Understanding the Decisions and Values of Stakeholders in Health Information Exchanges: Experiences From Massachusetts

Abstract

Objectives. We studied how health information exchange systems are established by examining the decisions (and thus, indirectly, the values) of key stakeholders (health care providers) participating in a health information exchange pilot project in 3 Massachusetts communities. Our aim was to understand how these kinds of information exchanges can be made viable.

Methods. We used semistructured interviews to assess health care providers' decision-making processes in selecting technical architectures and vendors for the pilot projects to uncover their needs, expectations, and motivations.

Results. Our interviews indicated that, after extensive evaluations, health care providers in all 3 communities eventually selected a hybrid architecture that included a central data repository. However, the reasons for selecting this architecture varied considerably among the 3 communities, reflecting their particular values. Plans to create a community patient portal also differed across communities.

Conclusions. Our findings suggest that, to become established, health information exchange efforts must foster trust, appeal to strategic interests of the medical community as a whole, and meet stakeholder expectations of benefits from quality measurements and population health interventions. If health information exchange organizations cannot address these factors, sustainability will remain precarious.

Health information exchange (HIE)—mobilizing health care information electronically across organizations within a given region or community—has been heralded by health care experts and government leaders as an essential component of programs intended to improve the quality and cost-effectiveness of health care delivery and facilitate health improvements at the population level.^1,2 Exchanges of clinical data will be especially important for improving public health and obtaining information that will enable early detection of disease outbreaks and make biosurveillance a reality.³

Despite considerable promise, few examples of well-functioning and sustainable HIEs exist to serve as models. A 2007 survey revealed that of more than 100 regional health information organizations, 25% were defunct and only 15 were of at least modest size and exchanging data across a range of populations.⁴ At most, 12 regional health information organizations were presumed to be self-sustaining and no longer relying on grants to support their operations. In this environment, according to one leader in the realm of HIE, “We desperately need, efficiently and expeditiously, to learn what works and what doesn't.”^5(pw596)

Learning “what works and what doesn't” in relation to HIE is challenging not only because of the dearth of successful models but also because knowledge of the needs, expectations, and motivations of stakeholders (physicians, their practices, and hospitals) regarding HIE is lacking.⁶ One way to better understand these stakeholder characteristics is to investigate qualitatively the factors that members of the medical community setting up an HIE consider in selecting the organization and technical architecture of their system.

When stakeholders make these decisions, their individual and collective values emerge both implicitly and explicitly. Decisions about technical architecture can therefore be viewed, in economists' terms, as “revealed preferences.” The final decisions and the reasons behind them provide a helpful window into stakeholders' perspectives, bringing the complex dynamics of creating an HIE into sharper focus and allowing one to infer the factors that contribute to achieving sustainability. This knowledge may also inform the policy debate over the potential for success of the current market-oriented approach to HIE, a paradigm that assumes that the market incentives of health stakeholders will be sufficient to motivate them to engage in HIE.

To explore these issues, we conducted a case study of HIE development efforts in 3 Massachusetts communities participating in the Massachusetts eHealth Collaborative (http://www.maehc.org) pilot, each of which had representatives of their health care providers choose their respective HIE technical architectures. The Massachusetts eHealth Collaborative, formed in 2004, was funded by a grant from Blue Cross Blue Shield of Massachusetts to promote statewide adoption of electronic health records and HIEs.⁷ As a demonstration project, the collaborative employed a request-for-proposals process to select the 3 Massachusetts communities that would take part in the project. The plan was to supply every physician with an electronic health record system and each medical community with HIE capabilities so that patient health data could be electronically transferred among independent practices and aggregated for public health reporting and population health management.

The communities selected were Northern Berkshire (in northwestern Massachusetts), Greater Newburyport (60 miles north of Boston), and Greater Brockton (25 miles south of Boston). The collaborative has covered all financial costs of the electronic health records and HIE for several years.

The HIE products are currently being installed in the communities and customized by the vendor companies chosen to implement each community's HIE architecture, representing the culmination of an arduous process of selecting vendors and technical architectures that occurred mostly in the summer of 2006. During that time, collaborative staff and medical community steering committees worked together to select the architecture and vendor that best met the requirements of each medical community. We investigated how the following 8 factors influenced decision-making regarding the structure and technical architecture of the HIEs: security, cost, complexity of implementation, performance, ability to measure quality of care, strategic goals, level of trust in the medical community, and stakeholders' desire for independence.

METHODS

We conducted semistructured interviews with members of the collaborative and with health care leaders from the 3 individual communities and also reviewed relevant documentation. Informants, chosen as a result of their participation in the vendor selection process, included chief information officers of hospitals and large medical practices, other executive officers of community provider organizations, active steering committee participants, and the collaborative staff who guided the project.

Interview questions were drafted on the basis of preliminary interviews conducted with the collaborative leadership and were tailored to each informant on the basis of her or his role in the process. Some questions were asked of multiple respondents to ensure consistency between responses from the members of the collaborative and members of the medical community. Each interview was 1 hour in duration and was conducted over the telephone (or, in 2 cases, in person).

The interviews focused on identifying the factors deemed by health care providers to be most critical in the selection of the overall structure and technical architecture of the HIE. To provide context, we also asked about decision processes. To remain consistent with the goals of the collaborative, we focused our information gathering on the exchange of health information across provider institutions as opposed to how each practice exchanged information internally. We also reviewed documentation that was used and generated in the process of selecting vendors and the specific HIE architectures. We interviewed 14 key informants in the summer and fall of 2007, approximately 12 months after their technical architecture decisions were made and before the completion of their HIE implementation.

From the interview notes, we reconstructed the process timeline and identified themes. We then considered the implications of the stakeholders' values in the context of the broader HIE movement.

RESULTS

Our interviews and relevant documentation indicated that 3 general architectures were considered by each medical community. The process of selecting among the architectures began with establishing general guidelines and a general structure for the HIE common to all communities and driven by the collaborative. The process continued with each individual medical community making its own decisions about the specific features of the HIE and the vendor that would provide the system and services.

Technical Architectures

The 3 general architectural alternatives that were considered were fully centralized, peer to peer, and hybrid (Table 1). These alternatives described the physical storage location of clinical data and the way data are shared among the members of the HIE network. This choice has many implications, which include the balance between privacy and the ability to measure quality and perform public health surveillance. The architectural choice may also affect the likelihood of the HIE's eventual success.

TABLE 1.

Characteristics of Health Information Exchange Technical Architectures: Massachusetts eHealth Collaborative

Characteristic	Fully Centralized	Peer to Peer	Hybrid
Centralized repository	Yes	No	Yes
Electronic health records in physician's office	No	Yes	Yes
Servers in physician's office	No	Yes	No
Network queries required	No	Yes	No

In a fully centralized architecture, all clinical data are stored in a single central repository; no data are stored locally in physician offices. All clinical data stored in electronic health records would be shareable among community physicians. This architecture is essentially a fully integrated electronic health record system similar to that found in an organization such as the Veterans Health Administration.

By contrast, the peer-to-peer approach—often called a “federated” model—involves no centralized repository. Rather, clinical data are stored at the physician practice that generated the data. Data exchanges occur when an authorized health care provider sends a query for a particular patient and then receives responses from sites within the HIE network that have data on the patient. This architecture requires physician practices to host and maintain data repositories as well as servers that can access the repositories and respond to queries.

A hybrid architecture combines aspects of the fully centralized and peer-to-peer approaches. A centralized repository is used, but this repository is a copy of a portion of the data that are stored locally at each physician practice so that physicians can still manage their own electronic health records as they do with the peer-to-peer approach. Each individual electronic health record “pushes” designated elements of new patient data to the centralized repository, which can then be read by other community physicians. Unlike the peer-to-peer approach, networkwide queries are unnecessary in the hybrid approach and are replaced by direct access to the centralized repository from any site in the HIE network. Other architectures beyond these 3 options are possible but were not seriously considered by any of the project communities. (A description of relevant technical architectures can be found at http://toolkit.ehealthinitiative.org/technology/principlesaddendum4.mspx.)

Selection Process and Final Decisions

The collaborative organization guided much of the process by working with physician councils and privacy and security councils and by educating the medical community steering committees; however, each medical community ultimately made its own decision. Engaging the stakeholder groups resulted in broad medical community support for the process and decisions.

In selecting HIE vendor companies, the collaborative issued a request for proposals and assembled a task force to oversee the selection process. (The request for proposals is available at http://www.maehc.org. To promote interoperability, all electronic health record products supported by the collaborative are certified by the Certification Commission for Health Information Technology, and the HIE vendor is expected to comply with technical specifications from the Massachusetts Health Data Consortium's MA-SHARE [Simplifying Healthcare Among Regional Entities] initiative.) The task force decided that the medical community members would not be given any cost information, except for estimates of what they would have to pay after the collaborative pilot ended, to avoid bias and prevent proprietary vendor cost information from being circulated. Ten request-for-proposal respondents were narrowed down to 4 finalists with the assistance of a technology review committee.

Each medical community's experience with the final selection process was different. In addition, although each of the communities selected the same approach in the end, the reasons varied.

Northern Berkshire.

This small community in the Berkshire Mountains of western Massachusetts is home to roughly 15 physician practices. The Northern Berkshire health care providers settled on a vendor and technical architecture without extensive debate. They had agreed to adopt the same electronic health record vendor for every physician practice in the community, and thus it was sensible to have that same vendor provide the HIE product. The vendor had been selected with the assurance that it could implement the medical community's architecture of choice: the hybrid approach.

Greater Newburyport.

The Newburyport medical community's decision for technical architecture, a hybrid approach, was obvious and unanimous because the medical community's goal of measuring quality of care was best served by a centralized repository, which allows more efficient quality analysis than a decentralized approach. Selecting a vendor, however, proved more challenging. The decision was finally resolved when the members elected to choose a vendor selected by one of the other communities to minimize implementation challenges.

Greater Brockton.

Whereas the Northern Berkshire and Newburyport medical communities settled on their technical architecture early, Brockton—a medical community with a diffuse and competitive health care milieu formed out of 6 large provider institutions and many smaller physician offices—deliberated the architecture question extensively. Although it initially leaned strongly toward the peer-to-peer approach, the medical community eventually selected the same architecture as the other 2 communities: the hybrid approach.

Factors

With the exception of cost, all 8 of the factors that we planned to investigate were considered by the health care providers; cost was not a factor because the collaborative paid the HIE construction fees and because ongoing expenses were comparable.

Performance.

Performance concerns for the peer-to-peer approach were a high priority for all of the communities. Integrating HIE into clinical workflows was expected to be challenging even when performance levels were high. Providers worried that delays would exacerbate workflow challenges and impair usability.

Complexity.

The relatively high complexity of the peer-to-peer approach was expected to delay the implementation of the HIE, particularly in Brockton. That medical community's selected vendor emphasized the technical complexities and difficulties of implementing peer to peer, which would involve more than 50 servers (1 in each practice), and argued for the hybrid approach.

Security.

Security concerns were paramount in each medical community, but the steering committees in Northern Berkshire and Newburyport did not view any architecture as possessing inherently superior security. However, the Northern Berkshire steering committee members believed that by preventing a physician's electronic health record system from accepting queries and only “pushing” the data to another location, as is the case with the hybrid approach, some “gut-level” security would be created. This eliminated the peer-to-peer approach. Initially, the Brockton community believed that peer to peer would be more secure but then decided that securing the 50 servers required would be riskier.

Quality of care.

Health care providers in all 3 communities were very motivated to use the HIE for measuring quality of care and analysis, which, they reasoned, would be more easily accomplished through a hybrid approach than through the peer-to-peer approach, because only a centralized repository would yield sufficient efficiency. The Newburyport medical community was unique among the collaborative communities in that its physicians and hospital institute collective contracting and pay-for-performance quality programs under a single organization. Newburyport viewed the HIE as an opportunity to strengthen this organization's care improvement and negotiating capabilities.

The Brockton medical community, unlike the other communities, had not emphasized community-wide quality analysis but was interested in quality measurement for individual organizations. Still, the ability to conduct quality analyses across an individual organization favored a hybrid approach. Although none of the communities placed a specific priority on issues such as disease outbreak detection, all hoped that the ability to measure quality of care would allow the creation of more value-added programs such as referral management and patient matching with community specialists.

Strategic goals.

Health care providers made architecture decisions to account for the strategic goals of both the medical community as a whole and the health care providers individually. Each medical community had the goal of improving health care quality. In addition, Newburyport's medical community strategy included providing more patient-centered care through a patient portal—a Web site that allows patients access to their medical information—for which the centralized repository of the hybrid model was viewed as more conducive than the peer-to-peer model.

In Brockton, individual provider institutions initially argued for the peer-to-peer approach in support of their strategic goals in an interesting way: at least 1 of the larger provider organizations considered using the community HIE to integrate its own providers with each other, thereby helping the organization achieve its corporate goals of providing better continuity of care and reducing costs. Although the organization wanted to use the service of the HIE, it also wanted to avoid undue dependence on an external organization. The medical community perceived the peer-to-peer approach as involving less dependence on the community project than a hybrid approach; the reason was that if the provider institution were to separate from the community HIE, it could still leverage the peer-to-peer exchange mechanism for its own providers, forming a smaller, internal network. The hybrid approach would bind the institution to the community project.

The hybrid approach, however, also conferred strategic advantages on Brockton's health care providers, possibly lending assistance to both individual institutions (through quality measurement) and the medical community as a whole. The medical community could potentially benefit from the technical architecture's scalability. If health care providers in neighboring communities were willing to pay a fee to access Brockton's HIE, expanding to those communities could help achieve sustainability. The hybrid approach was decidedly more scalable than was the peer-to-peer model, which would have involved more complexity and decreased performance with each additional node.

In contrast to the Newburyport medical community's inclusion of a patient portal, a patient portal was not a part of Brockton's architectural decision or community strategy and was considered an added benefit. The Brockton medical community has yet to decide whether it will implement a community-wide patient portal.

Trust.

The level of trust among physician practices varied at the beginning of the pilot, but in each medical community trust seemed to increase as a result of participation in the HIE. Northern Berkshire and Newburyport health care providers had strong bonds of trust from the outset, and these bonds continued into the project. At the start of the pilot, Brockton's health care providers, despite having a history of competition, informally agreed not to use technology to compete inside the community for patients, forming the basis of more trusting relationships.

Because of these bonds of trust, all of the participating providers were willing to make their patients' data available for the HIE, and such data are considered by some health care providers across the country to be a competitive asset.⁸ Plans in every community to have in place clear policies and procedures for data access, including monitoring and sanctions, also increased trust and willingness on the part of health care providers to exchange data.

Desire for independence.

Despite growing trust among health care providers, a completely centralized approach was never seriously considered in any of the communities because physicians wanted to operate their electronic health records independent of a centralized organization. The Northern Berkshire medical community briefly considered a completely centralized approach but ultimately decided against it because it would have involved commingling patient data from different physician practices, which are separate legal entities. For that reason, the collaborative would have disallowed a centralized repository even if it was favored by a medical community.

Summary.

The Northern Berkshire medical community's selection was influenced primarily by performance and the ability to measure quality of care. In Newburyport, the ability to measure quality of care and the ability to create a working patient portal were the dominating factors. The Brockton medical community's deliberations involved almost all of the factors we investigated; in the end, the ability to measure quality of care and the potential for future sustainability led the medical community to favor the hybrid approach. Each medical community was found to have gained considerable trust as a result of taking part in the collaborative pilot.

DISCUSSION

Health care providers in all 3 of the communities involved in this pilot project selected the hybrid approach for their HIE technical architecture, but they differed somewhat in their reasoning. Choosing a technical architecture was much more than just optimizing technical variables; rather, it was deeply connected to the values of the key stakeholders in their particular circumstances.

In each pilot community, every willing health care provider received an electronic health record system, was invited to participate in the HIE, and had representation on the steering committee; thus, each individual medical community's technical architecture decision may reflect the values of the entire medical community. By contrast, many regional health information organizations begin as a small number of large institutions that design the infrastructure without involvement from local health care providers to which they may expand. Therefore, this study, although it may illuminate the values of medical communities embarking on HIEs, may not be representative of organic regional health information organizational efforts in other communities.

Two other studies have also used qualitative methods to evaluate specific HIEs. A case study of the Indiana HIE described a working exchange that consists of 5 health systems but only some of the office practices in the community.⁹ In contrast to our study, it described how to build an HIE organically without the inclusion of all community practices; it did not directly address the values of key stakeholders.

A study of the Santa Barbara County Care Data Exchange evaluated reasons for that effort's demise, claiming that the experience “illustrates the danger that in some communities, unfavorable short-term private value propositions for simple HIE services may delay more advanced HIE services with greater potential medium- and long-term private/societal payoff.”^10(pw578) By elucidating the perspectives of key stakeholders, our study advances this discussion of value propositions and helps to identify those that may be favorable in both the short and long term.

Policy Implications

The stakeholder's reasoning surrounding many of the factors we studied offers a view into their perspective and centers on 3 aspects: trust, strategic interests, and benefits from quality measurements. How effectively an HIE addresses these aspects may largely determine its ability to become established and achieve long-term success.

The experiences of the collaborative communities illustrate the importance of trust for HIE. Both overcoming security concerns and stakeholders' desire for independence were dependent on community trust. The trust that was created in Brockton—probably the most representative of the US health care system because of its large, diffuse, and competitive health care market—is an optimistic sign that competing providers may be able to collaborate.

Trust is necessary, but not sufficient, for an HIE. In a 2005 study of the factors that contribute to efficient and successful use of information technology in various industries, Bower recommended that the government “make policy decisions that turn [health care information technology] into a competitive weapon.”^11(p67) Transforming health care information technology into a competitive weapon while still retaining trust is a fundamental challenge for HIEs, especially when retaining trust involves renouncing the use of information technology as a competitive weapon.

Our study shows that trust and competition can coexist if the unit of competition is the medical community and if participation in an HIE—rather than control of patient data—forms the basis of competition. If used to attract new patients, for example, HIEs would provide a strategic advantage for the medical community as a whole. Our findings suggest that the size of a medical community will affect its propensity for acting in its own strategic interests, and a very large medical community may be especially challenging.

The success of an HIE may also depend on how effectively quality measurement is addressed. Quality measurements might motivate health care providers to remain engaged in an HIE by facilitating pay-for-performance programs, enhancing providers' reputations, or producing information to better match patients' needs with those of community health care providers. However, communities in which health care providers are less willing to have the quality of their care measured may view this capability as an obstacle rather than a motivator.⁸ Although traditional public health interests may not be high on the priority lists of a medical community, the technical architecture of an HIE will have a major impact on the ability to obtain key public health data, and public health providers should try to come to the table earlier than they have with most regional health information organizations.

The ultimate success of the market-oriented approach to HIEs is yet to be determined. Our study provides evidence that an HIE's chances of success can be maximized by policies and programs that foster trust, appeal to strategic interests, and provide quality measurement benefits. Other business models that involve payments from patients directly for use of personal health records or from drug or medical device companies that use the data for marketing have been proposed but have not yet been fully tested.^12,13 If the benefits of quality measurement are substantially less than the costs to physicians of participating in an HIE, the market-oriented approach may fail, and mandates or direct government funding may become necessary for widespread HIE to become a reality.

Limitations

Although we have identified some of the needs, motivations, and expectations of health care providers regarding HIEs, it would be premature to draw strong conclusions about “what works and what doesn't” because the collaborative project assessed here is not yet fully operational. The information was collected for a specific decision point before implementation and therefore does not capture changes in stakeholders' perspectives over time.

Another possible limitation of this study is that the methodology relied on stakeholder interviews instead of witnessing the selection process firsthand. Informants may have been reluctant to discuss conflicts and negative experiences. Also, patients were not included in the study. Finally, the results may not be representative of other communities, especially because the costs of electronic health records and the HIEs were borne by the collaborative and because the communities were volunteers rather than a random sample.

Conclusions

The experiences of the 3 collaborative pilot communities provide a revealing characterization of the perspectives of key HIE stakeholders. One overarching insight is that it appears unlikely that the health care providers in the pilot communities would have moved to develop an HIE rapidly without assistance from the collaborative. Important aspects of the stakeholders' perspectives included community-wide trust, strategic interests of individual health care providers and the medical community as a whole, and benefits derived from measuring quality of care.

Compelling reasons led all 3 communities to select a hybrid approach; whether other communities would ultimately arrive at the same conclusion is a testable hypothesis. In any case, it remains to be seen whether this effort or any other HIE in the United States can provide direct benefits from quality measurements or other activities. Without such benefits, the sustainability of HIEs and the accompanying increases in clinical data available for public health practice and research may remain precarious.