Hospital’s adoption of multiple methods of obtaining outside information and use of that information

Abstract

Objective

Hospitals have multiple methods available to engage in health information exchange (HIE); however, it is not well understood whether these methods are complements or substitutes. We sought to characterize patterns of adoption of HIE methods and examine the association between these methods and increased availability and use of patient information.

Materials and Methods

Cross-sectional analysis of 3208 nonfederal acute care hospitals in the 2019 American Hospital Association Information Technology Supplement.

Results

The median hospital obtained outside information through 4 methods. Hospitals that obtained data through a regional HIE organization were 2.2 times more likely to also obtain data via Direct using a health information service provider (HISP) than hospitals that did not (P < .001). Hospitals in a single electronic health record (EHR) vendor network were no more or less likely to participate in a HISP or HIE. Six of 7 methods were associated with greater information availability. Only 4 of 7 methods (portals, interfaces, single vendor networks and multi-vendor networks but not access to outside EHR, regional exchange or Direct using a HISP) were associated with more frequent use of information, and single vendor networks were most strongly associated with more frequent use (odds ratio = 4.7, P < .001).

Discussion

Adoption of some methods was correlated, indicating complementary use. Few methods were negatively correlated, indicating limited competition. Although information availability was common, low correlation with use indicated that challenges related to integration may be slowing use of information.

Conclusion

Complementarities between methods, and the role of integration in supporting information use, indicate the potential value of efforts aimed at ensuring exchange methods work well together, such as the Trusted Exchange Framework and Common Agreement.

INTRODUCTION

Healthcare organizations can choose from several methods to obtain interoperable health information from sources outside of their organization.^1,2 However, each method does not generally facilitate exchange with organizations using other methods or meet all use cases. In consequence, organizations have adopted multiple methods to obtain and share information from outside organizations,² with the median hospital having adopted 4 methods in 2017.³

A wide variety of methods for obtaining information are available, so that health systems can choose among methods that match their needs. Some available methods, such as an interface connecting 2 health systems, reflect one-to-one exchange among trusted partners. Other methods involve intermediaries that facilitate exchange enabling many-to-many connections.⁴ Among these, regional health information exchanges (HIEs) and Direct Messaging via health information service providers (HISPs) were both strongly supported by the 2009 HITECH Act through the State HIE program and EHR Incentive Program.^5,6 Somewhat more recently, EHR vendor-based exchange, most prominently Epic Systems’ Care Everywhere tool, and national networks, including the CommonWell Alliance, have grown in prominence and use.⁷

Although most hospitals use multiple methods of HIE, there is little empirical evidence on how these methods inter-relate and how their adoption independently relate to greater information availability and use by providers.³ Market dynamics and the use cases that each method best fulfills are likely to influence which methods are adopted together—serving as complements—and which are alternatives or substitutes.⁸ Although a variety of methods exists that in theory could be used together, some commonly used methods represent different paradigms or approaches to exchange that warrant further examination.

For instance, Direct Messaging and regional HIEs may be adopted together and serve as complements to each other for a number of reasons. HIEs and Direct Messaging may address different use cases, both represent “open” approaches (nonenterprise) that enable exchange between varied health care organizations that choose to participate, and HIEs may serve as HISPs to facilitate Direct Messaging.^9–11

In contrast to these more open forms of exchange, one-to-one exchange methods, such as interfaces and portals, and vendor-based exchange represent relatively closed approaches wherein organizations either choose to exchange with one another for strategic reasons or are connected by a choice of common EHR developer. Because these approaches represent a shared paradigm of exchange and often rely on the capabilities of EHR vendor to enable exchange, they may represent a separate, alternative approach to HIE. These “closed” approaches may serve as substitutes or alternatives to those open approaches.¹² In fact, there is evidence to suggest that HIEs perceive EHR developers as competitors.¹³

Examining relationships between methods and their independent value could be useful to health system leaders, who select among the diverse approaches to engaging in HIE and may lack information on which approach will be most impactful in facilitating exchange.^11,14 Examining the use of multiple methods and their impact on information availability and use also has implications for health IT policy and an important component of the 21st Century Cures Act: the Trusted Exchange Framework and Common Agreement (TEFCA). TEFCA is aimed at simplifying exchange by linking networks together such that participation in one network enables connections to other participating networks.¹⁵ Information on the relationship between methods of exchange could inform implementation of TEFCA among hospitals. If network methods are often adopted as substitutes, TEFCA may be most impactful by connecting hospitals on different networks. If instead network methods are adopted as complements, TEFCA may offer benefit to hospitals by simplifying the use of multiple methods. In either scenario, these data serve as a useful baseline to understand how the field evolves as TEFCA facilitates exchange across networks that participate in it. And because TEFCA does not directly address all methods of exchange, it may further shape the perceived need and value of the methods it includes and excludes.

Given widespread adoption of multiple methods of HIE and the upcoming changes that TEFCA might create, we sought to answer 3 questions about how multiple methods of HIE are adopted and used. First, how commonly are different methods of obtaining health information adopted, and how often are they used when adopted? Second, how does adoption of an individual method of obtaining health information relate to adoption of other methods—in other words, are there sets of complements or substitutes that could inform health system leaders’ and policymakers’ decisions, such as those suggested above? Third, how does the adoption of each method independently relate to the likelihood that health information is routinely available to providers, that information is integrated into the EHR, and that the information is often used?

METHODS

Our data include all nonfederal acute care hospitals that responded to the 2019 American Hospital Association (AHA) Information Technology Supplement, the most recent data available when the study began because no survey was fielded in 2020.¹⁶ The IT Supplement was completed by hospital chief information officers (or their delegates) and includes a variety of data on how hospitals adopt and use information technology, including the methods used to receive and query information. The overall response rate for the IT survey in 2019 was 54%. Additional hospital characteristics were drawn from the 2019 AHA Annual survey and Medicare cost report.^16,17

Methods of obtaining information

We examined the methods used by hospitals to electronically query and receive information. The AHA IT supplement survey separately asks about 6 electronic methods for querying for patient health information and 7 electronic methods for receiving information. The 6 methods for querying are (1) “Provider portals that allow you to view records in another organization's EHR system,” (2) “Interface connection between EHR systems (eg, HL7 interface),” (3) “Access to other organizations' EHR system using login credentials,” (4) “Regional, state, or local health information exchange organization (HIE/HIO). Not local, proprietary, enterprise network,” (5) “EHR vendor-based network that enables record location within the network (eg, Care Everywhere),” and (6) “EHR connection to national networks that enable record location across EHRs in different networks (eg, CommonWell, e-health exchange, Carequality).” The same 6 methods are listed as methods to receive information, along with a seventh method, (7) “HISPs that enable messaging via DIRECT protocol,” which is excluded from the prior list because it does not support querying for information. We characterized the first 3 methods as facilitating “one-to-one” exchange and the latter 4 as facilitating “many-to-many” exchange because of the presence of an intermediary facilitating network connectivity.

To measure the use of each method for querying, the survey asks, “How often are the following electronic methods used to search for (eg, query or auto-query) and view patient health information from sources outside your organization or hospital system?” and response choices are Often, Sometimes, Rarely, Never and Don’t know. To measure use of each method to receive information, the survey asks, “When a patient transitions from another care setting outside your organization or hospital system, how often does your hospital use the following methods to RECEIVE a summary of care record?” and lists the same set of response options.

In our analysis, we counted each method as adopted if it was at least rarely used to receive or query for information at the hospital. We also assessed the frequency that respondents indicated each method was often used for either querying or receiving information among those that used it at least rarely.

Perceived information availability and use

To measure the relationship between adoption of each method of obtaining information and 3 factors related to information use, we used 3 survey measures reflecting hospital leaders’ perceptions of the availability and use of information. The first measure reflected whether information is available to providers and asks, “Do providers at your hospital routinely have necessary clinical information available electronically (not e-Fax) from outside providers or sources when treating a patient that was seen by another healthcare provider/setting?” The response options were “Yes,” “No,” and “Do not know,” We created a dichotomous variable by comparing “Yes” to “Do not know” and “No”.

The second measure reflected whether information was integrated into the EHR, a key factor in whether information is used. The associated survey question asked, “Does your EHR integrate the information contained in summary of care records received electronically (not eFax) without the need for manual entry?” The response options were “Yes, routinely,” “Yes, but not routinely,” “No,” “Do not know,” and “NA.” We created a dichotomous variable by comparing “Yes, routinely” to all other response categories.

The third measure reflected whether information was not just available but actually used by providers. The associated survey question asked, “How frequently do providers at your hospital use patient health information received electronically (not e-Fax) from outside providers or sources when treating a patient?” The response options were “Often,” “Sometimes,” “Rarely,” “Never,” and “Do not know.” Respondents who indicated “Do not know” were counted as “Never.” We retained the ordinal form of the measure for analysis to represent wide variation across response categories.

Hospital characteristics

We included a number of hospital characteristics that might be associated with adoption and use of specific methods of information exchange as well as the availability and use of information. We included measures of hospital capabilities drawn from the AHA survey including hospital size, teaching status, multi-hospital system membership, urban/rural location, ownership status, and the presence of a cardiac ICU. These characteristics have been previously used to control for hospital size and scale in a variety of studies on interoperability and exchange among hospitals.^1,2,18 Each of these is likely to be related to hospital’s IT strategy and information needs. We also included hospital operating margins drawn from the Medicare cost report, which reflects the financial wellbeing of the hospital and may relate to their ability to support a robust IT infrastructure.

Analytic plan

We first summarized the proportion of hospitals that had adopted each method of obtaining information and the proportion of hospitals that often used those methods when adopted.

Next, to measure the relationship between methods, we created Poisson regression models that estimated the likelihood that a hospital had each method and included as regressors dichotomous variables indicating whether each other method was adopted by the hospital as well as hospital characteristics. Poisson models with binary outcomes produce estimates of relative risk ratios, rather than odds ratios and can therefore more easily facilitate interpretation. Positive coefficients in these models would indicate that methods are adopted as complements or otherwise facilitate connectivity with one another. Negative coefficients would indicate that they are adopted as substitutes in that hospitals with a given method were less likely to have the other.

As an alternative approach to summarizing these relationships, we used exploratory factor analysis (EFA) to identify methods of obtaining information that were clustered together and so loaded on the same factor. We used Promax oblique rotation and report both 1 and 2 factor solutions. Methods that load on the same factor were more commonly adopted together; however, EFA is limited because it does not easily accommodate the identification of relationships controlling for potential sources of bias. This is a challenge because hospital characteristics are likely related to the patterns of electronic methods adopted by hospitals to obtain patient health information. Nevertheless, results of EFA can reinforce insights from the regression-based approach.

Finally, we assessed the association between adoption of each method and the likelihood that information was available to providers, integrated into the EHR, and used by providers. To estimate the likelihood that information was available to providers and was integrated into the EHR, we created logistic regression models that included dichotomous variables for whether each method was adopted by the hospital as well as hospital characteristics. To estimate the frequency of use, we used an ordinal logistic regression because the outcome was not dichotomous and included the same set of regressors as above.

All analyses were weighted using nonresponse weights based on hospital size, system membership and location, and conducted in Stata 15.

RESULTS

Prevalence of adoption and use

Our sample included 3208 non-federal acute care non-federal acute care hospitals. On average, hospitals had adopted 4.3 methods of obtaining information and used 2.0 methods often. Provider portals to other EHR systems and regional/state/local HIEs were the most commonly adopted method of obtaining information, adopted by 66.8% (95% CI: 65.1%–68.6%) and 65.2% (95% CI: 63.4%–67.0%) of hospitals, respectively (Figure 1). However, adoption across methods was relatively consistent with no method adopted by fewer than 50% of hospitals.

Figure 1.

Frequency of adoption and use of methods of obtaining outside information. N=3208 non-federal acute care hospitals in the US States and DC. Adoption defined as at least rarely using a method. Methods often used defined among hospitals that reported at least rarely used the method. Sample weights adjusted for response bias based on hospital size, system membership, and location.

Single EHR vendor networks and national networks were the most frequently used methods by those that had adopted each, with 63.8% (95% CI: 61.4%–66.2%) and 55.8% (95% CI: 53.4%–58.2%) of adopters reporting often using these methods, respectively. Though most frequently adopted, provider portals were only reported as often used by 34.6% (95% CI: 32.5%–36.7%) of respondents. Similarly, though often adopted, regional HIEs were only reported as often used by 47.0% (95% CI: 44.9%–49.2%) of respondents, least among the 4 methods of many-to-many exchange.

Relationship between methods for obtaining information

We observed substantial positive associations between several HIE methods in multivariate analysis.

Predictors of direct messaging via a HISP

First, hospitals that participated in regional HIEs were much more likely to also obtain information through Direct Messaging via a HISP (relative risk ratio = 2.19; 95% CI: 1.94–2.48; Figure 2); hospitals participating in national networks were modestly more likely to use Direct via a HISP (RR = 1.34; 95% CI: 1.24–1.45) and hospitals with single vendor networks were only slightly more likely to use Direct via a HISP (RR = 1.13; 95% CI: 1.05–1.22).

Figure 2.

Association between adoption of each intermediary HIE method of obtaining information and adoption of other methods. Note: Relative risk ratio derived from Poisson regression model including the 6 other methods of exchange, hospital size, multi-hospital system status, urban/rural location, teaching status ownership, and presence of cardiac ICU and operating margin. Sample weights adjusted for response bias based on hospital size, system membership, and location. HIE: health information exchange.

Predictors of regional HIE participation

Similarly, hospitals that obtained information through Direct Messaging via a HISP were much more likely to also obtain information through a regional HIE (RR = 1.68; 95% CI: 1.55–1.82), and hospitals in national networks were also more likely to use a regional HIE (RR = 1.45; 95% CI: 1.35–1.56). Hospitals with single vendor networks were again only slightly more likely to use an HIE (RR = 1.07; 95% CI: 1.00–1.14). Associations between the 3 methods without an intermediary and either Direct Messaging or HIE use were small.

Predictors of single vendor network participation

In contrast to those patterns, hospitals were substantially more likely to obtain information through a single vendor network if they also obtained information via provider portals and interface connections (RR = 2.0; 95% CI: 1.76–2.38 and RR = 1.74; 95% CI: 1.58–1.92). Hospitals participating in national networks were also substantially more likely to obtain information through a single vendor network (RR = 1.65; 95% CI: 1.51–1.81). The relationships between HIE use and single vendor networks and Direct Messaging and single vendor networks were smaller but still positive (RR = 1.32; 95% CI: 1.17–1.4 8; RR = 1.23; 95% CI: 1.12–1.35, respectively).

Predictors of national network participation

Finally, hospitals that participated in Direct Messaging via a HISP, regional HIEs, and single vendor networks were more likely to participate in a national network (RR = 1.48; 95%: 1.34–1.64; RR = 2.18; 95% CI: 1.87–2.54; RR = 1.64; 95% CI: 1.50–1.80). Associations between the 3 methods without an intermediary and use of national networks were small.

Exploratory factor analysis

In EFA, we identified patterns of adoption that largely paralleled these associations. The data supported identification of either 1 or 2 factors. All items loaded on the single factor in the single-factor solution, consistent with some hospitals adopting more methods due to their overall size or scope, and that relationship is not accounted for in this approach as in the regression (Table 1). In the 2-factor solution, the 4 network-based methods of exchange loaded on the first factor while the 3 methods without intermediaries loaded on the second factor. The single developer network method loaded equally on both factor and was the only network method that loaded on the factor comprised primarily by the methods without intermediaries.

Table 1.

Exploratory factor analysis

	One-factor solution	Two-factor solution
	Factor 1	Factor1	Factor2
Variance explained	2.67	2.37	2.28
One-to-one methods
Provider portal	0.67	0.07	0.67
Interface to outside HER	0.57	0.11	0.51
Direct EHR access	0.50	0.06	0.48
Many-to-many methods
Direct via HISP	0.62	0.59	0.08
Regional HIE	0.64	0.61	0.08
Single EHR vendor network	0.67	0.37	0.35
National network	0.63	0.61	0.06

Note: Results derived from factor analysis and Promax oblique rotation with power 3.

EHR: electronic health record; HIE: health information exchange; HISP: health information service provider

Bold value are above the 0.30 threshold for factor loading to indicate that a given variable was associated with each factor.

Relationship with information availability, integration, and use

All 4 “many-to-many” methods of obtaining information were associated with greater likelihood of information being available to providers (Table 2). Relative risk for Direct Messaging via HISPs was 1.39 (95% CI: 1.25–1.54); for regional HIEs the relative risk was 1.19 (95% CI: 1.07–1.32); for single vendor networks the relative risk was 1.37 (95% CI: 1.25–1.49); and for national networks was 1.28 (95% CI: 1.17–1.40). These risk ratios are not statistically distinguishable (P > .05 in all comparisons).

Table 2.

Association between adoption of each method and information availability and integration

Variables	Is information routinely available? Relative risk (Standard error)	Is information integrated into EHR? Relative risk (Standard error)
One-to-one exchange
Provider portal	1.44**	1.01
	(0.09)	(0.08)
Interface to outside HER	1.05	1.20**
	(0.04)	(0.07)
Access to outside EHR	1.00	0.79**
	(0.04)	(0.04)
Many-to-many exchange
Direct via HISP	1.39**	1.11†
	(0.07)	(0.07)
Regional HIE	1.19**	1.12†
	(0.07)	(0.07)
Single EHR vendor	1.37**	3.38**
	(0.06)	(0.28)
National network	1.28**	1.24**
	(0.06)	(0.07)
Observations	2941	3053

Note: Results derived from Poisson regression models with binary outcomes. Coefficients represent relative risk ratios. All models include hospital size, multi-hospital system status, urban/rural location, teaching status ownership, and presence of cardiac ICU and operating margin. Sample weights adjusted for response bias based on hospital size, system membership, and location.

EHR: electronic health record; HIE: health information exchange; HISP: health information service provider.

**P < .01; ^†P < .10.

However, among these 4 methods, only single vendor networks and national networks were associated with greater relative risk of information being integrated into the EHR. The association between single vendor networks and integration (RR = 3.38; 95% CI: 2.87–3.98) was very large and substantially greater than the association between national networks and integration (RR = 1.24; 95% CI: 1.11–1.40).

Similarly, among the many-to-many exchange modalities, only the single vendor networks and national networks were associated with higher reported rates of information being used by providers (Table 3). Once again, the association between single vendor networks and use of information (odds ratio = 4.68; 95% CI = 3.53–5.18) were substantial and larger than the association between national networks and use of information (odds ratio = 1.83; 95% CI = 1.55–2.16).

Table 3.

Association between adoption of each method information use

Variables	How often is information used? Odds ratio (Standard error)
One-to-one exchange
Provider portal	1.76**
	(0.21)
Interface to outside EHR	1.32**
	(0.13)
Access to outside EHR	1.12
	(0.11)
Many-to-many exchange
Direct via HISP	1.13
	(0.11)
Regional HIE	0.88
	(0.09)
Single EHR vendor	4.68**
	(0.47)
National network	1.83**
	(0.16)
Observations	2996

Note: Odds ratios derived from ordinal logistic regression model (with an outcome with 4 levels: often, sometimes, rarely, and never) producing odds ratios. All models include hospital size, multi-hospital system status, urban/rural location, teaching status ownership, and presence of cardiac ICU and operating margin. Sample weights adjusted for response bias based on hospital size, system membership, and location.

**P < .01.

Of the 3 one-to-one methods, only interface to an outside EHR was positively related to greater integration (RR = 1.20; 95% CI; 1.08–1.33). Adoption of both an interface to an outside EHRs and a provider portal also related to higher reported rates of information being used by providers (OR = 1.32 95% CI: 1.09–1.60; and OR = 1.76; 95% CI: 1.39–2.23).

DISCUSSION

Using data from a national survey of hospitals, we observed predominantly complementary relationships between adoption of diverse methods of obtaining information from outside sources. And while each method connecting hospitals to many other providers was independently associated with greater reported information availability, only a subset of those methods was associated with higher reported information use. These findings suggest that while each method led to increased information, certain methods may be easier to use and provide information that may be more valuable.

Across the 7 methods of obtaining information, we observed a negative correlation between how frequently the method was adopted and how frequently it was used. This was particularly true for the 2 most commonly adopted methods: the use of provider portals and access to outside EHRs. Notably, both methods are likely outside of healthcare professionals’ primary EHR. The requirement to leave their workflow and (potentially) log into a separate system may pose a serious barrier to use.^19–21 The same lack of integration may drive the relatively low use of regional HIEs, which are often contained in portals separate from the EHR.^22,23

In contrast, the 2 least frequently adopted methods—single EHR vendor networks and national networks—were the most frequently used when adopted. Both of these methods are likely closely tied to the EHR: Single vendor networks are most prevalent among leading EHR vendors and some prominent national network emerged from an alliance between specific EHR vendors.^21,24,25 This close tie between method of obtaining information and EHR platform likely facilitates easier integration and related usability improvements that facilitate easier use of information.

These trends persisted in adjusted analysis. All 4 methods that included an intermediary and facilitated connectivity to many others were associated with a greater likelihood of reported information availability. However, only 2 of the 4 methods were associated with a greater likelihood of integration with the EHR (single EHR vendor networks and national networks) and those same methods were related to more frequent information use. These findings reinforce the crucial importance of “last mile” issues in driving the value of information exchange: without usable, integrated methods of obtaining information, available information may not be used.^26,27 Although our analysis does not capture other dimensions that might promote the use of exchanged information, we suspect that issues beyond integration—such as the quality of data and the ease of finding specific information—may also underline the greater apparent use of the single EHR vendor networks and national networks.

Despite concerns about competition between these methods, we observed few negative correlations between methods of obtaining information. Instead, we observed a few tightly related sets of methods in which adoption of methods were highly correlated and these patterns followed our observations of market dynamics.¹³ The first set was comprised of Regional HIEs and Direct via HISPs.⁹ Regional HIEs and Direct represent open network based approach to exchange, fulfill complementary use cases, and support for both is often provided by the same organizations.²⁸ The second set was comprised by single vendor networks, provider portals and interface connections, each of which was only weakly related to the first set. These methods may reflect a more closed approach to exchange: both portals and interface connections can be established only with specific trusted partners, and single EHR vendor networks are constrained to the EHR vendor’s network.⁴ We also suspect that differences in these connections reflect how actively EHR vendors enable specific methods, an important market dimension that we do not fully explore here and which might warrant further investigation. Participation in national networks was substantially related to participation in all network-based methods. This likely relates to how connectivity to national networks is achieved: in many cases hospitals are connected to the national network by their HIE or EHR vendor.

The cross-cutting position of national networks may have implications for TEFCA. On one hand, this relationship may indicate that even before its enactment some of the goals of TEFCA have been achieved because 2 hospitals that, for instance, are each in different regional HIEs may nevertheless be connected because both HIEs participate in the same national network. Alternatively, this correlation may indicate that the substantial portion of hospitals in HIEs or single vendor networks that did not indicate participating in a national network may be ready to move towards that national approach if an overarching framework provides further ease or incentive to them, their vendor and/or their regional HIE.

It will be important to monitor the impacts of TEFCA on the number of hospitals adopting multiple methods and the frequency of use of each method. TEFCA may reduce the need to use these complementary methods by enabling broader access to exchange partners through the use of one method. Thus, over time, the number of methods used by hospitals may diminish or the frequency of using methods not connected to TEFCA may decrease. This may represent increased value by reducing the cost associated with adoption and maintenance of methods that do not independently relate to greater use of information.

It is challenging to predict what our findings imply for methods of exchange not directly addressed by TEFCA. For example, TEFCA as currently planned supports message delivery (or “push” messaging) between planned, large coordinating networks, referred to as qualified health information networks (QHINs). However, it does not specify what happens within-QHIN. Direct Messaging could remain the preferred method to push messages within individual QHINs. Furthermore, Direct may continue to provide access to other providers that are not part of an HIE or any TEFCA participating entity. TEFCA may result in increased use of Direct by further incentivize use of its complements, such as HIEs. It is not clear if this same dynamic holds for the use of provider portals, an interface to outside EHRs or access to outside EHRs, which commonly adopted but are also not included in TEFCA and do not as closely correlate to network participation in the data. Follow-on work should assess dynamics related to participation in each method as the landscape continues to mature.

Our study is subject to a number of limitations. First, this analysis is cross-sectional and cannot support causal inference. Nevertheless, by demonstrating the relationship between various methods and between those methods and information availability, our study points to important mechanisms that may facilitate the use of information from outside sources to inform clinical decisions. Second, our measures related to the frequency of information availability and use are from the perspective of hospital administrators and it is possible that they do not directly reflect the experience of healthcare professionals. It is also possible that some underlying challenges to using multiple methods are not reflected in simple data on frequency of use, and we do not capture the overall cost of adoption to the organization or other problematic dimensions in this analysis. Third, methods of obtaining information are closely correlated and in adjusted models that correlation could create a downward bias in the estimation of the benefit of participating in any one method. For instance, participation in a regional HIE is related to participation in a national network and the relationship between regional HIEs and information use may be positive and statistically significant if the national network measure was omitted from the regression model; however, exclusion of national networks may create an upward bias (because not all of the benefit of a national network is attributable to the regional HIE). Given this dynamic, these results represent the independent effect of each method but may underestimate the total effect because methods are tied together.

CONCLUSION

Newer methods of exchange, such as national and EHR vendor networks, do not appear to be substitutes to other methods but rather are adopted as complements or unrelated to the decision to adopt well-established modes of exchange. Each network-based measure independently contributed to a greater likelihood of having information available for healthcare professionals. However, only a subset of methods that also enabled greater integration of data into EHRs were associated with greater use of that information. This highlights the essential role of facilitating the usability of available information. As TEFCA is implemented, future work should examine whether these complementary methods become substitutes, and thereby reduce the need for using multiple methods and whether TEFCA leads to information that is more widely available, integrated, and used.

FUNDING

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sector.

AUTHOR CONTRIBUTIONS

JE and VP conceived of the presented study and devised analytic plan. JE carried out analysis and data management. JE and VP collaborated on interpretation of results. JE took the lead in writing the manuscript. VP provided critical revisions.

CONFLICT OF INTEREST STATEMENT

None declared.

Data Availability

Data for this study is available, for a fee, from the American Hospital Association. https://www.ahadata.com/aha-data-resources