Electronic Health Records in Ambulatory Care — A National Survey of Physicians

Survey Development

The survey was developed by the investigators, with guidance from a consensus panel of experts in the fields of survey research, health-information technology, and health care management and policy and from representatives of hospital and physician groups and organizations. The development of the survey was also informed by focus groups and interviews with physicians and chief information officers and by a systematic review of previous surveys that were focused on the adoption of electronic health records.⁴

The survey was approved by the institutional review board at Massachusetts General Hospital and by the federal Office of Management and Budget. The investigators drafted the manuscript and had complete independence in developing the survey, collecting and analyzing the data, and reporting the results.

Developing a Measure of Adoption

Table 1. Table 1. Survey Items Defining the Use of Electronic Health Records.

On the basis of advice from the expert panel, the investigators defined the key functions that constitute an outpatient electronic health record and asked respondents to describe the availability and use of those functions. The investigators began with the Institute of Medicine's framework that defines possible functions of an electronic health record.⁷ Using a modified Delphi process, the panel reached consensus on functions that should be present to qualify the system as a “fully functional” electronic health record.² These functions generally fall into four domains: recording patients' clinical and demographic data, viewing and managing results of laboratory tests and imaging, managing order entry (including electronic prescriptions), and supporting clinical decisions (including warnings about drug interactions or contraindications). Physicians were asked whether their main practice site had a computerized system for each function (Table 1).

Recognizing that relatively few physicians might have fully functional electronic health records and that less complete electronic records might nevertheless convey benefits for patients' care, the investigators defined a minimum set of functions that would merit the use of the term “electronic health record,” calling this a “basic” system (Table 1). The principal differences between a fully functional system and a basic system were the absence of certain order-entry capabilities and clinical-decision support in a basic system. The survey assessed physicians' access to various functions and whether the functions were used. However, since the overwhelming majority of physicians said they used most available functions, we primarily report findings on the availability of electronic health records in the office setting.

Survey Sample

We identified all U.S. physicians who provide direct patient care from the 2007 Physician Masterfile of the American Medical Association (AMA). We excluded all doctors of osteopathy, residents, physicians working in federally owned hospitals, those with no listed address, those who requested not to be contacted, and those who were retired. From the resulting list, we randomly selected 5000 physicians for inclusion in the sample.

Of these 5000 physicians, 516 were ineligible to participate in the survey because they were deceased, retired, out of the country, practicing in a specialty that was not included in the survey (i.e., radiology, anesthesiology, pathology, or psychiatry), had no known address, or were not providing care to patients. Of the 4484 eligible respondents, 2758 completed the survey, which yielded a response rate of 62%. A copy of the survey appears in the Supplementary Appendix, available with the full text of this article at www.nejm.org.

Survey Administration

RTI International administered the survey between September 2007 and March 2008. Physicians received an initial mailing that included a cover letter, the survey, a postage-paid return envelope, and a check for $20. Nonrespondents received reminders by mail and telephone. In January 2008, nonrespondents received another reminder and a $40 check to encourage participation.

Statistical Analysis

Table 2. Table 2. Characteristics of Survey Respondents and Their Practices.

All statistical analyses were conducted by researchers at Massachusetts General Hospital. We compared the characteristics of respondents with those in the AMA Masterfile using two-tailed chi-square tests with the use of SAS software, version 9.0 (Table 2).⁸ The respondents were more likely to be male than would be expected on the basis of national statistics. We adjusted for possible nonresponse bias as a result of this difference by creating a weight equal to the inverse of the response probability for men and women and used this weight in all the multivariate analyses.

We examined the univariate and bivariate relationships in the data. On the basis of these analyses, we applied a cumulative logit model, using SUDAAN, version 9.0.1 (RTI International),⁹ to evaluate the association between the characteristics of physicians (sex, race and ethnic background, number of years in practice, and medical specialty) and their practices (practice size, practice setting, location, and region of the country) with the availability of electronic health records, which was treated as an ordinal variable. From this model, we obtained percentages¹⁰ and the accompanying standard errors of availability of electronic health records, with adjustment for the characteristics mentioned above.

Second, we performed logistic-regression analysis to assess whether the availability of electronic health records was associated with a report by respondents that an electronic-records system had a positive effect on certain aspects of their practice. The third analysis assessed whether physicians were satisfied with their electronic records. The fourth analysis examined the barriers to and facilitators of adoption. These analyses were restricted to physicians who reported having access to a basic system or a fully functional system; the analyses were adjusted for significant characteristics of physicians and their practices.

Survey Respondents

Four percent of respondents reported having a fully functional electronic-records system, and 13% reported having a basic system. Of the small number of respondents who had a fully functional system, 71% reported that their system was integrated with the electronic system at the hospital where they admit patients, as compared with only 56% of respondents with a basic system (P=0.006).

Among the 83% of respondents who did not have electronic health records, 16% reported that their practice had purchased but not yet implemented such a system at the time of the survey. An additional 26% of respondents said that their practice intended to purchase an electronic-records system within the next 2 years.

Factors Associated with Availability

Table 3. Table 3. Rates of Adoption of Electronic Health Records by Physicians, with Adjustment for the Characteristics of the Physicians and Their Practices.

In multivariate analyses, having an electronic-records system was significantly associated with several characteristics of both individual physicians and their practices (Table 3). Electronic-records systems were more prevalent among physicians who were younger, worked in large or primary care practices, worked in hospitals or medical centers, and lived in the western region of the United States. Rates of adoption did not differ significantly among providers serving a high proportion of minority patients or patients who were uninsured or receiving Medicaid, as compared with other physicians (data not shown).

Frequency of Use

Among the 4% of doctors with a fully functional electronic-records system, 97% reported using all the functions at least some of the time. Among the 13% of doctors with a basic system, more than 99% reported using all the functions at least some of the time.

Other Capabilities

Table 4. Table 4. Rates of Response Regarding Functions of Electronic Health Records and Their Effects.

Physicians with electronic health records were asked to report the extent to which these systems allowed patients to do each of the following online: view and make changes to their medical records and request prescription refills, appointments, and referrals. Physicians with fully functional electronic-records systems were significantly more likely than those with basic systems to have each of these functions (Table 4). Enabling patients to request a prescription refill online was a prevalent function for both basic systems and fully functional systems.

Effect on Practices

Figure 1. Figure 1. Rates of Positive Survey Responses on the Effect of Adoption of Electronic-Health-Records Systems.

Percentages were calculated with the use of a multivariable logistic-regression model. Variables included were medical specialty, the number of years since graduation, the number of physicians in the practice, clinical setting, and geographic region.

Figure 1 shows the percentages of respondents reporting positive effects of electronic health records on various aspects of their practices. Among the small number of respondents who had fully functional electronic-records systems, most physicians reported the positive effects of the system on the quality of clinical decisions (82%), communication with other providers (92%) and patients (72%), prescription refills (95%), timely access to medical records (97%), and avoidance of medication errors (86%). Furthermore, 82 to 85% reported a positive effect on the delivery of long-term and preventive care that meets guidelines. For physicians with basic systems, the magnitudes of effects were generally smaller. Results were adjusted for the characteristics of physicians and their practices.

Respondents also reported on whether the use of electronic health records had assisted in the care of patients in several specific ways (Table 4). Most of those with fully functional systems reported averting a known drug allergic reaction (80%) or a potentially dangerous drug interaction (71%), being alerted to a critical laboratory value (90%), ordering a critical laboratory test (68%), and providing preventive care (69%). Physicians with basic electronic-records systems reported having the same effects but less commonly than did those with fully functional systems.

Physician Satisfaction

A large majority of physicians reported being satisfied with their electronic-records systems overall (93% for fully functional systems and 88% for basic systems, P=0.20) and with the ease of use of the system when providing care to patients (88% and 81%, respectively; P=0.11). Physicians with fully functional electronic-records systems were significantly more likely to be satisfied with the reliability of their system than were those with basic systems (90% and 79%, respectively; P=0.01). Here again, results were adjusted for the characteristics of physicians and their practices.

Barriers to Adoption

Table 5. Table 5. Barriers to the Adoption of Electronic Health Records.

Among physicians who did not have access to an electronic-records system, the most commonly cited barriers to adoption were capital costs (66%), not finding a system that met their needs (54%), uncertainty about their return on the investment (50%), and concern that a system would become obsolete (44%) (Table 5). Physicians with electronic health records tended to highlight the same barriers but less frequently than did nonadopters.

Facilitators of Adoption

Among all respondents, the factors that were most frequently cited as facilitators of adoption were financial incentives for the purchase (55% among physicians with no electronic health records and 46% among those with electronic health records, P=0.001) and payment for use of an electronic-records system (57% and 52%, respectively; P=0.04). About 40% of respondents with and without an electronic-records system also reported that protecting physicians from personal liability for record tampering by external parties could be a major facilitator of adoption.

On the basis of a large, representative sample of U.S. physicians and clearly specified, replicable definitions of electronic-records systems that were developed by a panel of experts, our study indicates that electronic health records are available in the office setting to only a small minority (17%) of U.S. physicians at present. Only 4% of physicians have what the expert panel considered a fully functional electronic-records system.

Previous studies have shown that the practice setting (and especially the size of the group) had a significant influence on the adoption of electronic health records in the United States, findings that our results confirm.^5,6,11 After adjustment for other characteristics of physicians and their practices, we found that physicians who practice in groups of more than 50 were three times as likely to have a basic electronic-records system and more than four times as likely to have a fully functional electronic-records system as were physicians in groups of 3 or fewer. However, even in large groups, only a small minority (17%) had a fully functional system, and 49% had no electronic-records system at all.

Subjective reports by respondents about the influence of electronic health records on the quality of their practice and clinical decisions and about their satisfaction with the system are encouraging. The proportion of respondents reporting positive effects was generally larger for fully functional systems than for basic systems, a finding that is consistent with the hypothesis that more capable systems offer greater benefits. However, the possibility of bias among respondents, especially greater receptivity to and facility with electronic health records among early adopters, cannot be excluded. The quality and cost effects of electronic health records need to be confirmed by direct studies of clinical outcomes. Considerable controversy continues about the overall effect of electronic health records, and further research needs to clarify the effects of this technology on our health care system.¹

It is also encouraging that a large majority of respondents reported overall satisfaction with their electronic-records system. However, approximately 20% of physicians with basic systems expressed reservations about the ease of use and reliability of their systems. Improving the usability of electronic health records may be critical to the continued successful diffusion of the technology.

Even though we used definitions and methods that differed from those used in previous studies of electronic-records systems, it is possible, within limits, to compare our findings with those of other studies. For example, in 2006, the National Ambulatory Medical Care Survey (NAMCS) showed that 9.3% of respondents had adopted systems similar to (though not exactly the same as) our current definition of a basic electronic record.² Applying the NAMCS definition, we found that 14% of our respondents reported having an electronic-records system. This finding suggests that the number of physicians with some type of electronic-records system has increased in the past year. The function-based approach that we used to measure the availability and use of electronic health records will enable future researchers to gauge progress in the adoption of such systems on the basis of alternative definitions, including that used by NAMCS.

Our study and others^1-3 serve to underscore both the potential benefits of electronic health records and the low current availability of this technology. The combination of these findings suggests that the U.S. health care system faces major challenges in taking full advantage of electronic health records to realize its health care goals. President Bush has proposed that electronic health records should be widespread in the U.S. health care system by 2014, and both of the likely presidential candidates have prominently featured the diffusion of electronic health records in their health care proposals.^12-14 Indeed, recent Medicare cost-containment proposals included incentives for the adoption of health-information technology by physicians as a means of spurring greater use.¹⁵ Our data suggest that such incentives could be important facilitators of adoption. However, the cost of achieving widespread adoption of electronic health records in the United States could be high, probably in the tens or hundreds of billions of dollars,^1,16-19 and whether any future federal administration will find the necessary resources is uncertain.

In their efforts to spur adoption of electronic health records, policymakers may benefit from studying the experience of other Western countries, which seem to have been much more successful (despite significantly lower overall national health expenditures) at encouraging the adoption of health-information technology by physicians. Data from 10 Western industrialized nations suggest that a large majority (often more than 90%) of primary care physicians currently use computers in their office practices.²⁰ These countries seem to have achieved these results using a variety of interventions — public and private, economic and noneconomic — that may offer guidance to future actions in the United States.²⁰

Certain limitations of our study should be taken into account. Like all surveys, ours was subject to potential problems of response bias. It is possible that physicians who responded to our survey had a greater interest than did nonresponders in the subject of electronic health records. Although we adjusted for potential nonresponse bias, our data may overestimate actual rates of adoption of electronic health records. Another reason to be cautious about the reports is that the estimates of the effect of these systems on quality of care and satisfaction are based on a small number of respondents with a large margin of error, especially for the fully functional electronic-records systems. As already noted, by virtue of having electronic health records at this stage in their diffusion, the respondents with these systems are probably different from respondents without them. This limitation, coupled with the small number of adopters in our study, suggests that any extrapolation of the benefits and satisfaction with electronic health records reported by respondents should be done with caution.

In discussions about health-information technology, our study informs the debate by providing benchmark information about the levels of adoption of electronic health records by U.S. physicians as of late 2007 and early 2008. Further studies that use clear, similar definitions of electronic health records and representative samples of physicians will be necessary to inform the development of policies with regard to electronic health records in our health care system.