records and an InterSNO Bridge (ISB) to talk to other SNOs. Organizations that want to share information across the NHIN must be members of a SNO, maintain an RLS and build an ISB¹².

Based on this framework, for the geographic SNOs, CITL considered two tiers. Tier 1 represents the costs for coordinating and governing data exchanges among SNOs (Table 1). There is little to no hardware or software required in this tier, but it does include the directories, in essence the white and yellow pages for the SNOs. In addition, this tier contains the cost of the governance to manage data sharing from SNOs through the NHIN, typically the costs for contract negotiation, data sharing policy development, and overall management of information exchange policies and procedures. Though important, CITL did not include communication costs in the model, which are in essence the funds required to support the HIE traffic for a SNO through a common Internet service provider.

Table 1.

Components of NHIN Cost Model

		Cost Components
Geographic SNOs	Tier 1	Directories Governance Interfaces
	Tier 2	Record Locator Service InterSNO Bridge (ISB)
Non-Geographic SNOs		InterSNO Bridge (ISB)

For the second Tier, CITL considered the costs for connecting a SNO to the NHIN (Table 1). These costs include the RLS, the ISB, and the interfaces from the NHIN to the SNOs. The RLS is a service that queries the locations of patient records within a SNO¹³. Each SNO has its own RLS. Each SNO would have an ISB, which would be its single gateway for channeling all requests and responses from other SNOs¹¹. The interfaces for this bridge may be built from scratch, modified, or re-deployed from existing intra-SNO interfaces.

For the non-geographic SNOs, only the cost of the ISB was considered since non-geographic SNOs already have existing independent infrastructures in place (Table 1).

CITL assumed that all SNOs were ready for data exchange internally prior to connecting to a national network and therefore did not cost out what the SNOs would need to spend themselves in order to make them ready to exchange data. Thus the only costs were to connect the SNOs to the NHIN itself. For the purpose of this study, we modeled only the base national network without the redundant systems and networks needed for disaster recovery purposes.

NHIN Cost Model

CITL created the NHIN cost model as a stochastic simulation model using Analytica decision software¹⁴. The costs used in the model were based on the actual expenditures of IHIE, MA-SHARE and Mendocino-HRE in their development of a prototypical NHIN-style connection for their SNOs. Cost components for both tiers were compiled into a standardized form and sent to these communities, who were asked to provide initial and annual costs of their respective HIE efforts, as well as demographic data for their markets.

Because there is no existing example of interconnectivity of non-geographic SNOs, our expert panel suggested using IHIE’s experience in connecting distant, geographically separated regions to estimate the corresponding marginal cost for non-geographic SNOs.

Cost Projection

Once costs were determined, CITL needed to extrapolate those costs out to the nation. CITL used data from the U.S. Census Bureau on population density of the U.S. The Bureau has defined ‘metropolitan’ (MetroSA) and ‘micropolitan’ (MicroSA^†) statistical areas in the country, with the former being an urban area with a population of ≥50,000 and the latter containing an urban area with a population of 10,000–50,000¹⁵. They also define a third statistical area, the ‘Combined Statistical Area’ (CSA), defined as two or more MetroSAs.

The Census Bureau maintains statistics on the numbers of each of these statistical areas in the country. With this data, CITL estimated the number of MicroSA SNOs, MetroSA SNOs, and CSA SNOs in a NHIN topology that would provide nearly complete coverage of U.S. population. Population statistics from IHIE, MA-SHARE, and Mendocino were then used as prototypes against these statistical areas with Mendocino serving as the prototype for the MicroSA SNO, IHIE as the prototype for the MetroSA SNOs, and MassShare as the prototype for the CSA SNO (Table 2). Thus, CITL was able to estimate that a NHIN encompassing 77.5% coverage of the U.S. population would be composed of 40 CSA SNOs, 24 MetroSA SNOs, 364 MicroSA SNOs, and, as estimated by our experts, 15∼25 non-geographic SNOs.

Table 2.

Population Extrapolation

Statistical Area (SA)	Prototype Community	Population per SA	Numbers of Communities
CSA	MA-SHARE	4,000,000	40
MetroSA	IHIE	2,000,000	24
MicroSA	Mendocino HRE	67,000†	364
Non-Geographic	IHIE ISB-only	N/A	15–25

While the MicroSA SNOs comprised more than 80% (364/443) of the total geographic SNOs, they only cover approximately 10% of the total population (Table 3). This reflects the U.S. health care landscape, in which there are large numbers of small health care facilities in rural areas. Of note, this reality presents particular challenges for a comprehensive national health care infrastructure, due to the number of potentially diverse organizations that would have to invest resources to allow them connectivity to a national network.

Table 3.

Population Distribution of Geographic SNOs.

Statistical Area	Prototype Community
CSA	69%
MetroSA	21%
MicroSA	10%

Results

NHIN Cost Model Results

Using the methodology and data sources outlined above, the CITL model projected an initial deployment cost of $97 million and an annual maintenance cost of $41 million for establishing healthcare information exchange between existing SNOs across the country. Total costs are summarized in Table 4. The $97 million deployment cost included more than $22 million for CSA SNOs, nearly $9 million for MetroSA SNOs, more than $65 million for MicroSA SNOs, and less than $1 million for non-geographic SNOs.

Table 4.

Total NHIN Costs by Type of Community

Statistical Area	Population	Initial	Annual
Combined SA	160,000,000	$22,120,000	$4,360,000
MetroSA	48,000,000	$8,520,000	$3,240,000
MicroSA	24,388,000	$65,156,000	$33,124,000
Non-Geographic	N/A	$800,000	$100,000
National Totals	232,388,000	$96,596,000	$40,824,000

As shown in Table 5, on a per-capita basis, the deployment cost ranged from $0.138 per-capita to $2.672 per-capita with the lower cost projections possibly reflecting an economy of scale in building health care infrastructures for dense population centers.

Table 5.

Per-Capita Costs by Type of Community

Statistical Area	Population	Per-Capita Initial	Per-Capita Annual
Combined SA	160,000,000	$0.138	$0.027
MetroSA	48,000,000	$0.178	$0.068
MicroSA	24,388,000	$2.672	$1.358

Discussion

When considering the distribution of initial cost by SNO type (Table 6), the MicroSA SNOs are again noted to consume nearly two-thirds of the total cost. This disproportionate consumption of resources by micropolitan areas will need to be addressed aggressively in order to implement a successful NHIN. It suggests a need for an alternative approach for these areas to exchange data that is not solely based on geography.

Table 6.

Distribution of Initial Cost by SNO Type

Statistical Area	Prototype Community
CSA	22.9%
MetroSA	8.8%
MicroSA	67.5%
Non-Geographic	0.8%

Figure 1 shows the percentage of expenditures on different cost components for each of the three types of SNOs. It is noted that MicroSA SNOs spend a higher percent of their budget on record locating services (RLS) as compared with CSA and MetroSA SNOs. This suggests a target for federal intervention in lowering the cost of RLS software by enforcing common data interchange standards to support rural SNO efforts. Indeed, open source has been noted as one element to facilitate the creation of the NHIN¹⁶.

Figure 1.

Distribution of Cost Components by SNO Type

Limitations

A possible limitation of the method used for the cost model is whether the cost data from our three markets are truly generalizable to other communities. However, the costs used for the model are not estimates, but real costs incurred by these communities to build their SNOs. They therefore should provide a realistic picture of the cost of building the infrastructure for connectivity to a national network. In addition, this may be an overestimate of costs since they were extrapolated linearly and thus do not account for any decreasing cost as the NHIN community gains more experience.

Due to different characteristics of the three communities, the associated cost estimates between the three varied tremendously. For example, directories and governance costs, such as legal fees and contract costs, varied by the size of the community. Many factors, such as commercial versus open-source software or unique care patterns of each individual market, could affect the variations in costs of the three markets. However, due to small sample size, we were unable to parameterize the cost estimates based on factors other than population density.

Finally, for this analysis CITL estimated the marginal cost for interconnecting the SNO/RHIOs within the NHIN. As such we did not include the costs that healthcare organizations would incur in order to make their individual systems interoperable. While it is the stated goal of Connecting for Health’s Common Framework to minimize adoption cost by building on existing infrastructure and requiring adherence to only a small set of standards, the total cost to achieve interoperability for the nation will still undoubtedly be significant^{7, 8}. However, the large cost of achieving HIT interoperability is inevitable if the current administration’s goal of providing the majority of Americans with electronic health records by 2014¹⁷ is to be achieved.

Conclusion

Using the technical cost assumptions from the architecture described by the Connecting for Health Common Framework, standard modeling techniques, and real cost data from three organizations engaging in healthcare information exchange, CITL projected costs of connecting local HIE networks to form a national network. The CITL model projected an initial deployment cost of $97 million and an annual maintenance cost of $41 million for HIE across the NHIN.