|Home | About | Journals | Submit | Contact Us | Français|
With a focus on improving the quality of patient care, the President George W. Bush called for electronic health records (EHRs) for all Americans by the year 2014; however, recent estimates for EHR adoption in the ambulatory care environment are just over 10 percent. The objective of this study was to determine the individual characteristics and the social and technical factors that may contribute to physician acceptance of EHRs. This first paper of a two-part study presents a framework grounded in Diffusion of Innovations theory and the Technology Acceptance Model, which was developed using case study and survey methods, and tested using structural equation modeling (SEM). Model variables explained over 73 percent of the variance in attitude toward EHRs, and acceptable model fit was achieved. Lack of user acceptance continues to impede diffusion of EHRs, and this analysis supports the impact that leadership and other organizational changes have on user adoption.
With a heightened awareness of medical errors and an increased focus on improving the quality of patient care, President George W. Bush called for electronic health records (EHRs) for all Americans by the year 2014.1,2 While the EHR has been in development for nearly three decades, few providers have yet realized a fully integrated electronic health record environment.3–8 Recent figures estimate EHR adoption in the ambulatory care environment to be 13 percent for a basic system and a mere 4 percent for a fully functional EHR system.9 The American Recovery and Reinvestment Act of 2009 (ARRA) includes $17.2 billion for financial incentives to physicians and hospitals through Medicare and Medicaid to accelerate adoption of health information technology. It also includes penalties for providers failing to adopt by 2015.10 Some speculate that even with incentives, critical mass adoption by the prescribed deadline is unlikely, and the informatics literature warns that hasty deployment of health information technology may result in implementation failure or unintended consequences.11–13 The literature also provides evidence of clinical system implementation failures due to lack of adoption by users.14–16 Because physicians are the key coordinators and providers of patient care, their acceptance of an EHR application will determine the overall success of a product's implementation.17–19
The purpose of this study is to confirm a hypothesized correlational model for predicting physician attitudes toward EHR adoption. Using case study and survey methods, this first part of a two-part study presents a framework for examining physician perceptions prior to implementation. The second part of the study will provide an analysis of the participants and their perspectives to gain deeper insight into the antecedents of EHR adoption attitudes. This research is important because an understanding of the reasons why a system may succeed or fail is crucial for a successful implementation.20,21
In order to achieve nationwide interoperability and realize the benefits that EHRs can provide, physician adoption rates must be increased substantially. However, implementing the right system the right way is essential for ensuring project success and protecting patient safety. Nearly 75 percent of all large health information technology (HIT) projects fail, as well as 30 percent of EHR implementations.22,23 An understanding of the factors associated with physicians' acceptance will allow organizations to better assess system readiness and facilitate successful implementation.24, 25 Much of the published research on physician attitudes focuses on satisfaction with clinical applications already in use. This research examines physician attitudes prior to EHR implementation in an academic healthcare system. A variety of disciplines contributes to information system evaluation; the current study draws upon Diffusion of Innovations theory and the Technology Acceptance Model (TAM).26, 27
Lorenzi et al. classified key factors associated with successful implementations and user acceptance.28 These include factors at the organizational level, the group level, and the individual level. The model in this study was developed to address usage determinants at all three levels.
While not specific to information technology, Diffusion of Innovations research examines the social processes surrounding changes that occur when an innovation—a new idea, practice, or object—is introduced into an organization.29 Healthcare organizations are complex social systems comprising individuals with varying backgrounds, experiences, and values. It is important to understand how these social factors influence adoption attitudes. The Technology Acceptance Model (TAM) (Figure (Figure1)1) focuses exclusively on factors that determine users' behavioral intentions toward using a new computer technology, specifically perceived usefulness and perceived ease of use.30
It hypothesizes that a user's intended behavior predicts actual system use. This theory suggests that external variables, such as human and social factors, indirectly determine an individual's attitude toward technology acceptance by influencing perceived usefulness and perceived ease of use.31, 32
Few empirical studies have evaluated physician attitudes toward EHR adoption prior to implementation. Dansky et al. found perceived usefulness, computer experience, patient care values, and organizational support to positively impact attitudes.36 Gadd and Penrod assessed physician attitudes prior to and after EHR implementation.37, 38
Findings indicated perceived usefulness to be the significant predictor before and after implementation, along with concerns regarding patient privacy, interference with physician-patient rapport, workflow, efficiency, and autonomy. A study by van der Meijden et al. found computer experience to be the major predictor of acceptance, with age being nonsignificant.39
Poor training and absence of computer skills were perceived to be adoption barriers in a recent survey of general practitioners.40 In a study of medical group practices, Gans et al. observed “people barriers” to be significant obstacles to EHR adoption.41
The purpose of the study was to test the hypothesized extended TAM shown in Figure Figure22 to determine which factors contribute to physician acceptance of an EHR system. Research questions are as follows:
This case study was conducted at the University of Mississippi Medical Center (UMMC), an academic-based healthcare system in Jackson, Mississippi. This site was selected because it was in the process of choosing an EHR application that would eventually be implemented in all physicians' offices in the system.
Data were collected between August and December 2007. A self-reporting online questionnaire was distributed to 802 UMMC faculty, fellow, and resident physicians. Three follow-up e-mail reminders were sent to nonresponders.
The research instrument was based upon a survey developed by Aldosari and incorporates some additional questions developed by Cork, Detmer, and Friedman.42,43 Both have been validated and tested for reliability in prior studies. Section 1 elicited general information about the respondents. Sections 2 through 9 collected data regarding eight constructs (unobserved variables composed of multiple survey items):
Section 10 gave respondents an opportunity to provide comments. All questions, except those in the general information and comment sections, captured responses via a five-point Likert scale with responses ranging from “strongly disagree” to “strongly agree.” A more in-depth description of the variable constructs will be provided in the second part of the study. Copies of the survey are available from the authors.
A total of 239 usable responses was received, resulting in a net response rate of 29.8 percent. Data were imported into SPSS 16.0 from the online survey application. Using the SPSS Syntax Editor, scales were created for the eight constructs by averaging the participant's responses (i.e., one to five) for all question items for each construct. Construct variables in this study include the social and technical factors shown in Figure Figure22.
A preliminary correlation matrix revealed no significant correlations between the individual physician characteristics and the construct variables, as hypothesized in research question 1. Consequently, individual physician characteristics were excluded from the final predictive model. Descriptive analysis of the participants and the social and technical factors will be reported in the second part of the study.
Structural equation modeling (SEM) was used to analyze the predicted paths between model variables. SEM is the preferred approach for analyzing interactions between multiple independent and dependent variables, such as those used in our model.44,45 The proposed conceptual path model (Figure (Figure3)3) was created using AMOS 16.0 structural equation modeling software and includes the following sociotechnical construct variables: management support (“Support”), physician involvement (“Involve”), adequate training (“Train”), physician autonomy (“Autonomy”), doctor-patient relationship (“Dprelation”), perceived ease of use (“Ease”), perceived usefulness (“Useful”), and attitude about EHR use (“Att”).46
The existing data set was imported directly into AMOS from SPSS. Path (regression) coefficients were estimated for each of the proposed paths to determine the strength of the relationships. During model testing, some unpredicted paths emerged and were added to the result model presented in Figure Figure4.4. New paths include direct effects from physician involvement (“Involve”) and physician autonomy (“Autonomy”) to attitude about EHR use (“Att”).
In this model, independent variables may directly or indirectly affect dependent variables. Standardized estimates allow the researcher to evaluate the relative contribution of each predictor variable to each outcome variable, as well as to compare across groups.47, 48 Standardized direct effects are illustrated in Figure Figure4.4. Larger path coefficients indicate stronger correlations. To gain an understanding of a variable's overall impact on a dependent variable, it is necessary to examine its combined direct and indirect, or total, effects. Standardized total effects are reported in Table Table11.
The variable with the strongest combined total effects on perceived ease of use (“Ease”) is management support (“Support”) (see Table Table1).1). Doctor-patient relationship (Dprelation) had a significant negative influence on perceived ease of use (-.23) due to the negative content of the questions in the doctor-patient relationship construct. Items in this construct asked whether respondents felt that the EHR would diminish the patient's confidence in the physician, threaten the physician's credibility with patients, or decrease patients' satisfaction with the quality of their healthcare. As a physician's perception of the EHR's ability to inhibit the doctor-patient relationship increases, his or her perceived ease of use decreases. Physician involvement (“Involve”) also had significant total effects on perceived ease of use (“Ease”), while adequate training (“Train”) was not found to be statistically significant.
Perceived ease of use (“Ease”) had the strongest total impact on perceived usefulness (“Useful”) with a standardized coefficient of .55. Doctor-patient relationship (“Dprelation”) had a significant negative influence on perceived usefulness (–.33), again due to the negative content of questions in the doctor-patient relationship construct. Management support (Support), physician involvement (Involve), and adequate training (“Train”) had minimal overall impact on perceived usefulness (“Useful”).
Perceived usefulness had the strongest impact (.63) on attitude about EHR use, with physician involvement (.47), perceived ease of use (.34) and doctor-patient relationship (–.21) making noteworthy contributions. Perceived ease of use did not directly impact attitude about EHR use as hypothesized.
Figure Figure44 also shows the R2 values (explained variance) for each of the dependent variables. Together, management support (“Support”), physician involvement (“Involve”) and doctor-patient relationship (“Dprelation”) accounted for 30 percent of the variance in perceived ease of use (“Ease”). Perceived ease of use (“Ease”) and doctor-patient relationship (“Dprelation”) explained 46 percent of the variance of perceived usefulness (“Useful”). The model reflects that 73 percent of the variance of attitude about EHR usage (“Att”) is captured by the independent variables in the model. R2 values are summarized in Table Table22.
Using AMOS, the research model was tested for goodness-of-fit using three popular model fit indices. Obtained values were within recommended limits (Table (Table3),3), indicating good fit.
In a comparison of well-known information technology acceptance models, Venkatesh et al. identify common limitations of most prior research studies.49 These studies have been conducted with simple rather than complex information technologies, and the subjects have been students. The research usually was conducted after users had been exposed to the system and had already chosen to adopt or reject the technology. Most prior studies have also tested the models in contexts where use is discretionary rather than mandated. This last issue could be problematic in the healthcare environment, as clinicians will ultimately be required by the U.S. government to use health information technology. This research proposes a framework for predicting physician acceptance of EHRs and was tested using a case study. While the TAM has been successfully used to predict attitudes toward technology adoption, the model's explanatory power has varied in prior research. Some studies have reported R2 (explained variance) values as low as 37 percent, while others have been closer to 100 percent.50 The variables in this study explained 73 percent of the variance in attitude, which suggests it is an appropriate mechanism for assessing preadoption perceptions.
In this study, none of the physician characteristics correlated with the social and technical variables in the model. These findings could be reflective of a homogenous sample and are consistent with several prior studies.51–58 The majority of respondents (67 percent) in the study were under the age of 40, which could signify a broad exposure to computers prior to their medical practice experiences. In addition, virtually all of these subjects had prior experience with the Veterans Affairs (VA) hospital's Computerized Patient Record System (CPRS).
The second and third research questions focus upon the impact of the social and technical variables on attitude about EHR use. Path analysis revealed a strong positive relationship between management support and perceived ease of use, which is consistent with previous findings.59 Perceptions of organizational leadership relate to management's ability to provide adequate time and resources for EHR implementation.60–62 Respondents' written comments suggest an expectation of management to ensure availability of adequate workstations, provide training and support, and resolve technical problems in a timely manner. They also expect management to incorporate their feedback regarding system use.
As hypothesized by the TAM, perceived usefulness was highly correlated with attitude about EHR use and was its strongest predictor. An EHR system must provide clear benefits to the medical staff.63,64 Systems often fail because they support the values of management, not the values of staff and users.65 In a survey conducted by the American Medical Association in 2001, only 13 percent of physicians responded that EHRs would make it easier to practice medicine or to manage a medical practice.66 A more recent study found physicians to be dissatisfied with the currently available EHR software applications, and many felt the products disrupted workflow and caused additional problems.67 Successful EHR implementations have been associated with a focus on improving clinical processes and solving clinical problems with information technology.68 Addressing physicians' immediate needs rather than emphasizing future predicted benefits of system use is critical in achieving EHR acceptance.69 Ongoing evaluation and modification based on medical staff feedback is key for continued use of the EHR.70
User involvement in the system selection and implementation process can foster development of user ownership.71,72 SEM analysis revealed a positive correlation between physician involvement and perceived ease of use. Comments indicated that physicians should be responsible for product selection because of their innate understanding of clinical workflow. It is important to them that the system be compatible with clinicians' practice patterns. A number of concerns were noted regarding computerized documentation, and there was an overall apprehensiveness related to use of inflexible data entry templates and online forms. The literature warns that use of structured data entry tools may result in a loss of contextual meaning of patient information and may also compel physicians to cut and paste text from previous documents.73 An unexpected finding was the discovery of a strong positive relationship between physician involvement and attitude about EHR use. This unmediated direct effect indicates that physicians' attitudes are affected by their perceptions of involvement, regardless of the usability or utility of the system selected.
There was a strong negative direct relationship between autonomy and attitude about EHR use. This relationship was not hypothesized but is consistent with previous studies.74,75 Gadd and Penrod found that perceptions of the system's impact on physician autonomy was one of the top concerns physicians had about using EHRs.76, 77
This point was evident prior to EHR implementation and increased in a postimplementation satisfaction study. The literature reports technology-related adverse events where information systems have actually increased error rates and resulted in unintended consequences.78–83
The current study was conducted prior to the Joint Commission's increased scrutiny on safe implementation of HIT.84 It is likely this factor will be a stronger consideration in subsequent studies.
This case study is limited to one large healthcare system, and results may not be reflective of attitudes found in other physician populations. The study is also constrained by the use of an anonymous survey for data collection. The small sample size is a shortcoming of the subjects' willingness to participate.
Limitations of structural equation modeling include the use of a model development process to improve goodness of fit, which is sometimes referred to as a “post hoc” procedure for hypothesis formulation. Another limitation of SEM is the use of goodness-of-fit measures to accept or reject a proposed model. These measures can inform the researcher whether a model is acceptable but cannot tell whether it is a superior model. Because SEM analysis requires a complete data set, the process of imputing missing data could potentially influence analysis as well.
Follow-up studies with focus groups, user interviews, or observations would provide a more detailed understanding of physicians' needs. Future research could also address additional user groups within the healthcare system, such as nurses, administrators, or clerical staff. The EHR acceptance model might be tested in other venues to determine if attitudes vary by care setting. A postimplementation study on actual system use could be performed to determine if preimplementation attitudes accurately predicted true behaviors.
Some of the most interesting findings that emerged from this study are related to perceptions of the EHR's impact on clinician workflow and efficiency. Related factors include time required to document, loss of data granularity collected in patient records, and the need for sufficient training, hardware, and technical support. Postimplementation usability studies could be conducted to gain a better understanding of the EHR's overall impact on the physicians' workflow and productivity.
Different specialties may use the system in various ways. Observations, interviews, or focus groups could further investigate if and how documentation templates are being customized and if physicians are using workarounds for completing documentation. Further study might also compare the use of templates to other data entry formats, such narrative documentation, digital dictation, or data capture via handheld devices. The results would be useful to EHR vendors as well as other healthcare systems that are working through data entry challenges.
The national push for EHR adoption is accelerating; however, the focus should now be on successful implementation.85 The complexity of EHR systems, as well as the healthcare environment, cannot be underestimated. Results from this study highlight the need for strong physician leadership and management support in the EHR selection and implementation process. By assessing the information needs of physicians and other EHR users, HIM practitioners can help develop criteria for evaluating and selecting EHR systems specific to their users' needs. This study revealed an overwhelming need for flexible, customizable EHR products. In general, commercial EHR system development is still quite immature, and often healthcare vendors welcome input from the user community.86
HIM and information technology (IT) practitioners, in conjunction with medical staff leaders, should recommend hardware and software functionality to developers based upon workflow requirements and user needs. HIM practitioners can assist physicians in the selection or design of user interfaces to improve ease of use, and can advise developers on the need for diverse modes of data entry and flexible documentation tools. HIM and IT professionals must consistently work in harmony with clinicians and other users in order to promote initial and long-term EHR adoption.
The authors would like to thank Dr. Scott Stringer for providing access to the study site, as well as Dr. James Gill and his colleagues in the Family Medicine Residency Program at Christiana Care Health Services for serving as a pilot testing site.
Mary E Morton, University of Mississippi Medical Center's School of Health Related Professions in Jackson, MS.
Susan Wiedenbeck, College of Information Science and Technology in Philadelphia, PA.