Electronic health record developer market segmentation contributes to divide in physician interoperable exchange

Abstract

Objectives

To assess whether previously observed differences in interoperable exchange by physician practice size persisted in 2019 and identify the role of 3 factors shaping interoperable exchange among physicians in practices of varying sizes: Federal incentive programs designed to encourage health IT use, value-based care, and selection of electronic health record (EHR) developer.

Materials

Cross-sectional analysis of a 2019 survey of physicians. We used multivariable Poisson models to estimate the relative risk of interoperable exchange based on the size of the practice accounting for other characteristics and the mediating role of 3 factors.

Results

Seventeen percent of solo practice physicians integrated outside data relative to 51% of large practice physicians. This difference remained substantial in initial multivariable models including physician characteristics. When included in models, Federal incentive programs partially mediated the relationship between practice size and interoperable exchange status. In final models including EHR developer, developer was strongly associated with both exchange and integration while practice size was no longer an independent predictor. These trends persisted when comparing practices with 4 or fewer physicians to those with 5 or more.

Discussion

Public and private initiatives that increase the benefits of interoperable exchange may encourage small practices to pursue it. Technical and policy changes that reduce the costs and complexity of supporting exchange could make it easier for small developers to advance their capabilities to support small practices.

Conclusion

Addressing the gap between small and large practices will take a 2-pronged approach that targets both small EHR developers and small practices.

INTRODUCTION

Interoperable exchange of clinical information between healthcare organizations may offer a substantial benefit to physicians practicing in independent small and solo practices.^1,2 Whereas patients treated by large, integrated practices may receive all care within the organization, patients of small practice physicians are likely to receive care from physicians or healthcare organizations outside of the small practice.³ Physicians in small practices may, therefore, need to rely on interoperable technology to a greater extent to receive important clinical information in a timely and useable manner.

Despite this potential value, small practice and independent physicians have been observed to adopt EHRs and engage in interoperable exchange at lower rates than other physicians.^4,5 Similarly, previous research has observed substantial gaps in health information technology adoption and use across different types of hospitals.⁶ However, it is not clear whether similar gaps in interoperable exchange remain prevalent across physician practices of varied sizes or whether ongoing trends in technology and policy promoting interoperability have reduced or eliminated such differences.

Three important trends may shape the magnitude of differences in interoperable exchange by physician practices of varied sizes. The first trend is physician participation in incentive and performance programs, which increasingly prioritize interoperability. For instance, the Centers for Medicare & Medicaid Services’ EHR Incentive Programs for eligible hospitals and the related category in the Merit-Based Incentive Programs for eligible clinicians were renamed “Promoting Interoperability” beginning in 2018 to reflect the new priority of the programs. This focus is also a key part of a second trend, the increasing prevalence of value-based care (VBC) programs. VBC programs are led by both public and private entities and often include requirements to use interoperable technologies. The value-based paradigm is also thought to motivate interoperability by bringing a focus to high-quality care and cost savings, which may be achieved through greater information sharing.^6,7

These programs could reduce gaps in physician engagement in interoperable exchange by providing support or incentives to a wide range of physicians or perhaps specifically targeting those lagging behind.^8,9 However, challenges related to the administrative complexity of participating in incentive programs and VBC programs may reduce participation among physicians in small practices and thereby unintentionally increase differences in technology adoption across practices.^10–14

A third important trend related to interoperability is the growing importance of features offered by certain EHR developers in driving interoperability. While differences by developer have likely always existed, the emergence of developer-specific networks that support interoperable exchange between customers of those developers and the growth of national networks that facilitate interoperable exchange may lead to larger differences in how well specific EHR developers support interoperability.^15–17 Large, market-leading developers co-led the establishment of some national networks—for example, Cerner’s, athenahealth’s, and Allscripts’ roles in founding the CommonWell Health Alliance—which have become an important method of health information exchange in recent years.^18,19

The growth of these networks may have different impacts on some physicians because some developers offer more robust vendor-specific networks than others and some developers facilitate connectivity to national networks. The EHR developer market is substantially segmented, with some developers offering more expensive “enterprise” solutions aimed explicitly at large organizations and other developers offering lighter-weight products targeted at small practices.²⁰ The evolving differences in EHR developers’ support for interoperable exchange may, therefore, further the gap between small and large practices.

Given these trends, we sought to assess whether the rate of interoperable exchange by physicians varied by practice size in 2019 and whether differences in participation in federal incentive programs, VBC participation, and EHR developer market segmentation played a role in explaining this variation.

METHODS

Data

Data for this study are from the 2019 National Electronic Health Record Survey (NEHRS). The NEHRS sampling frame includes all office-based physicians principally engaged in patient care who are not federally employed. The sampling frame is based on the American Medical Association and American Osteopathic Association Master Files and reflects data contained in those sources.^21,22 It also excludes traditionally nonpatient facing physician specialties (anesthesiology, pathology, and radiology), physicians older than 85 years of age and physicians practicing in hospital outpatient departments. The total population of physicians in the sampling frame includes 301 603 physicians and sampling weights were developed to facilitate estimates reflecting the characteristics of this population. The survey response rate was 39.0%.

Interoperable exchange

While a number of definitions of interoperability exist, including the definition established in the 21st Century Cures Act, we limited our focus to measuring whether a physician was engaged in the electronic exchange of patient health information with outside healthcare organizations. To do so, we created a composite measure that combined responses to 3 questions which respectively asked whether physicians received, found, or sent information to outside organizations. See Supplementary Appendix for items used to develop these and other composite measures.

We also included a single binary measure capturing whether patient health information that was received electronically was integrated into the EHR. Integration represents a more advanced component of interoperable exchange that facilitates more frequent and diverse use of information within the EHR and clinical workflow.

Federal incentive program and VBC participation

Three questions about participation in programs incentivizing EHR use are included in the NEHRS survey. The questions reflected participation in the Medicare Merit-Based Incentive Payment System, Medicaid EHR Incentive Program, or a Medicare Advanced Alternative Payment Model. We created a dichotomous composite measure that combined responses to these 3 questions.

Similarly, 3 questions about VBC participation (participation in a Patient-Centered Medical Home, Accountable Care Organization, or Pay-for-Performance Program) are included in the NEHRS survey. We measured these separate from Federal incentive program participation because VBC programs are often supported by commercial payers and do not necessarily directly incentivize engagement in interoperable exchange. We again combined responses to these 3 questions to create a composite binary measure.

EHR developer

In their response to the NEHRS, physicians indicated which of 14 specific developers’ EHRs they used, selected “Other” or indicated that they did not use an EHR. To parsimoniously capture market dynamics, we grouped EHR developers into 4 groups according to ambulatory market share as measured by responses to the NEHRS survey: the 5 market-leading developers (Epic, eClinicalWorks, athenahealth, Cerner, and Allscripts), the next 5 developers (NextGen, Practice Fusion, Greenway, GE, and Modernizing Medicine), all other EHRs, and a final group for those that did not use an EHR.

Physician practice size and other characteristics

Physician practice size was coded as 5 groups (solo practice, 2–3 physicians, 4–10 physicians, 11–50 physicians, and 50+ physicians). Several additional covariates hypothesized to relate to both the incentive or ability to engage in interoperable exchange and practice size were included in multivariate models. These included specialty category (primary care, medical specialist, and surgical specialist); age category (under or over 50); number of midlevel providers (MLPs) (0, 1–3, 4+); practice ownership (combined into 4 groups: (1) physician or physician group, (2) health system [respondents indicated ownership by “Medical/Academic health center,” “other hospital” or], (3) “Community Health Center,” or (4) other [respondents indicated ownership by “insurance company, health plan or HMO,” “Other health care corporation,” or “Other”]); participation in an independent practice association (IPA); whether the physician treated Medicare patients; and the percent of the physician’s patient panel comprised by Medicaid patients (0%; 1–24%, 25–50%; >50%). In some cases, missing data in covariates were imputed to retain the overall sample size. Full details are available in the Supplementary Appendix.

Analytic approach

We first assessed differences in interoperable exchange status by physician practice size category. Sampling weights were used in all cases.

We next sought to assess whether federal incentive program participation, VBC participation, and EHR developer choice mediated the relationship between practice size and interoperable exchange using a classical mediation model. We first assessed whether practice size was related to each of the 3 potential mediators in multivariable Poisson regression models that included practice size and other physician characteristics. To simplify interpretation, we created a binary indicator of whether the physician used 1 of the top 5 market leaders or not. The coefficients in Poisson regression models on binary outcomes can be interpreted as relative risk ratios.²³

We next sought to assess the direct association between practice size and interoperable exchange. To do so, we implemented 2 multivariate Poisson regressions predicting whether the physician was able to find, send or receive information electronically and whether the information was integrated into the EHR. These models included practice size and all other covariates listed above. We then replicated these models with the addition of 2 variables reflecting Federal incentive program and VBC participation and observed both their direct association with interoperable exchange status and changes in the association between practice size and interoperable exchange. Finally, we added categories of EHR developer to the models and again observed direct effects and changes in the relationship between practice size and interoperable exchange status.

To validate the mediating effect, we compared the coefficients on practice size across models including the potential mediators. To do so, we treated each model as seemingly unrelated estimates and compared differences of a linear combination. Statistically significant differences in these coefficients would indicate that Federal incentive program participation, VBC participation, and EHR developer were acting as mediators on the relationship between practice size and interoperable exchange status.

We performed 3 robustness checks. We first replicated these models while excluding physicians that did not use an EHR to ensure that our results were not driven solely by differences in whether physicians had adopted an EHR at all. We next replicated the mediation models including all developers as individual fixed effects to observe whether idiosyncratic differences in developers not directly related to market share produced a different result. Finally, we examined whether our findings extended to small practices generally by comparing practices with 4 or fewer physicians and practices with 5 or more physicians.

RESULTS

The sample includes 1524 physicians that responded to the 2019 NEHRS. Weighted descriptive statistics are shown in Supplementary Appendix Table S1. Forty-eight percent of physicians were in primary care, 76% were over 50 years of age, 25% were in solo practice while 11% were in practices with 50 or more clinicians, and 62% of physicians were in practices owned by physicians or physician groups.

Among this population, our weighted estimate is that 65% of physicians (95% confidence interval [CI] 61%–68%) were able to electronically send, receive, or query for information from outside sources (Table 1). In comparison, 29% of physicians (95% CI 26%–33%) were able to integrate information into the EHR. Overall, 58% of physicians indicated participating in a federal incentive program (95% CI 54%–62%), and 49% (95% CI 45%–53%) indicated participating in VBC. Forty-two percent of physicians used an EHR from 1 of the 5 market-leading developers (95% CI 38%–46%); 21% of physicians used an EHR from the 6th to 10th market-leading developers (95% CI 17%–24%); 27% from another vendor (95% CI 23%–31%) and 10% did not use an EHR (8%–12%).

Table 1.

Physician interoperability, incentive program participation, and EHR developer characteristics

	%
Interoperability
Send	34
Receive	35
Query	49
Send, receive, or query	65
Integrate	29
Federal incentive program participation
Any program	58
Value-based payments (VBC) participation
Any program	49
EHR Developer
1st–5th market position (Epic, eClinicalWorks, athenahealth, Cerner, and Allscripts)	42
6th–10th market position (NextGen, Practice Fusion, Greenway, GE, Modernizing Medicine)	21
Other	27
None	10

Note: n = 1524 physicians.

There was substantial variation in interoperable exchange by practice size (Figure 1). Forty-four percent of solo practitioners were able to send, receive, or query for information relative to 71% of physicians in practices of 4–10 physicians and 76% of physicians in practices of more than 50 physicians. Differences in ability to integrate were notable across practice size categories: 17% of solo practitioners reported being able to integrate data, compared to 30% of physicians in practices of 4–10 physicians, and 51% of physicians in practices of greater than 50 physicians.

Figure 1.

Rates of interoperability by practice size.

Note: n = 1524 physicians.

In multivariable analysis, physician practice size was associated with each mediator (Table 2). Relative to solo practitioners, physicians in groups of 50 or more physicians were 88% more likely to participate in a federal incentive program (relative risk [RR] 1.88, 95% CI 1.43–2.49), 110% more likely to participate in VBC (RR 2.10, 95% CI 1.57–2.82) and 145% more likely to use 1 of the top 5 developers (RR 2.45, 95% CI 1.77–3.41). Additional details on physician characteristics and developers are available as cross-tabulations by 4 developer groups and by each individual developer in Supplementary Appendix Tables S2 and S3, respectively.

Table 2.

Relationship between physician characteristics, federal program participation, value-based payment participation, and EHR developer market position

	Federal incentive program relative risk ratio	95% CI	Value-based payment relative risk ratio	95% CI	Top 5 market position developer relative risk ratio	95% CI
Practice size (omitted: solo practice)
2–3 physicians	1.30**	1.02–1.65	1.29*	0.99–1.69	1.47**	1.08–1.99
4–10 physicians	1.46***	1.17–1.83	1.46***	1.14–1.86	1.73***	1.31–2.29
11–50 physicians	1.60***	1.24–2.07	1.75***	1.33–2.31	1.80***	1.31–2.48
50+ physicians	1.88***	1.43–2.49	2.10***	1.57–2.82	2.45***	1.77–3.41
Specialty (omitted: primary care)
Medical specialty	0.99	0.84–1.18	0.56***	0.46–0.69	0.73***	0.59–0.90
Surgical specialty	0.96	0.82–1.13	0.63***	0.53–0.76	0.76***	0.62–0.92
Age (omitted: under 50)
Over 50 years	0.89	0.77–1.03	0.95	0.81–1.11	1.02	0.86–1.21
Midlevel providers at location (omitted: 0)
1–3 MLPs	1.10	0.92–1.31	1.14	0.94–1.38	1.29**	1.04–1.60
4+ MLPs	1.19*	0.97–1.46	1.10	0.88–1.37	1.14	0.89–1.47
Ownership (omitted: physician or physician group)
Health system	0.95	0.80–1.13	1.10	0.91–1.32	1.45***	1.19–1.76
Community health center	0.97	0.71–1.32	1.25	0.91–1.72	0.94	0.62–1.42
Other	0.77**	0.62–0.95	1.05	0.86–1.30	1.27**	1.01–1.58
Independent practice association affiliation (omitted: no)
Yes	1.07	0.91–1.26	1.54***	1.30–1.81	1.29***	1.07–1.56
Do not know	0.85*	0.71–1.02	1.03	0.85–1.25	1.14	0.93–1.39
Treat Medicare (omitted: no)
Yes	1.41***	1.13–1.78	1.01	0.82–1.24	0.91	0.72–1.15
% Medicaid (omitted: 0%)
1–24	1.72***	1.27–2.32	1.22	0.92–1.61	1.43**	1.03–1.99
25–50	1.73***	1.27–2.35	1.06	0.80–1.42	1.24	0.88–1.75
>50	1.88***	1.30–2.72	1.19	0.84–1.70	1.11	0.73–1.70
Constant	0.20***	0.14–0.29	0.30***	0.21–0.44	0.16***	0.11–0.25
Observations	1524		1524		1524

Note: *P < .10, **P < .05, ***P < .01. Results derived from multivariable Poisson regression models.

Electronic exchange among office-based physicians

In models without hypothesized mediators, the largest practice size was associated with a 41% greater likelihood of being able to send, receive, or query for information from outside organizations (Figure 2, Panel 1; RR 1.41, 95% CI 1.11–1.80; Full results available as Supplementary Appendix Table S4). This association is reduced in magnitude to 31% (RR 1.31, 95% CI 1.03–1.68) in models including federal incentive programs and VBC programs (test of difference P = .01). Federal incentive program participation was marginally independently associated with the ability to send, receive, or query (RR 1.11, 95% CI 0.98–1.26) as was VBC participation (RR 1.13, 95% CI 1.00–1.28). When the EHR developer was also included in the regression model, the largest practice size was no longer more likely to have the ability to send, receive, and query relative to the smallest (RR 1.14, CI 0.89–1.53; test of difference P < .001). The association between EHR developer market position and the ability to send, receive, or query is substantial: relative to physicians using the market-leading developers, physicians that used a developer other than 1 of the top 10 developers were 26% less likely to report being able to send, receive, or query (RR 0.74, 95% CI 0.62–0.87).

Figure 2.

Association between physician characteristics and interoperability, with mediation by federal program participation and EHR developer.

Note: n = 1524 physicians. Results derived from sequential multivariable Poisson regression models. Initial models included only physician and practice characteristics. Policy-related variables were included in the second step. Developer market segment variables were included in the third step.

Integration of data received from external sources

Without mediators, practice size was associated with a 109% greater likelihood of being able to integrate information from outside organizations (Figure 2, Panel 2 RR 2.09, 95% CI 1.28–3.38). That association was reduced to a relative risk of 1.77 (95% CI 1.06–2.94) when Federal incentive program participation and VBC program participation were included (test of difference P = .01). Federal incentive program participation was associated with a greater likelihood of being able to integrate (RR 1.40, 95% CI 1.05–1.87), but VBC participation was not. When EHR developer is included in the regression model, the association between practice size and integration was no longer statistically significant (RR 1.43, 95% CI 0.88–2.35; test of difference < 0.001). Just as with the send, receive, or query mediated model, the association between EHR developer market position and integration ability is substantial: physicians that used an EHR product other than 1 offered by 1 of the top 10 developers were 54% less likely (RR 0.46, 95% CI 0.31–0.69) to integrate information as physicians using 1 of the top 5 developers. This mediation model is depicted in detail in Figure 3.

Figure 3.

Illustration of mediation model. This figure highlights key mediation processes displayed in Table 2 and Figure 2. For simplicity, only the relationship between largest (vs smallest) practice size, participation in a Federal Incentive Program, and choice of EHR developer are presented, and only the relationship between these variables and Integrating Information is shown. *indicates statistical significance at P < .05.

When we replicated these models among only physicians that used an EHR, we found results were of a consistent direction and marginally reduced magnitude (Supplementary Appendix Table S5). When we used individual EHR developers rather than categories, we observed a more pronounced mediation effect. In mediated models using developer fixed effects, the difference between the smallest and largest practice size in ability to send, receive, or query for information was reduced to 7% and in integration ability to 38%. Neither coefficient was statistically significant (Supplementary Appendix Table S6). Last, when we re-examined these associations by using 2 general categories of practice size (4 or fewer vs 5 or more), results were directionally consistent and with identical patterns of statistical significance, but the magnitude of differences by practice size was reduced (Supplementary Appendix Table S7).

DISCUSSION

National survey findings in 2019 show the need for greater progress in interoperable exchange among office-based physicians, particularly among physicians working in smaller practices. Overall, 65% of physicians indicated they engaged in some type of electronic exchange and 29% reported the capability to integrate patient health information from external sources. Importantly, we observed substantially greater capabilities for physicians in large organizations to both send, receive, or query for information from outside organizations and to integrate that information into their EHR compared to those in smaller organizations. In seeking to understand the drivers of this difference, we assessed the mediating role of 3 trends: participation in Federal programs that incentivize information exchange and interoperability; participation in broader VBC initiatives; and choice of EHR developer product. We found that the federal incentive program and VBC participation played a modest role in explaining the differences in interoperable exchange status by practice size, while the EHR developer used was a key determinant.

In the available data, we observed a substantial mediating effect by EHR developer choice such that associations between practice size and engagement in interoperable exchange are deeply confounded by the choice of developer. Put another way, we found that physicians in practices of different sizes used EHRs from different developers, and this choice appears to be an important reason for differences in exchange and integration capabilities across practice sizes. We found clear differences in both the ability to exchange information and to integrate information across both practice size and developer. However, physicians in larger practices and those using market-leading developers had a substantially greater advantage in the ability to integrate information into the EHR—a more advanced function—than in the ability to exchange information with other organizations at all—a relatively low bar. Because the integration of data into the EHR and clinical workflow is important for subsequent usage of exchanged information, this difference in integration ability may lead to large gaps in benefits from the exchange. Even beyond integration, market-leading developers likely offer greater support for a range of more advanced interoperability, including connectivity to national networks, proprietary exchange networks, effective presentation of data, and workflow integration.²⁴ The trends we identify here indicate that there may be even more pronounced differences in the advanced capabilities available to providers than the differences in capabilities we describe.

A crucial question this raises is what factors led small practices to select, and continue using, developers with less support for interoperability and, as a corollary, how public policy or other actors might increase their motivation and ability to use an EHR that better supports interoperability. One possibility is that practices lack the means to select a developer with more robust interoperability features. Federal programs, such as the CMS Promoting Interoperability Performance Category within the Merit-based Incentive Program, link payment adjustments to interoperable exchange, which could stimulate investment in interoperable technology. However, to reduce the burden on small practices, this program includes features that exempt small practices and practices that do not meet thresholds related to the volume of services provided. As a result, our data indicate that this program, and other Federal incentive programs, achieve greater participation by physicians in large practices than in small practices: only 41% of physicians in solo practice reported participating in a program compared to 78% of physicians in larger practices, and participation in Federal incentive programs was associated with a greater likelihood of engaging in exchange and integration of data. These findings echo prior work showing the disproportionate burden of participation on smaller practices.^25,26

Exemptions provided under the program may be unlikely to address these issues and advance more widespread use of interoperable technology. Instead, they may result in limited incentives for small practices to engage in interoperable exchange by exempting them from financial incentives that apply to large practices, thereby widening the gap in interoperable exchange. More support for small practices, such as targeted financial incentives and technical support, may be needed to encourage the use of certified EHRs with more advanced interoperable exchange capabilities. Past examples of targeted programs that successfully encouraged health IT use among health care providers with fewer resources include the Regional Extension Center Program and the “adopt, implement, and use” portion of the Medicaid EHR Incentive Program which respectively directed support to small provider practices and Medicaid providers.^8,9

Beyond these direct financial incentives, policies, and technical improvements that ensure ease of transferring data between systems and minimize charges for providers to extract data from their current EHR to change developers could allow those dissatisfied with their current developer to seek another. A number of recent policy actions have sought to address these barriers. Information blocking rules seek to ensure that practices do not face excessive charges to extract data from their current EHR when seeking to change developers.²⁷ Additionally, EHR certification criteria make it easier to transfer large quantities of data between differing EHR systems.²⁸

A second possible reason physicians continue to use developers with limited interoperability support is that they lack the ability to easily view and compare which developers best support interoperability and so cannot choose a developer with robust interoperability. It is possible that in initially selecting a developer offering an EHR certified by the Office of the National Coordinator for Health Information Technology (ONC), certain assumptions were made about its out-of-the-box connectivity to health information networks, the extent to which it included supportive workflows, and that limited additional information is available to guide the choice of a developer with more interoperability support. Interventions to increase transparency about products, such as ONC’s EHR Reporting Program and Real-World Testing program could work to address this issue and increase awareness around a product’s interoperability features.

A third possibility is that practices have prioritized other features of the EHR, and given a limited set of choices, these considerations (for instance, cost, fit to the organization, fit to the practice’s style, or specialty) may outweigh differences in interoperability in practices’ choice of developer. In parallel, practices may devote resources related to quality and efficiency goals to other activities. For instance, currently, there is an assumption that VBC will motivate health care provider engagement in interoperability as a means to accomplish value-based goals, thereby driving the use of advanced interoperable technology. However, while larger practices were more likely to participate in VBC programs than smaller practices, VBC participation did not relate to engagement in interoperable exchange. This contrasts with other data in which participation in VBC programs did appear associated with greater levels of interoperability.⁷ Our findings indicate that healthcare organizations may be able to at least meet the base requirements for participating in value-based programs without substantially increasing information exchange performance.

Policy initiatives and strategies that increase the value of interoperability (as perceived by physicians) while making it easier for small developers to facilitate interoperability may help address the gaps we identified. Increased value could be derived from the greater availability of usable information as discrete, standardized data. Towards that end, ONC has updated the standards and certification criteria applicable to health IT so that it better supports and makes it easier to exchange discrete data elements as well as documents. The most recent ONC certification criteria for health IT require support for the United States Core Data for Interoperability (USCDI) standard and include technical requirements for application programming interface (API) functionality through the Standardized API for Patient and Population Services criterion, which references Release 4 of the Health Level 7 (HL7) Fast Healthcare Interoperability Resources (FHIR) standard. Future use of FHIR-based APIs for exchange among providers may also help small developers offer more advanced exchange capabilities at a lower cost. FHIR-based APIs provide a simplified means to access and exchange EHR data in a secure manner by enabling systems to request and obtain data without having to know how the data are stored within an EHR.

In the current system, connectivity to multiple networks and technologies is often necessary to facilitate impactful levels of interoperable exchange.^19,29 The complexity of the current state may be a substantial barrier for small practices and small developers with smaller market shares. Recent efforts to simplify exchange via networks could reduce this complexity, and associated costs, for small developers. The Trusted Exchange Framework and Common Agreement (TEFCA) should facilitate exchange with a variety of health care providers and other exchange partners by enabling participants of networks to exchange with each other more easily, reducing the need to join multiple networks. Small developers that might otherwise need to set up one-to-one interfaces to connect their customers to their exchange partners can enable their customers to join a single network through TEFCA to exchange data. Increased participation in such networks could also increase the value of interoperability since the size of networks is thought to dramatically increase their value.^30,31

Limitations

Our study is subjected to a number of limitations. First, the data employed in this analysis are cross-sectional and do not support rigorous causal inference. However, close correlations between variables appear to indicate important dynamics in the marketplace. In interpreting the data, we have suggested some causal mechanisms, but these warrant further investigation. Second, the data used in this survey may over-represent small practices and primary care because it is based on the AMA and AOA master files, and some specialists are likely not members of either or accurately represented in these files. It is therefore not certain whether overall estimates of rates are representative of the total population of physicians; however, the potential over representation of small practices is useful in the current analysis. Third, physicians may not be aware of all of the functions that their EHR support and may underestimate some exchange of information if they do not make use of the information. If a physician is not aware of the exchange of information, it is unlikely that they experienced benefits from the exchange. Therefore, it is possible that our data are closer to representing the perceived impact or benefit of interoperability rather than simple exchange. Finally, it is possible that the association between EHR developer and interoperability is confounded by omitted variables related to the overall quality of the EHR, investment in training, or interest in technology. It is therefore possible that our measure related to EHR developer is capturing a broader construct related to differences in investment in health IT by practices, but that construct would motivate similar conclusions.

CONCLUSION

Widespread adoption of interoperable capabilities by small, independent practices could facilitate clinical integration leading to improved quality.^32,33 However, in 2019 small practice physicians reported substantially lower rates of interoperable exchange than large practices. Differences in which EHR was used by different practice sizes were the most significant variable explaining observed differences in interoperability. And rather than reducing this difference, Federal incentive programs appear to have modestly contributed to widening differences because of limited participation among smaller practices. Public and private initiatives that increase the benefits and reduce the costs of facilitating interoperable exchange may also encourage small practices to pursue interoperable exchange while making it easier for small developers to advance their support.

FUNDING

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

AUTHOR CONTRIBUTORS

JE, WB, and VP contributed to the conception and design. JE drafted the work and performed analyses. WB and VP performed critical revisions.

SUPPLEMENTARY MATERIAL

Supplementary material is available at Journal of the American Medical Informatics Association online.

CONFLICT OF INTEREST STATEMENT

None declared.

DATA AVAILABILITY

The data underlying this article are available at https://www.cdc.gov/nchs/nehrs/questionnaires.htm.