Background: As more hospitals adopt Electronic Health Records (EHR), focus has shifted to how these records can be used to improve patient care. One barrier to this improvement is limited information exchange between providers. In this work we examine the role of EHR vendors, hypothesizing that vendors strategically control the exchange of clinical care summaries. Their strategy may involve the creation of networks that easily exchange information between providers with the same vendor but frustrate exchange between providers with different vendors, even as both Federal and State policies attempt to incentivize exchange through a common format. Methods: Using data from the 2013 American Hospital Association’s Information Technology Supplement, we examine the relationship between a hospital’s decision to share clinical care summaries outside of their network and EHR vendor market share, measured by the percentage of hospitals that have the same vendor in a Hospital Referral Region. Results: Our findings show that the likelihood of a hospital exchanging clinical summaries with hospitals outside its health system increases as the percentage of hospitals with the same EHR vendor in the region increases. The estimated odds of a hospital sharing clinical care summaries outside their system is 5.4 (95% CI, 3.29–8.80) times greater if all hospitals in the Hospital Referral Region use the same EHR Vendor than the corresponding odds for a hospital in an area with no hospitals using the same EHR Vendor. When reviewing the relationship of vendor market concentration at the state level we find a positive significant relationship with the percentage of hospitals that share clinical care summaries within a state. We find no significant impact from state policies designed to incentivize information exchange through the State Health Information Exchange Cooperative Program. Conclusion: There are benefits to exchanging using proprietary methods that are strengthened when the vendors are more concentrated. In order to avoid closed networks that foreclose some hospitals, it is important that future regulation attempt to be more inclusive of hospitals that do not use large vendors and are therefore unable to use proprietary methods for exchange. Keywords: Electronic Health Records, Interoperability, EHR vendors, Meaningful use * Correspondence: firstname.lastname@example.org Department of Engineering and Public Policy, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA 15213, USA © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Castillo et al. BMC Health Services Research (2018) 18:405 Page 2 of 12 Background that hospitals in competitive markets were less likely to Exchange of health information through interoperable share information. Furthermore, Miller and Tucker  systems is an essential goal as providers transition from found that hospitals that are part of larger health hard to digital copies of medical records . Interoper- systems are less likely to exchange information with able systems ensure that electronic health information hospitals outside their system. In sum, several studies can be used and exchanged without any special effort find that information exchange is a function of a hospi- from the sender or receiver through the use of a tal’s strategic reasons for sharing [16–20]. common language . Without interoperable systems, Although most research has looked at the characteris- the full potential benefits of adopting Electronic Health tics of hospitals associated with information exchange, Records cannot be achieved . recently more focus has been directed toward vendors The Health Information Technology for Economic and and how they may also use information exchange Clinical Health (HITECH) Act enacted in 2009 , and strategically . While EHR products must be capable the associated State Health Information Exchange of transforming from proprietary architectures to the Cooperative Program , provide monetary incentives to semantics and structure used in CCR or CCD formats at eligible providers and hospitals to support the adoption least once during the certification process, there are still of EHRs and health information exchange. To receive reports of both cost and technical barriers for the these incentives, hospitals and physicians must meet process of exchanging clinical care summaries [22, 23]. usability criteria also known as meaningful use (MU) Hence, EHR vendors could knowingly and unjustifiable objectives (core and menu) that ensure EHRs are used interfere with health information exchange by engaging to support health policy priorities . EHR products that in what is known as information blocking . are purchased through the Meaningful Use incentives We hypothesize that vendors can leverage proprietary are certified by the Department of Health and Human software to make it easier for hospitals to share clinical Services . While certification criteria changed the sup- care summaries with other hospitals that have the same ply side of the EHR market, the stated incentives allowed software while making it challenging to share with for a greater demand for EHR certified products . hospitals that use a different EHR vendor. This imposes A core objective of Meaningful Use’s second stage is the costs on hospitals that need to share information and exchange of clinical summaries when patients transition creates an incentive for them to adopt the dominant between providers. Until the year 2014, certification vendor. Specifically, research has shown that hospitals that requirements stated that both the Continuity of Care use dominant vendors likely face fewer technical obstacles Record (CCR) and the Continuity of Care Document and engage in a higher number of HIE activities . (CCD) standardized formats could be used for said ex- Empirical work in other domains supports this. For change . Current regulation, published in 2015, requires example, Shapiro and Varian find that there are network that vendors demonstrate that they are able to use the externality benefits to being connected to a larger com- second version of the Consolidated Clinical Document munication network . The value of connecting to a Architecture (C-CDA 2.1) markup standard. Aside from network depends on the number of others that are the format standard, MU incentives foster an open already connected, which means that from the perspec- approach to health information exchange, allowing for tive of a user, being connected to a bigger network is direct exchange among EHR vendors as well as enterprise better . When an EHR vendor has a large closed and community solutions . network, in this case a large number of hospitals that Despite these efforts to promote interoperability and use its product, a user will have an easier time sharing meaningful use, information exchange has remained information. This creates a competitive advantage for below expectations set by HITECH [1, 11, 12]. Qualita- the EHR vendor that provides the closed network. tive and quantitative studies identify several operational Even if policy incentivizes the exchange of clinical care and economic barriers to health information exchange. summaries, there is significant variation in the use of Qualitative studies have found that operational barriers HIE across EHR vendors. Some vendors have been at include the use of information as a competitive advan- the forefront by facilitating exchange through private tage, lack of cost-efficiency, limited incentives for staff proprietary networks. The most prominent of these and diminished trust of other providers [13–15]. Quanti- networks is Care Everywhere, a system incorporated into tative studies have shown that certain hospital character- EPIC EHR products since 2005 . Although Care istics are related to the probability that hospitals Everywhere is meant to be able to connect to EHR exchange health information with other hospitals. For systems from other vendors, it is most successfully used example, Adler-Milstein and Jha  found that hospi- to connect with other EPIC users . Additionally, tals with a larger market share within a region were connection even within the Care Everywhere network more likely to participate in information exchange but may require additional customization . Castillo et al. BMC Health Services Research (2018) 18:405 Page 3 of 12 The present study aims to identify the effect of vendor 4451 hospitals). For our first model we drop 311 hospi- choice and vendor network size on whether a hospital tals that are not able to generate summary of care reports participating in the exchange of clinical care records for transitions of care. We also remove the summaries. As a measure of the size of an EHR vendor’s hospitals for which we have no information relevant to network we will use EHR vendor market share and our main variables of analysis (electronically exchange market concentration. We hypothesize that the probabil- clinical care summaries, EHR vendor, use of common ity of a hospital engaging in the exchange of clinical care format and ability to exchange with other EHR summaries with another hospital outside its health vendors). Finally, we drop regions with less than three systems is associated with the market share of the EHR hospitals and are left with a sample of 1871 acute care vendor in the immediate region where exchanges are hospitals. Detailed characteristics of our final sample more likely to occur. are included in the Additional Files used in the logistic To determine this association, we use a logistic regression. At the state level we aggregate the data from regression model at the hospital level using the re- this survey to create indicators for the percentage of sponse from each provider about whether they ex- hospitals that participate in HIE exchange and the change clinical care summaries outside of their system prevalence of EHR vendors in each state. as a dependent variable. EHR vendor market share is Additionally, we use data from the Healthcare Infor- measured by the percentage of hospitals that have the mation and Management Systems Society (HIMSS) same vendor in the hospital referral region (HRR) Analytics Database for the year 2012, which compiles where a hospital operates. These regions, or markets data on the Information Technology capabilities for for tertiary medical care, consolidate zip codes where 5467 hospitals. From this database we extracted each the majority of patients are referred to a specific hos- hospital’s affiliation to an Integrated Delivery System pital for cardiovascular surgery and neurosurgery . (IDS). An IDS is a healthcare organization that owns We expect that in HRRs where intra-vendor sharing at least two medical/surgical hospitals. In this analysis occurs the probability of a hospital engaging in infor- we will refer to an IDS as a health system. We also use mation exchange increases as the market share of this reports from the Centers for Medicare and Medicaid hospital’s EHR vendor increases. This effect is due to Services (CMS) that detail the EHR Products used for the increase of the number of opportunities to engage meaningful use attestation by eligible hospitals. Finally in intra-vendor exchange. We also test for differences we use reports from the Office of the National Coord- that might be unique to large EHR vendors that have inator for Health Information Technology (ONC) on established proprietary information networks, such as the status of the State HIE Cooperative Agreement EPIC, by testing the different interactions in logistic Programs in 2013. regressions for the three largest EHR vendors. A second analysis is done at the state level. The aim of Measures this second model is to further analyze the propensity to Dependent variables share in the context of state level policies that incentivize Information Exchange (IE) and percent of hospitals health information exchange. Our hypothesis is that a that Exchange Information (%IE) We use information higher EHR vendor market concentration, measured by exchange as our dependent variable, operationalized as the Herfindahl-Hirschman Index (HHI), is associated with the yes/no answers found in the AHA IT supplement a positive change in the percentage of hospitals that database to questions about whether each hospital elec- participate in information exchange within each state. We tronically exchanges/shares patient information such as also expect to find differences in the propensity of this laboratory results, medication history, radiology reports, exchange depending on the strategies adopted by each and clinical care summaries with providers outside their state to incentivize HIE. health system. We use the exchange of clinical summar- ies during transitions of care, which is the requirement Methods for Stage 2 meaningful use compliance, coded as one or Data zero for yes and no, respectively. According to the ONC, We use data from the 2013 American Hospital Associ- a clinical care summary includes basic clinical informa- ation (AHA) Annual Survey Information Technology tion regarding the care provided, such as medications, Supplement. The survey was distributed between upcoming appointments, or other instructions. It is November 2013 and February 2014 to the Chief shared with patients and clinicians in order to increase Executive Officers of U.S. Hospitals, who in turn may awareness of what occurred during office visits and can delegate the responsibility of completion to the institu- be used to assist care coordination. This variable was tion’s qualified IT personnel. The survey had a response used to determine a hospital’s indication of health rate for non-federal acute care hospitals of 61% (2737/ information exchange (IE) and was also aggregated to Castillo et al. BMC Health Services Research (2018) 18:405 Page 4 of 12 determine the percentage of hospitals that answered included a dummy variable that indicates if there is positively to sharing within a state (%IE), using as a de- more than one hospital from the same health system in nominator the number of hospitals on the final sample the HRR (System Hospital). The aim of this last indica- (a total of 2296 hospitals). tor is to account for different sharing policies between hospitals that are part of the same system and are in the Vendor Market Share (VMS) To operationalize vendor same region. Finally, we used dummy variables for the market share, we used data from the AHA IT supple- largest three EHR Vendors: Epic, Meditech and Cerner. ment database that requested the name of the hospital’s We test the efficacy of state programs to encourage HIE primary outpatient EHR/EMR tool. This data was by adding dummy variables indicating the availability checked and complemented with data from the CMS (RHIO) and use (RHIO ) of Regional Health Infor- PART Meaningful Use Attestation database, which has infor- mation Organizations (RHIO), organizations that bring mation on the outpatient EHR product used by eligible together health care stakeholders within a defined geo- hospitals that participate in the MU program. The indi- graphic area and govern health information exchange cator for EHR vendor market share (VMS) for each hos- among them . pital was calculated by determining the percentage of hospitals within a Hospital Referral Region that use the State-level variables same EHR vendor as the subject hospital. Information on the models used by states for informa- tion exchange was extracted from the ONC progress re- State EHR Vendor HHI (VendorHHI) To determine port on the State HIE Cooperative Agreement Program the EHR vendor market concentration in a state we . We coded variables on the availability of Direct and use the Herfindahl-Hirschman Index (HHI), the Query exchange if the state reported that each type of standard measure used by the U.S. Department of exchange was “broadly available”. Broadly available types Justice to determine market concentration . This of exchange include Directed Exchange (point-to-point indicator measures market concentration using the secure communication) and Query-based Exchange (pull relative size of the market share and distribution of transactions through a request) . We also coded thefirms in amarket. For our analysis we define variables for the strategic approaches each state used to market share as the number of final users (patients) encourage information exchange, including four categor- that will use a specific EHR Vendor. As a proxy for ies Elevator (rapid facilitator of Directed Exchange), thenumberof patientsweuse thenumberof bedsin Capacity Builder (assists regional exchanges through each hospital, giving more weight to larger hospitals. financial and technical support), Orchestrator (state level We then calculate the HHI index by squaring the network to connect regional exchanges) and Public percentage share of beds for each EHR Vendor at the Utility (provides HIE services directly) . state level (VendorHHI). Using the IDS indicator from the HIMSS analytics database we calculated an HHI index for Health Systems Hospital-level variables in a State (SystemHHI) also weighted by hospital beds. Other potential explanatory variables are extracted We also included a variable for the number of beds in a from the AHA IT Supplement database. We use an State (HospitalsState). indicator for a hospital’s capability to send clinical summary of care records in one of three formats Analyses (CL): Continuous Care Record (CCR), Clinical Docu- To determine the relationship between the probability of ment Architecture (CDA) or Continuous Care Docu- a hospital engaging in information exchange and EHR mentation (CCD). Also included is a variable that Vendor market share we used a logistic regression asks if the hospital’s EHR allows sending clinical care model. The basic bivariate model between the dependent summaries to unaffiliated hospitals using a different variable Information Exchange (IE) and our variable of EHR vendor (CS). interest Vendor Market Share (VMS) Concentration is Other hospital descriptive indicators, which have been represented by eq. 1. found significant in the literature, such as hospital size (Size), ownership (Ownership), rural or urban loca- PIðÞ EjVMS tion (Rural) and hospital HHI , are also included. logitðÞ IEjVMS¼ log 1−PIðÞ EjVMS Hospital HHI (HospitalHHI) at the regional level is ¼ β þ β VMS ð1Þ calculated by weighting hospital market participation in 0 1 an HRR with hospital size, using total beds as a proxy. To determine health system affiliation, we used the IDS indicator from the HIMSS analytics database and pIðÞ EjVMS¼ logistic equation Castillo et al. BMC Health Services Research (2018) 18:405 Page 5 of 12 We then added other explanatory variables found graphic representation of the logistic regression results in the literature to reduce possible omitted variable can be seen in Fig. 1, which presents the odds ratio re- bias (eq. 2). sults of the logistic regression with error bars represent- ing a 95% confidence interval. Additionally, in an effort to control for the different state level characteristics that px ^ðÞ might influence the likelihood of hospital sharing we log ¼ η^ðÞ IE ð2Þ 1−px ^ðÞ used a state fixed effects (Table 1) again finding the same positive relationship between VMS and information ex- change. The location of the hospital was determined by ^ ^ ^ ^ ^ the provided zip code address as HRRs are regions of η^ðÞ IE ¼ β þ β VMS þ β CL þ β CS þ β 0 1 2 3 4 service provision and therefore are not always within ^ ^ RHIO þ β RHIO þ β Ownership PART 5 6 state boundaries. ^ ^ ^ þ β Rural þ β Size þ β 7 8 9 When we control for State fixed effects, the estimated System Hospital þ β HospitalHHI þ ε odds of a hospital sharing clinical care summaries out- side their system is 5.4 times greater if all hospitals in We also include state fixed effects to control for local the HRR use the same EHR Vendor than the corre- characteristics that might impact information exchange sponding odds for a hospital in an area with no hospitals and analyzed the characteristics for the largest market using the same EHR Vendor. We include dummy vari- players by including dummy variables. ables for the ability to send documents in CCR or CCD For our second analysis we looked for an association format, if an EHR system allows for sending summary of between vendor concentration and the percentage of care records to another EHR vendor and the availability hospitals that exchange clinical care summaries within a of RHIOs. These three variables are significant in in- state. To test this association, we used a multivariate lin- creasing the likelihood of sharing clinical summaries; ear regression model represented by eq. 3. nevertheless, the effect of Vendor Concentration re- mains large in comparison. ^ ^ ^ The results remain stable as we include other control %IE ¼ β þ β VendorHHI þ β SystemHHI 0 1 2 variables that have been found relevant in the literature þ β HospitalsState þ ε ð3Þ such as ownership (non-profit versus for-profit), rural To this model we added dummy variable indicators versus urban location, normalized hospital size, hospital for state level policies to incentivize health information market concentration, and system affiliation. Of these exchange. only rural status and system affiliation were not statisti- cally significant. Results For-profit hospitals are found to be less likely to share Vendor market share and hospital information exchange information, which is consistent with the results found Logistic hospital level regression by Adler-Milstein and Jha  who hypothesize that a We find that for our 2013 dataset there was a positive hospital’s strategic decision not to participate in informa- relationship between the likelihood of sharing and a hos- tion exchange is an effort to minimize costs. We also pital’s EHR vendor market share within a HRR. A find that the measure of hospital market concentration Fig. 1 Odds Ratio for Independent Variables Predicting Probability that a Hospital “Shares Clinical Care Summary Outside their Health System” with error bars for a 95% interval Castillo et al. BMC Health Services Research (2018) 18:405 Page 6 of 12 Table 1 Adjusted odds ratio for hospitals with dependent Table 3 Differences in hospital characteristics of hospitals for variable “Shares Clinical Care Summary Outside their Health Epic, Meditech and Cerner System” with state fixed effects All EHR EPIC Meditech Cerner vendors Hospital shares clinical summary N = 1871 N = 407 N = 360 N = 338 Variables Odds ratio 95% confidence interval *** *** *** Hospital Small 47% 33% 44% 33% Vendor market share 5.37 (3.29, 8.80) size (< 100 beds) *** CCR or CCD (YES) 3.19 (2.00, 5.27) * *** *** Medium 40% 44% 51% 49% *** Allow other EHR Vendor (YES) 1.90 (1.47, 2.45) (100–399 beds) *** *** *** RHIO 1.23 (0.90, 1.67) Large 13% 23% 5% 18% *** (> = 400 beds) RHIO participation 1.56 (1.19, 2.04) *** *** ** Ownership Non-Profit 71% 88% 69% 78% Non-Profit ownership 0.69 (0.51, 0.94) *** *** ** For-Profit 8% 1% 14% 7% For-Profit ownership 0.57 (0.33, 0.96) *** ** *** Public 21% 11% 17% 15% Rural 0.84 (0.63, 1.11) Note: Using t-test for equality of means the significance levels for a two-tailed Number of beds 3.28 (0.94, 11.6) tests are *p < 0.1; **p < 0.05; ***p < 0.01 System hospital 0.87 (0.68, 1.11) Hospital HHI 0.62 (0.23, 1.67) Our results from the logistic regression show that * ** *** Note: p < 0.1; p < 0.05; p < 0.01 hospital size is positively related to information ex- change while being a For-Profit hospital is negatively is negatively related to the probability of participating in related to this variable. In Table 3 we see that hospitals the exchange of clinical care summaries, which suggests that use Epic as their EHR vendor are significantly larger that hospitals in more concentrated markets are less and less likely to have a For-Profit ownership model. likely to exchange information. This is consistent with being more likely to share infor- mation. The opposite is true for hospitals that use Medi- Differences between specific vendors tech, which are significantly less likely to be large and Three EHR vendors, Epic, Meditech and Cerner, to- more likely to be For-Profit. gether control 58% of the hospital market in our sam- We expect that these three vendors have the potential ple data. Of the pool of non-federal acute hospitals of exploiting the network effects of market concentra- that responded to the survey question, 39% had shared tion because of their large number of users. We ran clinical care summaries with outside hospitals. We find separate regressions to test the interactions between the that hospitals that use Epic exchange clinical care sum- main EHR vendors and the variable of interest. From maries significantlymorethanthe totalaverage,while Fig. 2 we find that there are important differences in the hospitals that use Meditech or Cerner do so signifi- coefficient of the key independent variable Vendor cantly less (Table 2). Thetypeof hospitals that chosea Market Share for each of the different EHR vendors. specific vendor also varies between the different EHR Although hospitals using Epic start with a higher pre- vendors (Table 3). dicted probability of sharing, the increase of market share in the HRR has an important positive effect. A similarly positive effect is found for hospitals using Table 2 Percent of hospitals that share clinical care summaries Meditech as their EHR Vendor. However, for hospitals outside their health system for the seven largest vendors that use Cerner find there are negative effects of having Vendor % share clinical n(N) p value care summary (two-tailed) other hospitals with the same vendor in the HRR. This suggests additional non-measured difficulties in informa- EPIC 73% 296(407) p < 0.01 tion exchange for Cerner users. Meditech 27% 97(360) p < 0.01 Cerner Corporation 32% 109(338) p < 0.01 Vendor concentration at the state level McKesson 30% 58(191) p < 0.01 Percent sharing within state and EHR vendor concentration CPSI 30% 48(160) p < 0.05 Our second analysis examines at the market dynamics of Siemens 40% 41(102) EHR vendors at the state level and state policies to incentivize information exchange. The percentage of Allscripts 26% 19(74) p < 0.05 sharing varies widely across states, with Florida, Illinois, *Using t-test for equality of means n = Number of hospitals that share clinical care summaries with hospitals Missouri, New Mexico, Oklahoma, Tennessee and Texas outside their system sharing significantly less than the global mean of 37% N = Total number of hospitals that use each EHR Vendor included in the database and that responded to the variable of analysis (see Additional file 1). The differences across states have Castillo et al. BMC Health Services Research (2018) 18:405 Page 7 of 12 Fig. 2 Predicted Probability that a Hospital “Shares Clinical Care Summary Outside their Health System” for Each of the Three Largest Vendors with 95% confidence interval been attributed to factors such as state-level privacy Our independent variable of interest, Vendor HHI, is regulation and information security practices [17, 35]. positively related to the percent of hospitals that partici- Another possible explanation is the different strategic pate in information exchange within a state. In our sam- approaches for information exchange prompted by the ple, as the market concentration of EHR Vendors in a incentives received through the State Health Information state increases there is also an increase in the percent of Exchange Cooperative Agreement Program. We find no hospitals that exchange clinical care summaries. This support for different strategies accounting for different value remains significant as we include control variables levels of sharing. However, this study does find that such as the market concentration of hospitals within a these differences could also be explained in part by system (column 2), the availability of Directed Exchange differences in the market concentration of EHR Vendors or Query Exchange within a state (column 3), and, the across states. Figure 3 shows the relationship between strategies used by the state as part of the State Health Vendor HHI and the total sharing within a state with Information Exchange Cooperative Agreement Program different colors for the dominant vendor in each state. (column 4). In this figure, the size of the point is proportional to In column 2 we see that including the System HHI percent market share of dominant vendor and trend variable has an effect on the marginal value of our key lines indicate linear relationship between Vendor HHI variable of interest. This result is consistent with the and Percent of Hospitals that Share Information. We fact that hospitals within a health system are likely to find that the three trend lines for each vendor mirror use a unique vendor and that the initial Vendor HHI the relationship found at the hospital level. However, effect might be related to health system concentration. because of the smaller sample at the state level for each Nevertheless, even correcting for this possible omitted vendor this relationship is only statistically significant variable bias, the Vendor HHI remains positive and for states in which the dominant vendor is Epic. significant. The negative nature of the coefficient on Our results show that there is a relationship between System HHI is consistent with previous research  EHR vendor concentration in a state and the percentage which showed that states with larger networks domin- of hospitals that participate in the exchange of clinical ating the market have a lower percentage of hospitals care summaries within a state. Table 4, column 1 shows that participate in information exchange outside their the result of the base bivariate linear regression model. health system. Castillo et al. BMC Health Services Research (2018) 18:405 Page 8 of 12 Fig. 3 State Sharing versus Vendor HHI with dominant EHR vendor Discussion outside their system 23% assert that they cannot ex- Through this analysis we have found empirical evidence change with hospitals using a different EHR vendor (des- that, among other factors, vendor market share and pite the fact that only 10% of hospital EHR systems concentration are related to the likelihood of a hospital don’t support CCD or CDA exchange standards), sharing clinical care summaries and the percentage of suggesting that exchange in this subgroup is happening hospitals within a state that exchange such information. directly between hospitals using the same EHR vendor. These factors remain important even when we take into Although we cannot conclude from the available data if account policies that incentivize information exchange exchange for the rest of the sample is taking place such as the requirement for the use of standardized through proprietary or standards-based methods, we can formats (CCR, CDA and CCD) and State level programs. presume that there are benefits to exchanging using While the capability to use a common format to send proprietary methods that are strengthened when the clinical care summaries is significant in increasing the vendors are more concentrated. These benefits may likelihood of participating in the exchange of these docu- include reduced technical difficulty and ease of access to ments, this ability is not enough to guarantee exchange specific interfaces, which might remain influential even outside a hospital’s network. In fact, 72% of hospitals if a hospital is technically able to exchange using stand- that do not share clinical care summaries with other ard formats. vendors are capable of using these common formats. When we control for each of the three largest EHR Furthermore, almost 30% of hospitals that can use CCD Vendors in the market we find relevant differences in and CDA continue to claim that they are not capable of the propensity for information exchange. We analyze the exchanging clinical care summaries with hospitals using interactions with these EHR vendors in our sample and a different certified EHR vendor. This supports the find that the positive relationship between HIE and notion that even when EHR systems are certified to market share is very strong for hospitals that use Epic. comply with this MU requirement, exchange with out- Hospitals using EHR vendor Epic are much more likely side vendors remains a challenge. to exchange clinical care summaries than the rest of the In this context, EHR vendor market share and concen- hospitals in our sample. Conversely, hospitals that use tration become relevant topics of analysis. Of the hospi- Meditech and Cerner are less likely to exchange this tals that exchange clinical care summaries with hospitals type of information. By promoting proprietary sharing, Castillo et al. BMC Health Services Research (2018) 18:405 Page 9 of 12 Table 4 State level linear regression with dependent variable “Percentage of Hospitals in State that Share Clinical Care Summaries” Dependent variable: Percentage that Share Clinical Summary (1) (2) (3) (4) ** *** *** *** Vendor HHI 0.349 0.657 0.636 0.639 (0.168) (0.188) (0.193) (0.202) ** ** *** System HHI − 0.714 − 0.900 − 0.948 (0.242) (0.292) (0.366) No. of Hospitals −0.209 −0.347 (0.145) (0.172) QE Statewide −0.045 (0.048) DE Statewide −0.016 (0.065) Elevator 0.073 (0.073) Public Utility 0.042 (0.083) Capacity Builder 0.116 (0.091) Orchestrator 0.081 (0.065) *** *** *** *** Constant 0.280 0.289 0.410 0.308 (0.061) (0.058) (0.099) (0.099) Observations 49 49 49 49 R 0.08 0.23 0.27 0.30 Adjusted R 0.06 0.20 0.19 0.19 Residual Std. Error 0.17 (df = 47) 0.15 (df = 46) 0.16 (df = 43) 0.16 (df = 41) ** *** ** ** F Statistic 4.3 (df = 1; 47) 6.9 (df = 2; 46) 3.3 (df = 5; 43) 2.6 (df = 7; 41) * ** *** Note: p < 0.1; p < 0.05; p < 0.01 larger players strengthen the network externality benefit by facilitating intra-vendor sharing in an effort to enlist of choosing an EHR from a large player. From these new users interested in sharing within its existing results we can infer that the availability of Epic’s Care network. Smaller practices and hospitals interested in Everywhere has important implications for hospitals exchanging clinical care summaries with larger hospitals looking to participate in information exchange. In fact, that use said EHR vendor would need to join the Epic becomes an interesting case study for the effects of network. The decision to choose a specific EHR product having a proprietary network for health information involves a lock-in factor because of the sizeable costs of exchange. Our analysis shows that Epic users might implementation. Not only does this make it unlikely that overcome some of the barriers for information exchange smaller hospitals could then change to a different when other hospitals in the same region use Epic. vendor, it may involve additional unforeseen costs that However, when there are no users nearby that use this could discourage them from implementing usable same EHR vendor the net benefits for exchange are exchange capabilities [10, 36]. diminished. This suggests that when removing the in- We have similar results at the state level. We find that centive of a geographically close Epic user for exchange, higher Vendor HHI is positively correlated with the per- additional customization could act as a deterrent for centage of hospitals within the state that share informa- developing further HIE capabilities [27, 28]. tion, even when controlling for different policies that Due to the competitive nature of the EHR market, a incentivize or hinder information exchange. The different larger player such as Epic could leverage its network size strategies applied through the State Health Information Castillo et al. BMC Health Services Research (2018) 18:405 Page 10 of 12 Exchange Cooperative Agreement Program (State HIE) do removed from the dataset. Both of these issues would not show a significant effect on the percentage of hospitals likely result in an overestimate of our measure of inter- that exchange clinical care summaries within a state. operability. Additionally, although we aimed to include Hence, in states with highly concentrated markets mea- most variables relevant to our analysis, there are other sured by the Herfindahl-Hirschman Index (where one or factors related to health information exchange that we two EHR vendors are used by the majority of the hospi- were not able to quantify for this analysis. For example, tals) there are more hospitals engaging in information we are not able to measure different security or privacy exchange. policies for different vendors that might facilitate or Part of the objective of the State HIE program was to deter information exchange. Similarly, although research fill HIE service gaps and build capacity for every eligible has found a relationship between state privacy policies provider . The fulfillment of this goal could be an and state information exchange practices, we were not important contribution toward overcoming some of the able to include a measurement of privacy legislation in limitations of vendor facilitated exchange and the pos- this study. It is possible that including indicators for sible failures of closed proprietary networks. Unfortu- state privacy regulation would have accounted for lower nately, our current research shows that none of the state levels of information exchange. Third, we were only able level strategies seem to be successful in reducing this to infer that EHR vendors in our analysis use proprietary effect. In states where there are less concentrated methods for exchange because we do not have detailed markets, none of the different implementations were information on the methods of information exchange for significant in incentivizing exchange. This might be a each hospital. Therefore, if a large percentage of hospi- symptom of misaligned incentives, as there have been tals are exchanging information through non-vendor reports of current regulation undermining the role of mediated methods or regional health information ex- community health information exchanges supported by changes, it is possible that some vendors offer an advan- State HIE by allowing EHR vendor mediated exchange tage for this type of sharing. Finally, all of our results that cuts out public exchanges . show association and not causality because of the nature As more hospitals transition to the second stage of of the sample and the method. meaningful use, data from recent years shows that simi- lar challenges for HIE persist. While the percentage of Conclusions hospitals that report that they have the capability to send Identifying the barriers for information exchange is a ne- clinical care summaries has increased, the percentage of cessary step to achieve the goals of the HITECH Act in hospitals that send them during transitions of care creating a more efficient and effective healthcare system. remains low. Data from Meaningful Use attestations Our research finds a relationship between the existence between 2014 and 2016 shows that a median hospital of dominant EHR networks and the exchange of clinical sends clinical care summaries electronically for 33% of care summaries, which has important policy implications transitions, while the use of Epic as an EHR provider as the meaningful use program continues to transition positively increases this probability . Furthermore, to future stages. In fact, there is some evidence that in- qualitative work evidences that the number of EHR formation blocking could be partly the result of vague providers in the market, and the need for different inter- policies that undermine public exchanges. faces to exchange clinical information between them, is Even though the current certification process for EHR still reported as an important barrier for HIE . A products requires the use of a common language, there recent survey of third party HIE organizations supports are several gaps that permit variability in its implemen- the issues of vendor information blocking, with half of tation. These gaps allow EHR vendors to implement those surveyed reporting that they had experienced information exchange capabilities in different ways. A information blocking by an EHR vendor . Finally, clear example is the implementation of Care Everywhere, vendor choice remains an important determinant in the which has been successful in increasing sharing among successful implementation of MU objectives [11, 39]. Epic users. Nevertheless, the existence of isolated net- works means that many hospitals are left out. In the Limitations case of Epic, this affects smaller and rural hospitals There are some important limitations to our results. disproportionally (only 21% of hospitals that use EHR First, data from the AHA IT Supplement is self-reported vendor Epic are rural which is significantly less than and has limited representativeness with a self-selected the sample mean). sample of 61% of the population. While this database In order to avoid proprietary exchange networks that has been validated for reliability against other sources, it foreclose some hospitals, it is important for the current does show some bias toward over reporting . It also regulation attempt to be more inclusive of hospitals that includes some responses that are inconsistent and were do not use large vendors and are therefore unable to use Castillo et al. BMC Health Services Research (2018) 18:405 Page 11 of 12 proprietary methods for exchange. Incentives could be Authors’ contributions AC and MS worked on the conception, design and acquisition of data. AC, tied to open exchange using previously defined stan- MS and AD worked on the analysis and interpretation of data. AC drafted dards rather than metrics that just measure if HIE oc- the manuscript. All authors read and approved the final manuscript. curs. For state level incentives, it might be necessary Ethics approval and consent to participate that state programs identify hospitals that are being left This research did not involve any human subjects, human material, or out of the exchange networks and offer technical and fi- human data. nancial support. In our analysis at the state level we find Data from the American Hospital Association Information Technology Survey Supplement Database used in this study was licensed through a Data no significant relationship between the percentage of License Agreement with Health Forum, LLC, a wholly-owned subsidiary of hospitals that participate in health information exchange the American Hospital Association, and is available for purchase through this and the policies implemented through the State Health entity. The data used in this study does not contain any personal identifiable information. Information Exchange Cooperative Agreement Program. Our research suggest that future state level policies Competing interests should take into account the different market conditions The authors declare that they have no competing interests. of EHR vendors in order to accommodate hospitals that may be left out of large proprietary networks. Publisher’sNote Finally, although our findings suggest the importance Springer Nature remains neutral with regard to jurisdictional claims in of a network where information is exchanged only published maps and institutional affiliations. among hospitals that use a specific EHR vendor within a Received: 19 April 2018 Accepted: 23 May 2018 region and a state, further research is necessary to valid- ate this relationship. Current information collection efforts only ask if information exchange occurs. More References 1. Furukawa MF, King J, Patel V, Hsiao C-J, Adler-Milstein J, Jha AK. work needs to be done to determine the methods of Despite substantial progress in EHR adoption, health information exchange, including interviews with hospital staff that exchange and patient engagement remain low in office settings. Health might give us some insight on if and why proprietary Aff. 2014;33:1672–9. [cited 2018 Mar 30]; Available from: https://doi.org/ 10.1377/hlthaff.2014.0445. methods of exchange are being used. 2. Institute of Electrical and Electronics Engineers (IEEE). Standards Glossary. [cited 2018 Mar 30]. Available from: https://www.ieee.org/education_ careers/education/standards/standards_glossary.html. 3. Kellermann AL, Jones SS. What it will take to achieve the as-yet-unfulfilled Additional file promises of health information technology. Health Aff. 2013;32:63–8. [cited 2018 Mar 30] Available from: https://doi.org/10.1377/hlthaff.2012.0693. Additional file 1: Additional Descriptive Statistics. Table A contains 4. Bell B, Thornton K. From promise to reality: achieving the value of an EHR. descriptive statistics of the data used and Table B contains state level Healthcare Financial Management. 2011;65:51–6. [cited 2018 Mar 30] data for information exchange. (DOCX 28 kb) Available from: http://link.galegroup.com/apps/doc/A252005722/ITOF?u= cmu_main&sid=ITOF&xid=0695c2e4. 5. Department of Health and Human Services. State Health Information Exchange. 2014 [cited 2018 Mar 30]. Available from: https://www.healthit. Abbreviations gov/policy-researchers-implementers/state-health-information-exchange. AHA: American Hospital Association; CCD: Continuity of Care Document; 6. Centers for Disease Control and Prevention. Meaningful Use. 2012 [cited CCR: Continuity of Care Record; CMS: Centers for Medicare and Medicaid 2018 Mar 30]. Available from: http://www.cdc.gov/ehrmeaningfuluse/ Services; EHR: Electronic Health Record; HHI: Herfindahl-Hirschman Index; introduction.html. HIMSS: Healthcare Information and Management Systems Society; 7. Department of Health and Human Services. Health information technology: HITECH: Health Information Technology for Economic and Clinical Health; standards, implementation specifications, and certification criteria for HRR: Hospital Referral Region; IDS: Integrated Delivery System; IE: Information electronic health record technology, 2014 edition; revisions to the Exchange; MU: Meaningful Use; ONC: Office of the National Coordinator for permanent certification program for health information technology. Federal Health Information Technology; RHIO: Regional Health Information Register United States of America; 2012 p. 54163–54292. [cited 2018 Mar Organizations; VMS: Vendor Market Share 30]; Available from: https://federalregister.gov/a/2012-20982. 8. Vest JR, Yoon J, Bossak BH. Changes to the electronic health records market in light of health information technology certification and meaningful use. J Funding Am Med Inform Assoc. 2013;20:227–32. [cited 2018 Mar 30] Available from: The authors acknowledge funding received for PHD research from the https://doi.org/10.1136/amiajnl-2011-000769. Government of Ecuador, which did not have any role in the design, collection, 9. Department of Health and Human Services. Test Procedure for §170.304 (i) analysis, and interpretation of data; in the writing of the manuscript; or in the Exchange Clinical Information and Patient Summary Record. 2010 [cited decision to submit the manuscript for publication. The authors would also like 2018 Mar 30]. Available from: https://www.healthit.gov/topic/certification- to acknowledge the contribution from the Dorenfest Institute for Health ehrs/testing-process-test-methods. Information who provided the HIMSS analytics database used for this analysis. 10. Vest JR, Kash BA. Differing strategies to meet information-sharing needs: publicly supported community health information exchanges versus health systems’ enterprise health information exchanges. Milbank Q. Availability of data and materials 2016;94:77–108. [cited 2018 Mar 30] Available from: https://doi.org/10. The data that support the findings of this study are available from the 1111/1468-0009.12180. American Hospital Association but restrictions apply to the availability of 11. Holmgren AJ, Patel V, Adler-Milstein J. Progress in interoperability: these data, which were used under license for the current study, and so are measuring US hospitals’ engagement in sharing patient data. Health Aff. not publicly available. Data are however available from the authors upon 2017;36:1820–7. [cited 2018 Mar 30] Available from: https://doi.org/10.1377/ reasonable request and with permission of American Hospital Association. hlthaff.2017.0546. Castillo et al. BMC Health Services Research (2018) 18:405 Page 12 of 12 12. Adler-Milstein J, Desroches CM, Furukawa MF, Worzala C, Charles D, Kralovec P, 33. Department of Health and Human Services. State HIE Implementation Stalley S, Jha A. More than half of US hospitals have at least a basic EHR, but Status. 2014 [cited 2018 Mar 30]. Available from: https://www.healthit.gov/ stage 2 criteria remain challenging for most. Health Aff. 2014;33:1664–71. [cited sites/default/files/statehietypologybrief.pdf. 2018 Mar 30] Available from: https://doi.org/10.1377/hlthaff.2014.0453. 34. Dullabh P, Hovey L, Ubri P. Evaluation of the state health information 13. Kruse CS, Regier V, Rheinboldt KT. Barriers over time to full implementation exchange cooperative agreement program. 2013 [cited 2018 Mar 30]. of health information exchange in the United States. JMIR Medical Available from: http://healthitgov.ahrqstg.org/sites/default/files/ Informatics. 2014;2:e26. [cited 2018 Mar 30] Available from: https://doi.org/ casestudysynthesisdocument_2-8-13.pdf. 10.2196/medinform.3625. 35. Adjerid I, Acquisti A, Telang R, Padman R, Adler-Milstein J. The impact of privacy regulation and technology incentives: the case of health 14. Wu H, Larue E. Barriers and facilitators of health information exchange (HIE) information exchanges. Manag Sci. 2016;62:1042–63. [cited 2018Mar30] adoption in the United States, System Sciences (HICSS), 2015 48th Hawaii Available from: https://doi.org/10.1287/mnsc.2015.2194. International Conference on System Sciences; 2015. p. 2942–9. [cited 2018 36. Yeager VA, Walker D, Cole E, Mora AM, Diana ML. Factors related to health Mar 30] Available from: https://doi.org/10.1109/HICSS.2015.356. information exchange participation and use. J Med Syst. 2014;38:78. [cited 15. Gold M, Mclaughlin C. Assessing HITECH implementation and lessons: 5 2018 Mar 30] Available from: https://doi.org/10.1007/s10916-014-0078-1. years later. Milbank Q. 2016;94:654–87. [cited 2018 Mar 30] Available from: 37. Lin S, Everson J, Adler-Milstein J. Technology, incentives, or both? https://doi.org/10.1111/1468-0009.12214. Factors related to level of hospital health information exchange. Health 16. Adler-Milstein J, Jha AK. Health information exchange among U.S. hospitals: Serv Res. 2018; [cited 2018 Mar 30] Available from: https://doi.org/10. who’s in, who’s out, and why? Healthcare. 2014;2:26–32. [cited 2018 Mar 30] 1111/1475-6773.12838. Available from: https://doi.org/10.1016/j.hjdsi.2013.12.005. 38. Adler-Milstein J, Pfeifer E. Information blocking: is it occurring and what 17. Miller AR, Tucker C. Health information exchange, system size and information policy strategies can address it? Milbank Q. 2017;95:117–35. [cited 2018 Mar silos. J Health Econ. 2014;33:28–42. [cited 2018 Mar 30] Available from: https:// 30] Available from: https://doi.org/10.1111/1468-0009.12247. doi.org/10.1016/j.jhealeco.2013.10.004. 39. Holmgren AJ, Adler-Milstein J, McCullough J. Are all certified EHRs created 18. Vest JR. More than just a question of technology: factors related to hospitals’ equal? Assessing the relationship between EHR vendor and hospital adoption and implementation of health information exchange. Int J Med meaningful use performance. J Am Med Inform Assoc. 2017;25:654–60. Inform. 2010;79:797–806. [cited 2018 Mar 30] Available from: https://doi.org/10. [cited 2018 May 22] Available from: https://doi.org/10.1093/jamia/ocx135. 1016/j.ijmedinf.2010.09.003. 40. Everson J, Lee S-YD, Friedman CP. Reliability and validity of the American 19. Vest JR, Campion TR, Kaushal R. Challenges, alternatives, and paths to Hospital Associations national longitudinal survey of health information sustainability for health information exchange efforts. J Med Syst. 2013;37: technology adoption. J Am Med Inform Assoc. 2014;21:e257–63. [cited 2018 9987. [cited 2018 Mar 30] Available from: https://doi.org/10.1007/s10916- Mar 30] Available from: https://doi.org/10.1136/amiajnl-2013-002449. 013-9987-7. 20. Adler-Milstein J, DesRoches CM, Jha AK. Health information exchange among US hospitals. Am J Manag Care. 2011;17:761–8. 21. Department of Health and Human Services. REPORT TO CONGRESS - Report on Health Information Blocking. 2015 [cited 2018 Mar 30]. Available from: http://www.healthit.gov/sites/default/files/reports/info_blocking_040915.pdf. 22. Sittig DF, Wright A. What makes an EHR “open” or interoperable? J Am Med Inform Assoc. 2015;22:1099–101. [cited 2018 Mar 30] Available from: https:// doi.org/10.1093/jamia/ocv060. 23. Damore JD, Mandel JC, Kreda DA, Swain A, Koromia GA, Sundareswaran S, et al. Are Meaningful Use Stage 2 certified EHRs ready for interoperability? Findings from the SMART C-CDA Collaborative. J Am Med Inform Assoc. 2014;21:1060–8. [cited 2018 Mar 30] Available from: https://doi.org/10.1136/ amiajnl-2014-002883. 24. Everson J, Adler-Milstein J. Engagement in hospital health information exchange is associated with vendor marketplace dominance. Health Aff. 2016;35:1286–93. [cited 2018 Mar 30] Available from: https://doi.org/10. 1377/hlthaff.2015.1215. 25. Shapiro C, Varian HR. Information rules: a strategic guide to the network economy. Boston: Harvard Business School; 1999. 26. Jacob JA. On the road to interoperability, public and private organizations work to connect health care data. JAMA. 2015;314:1213–5. [cited 2018 Mar 30] Available from: https://doi.org/10.1001/jama.2015.5930. 27. Garber S, Gates SM, Keeler EB, Vaiana ME, Mulcahy AW, Lau C, et al. Redirecting innovation in US health care: options to decrease spending and increase value. Rand Health Q. 2014;4:3. 28. Koppel R, Lehmann CU. Implications of an emerging EHR monoculture for hospitals and healthcare systems. J Am Med Inform Assoc. 2015;22: 465–71. [cited 2018 Mar 30] Available from: https://doi.org/10.1136/ amiajnl-2014-003023. 29. The Dartmouth Atlas Working Group. The Dartmouth Atlas of Health Care in the United States. American Hospital Association; 1996 [cited 2018 Mar 30]. Available from: http://www.dartmouthatlas.org/tools/glossary.aspx. 30. Department of Justice. Herfindahl–Hirschman Index. 2015 [cited 2018 Mar 30]. Available from: http://www.justice.gov/atr/herfindahl-hirschman-index. 31. Department of Health and Human Services. What is a regional health information organization (RHIO)?. Health Resources and Services Administration 2007 [cited 2016 Mar 21]. Available from: https://www. ruralhealthinfo.org/toolkits/networks/4/health-information-exchange-rhios. 32. Adler-Milstein J, Landefeld J, Jha AK. Characteristics associated with regional health information organization viability. J Am Med Inform Assoc. 2010;17:61–5. [cited 2018 Mar 30] Available from: https://doi.org/ 10.1197/jamia.M3284.
BMC Health Services Research – Springer Journals
Published: Jun 4, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
All the latest content is available, no embargo periods.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud