Health Information Exchange Among US Hospitals
Published Online: November 07, 2011
Julia Adler-Milstein, PhD; Catherine M. DesRoches, DrPH; and Ashish K. Jha, MD, MPH
The Health Information Technology for Economic and Clinical Health (HITECH) Act promotes the adoption of health information technology at an unprecedented level.1 The bulk of funding is devoted to helping physicians and hospitals adopt and meaningfully use electronic health records (EHRs), structured as incentives of up to $44,000 per provider for demonstrating meaningful use and then converting to penalties in 2015 if meaningful use is not achieved.1 Substantial funding was also directed to support electronic health information exchange (HIE) at state and regional levels.2,3 Health information exchange enables patients’ health information to follow them between delivery settings in order to support care coordination and avoid duplication of services. There is broad consensus that such connectivity is critical to improving care and reducing healthcare costs. Nationwide health information exchange is a key driver of the efficiency gains promoted by health information technology (IT) advocates, including the widely publicized $78 billion in annual savings.4 Therefore, the success of HITECH hinges in part on whether it can jump-start historically low levels of HIE.5
The HITECH Act does not specify how HIE should be achieved, and as a result, an array of options are emerging. They range from the federally initiated Direct Project, a set of standards that allow senders to push health information securely to known receivers, to marketbased HIE solutions that can be used to create an exchange network. Over the past decade, health information organizations (HIOs) have served as the primary mechanism by which the United States pursued clinical data exchange.6 Health information organizations offer a particular approach to achieving HIE by bringing together stakeholders with clinical data (eg, physician practices, laboratories, hospitals) in a given geographic region and setting up the infrastructure for HIE. Since many HIOs are well established and have substantial experience supporting HIE, they appear likely to serve as a foundational approach under HITECH. For examples of HIO-based plans, see Colorado, Indiana, or New York at the State Health Information Exchange Program Web site.7
Health information organizations face challenges on multiple fronts, including a lack of funding, concerns about privacy and security, legal and regulatory challenges, technical challenges, and stakeholder concerns about competitiveness.8 Health information organizations have also had to accommodate providers without EHRs, often by making data available on a portal that can disrupt work flow. The policy response has focused on securing funding—most recently in the form of large grants to states to increase HIE—as well as increasing EHR adoption and setting rules for data privacy, security, and technical standards.2 Providers’ concerns about the competitive implications of HIE, which have been documented in several case studies,9-12 have received far less attention. In case studies, hospitals report that patient data tie both patients and providers to their institution, conferring a competitive advantage that would be lost by participating in an HIO.9,10 Yet we know little about how broadly hospitals are engaged in HIE and whether certain key factors, especially those stemming from competitive concerns, are related to the decision to participate.
Therefore, we used nationally representative data to answer 3 questions. First, what proportion of US hospitals are exchanging clinical data with unaffiliated providers through an HIO (“engaging in HIE”)? Because the data were collected in the months after passage of HITECH, it offers a baseline against which national HIE progress can be measured. We focused on HIE taking place through HIOs because HIOs were the predominant strategy available when our data were collected and they are a key part of many states’ plans for expanding HIE.7 Second, are certain key hospital characteristics associated with the decision to engage in HIE? Specifically, we hypothesized that for-profit hospitals and hospitals with a small market share may be more concerned about loss of market share and therefore may be less likely to engage in HIE. Teaching hospitals, which routinely serve as referral hospitals, may be more likely to engage in HIE. Third, do features of healthcare markets where hospitals function affect the hospitals’ likelihood of engaging in HIE? Specifically, we postulated that hospitals in more concentrated markets or those with less fragmentation of hospital care would be more likely to engage in HIE.
We used national data from the IT supplement to the annual American Hospital Association (AHA) survey, which was administered during the spring and summer of 2009 to all acute-care hospitals, more than 95% of which were AHA member hospitals. 13 A total of 3725 member hospitals responded to the IT supplement survey (a response rate of 69%), and we limited our analytic sample to the 3101 acutecare, nonfederal hospitals located in the 50 states and the District of Columbia. We defined a marketas a hospital referral region (HRR), a designation developed by the Dartmouth Atlas to define healthcare delivery markets based on Medicare beneficiary travel patterns for tertiary hospital care.14
The AHA IT supplement included 2 questions related to HIE: (1) whether the hospital participates in a regional HIO (or a regional HIO-like effort) to share electronic patient-level clinical data and (2) whether the hospital electronically exchanges each of 5 types of data (eg, laboratory reports, clinical care records) with hospitals or ambulatory providers that are part of a different system. (Original survey questions are included in eAppendix 1, available at www.ajmc.com.) We merged the IT supplement data with results from the annual AHA survey to develop measures at the market level as well as capture additional hospital characteristics. The Area Resource File, Medicare Provider and Analysis Review File, and Dartmouth Atlas were used for additional market level measures.
Outcome Measures. We determined that a hospital engaged in HIE based on how it responded to the 2 HIE-related questions. Hospitals that (1) reported participating and actively exchanging data through a regional HIO (question 1) and (2) reported that they exchanged at least 1 type of clinical data with either hospitals or ambulatory providers that were part of a different system (question 2) were classified as participating in an HIO. Our approach helped ensure that the definition of HIO participation was consistent with the commonly held criteria that (1) HIE is actively occurring, (2) clinical data are exchanged, and (3) exchange takes place between organizations that are not part of the same system, which will be required under meaningful use.3,8 Given that most HIOs are relatively new,5 we did not require robust exchange (ie, multiple types of clinical data), though this is ultimately what is likely required to realize meaningful efficiency gains from HIE. We identified markets with HIO presence as those in which at least 1 hospital in the HRR was classified as participating. We sought to validate hospitals’ responses about participating in an HIO and exchanging data by examining a recently completed national survey of HIOs.5 In this survey, HIOs were asked to identify the hospital service area(s) in which they functioned and actively facilitated clinical data exchange. We compared these areas with the areas in which hospitals reported participating in an HIO on the AHA IT supplement. We found substantial overlap, giving us confidence in the validity of hospitals’ responses about HIO participation.
Hospital Characteristics. We examined 8 hospital characteristics, 3 of which were chosen a priori based on our hypotheses about which hospitals might or might not participate in an HIO: ownership (for profit, nonprofit private, or public), bed share in the market, and teaching status. The remaining characteristics were those that we thought might be directly related to participating or might confound the relationship between the 3 primary variables of interest (ownership, bed share, and teaching status) and participation: size, proportion of Medicaid admissions, whether the hospital was affiliated with a system, and whether the hospital had significant technological capability (a coronary care unit) and IT resources (an EHR).
Market Characteristics. We also examined a series of market-level characteristics. We created 2 measures to capture the competitiveness of the market. We measured market concentration using the Herfindahl-Hirschman Index. We measured market fragmentation as the proportion of discharged patients subsequently readmitted to a nonaffiliated hospital within the same market. Additional market-level measures included mean annual Medicare inpatient expenditures, mean proportion of hospitalizations from Medicare patients, urban/rural location, geographic region, managed care penetration, population density, percentage of the population that was uninsured, and per capita income. Each of these variables was chosen because we believed that it might be directly related to promoting HIE or might confound the relationship between the market variables of interest (market concentration and market fragmentation) and HIE. More details on why each measure was chosen and how it was defined are available in eAppendix 1.
Analysis. We first calculated an adjusted overall participation rate among all respondents in our sample. We used weights to adjust for potential nonresponse bias after finding modest but significant differences when comparing respondents and nonrespondents to the survey (eAppendix 2, available at www.ajmc.com). We next examined whether our hypothesized hospital characteristics were related to a hospital’s decision to participate in an HIO. We started with bivariate relationships and then built multivariable logistic regression models with hospital-level sampling weights and robust standard errors adjusted for clustering at the market level. Our analyses compared hospitals that participated in HIOs with those that did not within HRRs that had an HIO. Looking within the 165 HRRs (54%) with an HIO helped to ensure that hospitals had an opportunity to participate. We next examined whether our hypothesized market characteristics were related to hospital participation by adding marketlevel measures to the multivariable logistic regression model.
We ran a set of sensitivity analyses to assess the robustness of our findings. We assessed whether our results held when we used bed share at the system level within the market. In addition, we substituted our measure of the hospital Herfindahl-Hirschman Index for the hospital-system Herfindahl-Hirschman Index in our market characteristics model. These values could serve as better measures of competitive position and concentration in markets in which systems behave cooperatively. To assess whether unobserved market-level characteristics affected hospital-level results, we ran our hospital characteristics model using market conditional fixed effects. Finally, we used an alternative specification of the dependent variable that relied only on the AHA IT supplement question about hospital participation in a regional HIO and not the additional question about types of data exchanged. This approach likely included hospitals participating in regional HIOs focused on nonclinical data exchange or regional HIOs supporting data exchange between hospitals and other providers (eg, public health departments).
Characteristics of Hospitals in the AHA IT Survey
Hospitals in our sample were primarily small and medium in size (46% and 42%, respectively; Table 1). They were most often nonprofit hospitals (64%) and just over a quarter were teaching hospitals. Approximately half of the respondents were members of a hospital system and the majority were located in an urban area (73%). While more than a third of respondents had a cardiac intensive care unit, only 16% had at least a basic EHR.
Hospitals That Participate in HIOs We found that 10.7% of hospitals participated in a regional HIO. In bivariate analysis, for-profit hospitals were less likely than their nonprofit counterparts to participate (7% vs 18%, P <.001; Table 2). Hospitals with a more dominant market share were more likely to participate: while only 9% of hospitals in the lowest quartile of market share participated, 27% of hospitals in the highest quartile did so (P <.001 across the 4 quartiles). Finally, we found that major teaching hospitals were more likely to participate in a regional HIO than minor teaching hospitals or nonteaching institutions. Multivariate results were similar: for-profit hospitals were far less likely to participate in a regional HIO (odds ratio [OR] 0.35; P = .012), as were hospitals with a lower market share (Table 3). However, major teaching hospitals were no more likely to participate than minor teaching hospitals or nonteaching hospitals. We found 1 other characteristic that was associated with participation: whether the hospital had at least a basic EHR. Hospitals with an EHR were significantly more likely to participate (OR 1.76; P = .002).
Market Characteristics Associated With HIO Participation
Hospitals in the most concentrated markets (those in the top quartile of the Herfindahl-Hirschman Index) had 2.88 greater odds of participating (95% confidence interval [CI] 1.48-5.59) compared with the least concentrated markets (P = .015 across the 4 quartiles; Table 4). Hospitals in markets with higher Medicare spending were less likely to participate: those in the top quartile of spending had 66% lower odds of participating (95% CI 0.23-0.83) compared with the lowest spending markets (P = .018). We did not find a significant difference in hospital participation based on the degree of market fragmentation.
Our multivariate results did not meaningfully differ when we examined bed share at the system level in the model with hospital-level characteristics (eAppendix 4, available at www.ajmc.com). Our market-level results were also very similar when we examined the systemlevel Herfindahl-Hirschman Index: hospitals were still more likely to participate in more concentrated markets (eAppendix 5, available at www.ajmc.com). When we ran our hospitallevel multivariate model using conditional fixed effects at the market level, we generally found results consistent with our model that included specific market-level characteristics. The 1 exception was hospital bed share, which with the addition of market fixed effects was no longer associated with likelihood of participating in HIE (eAppendix 6, available at www.ajmc.com). Finally, our results were largely unchanged when we only required hospitals to report participating in a regional HIO, relaxing the requirement to exchange specific types of data (eAppendix 7, available at www.ajmc.com).
Our nation has embarked on an ambitious strategy to promote the adoption of health IT and electronic exchange of clinical data. Using nationally representative data, we found that only 1 in 9 US hospitals was sharing at least 1 type of clinical data through an HIO in 2009. Our study also suggests that key market forces are likely shaping hospitals’ decisions to participate in an HIO. Hospitals with a dominant market share may perceive participation as an opportunity, while for-profit hospitals and those with small market share have mostly opted out. Hospitals in more concentrated markets were more likely to participate, while hospitals in markets with high Medicare spending were less likely to participate. In addition, just over half the markets had an HIO, highlighting substantial gaps in HIE coverage at the start of HITECH.
We expected that competition might be an important force dampening the enthusiasm for regional HIO participation, but we were surprised at how prominently it appeared among both hospital and market characteristics. For-profit institutions as well as those with smaller market share had dramatically lower participation rates in regional HIOs. For-profit hospitals may be more sensitive to concerns about potential loss of market share than nonprofit hospitals, and those with a smaller market share may be less able to tolerate further patient erosion by participating in an effort that facilitates patients seeing other providers. Within more concentrated markets, hospital participation rates are higher, providing further support for the role of competition. These results suggest that concerns about competition and loss of market share may need to be addressed in order for HIE efforts to achieve broad penetration.
One interesting finding was that hospitals in higher Medicare spending regions were less likely to participate in clinical data exchange through a regional HIO. It is possible that greater exchange of clinical data has led to a reduction in Medicare spending, though we think this is unlikely because most of these HIOs are relatively new and the breadth of clinical data exchanged is relatively narrow. An alternative explanation is that low-spending regions have the kind of medical culture that fosters greater collaboration and sharing, which also drives willingness to share clinical data exchange in these communities. Unfortunately, our cross-sectional analyses do not allow us to establish causal relationships. As a result, the mechanisms underlying our results are largely speculative.
Although we are unaware of any large-scale empirical work focused on competition and providers’ choice to participate in HIE, there has been important work in this area. In case studies, Grossman and colleagues found that patients and their data were perceived to confer a competitive advantage that would be lost by participating in a regional HIO.10 This was particularly salient for hospitals who viewed clinical data as “a key strategic asset, tying physicians and patients to their\ organization.”10 Other case studies have found similar results.9,11,12 Our study extends these findings by empirically examining them in a national sample of hospitals.
Our findings have important policy implications. Given that the first stage of meaningful use only requires the capability to exchange clinical data, it is unlikely that the requirements in and of themselves will overcome issues of competitiveness that might be holding some providers back. States, the primary recipients of federal support to expand HIE, may want to consider how to address such barriers. Beyond existing efforts to promote EHR adoption, they could focus on facilitating exchange of data with the greatest financial benefits for providers (mainly laboratory and radiology tests). Such an effort would likely entice more providers to join in, although such an approach is not without costs: by narrowing its focus, these HIOs would simultaneously reduce the societal value of HIE. Whether HIE efforts would then be able to expand to include a more comprehensive set of data is unclear.
In future meaningful use regulations, the Centers for Medicare & Medicaid Services is signaling that it will require hospitals to exchange a robust set of clinical data, although it will likely leave the channels through which those data are shared up to individual providers. While such an approach might spur some hospitals to join state-sponsored HIOs and share data more widely, others might strategically share data with only a subset of providers in the market (eg, referring physicians, nursing homes), creating potentially greater market fragmentation with islands of HIE that effectively lock patients into a group of providers. Policy makers will likely need strong incentives, financial or otherwise, to overcome these market pressures that many hospitals feel. One approach may be stronger penalties than those currently tied to meaningful use for not engaging in broad exchange, which could act as an opposing competitive pressure by directly threatening hospitals’ financial viability. These stronger tactics may only be needed in the short term, however, because once robust HIE reaches a critical threshold, hospitals’ competitive position may be more harmed by opting out. The key challenge for policy makers is getting the country to this tipping point.
There are important limitations to our work. First, the AHA IT supplement is self reported and we were unable toverify accuracy of responses. However, we validated the data when we could by comparing markets with HIO presence as identified by the AHA IT supplement with an independent source of data on where HIOs operate.5 Second, although the response rate to the hospital IT supplement was high (69%), and we statistically adjusted for nonresponse, some degree of nonresponse bias likely remains. Our findings may also be affected by omitted variable bias. While we attempted to includea comprehensive set of variables, there are some that we were unable to measure. For example, we lacked a direct measure of hospitals’ financial health as well as the overall financial health of a market, which may influence hospital participation, although we included a set of proxy measures as covariates. The proxy measures did not provide consistent evidence of a relationship between participation and financial health of either the hospital itself or the market in which it operates.
More broadly, this is exploratory work in a rapidly changing, complex field. While our findings suggest that competition plays a role, a multitude of other factors that affect HIO participation need to be examined, and we were not able to determine the extent to which competition plays a role relative to other potential barriers. With limited empirical evidence on the benefits of HIOs, it may be that these organizations are still early in the innovation adoption cycle, and the majority of hospitals are waiting to see how early adopters fare. In addition, because we could not examine specific HIOs and their characteristics, we were unable to account for governance approach, technical architecture, types of stakeholders involved, and other considerations that undoubtedly influence hospital participation decisions.
Finally, we were not able to assess causality and suggest that future work should rely on longitudinal data in order to more directly assess the direction of the relationship between hospital and market characteristics and HIE. However, hospital characteristics are quite stable, and, given that most HIOs are relatively new,15 we think it is unlikely that they have had the opportunity to shift hospital market share or market competitiveness.
We report the first set of national data on the number of US hospitals participating in HIE through HIOs at the start of HITECH and found that approximately 1 in 9 hospitals were engaged. In this national sample, we also found that a hospital’s ownership status and its market share were associated with its decision to participate in HIE. At both the market and hospital level, our findings suggest that concerns about the competitive implications of HIE may be playing a role. It is unclear whether the potency of the current financial and regulatory incentives will be adequate to coax US hospitals to share their clinical data with other providers, many of whom may be perceived as competitors.
Author Affiliations: From University of Michigan School of Information and School of Public Health (JA-M), Ann Arbor, MI; Massachusetts General Hospital’s Mongan Institute for Health Policy (CMD), Boston, MA; Harvard School of Public Health (AKJ), Boston, MA.
Funding Source: None.
Author Disclosures: Dr Jha reports consultancies or paid advisory boards from Humedica Scientific Board and has also received grants from the Office of the National Coordinator at the Department of Health and Human Services, which has funded similar work. The other authors (JA-M, CMD) report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.
Authorship Information: Concept and design (JA-M, AKJ); acquisition of data (CMD); analysis and interpretation of data (JA-M, CMD, AKJ); drafting of the manuscript (JA-M, CMD, AKJ); critical revision of the manuscript for important intellectual content (JA-M, CMD, AKJ); statistical analysis (JA-M); obtaining funding (CMD); and supervision (AKJ).
Address correspondence to: Ashish K. Jha, MD, MPH, Harvard Medical School, Boston Veterans Affairs Hospital, Health Policy and Management, Harvard School of Public Health, 677 Huntington Ave, Boston, MA 02115. E-mail: email@example.com.
1. Blumenthal D. Stimulating the adoption of health information technology. N Engl J Med. 2009;360(15):1477-1479.
2. Blumenthal D. Launching HITECH. N Engl J Med. 2010;362(5): 382-385.
3. Blumenthal D, Tavenner M. The “meaningful use” regulation for electronic health records. N Engl J Med. 2010;363(6):501-504.
4. Walker J, Pan E, Johnston D, Adler-Milstein J, Bates DW, Middleton B. The value of health care information exchange and interoperability. Health Aff (Millwood). 2005;Suppl Web Exclusives: W5-10-W5-18.
5. Adler-Milstein J, Bates DW, Jha AK. A survey of health information exchange organizations in the United States: implications for meaningful use. Ann Intern Med. 2011;154(10):666-671.
6. Kuperman GJ. Health-information exchange: why are we doing it, and what are we doing? J Am Med Inform Assoc. 2011;18(5): 678-682.
7. State Health Information Exchange Program. Approved state plans. http://statehieresources.org/state-plans/. Accessed October 25, 2009.
8. Adler-Milstein J, Bates DW, Jha AK. U.S. regional health information organizations: progress and challenges. Health Aff (Millwood). 2009;28(2):483-492.
9. Grossman JM, Bodenheimer TS, McKenzie K. Hospital-physician portals: the role of competition in driving clinical data exchange. Health Aff (Millwood). 2006;25(6):1629-1636.
10. Grossman JM, Kushner KL, November EA. Creating Sustainable Local Health Information Exchanges: Can Barriers to Stakeholder Participation Be Overcome? Center for Studying Health System Change. Research Brief No. 2. http://www.hschange. org/CONTENT/970/970.pdf. Published February 2008. Accessed October 19, 2008.
11. Rudin RS, Simon SR, Volk LA, Tripathi M, Bates D. Understanding the decisions and values of stakeholders in health information exchanges: experiences from Massachusetts. Am J Public Health. 2009;99(5):950-955.
12. Fontaine P, Zink T, Boyle RG, Kralewski J. Health information exchange: participation by Minnesota primary care practices. Arch Intern Med. 2010;170(7):622-629.
13. Jha AK, DesRoches CM, Kralovec PD, Joshi MS. A progress report on electronic health records in U.S. hospitals. Health Aff (Millwood). 2010;29(10):1951-1957.
14. Dartmouth Atlas. Appendix on the Geography of Health Care in the United States. http://www.dartmouthatlas.org/downloads/ methods/geogappdx.pdf. Accessed October 25, 2009.
15. Adler-Milstein J, McAfee AP, Bates DW, Jha AK. The state of regional health information organizations: current activities and financing. Health Aff (Millwood). 2008;27(1):w60-w69.