Despite The Spread Of Health Information Exchange, There Is Little Evidence Of Its Impact On Cost, Use, And Quality Of Care

Transcription

1 By Saurabh Rahurkar, Joshua R. Vest, and Nir Menachemi Despite The Spread Of Health Information Exchange, There Is Little Evidence Of Its Impact On Cost, Use, And Quality Of Care doi: /hlthaff HEALTH AFFAIRS 34, NO. 3 (2015): Project HOPE The People-to-People Health Foundation, Inc. ABSTRACT Health information exchange (HIE), which is the transfer of electronic information such as laboratory results, clinical summaries, and medication lists, is believed to boost efficiency, reduce health care costs, and improve outcomes for patients. Stimulated by federal financial incentives, about two-thirds of hospitals and almost half of physician practices are now engaged in some type of HIE with outside organizations. To determine how HIE has affected such health care measures as cost, service use, and quality, we identified twenty-seven scientific studies, extracted selected characteristics from each, and metaanalyzed these characteristics for trends. Overall, 57 percent of published analyses reported some benefit from HIE. However, articles employing study designs having strong internal validity, such as randomized controlled trials or quasi-experiments, were significantly less likely than others to associate HIE with benefits. Among six articles with strong internal validity, one study reported paradoxical negative effects, three studies found no effect, and two studies reported that HIE led to benefits. Furthermore, these two studies had narrower focuses than the others. Overall, little generalizable evidence currently exists regarding benefits attributable to HIE. Saurabh Rahurkar is a doctoral candidate in health care organization and policy at the School of Public Health, University of Alabama at Birmingham. Joshua R. Vest is an assistant professor in the Department of Public Health, Division of Quality and Medical Informatics, at Weill Cornell Medical College, in New York City. Nir Menachemi (nmenachemi@ iu.edu) is a professor in and chair of the Department of Health Policy and Management, Richard M. Fairbanks School of Public Health, at Indiana University, in Indianapolis. Widespread health information exchange (HIE), defined as the transfer of electronic health information such as laboratory results, clinical summaries, and medication lists among organizations according to nationally recognized standards, 1 is believed to boost efficiency, reduce health care costs, and improve outcomes. 2 Thus, policy makers have sought to make the US health care system more patient centered and value based by providing various financial incentives to promote HIE adoption. 3 The Health Information Technology for Economic and Clinical Health (HITECH) Act of 2009 provided financial incentives through its meaningful use program to providers that participated in HIE. It also provided funds to states to support HIE activities. Additionally, HIE among various stakeholders is considered critical for the success of accountable care organizations, patient-centered medical homes, and bundled payment initiatives fostered by the Affordable Care Act (ACA). As a result, by 2013 approximately two-thirds of US hospitals and nearly half of physician practices were engaged in some type of HIE with outside organizations. 4,5 This represents a sharp increase in the adoption of HIE from 2008, when HIE participation among hospitals 5 was 41 percent and among physicians 4 was about 17 percent. Two systematic reviews have examined literature focused on HIE. Specifically, Patricia Fontaine and coauthors 6 reviewed studies relat- March :3 Health Affairs 477

2 ed to adoption of HIE in primary care practices emphasizing the benefits, barriers, and value to the practice. Their review included sixty-four articles, of which thirty-nine were peer reviewed. Only three articles examined HIE and outcomes. The authors concluded that improved access to outside test results and the facilitation of referrals were the benefits supported by the limited empirical evidence in the primary care setting. Following this, Ana Hincapie and Terri Warholak 7 reviewed five published studies with a focus on HIE and health care outcomes. Their study found some evidence for cost savings in the emergency department (ED). Most of the studies contained in both systematic reviews had been published prior to 2009, when the HITECH Act was signed into law. Given the recent growth of HIE among hospitals and other providers, the purpose of this study is to systematically review the contemporary literature examining the relationship between HIE and health care measures, such as costs, use of services, and quality. To expand upon the previous systematic reviews, we are interested in analyzing how the use of HIE in any health care setting (ED, hospital, primary care) affects any health care measure. Overall, our study adds value by identifying gaps in the literature, synthesizing and critically appraising the nature of existing studies, and recommending an agenda to guide future research. Study Data And Methods Search Strategy Our systematic review follows a protocol consistent with the Preferred Reporting Items for Systematic Reviews and Meta- Analyses (PRISMA) guidelines where appropriate. 8 We comprehensively searched the PubMed and Scopus databases (from January 1980 to May 2014) with HIE-related search terms including electronic health information exchange, health information exchange, health information interchange, electronic document exchange, information exchange, and electronic data exchange, together with search terms related to effectiveness: outcome evaluation, impact, effect, association, correlation, assessment, influence, relationship, examination, evaluation, and ramifications. Our search terms are consistent with previous work in this area that used similar terms to identify articles related to HIE. 6,7,9 Only empirical publications appearing in peer-reviewed English language journals were included. To focus on evidence-based empirical studies, we excluded letters to the editor, policy briefs, executive summaries of governmental reports, commentaries, summaries of future research plans, and nonpeer-reviewed Internet-based publications. The keyword search identified 274 articles, which were then subjected to the process outlined in online Appendix Exhibit A1. 10 The articles were first reviewed based on the title and abstract.we eliminated studies that did not focus on HIE and studies that did not evaluate HIE and the subsequent outcomes that result from HIE use. Two reviewers individually reviewed the articles to decide which studies were to be included with differences in judgment reconciled by consensus. In this first step, there was a deliberate attempt to be as inclusive (sensitive) as possible, and fifty-six articles (20 percent) were selected for further evaluation. Articles were included if they evaluated systems that involved electronic exchange of any patient-level clinical data including lab results, clinical summaries, and medication lists between organizations. Next, the resulting articles were reviewed with a deliberate focus on specificity that is, identifying all articles that both had an HIE focus and examined some discrete health care measure empirically. We refer to these measures as outcomes, given that they were operationalized by previous authors as the consequence or resulting effect of HIE.We recognize that our use of the term outcome is different from the Donabedian context, 11 but we do so in order to reduce confusion. At this step, we eliminated thirty-five more articles, bringing the number of included articles to twenty-one. Finally, to be as comprehensive as possible, we examined the reference lists of included studies for additional articles possibly missed by the database search. For each new article identified, we continued to examine reference lists until no new articles were found. At the end of this process, a total of twenty-seven articles were included in our analysis (see Appendix Exhibit A2). 10 Research Data Set The twenty-seven articles included studies that consisted of ninety-four individual analyses, given that several articles analyzed more than one outcome that fit our inclusion criteria. An analysis was considered discrete for each unique outcome evaluated (such as repeat diagnostic tests, imaging tests, and health care costs). Analyses involving identical outcomes but at different units of analysis (different hospitals, or different departments or wards in a hospital) were also considered discrete. For example, Nurit Nirel and coauthors evaluated three different types of redundant imaging tests, one type of lab test, repeat hospitalizations, and hospital visits without an overnight stay in two wards (medical and surgical) that both received the HIE intervention (see Appendix Exhibit A2). 10 Thus, this study analyzed six outcomes in each of the two wards, resulting in 478 Health Affairs March :3

3 The current state of the literature does not provide sufficient rigorous evidence for the benefits of HIE. twelve separate analyses extracted for our study. From each included analysis, we extracted various information including country of origin, type of study design used (cohort study, crosssectional study, randomized controlled trial, quasi-experimental study), and setting and population under study (primary care physicians, hospital, ED). In the context of the HITECH Act, we noted the year of each study and dichotomized the measure as either before or after 2009, the year of the act s passage. In addition, outcomes studied in each included analysis were grouped into the following categories: disease surveillance (for example, automatic reporting of diseases requiring public health notification), coordination of care (for example, communication between a patient s different providers), health care use (for example, hospital or ED readmissions, redundant lab tests), health care costs (for example, outcomes measured in dollars), patient experience (for example, patient satisfaction), and quality-of-care measure (for example, hemoglobin A1c levels or medication adherence). Lastly, for each discrete analysis presented in each article, we determined whether or not there was a significant beneficial relationship between HIE and the outcome under study. A relationship was deemed beneficial if the study found a statistically significant positive association for positive outcomes and a negative association for negative outcomes. Analytical Approach We first examined the distributions of various characteristics of the selected studies (study location, study design, study setting, year category, outcomes analyzed, and outcome effect). Next, using chi-square analysis or Fisher s exact test as appropriate, we investigated differences in study characteristics for the analyses that found a beneficial effect and those that did not. We regressed the variable measuring a beneficial effect on each of the study characteristics that were found in bivariate analysis to be associated with a beneficial effect at p<0:20. Lastly, to account for advances in measurement of health care outcomes over time, we also controlled for year of study. In the regression analysis, we controlled for the nested nature of analyses within articles by clustering standard errors within each article. We report our findings in the form of odds ratios and marginal effects. All analyses were conducted in STATA, version 13, and statistical significance, in bivariate and regression analyses, was considered to be p<0:05. Limitations The following limitations should be noted. First, only twenty-seven articles were included in our review. We recognize that in some cases, our sample lacks the statistical power to detect differences and even limited our ability to perform more complex statistical analyses that included more covariates. However, the relationships that we identified as the salient findings of this article are robust, and our statistical methods are less sensitive to sample size than other methods are. It is also possible that our search strategy missed some articles that should have been included. However, to minimize this, we used a snowball technique to identify articles from bibliographic lists of included articles. Additionally, given the wide variability in how previous authors have described their HIE systems, we were unable to capture measures of HIE usage, mode of access (such as single sign-in or portal), or HIE mechanism (push versus pull). Finally, there is heterogeneity in the HIE evaluated by the studies as a result of variation in the outcomes, time periods, and study settings analyzed. As such, we recognize this as a limitation from a generalizability perspective. Study Results Twenty-seven published articles were included in the systematic review that included ninetyfour individual analyses. A majority of the articles focused on HIEs located in the United States (70.4 percent) and were published after 2009 (74.1 percent) (Exhibit 1). Most studies used data from EDs (51.9 percent) or hospitals (25.9 percent). In terms of study design, twenty studies (74.1 percent) used a cohort study, five studies were RCTs) (18.5 percent), one study used a quasi-experimental study design (3.7 percent), and one study used a cross-sectional study design (3.7 percent). Of the ninety-four analyses examined, the most common outcomes focused on health care use (71.3 percent), followed by health care costs (11.7 percent), quality of care (6.4 percent), coordination of care (6.4 percent), patient experience (2.1 percent), and disease surveillance (2.1 percent). A total of fifty-four analyses (57.5 percent) found a beneficial effect of HIEs on the outcome they studied, with March :3 Health Affairs 479

4 Exhibit 1 Descriptive Statistics Of Health Information Exchange (HIE) Analyses Reviewed Variable Number Percent Study location United States Other Study design Cohort study Randomized controlled trial Quasi-experimental study Cross-sectional study Study setting Emergency department Hospital HIV patients Primary care practice Provider group Outcomes analyzed Health care utilization Health care costs Quality of care Coordination of care Patient experience Disease surveillance Year categories Outcome effect Beneficial effect No effect Adverse effect SOURCE Authors analysis of published peer-reviewed studies from January 1980 to May NOTES For study design, location, and setting and for year categories, N ¼ 27. For outcomes analyzed and outcome effect, N ¼ 94 analyses percent reporting no significant effect of HIE and 10.6 percent finding an adverse effect of HIEs on health care outcomes. Bivariate relationships between finding a statistically significant beneficial result and various study characteristics are presented in Exhibit 2. Analyses focusing on a health care utilization outcome (the most frequent outcome type examined) were less likely to report a beneficial effect from HIE (50.8 percent versus 74.1 percent). On the other hand, studies based in the United States were more likely to report a beneficial effect from HIE (66.1 percent versus 42.9 percent). Lastly, while cohort studies were more likely than other study designs to report a beneficial effect from HIE (75 percent versus 26.5 percent), randomized controlled trials (17.7 percent versus percent) and quasi-experimental studies (16.7 percent versus 63.4 percent) were significantly less likely to find a beneficial effect from HIEs. Exhibit 3 shows the total number and percentage of analyses that found a beneficial effect, by outcome type and study design. Analyses using an observational design, as opposed to an experimental design, were almost always more likely to report beneficial findings for each general outcome group, including health care use (75.6 percent versus 11.5 percent) and quality of care (80.0 percent versus 0.0) (see Appendix Exhibit A3). 10 In addition, analyses with observational designs were more likely to report a beneficial effect for each individual outcome type as well (such as hospital admissions and readmissions or number of imaging tests), compared to studies with an experimental design that reported a beneficial effect. Moreover, whereas the majority of observational studies that examined quality of care (80.0 percent of five studies), coordination of care (83.8 percent of six studies), patient experience (100.0 percent of two studies), and disease surveillance (100.0 percent of two studies) reported a beneficial effect (see Appendix Exhibit A3), 10 no experimental studies on these outcomes were present in the literature. However, with respect to analyses focused on health care costs, a majority of both observational studies (55.6 percent of nine analyses) and experimental studies (100.0 percent of two analyses) reported a beneficial effect (Exhibit 3). In multivariable analysis, study design was the only characteristic that was found to be significantly associated with finding a beneficial effect of HIE (data not shown; see Appendix Exhibit A4). 10 Compared to studies with observational designs, studies with a randomized controlled trial or quasi-experimental design were still significantly less likely to report a beneficial effect (odds ratio: 0.06; 95% confidence interval: 0.02, 0.19; p<0:001). All other variables in the model were not statistically significant. Findings from all six studies that used methods that are associated with stronger internal validity (randomized controlled trials and quasi-experiments) are shown in Appendix Exhibit A5. 10 Overall, three of these studies examined ED population (Overhage JM et al.; Lang E et al.; and Hansagi H et al.); one examined hospitals (Nirel N et al.); one examined primary care practice (Tierney WM et al.); and one (Proeschold- Bell RJ et al.) examined a population of HIV patients (Appendix Exhibit A2). 10 Only two studies reported that HIE had a beneficial effect. In the first of these, Tierney and colleagues reported that access to previous test results significantly reduced the number of tests (diagnostic blood and urine tests and radiology tests) ordered at one primary care practice affiliated with a general hospital in Indiana (see Appendix Exhibit A2). 10 Additionally, charges per visit as a result of tests went down to 480 Health Affairs March :3

5 $12.17 from $14.87 in the intervention group ( 18 percent) compared to $13.99 from $15.24 in the control group ( 8.3 percent) (see Appendix Exhibit A2). 10 The second was a quasi-experimental study by Nirel and colleagues that evaluated the effect of HIE on the use of services in one health care system in Israel (see Appendix Exhibit A2). 10 That study found that HIE was associated with significantly lower diagnostic and imaging tests on medical wards but not surgical wards. Importantly, one randomized controlled trial by Lang and colleagues that examined the number of redundant referrals to specialists from EDs reported an undesirable effect from HIE (see Appendix Exhibit A2). 10 Discussion A key conclusion of our systematic review is that there is a dearth of rigorous studies that link HIE adoption to clear benefits. Moreover, the scant high-quality evidence that does exist was conducted in disparate settings and evaluated different outcomes. Prior studies that used study designs having strong internal validity (that is, capable of identifying causal relationships) were significantly less likely than studies not using such designs to find a benefit from HIEs for all outcomes except health care cost measures, and even these might not be generalizable to the United States. Two of six such studies found beneficial effects largely as a result of a reduction in diagnostic and imaging tests, associated costs, or both, and these studies were based in a single clinic affiliated with an Indiana hospital (Tierney et al.) or in one health care system in Israel (Nirel et al.) (see Appendix Exhibit A2). 10 Moreover, one additional randomized controlled trial (Lang et al.) found that HIE was paradoxically associated with an increase in duplicative specialty consultations (see Appendix Exhibit A2). 10 Thus, despite the abundance of observational studies finding a beneficial relationship between HIE and outcomes, there is currently no strong evidence to suggest that HIE is causally related to any widespread generalizable benefits. This conclusion is potentially a function of at least three possibilities. First, HIE use in the United States is still in its infancy; thus, most studies are focusing on first-generation systems and HIEs in institutions where active usage is low. Widespread meaningful use as a result of the continued rise of HIE use may allow better evaluation of HIE adoption in future systems. This argument is consistent with the benefit identified in the randomized controlled trial in Indiana, which has a more mature and evolved HIE relative to other HIEs. 12 Exhibit 2 Frequency Of Reporting A Beneficial Relationship Between Health Information Exchange (HIE) And Various Study Characteristics Beneficial relationship observed Variable Number Percent p value Outcomes analyzed Health care utilization (n ¼ 67) 34 vs vs Health care costs (n ¼ 11) 7 vs vs Quality of care (n ¼ 6) 4 vs vs Coordination of care (n ¼ 6) 5 vs vs Patient experience (n ¼ 2) 2 vs vs Disease surveillance (n ¼ 2) 2 vs vs Geographic location United States vs. other 39 vs vs Study design Cohort study (n ¼ 60) 45 vs vs < Randomized controlled trial (n ¼ 17) 3 vs vs < Quasi-experimental study (n ¼ 12) 2 vs vs Cross-sectional study (n ¼ 5) 4 vs vs Study setting Emergency department (n ¼ 51) 34 vs vs Hospital (n ¼ 23) 9 vs vs HIV clinic (n ¼ 8) 4 vs vs Primary care (n ¼ 8) 6 vs vs Multiple settings (n ¼ 4) 1 vs vs Years of study 2008 or previous vs or later 11 vs vs SOURCE Authors analysis of published peer-reviewed studies from January 1980 to May NOTES Fifty-four of ninety-four analyses (57.4 percent) found a beneficial relationship between the existence of health information exchange (HIE) and the selected characteristic. In each row, the comparison group (denoted by vs. ) refers to all studies other than the category listed. For example, under Outcomes analyzed, health care use was the outcome in sixty-seven out of the total ninetyfour analyses. Of the sixty-seven analyses examining health care use, thirty-four (50.8 percent) found a beneficial effect. In this row, the reference category refers to the categories other than health care use that is, the remaining twenty-seven analyses that focused on outcomes such as health care costs, coordination of care, patient experience, and disease surveillance, of which twenty (74.1 percent) reported a beneficial effect. Based on the sample size, Chi-square test or Fisher sexacttest,as appropriate, was used to determine significance. The p value refers to whether or not there was a significant difference between the groups being compared for example, whether studies that analyzed health care use and found a beneficial relationship between HIE and health care use were significantly different from studies that studied any of the other outcomes (health care costs, quality of care, coordination of care, patient surveillance, disease surveillance) and found a beneficial relationship. Second, rigorously studying the benefits of HIE is extremely challenging in real-world situations where randomized controlled trials are frequently not feasible. This may be attributable to both the costs involved and the limited availably of identifying suitable control institutions, physicians, or patients (depending on the unit of analysis being studied). For example, given the high costs involved in HIE implementation, identifying a suitable control group is challenging because organizations committed to HIE may be unwilling to refrain from participation and serve as a control group for evaluation purposes. 13 Furthermore, randomized controlled trials usually require informed consent, which March :3 Health Affairs 481

6 Exhibit 3 Total Number And Percentage Of Analyses That Found A Beneficial Effect Of Health Information Exchange, By Outcome Type And Study Design Outcome type Health care utilization Percent finding a beneficial effect Observational a studies Hospital admissions (n ¼ 14) Hospital readmissions (n ¼ 5) Number of imaging tests (n ¼ 12) Repeat imaging tests (n ¼ 8) Number of lab and diagnostic tests (n ¼ 10) Repeat lab and diagnostic tests (n ¼ 5) Number of ED visits (n ¼ 4) Repeat ED visits (n ¼ 1) c 0.0 Length of hospital stay (n ¼ 2) c 0.0 Outpatient visits (n ¼ 2) c 0.0 Number of appointments (n ¼ 1) c Number of referrals (n ¼ 1) 0.0 c Repeat specialty consultations (n ¼ 1) c 0.0 Ambulatory care sensitive hospitalizations (n ¼ 1) 0.0 c Experimental b studies Health care costs Patient visit costs (n ¼ 5) Annual financial savings (n ¼ 3) c Costs of lab tests (n ¼ 2) Costs of radiology tests (n ¼ 1) 0.0 c SOURCE Authors analysis of published peer-reviewed studies from January 1980 to May NOTES For health care use, N ¼ 67; 31 of 41 (75.6 percent) observational studies found a beneficial effect, and 3 of 26 (11.5 percent) experimental studies found a beneficial effect. For health care costs, N ¼ 11; 5 of 9 (55.6 percent) observational studies found a beneficial effect, and 2 of 2 (100 percent) experimental studies found a beneficial effect. In addition to the outcomes shown here, six studies analyzed quality of care. Of these, four observational studies found a beneficial effect, while one observational study and one experimental study did not. Also, five of six observational studies analyzing coordination of care found a beneficial effect. Both of the two observational studies involving patient experience found a beneficial effect, as did both of the two observational studies involving disease surveillance. A table showing the complete results is included as Appendix Exhibit A3 (see Note 10 in text). ED is emergency department. a Observational studies include cohort studies and cross-sectional studies. b Experimental studies include randomized controlled trials and quasi-experimental trials. c No studies in this group. may be complicated depending on whether an HIE uses an opt-in or opt-out approach with patients. Moreover, while randomized controlled trials have strong internal validity, results from these studies may might always be generalizable because the effects are evaluated under idealized situations and do not always represent realworld scenarios. Quasi-experimental studies offer a compromise between randomized controlled trials lack of generalizability and the weak internal validity of simpler observational cohort studies. However, quasi-experiments are rare in the literature and may be difficult to carry out on the scale at which the effects of HIE occur. As such, many researchers opt for more easily executable observational cohort studies, which have the potential to overestimate benefits as a result of selection bias and confounding (we note that forty-one out of forty-three analyses since 2009 that found beneficial results came from cohort studies). Third, although it is still premature to conclude, given the scarcity of rigorous studies, it is possible that the theoretical and expected benefits of HIE may be over- or underestimated in real-world settings. Clearly, more research is needed before conclusions can be drawn. Our analysis therefore also provides a guide for future research. Researchers need to shift focus and identify opportunities for study designs that minimize selection bias and can estimate causal pathways more reliably. The enactment of federal incentives to promote electronic health records and HIE creates opportunities for quasi-experimental studies that can include difference-indifferences studies; studies that use an instrumental variable approach; or perhaps, although less desirable, studies that reduce selection bias with the use of propensity score adjustment. Given the findings of the current study, future studies using more rigorous designs should also focus on more diverse settings. Although we recognize the particular relevance of HIE in settings where rapid access to information is of critical importance (such as EDs), 14 currently most HIE studies focus on this setting. Future studies should focus on settings other than adult hospitals and EDs such as primary care, public health, pediatric inpatient, and long-term care settings. Such studies should consider patient populations not yet represented in the literature, including patients with chronic diseases (such as diabetes, asthma, cancer, congestive heart disease, and mental health conditions). In addition, as we noted in our Limitations section, given the wide variability in how previous authors have described their HIE systems, we were unable to capture measures of HIE usage, mode of access (single sign-in or portal), or HIE mechanism (push versus pull). Authors should be encouraged to include all such information in their future studies so that these important attributes can be assessed. Improvements in population health outcomes, a priority goal of the ACA, should also be examined in the context of HIE. Notwithstanding the existing literature, future studies should also broaden the use of outcome measures that focus on quality of care, coordination of care, and disease surveillance in the context of population health management. Future studies should continue focusing on costs and patient satisfaction measures, examining settings that have been previously unexplored. Finally, most studies currently rely on HIE that is measured in a way that does not capture the 482 Health Affairs March :3

7 true extent of HIE use. In our analysis, most studies measured HIE in terms of the different types of information exchanged and whether organizations exchanged data with other entities inside and outside their organization. Although this provides information on whether or not HIE occurs, it does not indicate the magnitude and the quality of the exchange, which in itself may be vital to evaluating the effect of HIE. Previous studies have provided evidence that actual technology usage may be related to better financial and quality performance of hospitals (Overhage JM et al.; see Appendix Exhibit A2). 10 This further supports the need to examine actual HIE usage when evaluating its effectiveness. Notably, few studies used track-log analysis, which identifies all instances of information access in an HIE; one of the rigorous studies that identified benefits from HIE in Israel (Nirel et al.) did so (Appendix Exhibit A2). 10 Including information on actual HIE usage in terms of the type of information (such as lab tests and imaging), the source (local or external), and the magnitude of the access in future studies may add valuable insight to our understanding of the effects of HIE on health care outcomes. Conclusion The current state of the literature does not provide sufficient rigorous evidence for the benefits of HIE. It is incumbent upon future research to increase the use of study designs capable of reducing selection bias and confounding and to include settings, populations, and outcome measures for which little research currently exists. Likewise, government agencies and exchange organizations should support the use of stronger evaluation designs by allocating more resources, including adequate funds, cooperation, and access to data. Strong designs with adequate resources will enable researchers to conduct the types of studies that can best inform policy. NOTES 1 National Alliance for Health Information Technology. The National Alliance for Health Information Technology report to the Office of the National Coordinator for Health Information Technology on defining key health information technology terms [Internet]. Washington (DC): Department of Health and Human Services; 2008 Apr 28 [cited 2015 Jan 15]. Available from: %20Terms%20Definitions%20 Final_April_2008.pdf 2 Blumenthal D. Launching HITECH. N Engl J Med. 2010;362(5): Office of the National Coordinator for Health Information Technology. Principles and strategy for accelerating health information exchange (HIE) [Internet]. Washington (DC): Department of Health and Human Services; 2013 Aug 7 [cited 2015 Jan 15]. Available from: acceleratinghieprinciples_strategy.pdf 4 Hsiao CJ, Hing E. Use and characteristics of electronic health record systems among office-based physician practices: United States, NCHS Data Brief Jan; (143): Swain M, Charles D, Furukawa MF. Health information exchange among US non-federal acute care hospitals: Washington (DC): Office of the National Coordinator for Health Information Technology; 2014 May. ONC Data Brief No Fontaine P, Ross SE, Zink T, Schilling LM. Systematic review of health information exchange in primary care practices. J Am Board Fam Med. 2010;23(5): Hincapie A, Warholak T. The impact of health information exchange on health outcomes. Appl Clin Inform. 2011;2(4): Moher D, Liberati A, Tetzlaff J, Altman DG. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med. 2009;6(7): e Vest JR, Jasperson J.What should we measure? Conceptualizing usage in health information exchange. J Am Med Inform Assoc. 2010;17(3): To access the Appendix, click on the Appendix link in the box to the right of the article online. 11 The outcomes we considered include what Avedis Donabedian called process (what is actually done in giving and receiving care) as well as outcomes (effects of care on the health status of patients and populations). To reduce possible confusion, we referred to both types of dependent variables as outcomes ; thus, health care outcomes were those that were believed or expected to improve following HIE adoption (such as repeat diagnostic and imaging tests, hospital readmissions, public health reporting, and health care costs). Donabedian A. The quality of care. How can it be assessed? JAMA. 1988;260(12): Indiana Health Information Exchange. About IHIE [Internet]. Indianapolis (IN): IHIE; 2014 [cited 2015 Jan 12]. Available from: Vest JR, Gamm LD. Health information exchange: persistent challenges and new strategies. J Am Med Inform Assoc. 2010;17(3): Shapiro JS, Kannry J, Lipton M, Goldberg E, Conocenti P, Stuard S, et al. Approaches to patient health information exchange and their impact on emergency medicine. Ann Emerg Med. 2006;48(4): March :3 Health Affairs 483