Laboratory Testing

Six studies addressed laboratory testing, with five finding a benefit of HIE in reducing overall tests, although estimates of impact on cost were mixed.^{44,46,50,52,70,71} Four of these studies took place in the ED setting, all showing some aspect of reduced testing and cost savings. Two studies found overall reduced laboratory testing, with one reporting an odds ratio (OR) of testing among patients for whom HIE was accessed to be 0.880 (95% confidence interval [CI], 0.828 to 0.935)⁴⁴ and the other noting 23 percent fewer lab testing procedures (statistical significance not reported) in a propensity-matched group of patients for whom HIE could have been used.⁵² A third study logged physician self-reports of laboratory testing averted with use of HIE in the ED, with savings over 3 months of $462 calculated from tests reportedly not ordered.⁷⁰ A fourth study found 96 instances of duplicate lab testing averted in 1,488 patient encounters that were retrospectively analyzed.⁷¹Two studies were conducted in ambulatory settings, against a backdrop of increased overall laboratory testing. One U.S. study found that after HIE implementation, there was a reduction in the rising rate of testing, without overall cost savings.⁵⁰ In contrast, a study in Finland found increased laboratory testing during the period of HIE implementation (19.0% for primary care physicians and 7.0% for specialist physicians per total patient appointments).⁴⁶ As with all retrospective studies, the four studies of laboratory testing could have been complicated by confounders, while the prospective study did not validate physician self-reporting of tests avoided or measure overall costs of care for the ED encounter or subsequent utilization.

Radiology Testing

Nine studies assessed radiology testing, with all but one reporting an association of reduced testing with HIE.^{39,40,44,46,50,52,69-71} Six of these studies also examined laboratory testing and are described previously,^{44,46,50,52,71,75} and three additional ED studies assessed only imaging.^39,40,69

The ED studies showed a variety of findings. One study found that for all radiologic imaging, there was reduction of head computed tomography (CT) imaging, (OR of 0.913, 95% CI, 0.842 to 0.991) as well as body CT imaging (OR 0.886, 95% CI, 0.828 to 0.948) but no significant changes in echocardiogram, chest x-ray, or ankle x-ray testing across 12 EDs.⁴⁴ Another study demonstrated 22 percent decreased diagnostic radiology ordering and 52 percent reduced CT scan ordering (statistical significance not reported) when HIE was used in the ED.⁵² Two additional studies assessed neuroimaging for headache³⁹ and repeat imaging for back pain in EDs.⁴⁰ For neuroimaging, HIE usage was associated with decreased diagnostic imaging (OR 0.38; 95% CI, 0.29 to 0.50) and increased adherence to evidence-based guidelines (OR 1.33; 95% CI, 1.02 to 1.73), although there was no significant change in overall costs. HIE usage was associated with reduced repeat imaging for back pain (OR 0.36; 95% CI, 0.18 to 0.71), but no change in cost due to higher use of CT scans with HIE access. A prospective case series study reported $161K in savings over 3 months through averted radiologic testing in EDs,⁷⁰ while a retrospective case series found 453 duplicate radiology testing in 1,488 patient encounters retrospectively analyzed.⁷¹

One cross-sectional study looked at repeat imaging in the ED in two states (California and Florida), finding reduced probability of repeat CT (-8.7%; 95% CI, -14.7% to -2.7%), ultrasound (-9.1%; 95% CI, -17.2% to -1.1%), and chest x-ray (-13.0%; 95% CI, -18.3% to -7.7%) ordering in hospitals that had HIE participation as reported in the Healthcare Information and Management Systems Society Analytics Database of hospital information technology (IT) functionality.⁶⁹

In ambulatory settings, one U.S. study showed no statistically significant reduction in the rate of radiologic testing.⁵⁰ However, a Finland-based study showed a reduction in radiologic testing (16.4% reduction for primary care physicians and 11.0% reduction for specialist physicians).⁴⁶

Hospital Admissions

Eight studies assessed the role of HIE in reducing hospital admissions, with inconsistent findings.^{41,44,53,54,56,65,70,72} Two studies (described above) found a reduction in hospital admissions and lower costs using methods previously described. The bulk of the $1.07 million annual savings due to HIE found in one study resulted from reduced admissions.⁴⁴ Another study also reported $118K in savings from averted admissions over a 3-month period.⁷⁰ Two studies in an Israeli health maintenance organization found that viewing the medical history via an electronic health record (EHR) decreased possibly redundant admissions, with even greater reductions when information was accessed using HIE.^41,72 A study in New York found that viewing information reduced odds of admission (OR 0.70; 95% CI, 0.52 to 0.95).⁵⁶

Other studies, however, found no benefit from HIE in terms of avoiding hospital admissions. An RCT of directed HIE in Canada providing family physicians electronic reports of ED visits versus paper-based reports resulted in no difference in hospital admissions or return visits to the ED.⁶⁵ Other studies found that HIE was associated with increased admissions for ambulatory-sensitive diagnoses⁵⁴ and a 28 percent increased rate of admissions, although such admissions had reduced length of stay with 771 fewer bed days per 1,000 health plan members over 16 months.⁵³

Two studies assessed HIE in reducing hospital readmissions. One study found that assessing information in an HIE implementation was associated with reduced odds of hospital readmission (OR 0.43; 95% CI, 0.27 to 0.70)⁵⁵ while another found that U.S. hospitals participating in HIE in 2007 did not have lower readmission rates for acute myocardial infarction, pneumonia, or heart failure.⁶⁸

Referrals and Consultations

Two studies, described previously, assessed HIE for reducing referrals and/or consultations. The prospective ED case series reported reduced consultations, leading to savings of $3,990 over 3 months.⁷⁰ The Finland-based ambulatory study, however, found that HIE was associated with increased referrals by primary care physicians (43.6%) and specialists (12.8%).⁴⁶

ED Cost

Another two studies addressed reducing overall ED costs per patient, with both finding reductions when HIE was available. One study found that an HIE system encompassing 12 EDs resulted in net annual savings (total savings minus operating costs) of $1.07 million, with reduced hospital admissions accounting for 97.6 percent of the reduction.⁴⁴ Another study found that for a propensity-matched group of patients for whom HIE could have been used, the group for whom HIE was used had $29 per ED visit less expenditures.⁵² Neither study reported overall ED expenditures, making it unknown what proportion of overall ED spending was impacted by HIE.

Public Heath Reporting

Three studies assessed HIE in public health settings, all of which were conducted in the United States.^42,47,49 Two examined the completeness of notifiable disease reporting data. One study compared usual (“spontaneous”) public health reporting with automated lab reporting through the HIE, finding a 4.4-fold higher rate of reporting for the HIE-based approach, with cases identified an average of 7.9 days earlier.⁴⁹ The other study showed equal or improved completeness of reporting for a variety of data fields in notifiable disease reports, although completeness was reduced for some fields (e.g., laboratory units of measure, normal range, and abnormal flag) due to inadequacies in the clinical data entering the HIE.⁴² Another study found that a public health HIE led to increased identification of needed followup care of 419 HIV patients and 85 percent of them having actual followup care.⁴⁷

Quality of Care

Three studies looked at the value of HIE in improving quality of care in ambulatory settings.^45,48,67 One study assessed a benchmark group of clinical quality measures believed to be amenable to HIE usage among users and nonusers of an HIE portal. Users of the portal had a higher proportion of physicians exceeding mean clinical quality measure performance at baseline (57% vs. 48%) that increased after the HIE became available (64% vs. 49%), with the increase for portal users before and after availability of the HIE statistically significant (p<0.001).⁴⁵ An RCT of HIE data used in a clinical decision support intervention was able to detect medication adherence problems in eight categories of drugs but did not show any benefit in improving adherence by patients in taking medications prescribed based on evidence-based guidelines.⁶⁷ Another study of six physician practices found improved documentation and delivery of preventive services for mammography screening (21.1% to 57.1%, p<0.01), colonoscopy screening (31.7% to 53.8%, p<0.01), pneumococcal vaccine administration (39.1% to 50.6%, p<0.01), and influenza vaccine administration (22.7% to 41.7%, p<0.01).⁴⁸ The study also found that medication reconciliation completion improved from 35.3 percent to 44.9 percent (p<0.001).

Other Aspects of HIE

Three studies assessed other aspects of HIE. One study found a 30 percent reduction in evaluation time for Social Security Disability claims.⁴³ Another found that HIE data led to a 20.3 percent increase in identifying frequent ED users compared with site-specific data.⁵¹ An additional study focused on hospital-based HIE, finding that communication and satisfaction (based on the Hospital Consumer Assessment of Healthcare Providers and Systems survey) were higher in hospitals that implemented HIE compared with those that proposed to implement HIE.⁶⁴

Although the risk of bias in most studies was low, the resulting evidence from them was mostly of low quality. This low strength evidence mostly favored the value of HIE in reducing resource use and costs, especially in the ED. However, these studies used mostly retrospective designs that cannot account for how HIE was used and its impact on the overall care of the patient beyond the immediate setting where it was used.

Participation in HIE, Types of Data Exchanged, Characteristics of Successfully Participating Organizations (United States–Wide Surveys)

Six studies used survey methods to investigate the frequency of data exchange and types of data exchanged across regional health information organizations (RHIOs) across the United States (Table 4).^25,78-81,87 Across these studies, between 138 and 207 organizations met the definition of a RHIO; while between 20 and 81 RHIOs provided data. These data, collected from 2006 through 2012, suggest that entities most commonly providing data are hospitals (83%), followed by ambulatory settings (60%); and that the entities most commonly receiving data were ambulatory settings (95%), followed by hospitals (83%), public health departments (50%), and payers (44%).⁸¹ Using survey data collected in 2007, Hessler, et al. focused on the exchange between RHIO and State and local public health departments, and found that of 138 public health agencies, 50 (36%) had no RHIO in their jurisdiction; 16 (12%) had no relationship with a RHIO, and 26 (40%) were exchanging information. Twelve of 20 RHIOs were exchanging information; seven of these (35%) with public health entities.⁸⁷ The types of data most frequently exchanged were laboratory test results (84% to 90%),^78,81,87 inpatient data (70%), medication histories (70%), and outpatient data (60%).^78,81 In 2008 and 2009, of 75 operational RHIOs, covering 14 percent of U.S. hospitals and 3 percent of ambulatory practices, only 13 supported the criteria for meaningful use criteria of the Health Information Technology for Economic and Clinical Health Act (3% of hospitals and <1% of ambulatory practices),⁷⁹ while by 2012, there had been a 61 percent increase in the number of operational RHIOs, from 75 to 119.²⁵

Table 4

Level of use and primary uses of HIE: participation in HIE, types of data exchanged, and characteristics of successfully participating organizations (United States–wide studies).

Two additional surveys were conducted by the eHealth Initiative ^73,74 One-hundred, ninety-nine of 315 identified HIE organizations completed the 2013 annual survey. These HIE entities were a mix of community-based, State-based, and health care delivery organizations. Results indicate there is no single dominant model of HIE. Ninety organizations use a ‘Direct’ standards-based protocol for securely exchanging data, mostly for transitions in care. Patient opt out was the most common consent model, although patient engagement remains low amongst organizations exchanging data. Eighty-four organizations had reached an advanced stage of operation or innovation; most took 2 years to become operational. Among organizations that responded in both 2011 and 2013, 27 more had reached stages 5 (operating), 6 (sustaining), or 7 (innovating) on the eHealth Initiative's maturity scale, in 2013. Hospitals and ambulatory care providers are the stakeholders most commonly providing/viewing data; independent laboratories also commonly provide data. Community and public health clinics commonly view data. HIE organizations are focusing on functionalities to support health care reform initiatives and advanced analytics.

The number of HIE organizations identified and that responded in 2014 was lower than in 2013, with 126 of 267 identified responding in 2014.⁷⁴ Again, there was a mix of community-based, State-based, and health care delivery organization-based HIE entities responding. Data were provided by hospitals, ambulatory care providers, laboratories, and community/public health clinics. Data were accessed by ambulatory care providers, hospitals, community/public health clinics, and behavioral or mental health providers. Findings suggest an 11 percent increase over 2013 in the proportion of organizations that have reached stage 6 (operating) or higher (106 organizations). Uses of HIE included support for an accountable care organization to improve patient outcomes, for a patient centered medical home, for a State Innovation Model, and for a bundled payment initiative. Results suggest data exchange is reaching a point of stability and acceptance, and that organizations are settling on a set of core services offerings.²⁶

Nine studies investigated HIE use retrospectively, using U.S.-wide survey data collected for other purposes, with an information technology add-on.^{26,91,107-111,113} Four of these used data from the AHA,^{26,107,108,111} two from the National Ambulatory Medical Care Survey, (NAMCS),^91,110 and one each from the Commonwealth Fund Health Policy Surveys,¹⁰⁹ the National Nursing Home Survey/National Survey of Residential Care Facilities,¹¹¹ and, another from the National Survey of Residential Care Facilities.⁹³

These studies investigated overall participation in HIE use. Results suggest that HIE use by hospitals has risen from 11 percent (2009)⁷⁸ to between 30 percent and 58 percent more recently.^108,109,111 Results from the recently released ONC brief suggest that more than three-quarters (76%) of non-Federal acute care hospitals electronically exchanged laboratory results, radiology reports, clinical care summaries, and/or medication lists with any outside providers in 2014. This represents an 85 percent increase since 2008 and a 23 percent increase since 2013. Close to seven in 10 hospitals (69%) electronically exchanged health information with ambulatory providers outside of their organization, representing a 92 percent increase since 2008 and a 21 percent increase since 2013.²⁶ Characteristics associated with higher use are nonprofit status, presence of an EHR system, larger market share, and larger practices.^107-109,111 Results from the NAMCS (2011) suggest that the majority of office-based physicians reported being able to both send and receive data; 64 percent of these exchanges were through an EHR vendor and 28 percent through a hospital system. Activities included viewing laboratory results and incorporating these into the EHR, and exchanging clinical summaries with patients. Primary care providers were more likely to use HIE than specialists.⁹¹ Results from the NAMCS (2013) suggest that 39 percent of office-based physicians reported having HIE capability with other providers or hospitals. Characteristics of higher HIE use were larger practice size, practice owned by a health-system (vs. physician owned), and multispecialty (vs. single specialty) practice.¹¹⁰ Data from the National Nursing Home Survey (2004) and the National Survey of Residential Care Facilities Survey, both from the Centers for Disease Control and Prevention, indicate that HIE use in these settings is low, with the consistent pattern of nonprofits enjoying wider use than for-profit entities.¹¹³ Finally, recent data from the National Survey of Residential Care Facilities suggest that 23 percent of residential care communities that use EHRs also exchanged health information. Nearly 25 percent could exchange with pharmacies and 17 percent with physicians.⁹³

Transfer of Records Between Integrated Delivery Systems

The VA and DoD use the Virtual Lifetime Electronic Record (VLER) system for eHealth exchange with the private sector, in the Nationwide Health Information Network (NwHIN) – a ‘network of networks’. This is a federated, pull (query-based) model for transfer of records between integrated delivery systems, using an opt in consent approach by patients. The NwHIN allows users to pull in data from other organizations (Table 5). In an early study, Bouhaddou et al. investigated the transfer of records across three integrated delivery systems in San Diego, California; the VA, DoD, and Kaiser Permanente Southern California. They found that 264 of 363 of patients (73%) who opted in and provided valid authorization could be correlated across integrated delivery systems.⁸² In a recent, much larger study, Byrne et al. enrolled 12 sites. Of the 64,237 veterans who provided authorization and opted in, less than 0.01 percent opted in and subsequently opted out. The proportion of data matched between exchange partners ranged from 12 percent to 88 percent. The highest matching rates were accomplished using social security numbers in the matching algorithm. Data were retrieved for 2,724 unique VA patients with the exchange partner, and for 1,764 unique VA providers reviewing exchange partner data.¹¹⁶

Table 5

Level of use and primary uses of HIE: transfer of records between integrated delivery systems.

Participation in HIE and Extent of Use: Regional or Statewide Initiatives

Nine studies described the use of HIE in the State of New York. Five of these used audit logs,^{45,97,98,102,106} two used surveys,^76,77 one used a database of clinical data,¹¹² and one geo-coding¹²⁵ (Table 6). Most of the HIE implementations are query-based. The studies of audit logs indicate frequent queries,^97,98 and an increasing proportion of physicians accessing HIE over time (33% to 43% over 18 months).⁴⁵ Separately, of 63,305 patients enrolled from three hospitals, an average of 238 clinical event alerts were provided per day to notify ambulatory care providers of inpatient or ED admissions for their patients; a total of 42,818 events were detected over a 6-month timeframe.¹⁰⁶ Primary HIE users varied by study. In one study, primary users were non-clinical staff in the outpatient setting and clinicians in the inpatient setting,⁹⁷ while in another, 86 percent of sessions were with staff in an ED.¹⁰²

Table 6

Level of use and primary uses of HIE: participation in HIE and extent of use, by regional or statewide initiatives.

Abramson et al. conducted three statewide surveys in New York, two in 205 hospitals⁷⁶ and the other in 632 nursing homes.⁷⁷ In each, they investigated participation in HIE and the exchange of data. In hospitals, their results suggest that between 2009 and 2012 the percent of respondent hospitals participating in HIE and exchanging data, increased from 23 percent to 79 percent. In 2012, institutions exchanged data more frequently with other hospitals (71%) and ambulatory care providers (69%), than with long-term care facilities (45%) and home health agencies (38%).⁹⁶ Among nursing homes 54 percent participate in HIE, with 31 percent of providers exchanging information outside the system. HIE use was highest when nursing homes had an EHR. The types of data exchanged were pharmacy (42%), labs (39%), and hospital data (39%). The seventh study was a retrospective database analysis of clinical data that described a geriatric care coordination program that used a Clinical Event Notification system to request information from nursing homes when patients were seen in the ED.¹¹² The authors suggested that use of the Clinical Event Notification functionality may have facilitated avoidance of 18 percent of hospital admissions, as these admissions lasted less than 48 hours. As not all studies described the type of HIE in detail, we were unable to draw any conclusions based on the type of HIE utilized. Finally, using a novel study design, Onyile et al. estimated the proportion of patients in the New York Clinical Information Exchange (now Healthix) system by mapping the most current zip code for each patient to the appropriate U.S. county. They found that 88 percent of patients in the system live within 30 minutes of New York's Times Square.¹²⁵

A series of five studies investigated HIE use in a query-based Central Texas HIE. I-Care is an HIE implementation comprised of hospital systems, public and private clinics, and governmental agencies operating federally qualified health centers.^{54,101,103-105} Four of these studies were conducted across several facility member sites, with a fifth study across two sites.¹⁰¹ For adult patients seen in the ED, use was low; in 57 percent of patients⁵⁴ and only 2.3 percent of encounters.¹⁰⁵ In a subset of two sites that did not have an EHR (but that mandated use of the HIE), the HIE was accessed in 21 percent of the encounters.¹⁰¹ Across these studies, HIE use was higher for those with a greater number of ED visits and hospitalizations,^54,101,105 older age, a greater number of chronic conditions,^101,105 females, and those with fragmented care.¹⁰¹ HIE use was lower for blacks and Hispanics, visits for alcohol use, injury, poisoning, an unfamiliar patient, and a busier than average day.¹⁰⁵ Similar results were found in the study that focused on children seen in the ED; use was greater for those less than 1 year old, who had more frequent encounters in the past, and a greater number of diagnoses. Use was lower if the patient was unfamiliar, or if the day was busier than average.¹⁰⁴ In a companion study that investigated how use of HIE varies by job type and organization in an indigent care setting, Vest et al. found that the most frequent users were those whose positions were administrative, followed by social services, physicians, nurses, public health professionals, and pharmacy professionals. The hospital was the workplace for 50 percent of users, followed by adult ED, ambulatory care, public health agency, mental health agency, and children's ED. Most clinical access took place in the ED and in public/mental health agencies. In the majority of use sessions, users accessed the system in a minimal fashion; almost all use was administrative.¹⁰³

Of the five studies conducted in the MSeHA, based in Memphis, Tennessee, three used audit-logs,^40,99,118 one was a cross-sectional survey,⁸⁶ and one used mixed methods.¹¹⁹ MSeHA is an HIE implementation that facilitates data exchange across EDs and community-based ambulatory clinics. In 2007, across these studies, HIE use was low, being used for 12.5 percent of the study population.⁴⁰ In another, HIE was viewed in the ED for between 3 percent and 10 percent of visits.⁹⁹ In a third, HIE was used for only 15 percent of return ED visits and 19 percent of return clinic visits; yet users reported the HIE provided additional information about histories and prevented repeat tests or procedures.¹¹⁸ In the separate cross-sectional survey of 151 users, 43 percent reported using HIE less than 1 hour per week, 39 percent between 1 and 4 hours, and 18 percent, greater than 4 hours per week.⁸⁶ In a separate study of workflow, nurses accessed HIE when prompted by patients about a recent hospitalization, while providers accessed HIE for reasons beyond simply identifying a recent hospitalization. HIE access occurred at various points of care. Workflow patterns evolved over time, due to revisions in access policies and staffing changes.¹¹⁹ Across these studies, use was higher when the HIE was accessed by nurses and clerks versus physicians.^99,118

Separately, Dixon et al. conducted an online survey of 63 infection preventionists in six states with HIE, to gauge the awareness and engagement of these preventionists in using HIE for public health surveillance. One-half of their respondents were unaware of their organization's involvement in HIE, and only 10 percent reported their organizations used the HIE.⁸³

Nine additional studies describe HIE use at the State-level, two studies each from Indiana and Minnesota, and one each from Wisconsin, North Carolina, Massachusetts, Northeastern Ohio, and Louisiana.^{84,85,88,90,92,100,120,123,124} These studies used data from 2005¹²³ through 2013.⁹⁰ Methods of data collection included surveys,^{84,85,88,90,92,120} interviews,^85,124 focus groups,^123,124 and audit-logs.^100,120 Each study makes a useful contribution to the HIE literature.

In an Indiana study of a coordinated antibiotic-resistance infection tracking, alerting, and prevention system, of the several thousand patients for whom email alerts were sent, approximately one-quarter were identified as having had documentation in a different hospital system of a previous infection with methicillin-resistant staphylococcus aureus or vancomycin-resistant enterococcus. Capture of this type of laboratory data was found useful.⁸⁸ Other Indiana investigators found real-time alerting helpful in prompting followup,⁹² as did investigators in Louisiana.¹²⁴ Patients were generally accepting of data sharing, as long as patient benefit was evident.¹²⁴ In a study of small practices (<20 physicians) in Minnesota, results revealed that no practice was fully involved in a regional HIE and that HIE was not part of most practices' short-term strategic plans.⁸⁵ In a study more recently conducted in Minnesota, intended to monitor progress toward meeting the legislative requirement that all health care providers have an interoperable EHR by January 2015, investigators found that over one-half of respondents exchanged data with affiliated or unaffiliated hospitals.⁹⁰ The Tripathi et al. study was unique in that researchers conducted focus groups with patients who lived in three communities that piloted the Massachusetts HIE. All three communities agreed to share all EHR data except text notes, consult letters, and scanned reports. Consumer opt in was the preferred consent method, as it is in VLER. Strategies identified to drive consumer opt in included educating patients and providers about the enhanced convenience and lower costs of HIE.¹²³ Lobach et al. investigated the impact of the HIE on sentinel events for Medicaid patients in Durham County, North Carolina. In an analysis of almost 12,000 patients enrolled, they found that 19 percent experienced a sentinel event over a 6-month period. They concluded that the HIE was useful in population health management using HIE.¹⁰⁰ In a description of HIE implementations in Wisconsin, Foldy found that 78 percent (21 of 27) of organizations had HIE projects, some operational, others planned. Most were surveillance systems, delivering data to central registries, but a growing number served clinicians and patients.⁸⁴ Kaelber et al. investigated HIE use in the Northeast Ohio Public health care system, Care Everywhere. Of the 18 percent (74 of 412) of physicians who responded to the survey, approximately one-third of ED physicians, one-fifth of primary care physicians, and one-tenth of specialty care physicians used HIE. Use was highest when patients were older, with more comorbidities, Medicare/Medicaid insured, or black.¹²⁰ These results reflect the variation in the implementation and impact of HIE, providing data that are not necessarily generalizable to other settings. These data suggest that small practices are not adopting HIE, while larger health systems are. They further suggest that HIE may be useful in exchanging data in the ED, and for surveillance of infectious diseases, that patients and providers view HIE favorably, and that patients can and do ”buy-in” to the concept of HIE when the benefits are evident.

Extent of Use, Types of Information Exchanged, and Adoption in International or Multinational Settings

Six studies that evaluate the use of HIE in non-U.S. settings met our inclusion criteria, one in Australia,¹¹⁴ one in South Korea,⁸⁹ one in Scotland,¹²² one in England,¹²¹ two in Finland^61,115 (Table 7). Three multi-country studies,^94,95,117 two that included data from the United States,^95,117 comprise the last three studies in this group. Lee et al. found that the data most commonly transmitted differed by setting. From the hospital it was working diagnosis; from the clinic, it was clinical findings. The most useful data were laboratory or imaging data.⁸⁹ Silvester and Carr found that commitment and interest in adoption increased over time.¹¹⁴ Mäenpää et al. also found a steady increase in uses over time by physicians, nurses and administrative staff. ¹¹⁵ Maass et al. conducted a unique time-motion study of HIE-facilitated care of 20 diabetic patients, and found that of 20 visits, four involved use of HIE, with one facilitating a faster treatment decision and three providing access to the most recent test results.⁶¹ Investigating use in the National Health System in Scotland¹²² and England,¹²¹ Pagliari and Greenhalgh, respectively, both found use to be relatively low, although Pagliari's study is now older (2004). Finally, Jha et al. assessed HIE adoption by physicians and hospitals in six developed countries (United States, United Kingdom, Canada, Germany, the Netherlands, Australia, and New Zealand), and reported varying results, but they did find generally low use due to a variety of identified barriers that prevented fuller adoption. In the United States, fewer than 12 percent of organizations were exchanging data on less than 1 percent of involved populations.¹¹⁷ In a more recent study conducted in Australia, Canada, France, Germany, the Netherlands, New Zealand, Norway, Switzerland, the United Kingdom, and the United States, Schoen found that the percent of primary care physicians reporting HIE capabilities ranged from a low of 14 percent in Canada to a high of 55 percent in New Zealand; use in the United States was reported to be 31 percent.⁹⁵ In a study that included the 27 European Union countries plus Croatia, Iceland, Norway, and Turkey, Codagnone used a factor analysis to create a composite metric that ranged between 0 and 4 to measure the extent of exchange of health information.⁹⁴ The metric suggested low to moderate use, with an average score across the 31 countries of 1.88. These early reports suggest that HIE in developed countries was in the initial stages of use in the early years of the 21^st century, and is increasing slowly over time.

Table 7

Level of use and primary uses of HIE: extent of use, types of information exchanged, and adoption in international or multinational settings.

HIE Satisfaction

Satisfaction with HIE, a measure of usability, was examined in one cross-sectional study¹³⁴ and one before-after study ¹³⁸ One additional cross-sectional study that stratified satisfaction by types of HIE is described later.⁵⁸ Using a pre-post survey study design (n=29), physicians at one clinic and five AIDS service organizations in North Carolina reported increased satisfaction after the Carolina HIV Information Cooperative (CHIC) RHIO was implemented.¹³⁸ Participants reported improved satisfaction with ease of data exchanged and improved patient care after using CAREWare software. The respondents also perceived that CAREWare was a good use of resources. They also reported improved relations with HIV care partners after implementing the RHIO. By contrast, before implementation, the providers had high expectations for how exchanging information would affect their work and reported some unmet expectations afterward.

In a second study on satisfaction of HIE users in Adelaide, South Australia,¹³⁴ users who embraced the use of the data exchange integration tools were significantly more likely to rate integration higher than those who were not using it as often (p<0.001). This result echoes a more recent study that found frequent users are more pleased with the usability of an HIE system than infrequent users.⁸⁶ The response rate for the Massy-Westropp study was 24 percent (55 of 132). While both satisfaction studies^134,138 provide descriptive evidence from surveys that users were satisfied with usability, neither provided sufficient details in the methods sections to eliminate bias or a comparison that would enable generalization.

Usability of HIE by Type

We also examined whether certain functionality (direct exchange or push vs. query-based exchange) was more usable. Directed exchange is provider-to-provider electronic exchange of patient information to coordinate care.³² In this type of exchange, the data are electronically sent to the recipient's EHR or clinical inbox.⁵⁸ In query-based systems, the user accesses an exchange system, queries for information (e.g., ED, hospital admissions, or discharges) on a particular patient and pulls data from multiple health care organizations.⁵⁸ This is important particularly for unplanned care (e.g., patient comes into the ED).¹³ We also attempted to evaluate usability by type of architecture (e.g., whether the query-based system used a centralized or federated model). However, few publications provided this level of technical detail to make a comparison. Additionally, the authors used a variety of terms and descriptions which made it difficult to classify usability by architecture. When the authors provided detail on architecture, it was included in Table 8.

Only one cross-sectional study evaluated clinician satisfaction with exchanging health information using query-based (pull) or direct exchange (push).⁵⁸ In this comparison study, clinicians had access to “pushed” health data (laboratory and radiology) through certified EHRs; physicians who ordered tests could designate other physicians to receive the test results. The physicians in this study could also query (pull), using a secure web portal, for test results, patient demographics and transcribed reports provided by physicians, hospitals, laboratories and radiology centers across the greater Buffalo and Rochester areas of New York. More providers reported using electronically pushed data exchange (80%) than pulled exchange of health information (53%). A greater proportion of physicians reported using pushed data exchange always or most of the time (68%) compared with pulled exchange (19%, p=0.001). The physicians were more satisfied when data were pushed than pulled (p<0.05).

In summary, we found insufficient data to compare usability by type or architecture of the electronic data exchange.

Addressing Lack of Critical Mass

Concern was expressed in several studies about the need for a critical mass of users and populated patient information.^{82,116,127,129,130,132,135,136} Underlying reasons for lack of critical mass can include several reasons (e.g., the providers aren't electronically exchanging the data or patients have not consented). Patients concerned with privacy and security may not understand the benefits and/or may not consent to have their data shared with other providers. Even when they do consent, they may not be properly matched to existing data.¹³² Also, match rate can vary by population and setting; for example, the match rate for providers practicing in a homeless center was lower, but the match rate for ED physicians was higher.¹³² Some contributing providers reported legal concerns for sharing patient data and may choose to not participate. The end result was that providers searching for patient information may grow frustrated at taking the time to search and stop using the system.

To increase the critical mass, several approaches have been suggested. These include addressing concern about privacy, careful consideration about the consent process, and a process for educating patients.^{58,82,116,118,127,134,138} To address patient and provider concern about privacy, create clear understanding about privacy and data sharing among all stakeholders (providers, patients, nonclinician partners) prior to implementation.¹³⁸ In planning for electronic health exchange, several authors noted the importance of deciding whether to have opt out or opt in consent process for patients.^{58,82,99,116,118,119,127,129,136} Of veterans interviewed, 90 percent were positive about the VLER HIE system. At the same time, 81 percent felt each person should have a choice to opt in and the default should not be automatic participation.¹¹⁶ Opting in protocols seem to yield a high patient participation rate (93% to 97%).^58,127,136 When age is considered, older patients opt in more often than younger patients.⁸² The percentage of consented patients can be increased with a workflow that includes front staff members being trained to educate and consent patients as they first arrive.¹²⁷ Additionally, patient awareness of provider use of the HIE may increase patients perception of the benefits of electronically exchanged data. Patients in the VA reported being unaware that providers were using the VLER system to access information outside of the VA.¹¹⁶ The authors noted that the user interfaces of the VLER are not visible by patients because the display faces the providers. We identified one organization that used an opt out protocol (MSeHA).^118,119 Patients had the option to opt out at every encounter. The opt out rate was 1 to 3 percent,¹¹⁸ which is slightly better than programs with an opt in protocol that lose 3 to 7 percent of patients who do not consent.^58,136

Addressing Inefficient Workflow in Electronically Exchanging Data

Often the workflow was inefficient to providers attempting to exchange health information.^{62,116,118,127,128,130,136,138} Users complained that additional logins and policies against proxy users increased the time the provider needed to access the patient information.

Sites with proxy users (registrars, nurses, clerks, and other physicians) who accessed the system and then provided the information to the attending physician had the highest access rates.^99,118 Proxy use was described as a way to save provider time or address needs of limited users without privileges.^130,132,136 Additionally, some organizations made it difficult to get privileges to access exchanged data so those with privileges were called upon to look up information for those without.¹³⁰

An ethnographic qualitative study of the MSeHA identified two role-based workflow models: physician-based and nurse-based.¹¹⁹ These investigators completed 180 observation hours of six EDs and eight ambulatory clinics using the MSeHA exchange system, informational interviews during observation, and nine semi-structured interviews. In the nurse-based model, if a patient mentioned a recent hospital visit, the triage nurse or medical assistant would search for data primarily looking for summary documents related to recent hospital visits, such as a discharge summary, but rarely searched for other medical history. The nurse then printed off the information for use by the provider. In the provider-based model, physicians and nurse practitioners searched for electronically exchanged information for more reasons than hospital visits. These providers browsed online medical history for purposes of decisionmaking. Finally, another study of the MSeHA reported that use dropped significantly after a new policy prohibited registrars from searching the system at the start of a visit.¹¹⁸ Initially registrars would print off a summary sheet of available data. Providers then queried the system, based on the summary sheet. When a new policy came in place prohibiting registrars and nursing team members from accessing the system for security reasons, use dropped significantly.

During implementation several other strategies were mentioned related to changing current workflow: providing training and enough technical support to support the new workflow,^86,116 addressing needed culture change,¹³⁰ and having champion users.^99,127 One physician expressed in an interview that exchanging data is a change from practice. Physicians “get bogged down [with exchanging health information] and just want to see patients”.¹³⁰ Introducing new technology requires addressing the need for change and the resistance that may exist. These studies also encouraged sites to manage expectations upfront¹³⁸ and have a pilot implementation prior to launch so users aren't disappointed.^118,132

Addressing Poorly Designed Interface and Update Features

Several design features of the HIE created barriers to use.^{62,116,127,128,136} While HIE users understood why textual notes were not exchanged for confidentiality reasons, this lack of context made the information less valuable.¹³⁶ While some users wanted more information, other users wanted shorter reports to avoid having to scroll up and down, click on many pages or go to another task. Some complained that the exchange contained too much information that was not filtered enough to be meaningful for providers.^127,128 They reported that reading a paper report was much faster than reviewing the exchanged information.¹²⁸ This finding was echoed by another study that recommended the main findings should be sent first in a brief report.⁶² The design features could be addressed better at the implementation phase by including more providers during the design phase.¹²⁷ Another facilitator is to continually monitor the quality and usability of the exchanged data to meet standards and the needs of the users.¹¹⁶ Similarly, as more patient data and more types of data were exchanged, users reported that their system response slowed suggesting the need to continually review (and reduce) what was being exchanged.¹¹⁶

Some users expressed concern with how quickly the patient information was updated and found it more efficient to go directly to the partnering clinic or hospital for information than to rely on current information in the exchange.¹²⁸ Systems that automatically integrate with the providers' EHRs may reduce this concern and also reduce need for users to have to login into multiple systems.^62,130

Facilitators

Seven of the eight categories of facilitators for implementation identified in the literature (below) are predominately “internal” factors, concerned with the characteristics of the HIE or its components, while only one category, external policy, addresses the environment for the HIE.

Barriers

Barriers to HIE implementation cited in the research are not simply the inverse of the facilitators. While there is some overlap in the categories, the barriers cited include more external, environmental factors. The seven categories of barriers are included in Table 11.

Subgroup Differences

During our review we attempted to abstract data from the included studies that would allow us to determine if the barriers and facilitators to implementation varied by type of HIE, health care settings, and systems or IT system characteristics. Most publications did not include this information so we were not able to consistently identify any differences.

We also considered that implementation might change over time as HIE becomes more common and as new HIE efforts could benefit from the experience of early adapters. At this time we do not see any notable changes, but this may be to the relatively short time period (less than a decade) covered by the included studies. While the hardware and software that make HIE possible have changed significantly in less than a decade, organizational change and clinical practice patterns have historically changed more slowly.