Financial Effects of Health Information Technology: A Systematic Review
Published Online: November 26, 2013
Alexander F. H. Low, MBA; Andrew B. Phillips, RN, PhD; Jessica S. Ancker, MPH, PhD; Ashwin R. Patel, MD, PhD; Lisa M. Kern, MD, MPH; and Rainu Kaushal, MD, MPH
Health information technology (HIT) is an important lever with which to improve the quality and efficiency of the healthcare system.1,2 The federal government’s belief in the importance of HIT motivated a commitment of up to $30 billion in funding for HIT as part of the Medicare and Medicaid Electronic Health Record (EHR) Incentive Program and related efforts through the American Recovery and Reinvestment Act of 2009.3
Despite the promise of HIT, there has long been debate about its financial effects, both on individual providers and payers on the microeconomic level and on the US healthcare system at the macroeconomic level.4,5 This question has become especially significant given the great interest within the healthcare industry and among policy makers in finding ways to control growing healthcare costs. State and federal governments are the nation’s largest healthcare payers and have invested heavily in HIT. As a result, they have a strong interest in understanding its financial effects.
To date, several groups of researchers have reviewed the literature to understand the quality and efficiency effects of HIT in general or of specific types of HIT, including EHRs, computerized physician order entry (CPOE), and clinical decision support (CDS).5-13 Notably, Chaudhry and colleagues conducted a systematic review of articles published between 1995 and January 2004 to assess the effects of HIT on quality, efficiency, and cost.6 Two subsequent reviews, conducted by Goldzweig and colleagues8 (June 2004 to June 2007) and Buntin and colleagues13 (July 2007 to February 2010), updated that research, though each explored new themes. In 2008, the Congressional Budget Office assessed evidence on the costs and benefits of HIT to offer guidance for the federal government’s HIT strategy.5
Although several of the above articles explored HIT’s effect on cost, none compiled data on the financial effects of HIT in a systematic way. In at least 2 cases the authors cited a paucity of articles addressing HIT’s effect on costs.6,8 In addition, previous articles have not compared the financial effects and their mechanisms across different types of clinical settings and technologies. We systematically reviewed the literature to characterize the existing data on the financial effects of HIT and considered the implications for HIT’s effect on healthcare spending.
We limited our review to articles investigating the effects of 4 types of HIT applications used by healthcare providers in the delivery of care: EHRs, CPOE, CDS, and health information exchange (HIE). These applications were chosen because they have been the subject of the bulk of the debate about the potential beneficial effects of HIT and are central to the meaningful use criteria for the Medicare and Medicaid EHR Incentive Program. We used well-established definitions for EHR, CPOE, and CDS, most notably documented in an article by Blumenthal and colleagues on HIT.14 We defined HIE as systems or applications that connect HIT systems maintained by separate healthcare providers, payers, and other stakeholders, thus allowing providers to share electronic information about common patients.
In addition, our inclusion criteria required that studies (1) explore a financial effect as a principal outcome measure (alone or in combination with other outcomes); (2) quantify the effect in monetary terms for 1 or more stakeholders (articles reporting other related measures, such as length of stay or other types of utilization, were excluded unless the effect was explicitly measured in monetary terms); (3) present primary research rather than a compilation or review of existing literature; (4) be published in an English-language, peer-reviewed journal since 2000; and (5) be set in the United States, since we reasoned that the unique characteristics of this country’s healthcare system—specifically those impacting the financing, adoption, and use of HIT systems—would render foreign studies’ findings less relevant to our objectives. Finally, in the event authors had written more than 1 qualifying study on the effect of the same HIT application on a similar setting or population, we included only the most recent article.
Study Identification and Selection
Our search for candidate articles consisted of 2 phases. In the first phase, we searched PubMed in February 2012, using the intersection of 2 lists of search terms: the first related to HIT applications and the second related to financial or economic effects (eAppendix A, available at www.ajmc.com). In the second phase, we identified other relevant studies by reviewing the reference lists of the included articles as well as additional policy articles and literature reviews.
The team developed a list of data items to be extracted from the articles (eAppendix B, available at www.ajmc.com). Key data fields included the following:
Health Information Technology Application. We documented the primary HIT application under investigation: EHR, CPOE, CDS, HIE, or multiple. Because nearly all CPOE applications in the literature included CDS and because most CDS applications were part of a CPOE application, we merged those 2 categories into 1 category. Otherwise, when an article included more than 1 of these applications, we assigned the article a single designation based on the emphasis of the article.
Clinical Setting. We documented the primary clinical setting for the study. These included: emergency department (ED), inpatient, outpatient, or multiple.
Study Design Classification. We classified each study according to its design. A rating of 1 indicated experimental studies, including randomized controlled trials. A rating of 2 indicated observational studies with concurrent control groups. A rating of 3 indicated observational studies with historical controls. A rating of 4 was given to case studies, case series, or other reports in which no control group was included or experimental design described. A rating of 5 was assigned to quantitative simulations where outcomes were modeled based on inputs such as literature review, expert analysis, and projections.
Financial Outcomes. We categorized the studies’ financial outcomes as follows: (1) there was a positive effect for the stakeholder(s) of interest in the study; (2) there was a neutral or mixed effect for the stakeholder(s); or (3) there was a negative effect for the stakeholder(s). In addition, we classified each study according to whether or not it documented the costs of the HIT intervention to stakeholders as well as the benefits.
Stakeholder Perspective. Finally, we documented to which stakeholder(s) the financial benefit or loss accrued. The 4 stakeholder categories were (1) provider, (2) payer, (3) consumer, and (4) the community, society, or health system at large. It was not always clearly stated which stakeholder accrued the benefit or loss, so in those instances we used our best judgment.
Three different team members, including 2 health services researchers and a policy analyst, were responsible for reviewing and extracting the data items from each article. All extracted data were discussed and discrepancies were resolved through consensus.
We calculated counts and percentages for each category of interest and graphed the principal outcomes for all articles meeting our inclusion criteria. We also conducted 2 sensitivity analyses. In the first sensitivity analysis, we assessed whether our results would differ if we included only articles with study design ratings of 1, 2, or 3 (experimental studies or observational studies with concurrent or retrospective controls), because those studies might be considered more reliable than the others. In the second sensitivity analysis, we assessed whether our results would differ if we included only articles that documented both the cost and benefit of the HIT intervention, because we considered this documentation to be indicative of at least a basic rigor in the economic analysis.
We reviewed 4600 search results. Based on their abstracts, we selected 96 (2%) for full-text review and added 24 others that we identified from the bibliographies of other articles on this topic. Of these 120 articles, 57 articles (48%) met our inclusion criteria and were the basis for our analysis. (eAppendix C, available at www.ajmc.com).
Characteristics of the Included Studies
The 57 articles were very heterogeneous. For example, they ranged from studies of single CDS rules to multifunctional EHRs; from studies set in single solo or small practices to studies of the US healthcare system; from studies lasting a few months to those covering multiyear spans; and from studies where financial effects were the only outcome under investigation to those where the financial effect was one of several outcomes.
PDF is available on the last page.