Having been a serious computer geek in an earlier phase of my life, I always thought that having an electronic medical record (a.k.a. an electronic health record or EHR) would be a wonderful thing. I could access data from anywhere with no lost charts; search for text (e.g., to find when I last discussed knee pain with a patient); identify patients at need for preventive services; assess how my practice was doing on chronic disease management and identify patients in need of closer follow-up; quickly send prescriptions electronically and have dangerous interactions or other potential problems identified; and maybe even enter history and physical data more efficiently.

However, now that I am experiencing the third such system of my career, I have to say that I’ve been disappointed in virtually all of these hopes. As a former programmer, I have been appalled by the frequently abysmal quality of the programming of these systems, betraying both a lack of any central design and oversight of the user interface and an absence of understanding of the typical tasks and workflow in a primary care office.

For example, the current system I am using opens a blank screen with no functionality or information when I first open a chart. Most screens do not show age, date of birth, or medical record number. Despite the epidemic of childhood obesity, it provides no height-weight comparison for children under 2 because BMI computation for children under 2 is not a Stage 1 “meaningful use” requirement, though this will have to change if the Stage 2 meaningful use proposed rule provisions about BMI are adopted. Many free text boxes allow so few characters as to be essentially useless. Who really thinks 100 characters is enough to explain all of a patient’s lab results when sending a note?

Approving 12 months of refills when I receive an electronic refill request typically takes a combination of 14 mouse movements, clicks, and keystrokes – as opposed to four if it were implemented efficiently. The list of items needlessly making it more difficult to provide efficient and effective care would cover many pages. These might seem like issues that could be present in version 1 of a system and then promptly fixed, but we currently have version 5.6.

Shared Frustrations

I wish I could say my experiences were atypical, but in fact they seem to be representative of what many of the primary care colleagues I speak with, working in a wide variety of settings, encounter with EHRs. If anything, my facility with computers and skill as a touch typist have made things easier for me than many others. My workday seeing patients is at least an hour longer than it was formerly. Many colleagues spend substantial amounts of their evenings and weekends “VPNing” into their EHRs to catch up on documentation, billing, and other tasks.

I have yet to see an article discussing the “ROI on an EHR” account for the substantial increase in time these systems require from many users. Perhaps this is because turning highly-trained providers into unpaid medical transcriptionists counts as a savings to improve the ROI since out-of-hours provider time is “free.” However, there is a very real cost to increasing provider dissatisfaction and burnout.

Weak Evidence On Provider Satisfaction And Benefits For EHRs

Am I just crankier and more demanding than most physicians? Systematic data assessing provider satisfaction with their EHRs are rather scarce. A national survey of physicians providing direct patient care conducted in 2007-8, prior to the major federal push to adopt EHRs, found only 4 percent of respondents reported having a “fully functional EHR,” while another 13 percent reported some functionality that was classified as a “basic EHR.” Large practices with more than 50 physicians were much more likely to have a “fully functional EHR.” The authors reported that 93 percent of the 117 respondents with “fully functional EHRs” reported being satisfied with their systems overall and 88 percent with ease of use when providing care to patients, but cautioned about extrapolating from these findings.

In contrast, responses from family physicians to a recent volunteer-based survey carried out by the journal Family Practice Management suggest that provider satisfaction with outpatient EHRs may be rather low, with only about half of respondents indicating they were satisfied overall, even fewer being satisfied with vendor training and support, and only 38% agreeing that they would purchase their system again.

There are a variety of possible explanations for the differences between the results of these two surveys. The first report, while based on a survey of a nationally-representative sample, is several years old and represents the opinions of a small number of physicians in “early adopter organizations” that chose to develop or buy an EHR without inducements. This sample also presumably had substantial representation from specialists, whose narrower domains of care are more amenable to “point and click,” templated entry than primary care. The latter report likely has a substantial volunteer bias and allowed only family physicians to respond, but has a much larger sample of later adopters purchasing a wide variety of systems in response to federal incentives.

Given the steep learning curves of most EHRs, one would expect recent adopters to have more negative views than more experienced users. Some of this is probably unavoidable. However, humans are remarkably adaptable and can accommodate even to very unpleasant circumstances. As case in point, one of the physicians in my clinic who was a strong proponent of purchasing the system we obtained had used it in residency and knew no other. Having since left for another setting that uses another EHR, this physician has stated, “I had no idea how bad X was until I started using Y.” So one should not assume that reasonable user satisfaction after a year or two of experience with an EHR implies a good product as opposed to human adaptation to unchangeable, difficult circumstances. If we really want to know how good or bad an EHR is for users, we need comparative studies evaluating initial usability, ease or difficulty of the learning curve, and expert usability.

Surely, the federal government must be pushing hard for providers to adopt these systems because they yield dramatic improvements in health care. Alas, the evidence of EHR benefits is rather weak. A recent study has suggested that having laboratory and radiology results available in an EHR may increase, not decrease, costs. A systematic review highlighted that a substantial amount of the limited, extant literature came from a few settings with “home-grown” systems, while relatively few studies have been published describing commercial EHR systems.

A review from the Office of the National Coordinator for Health Information Technology (HIT) reported that 92 percent of recent articles on health information technology “reached conclusions that were positive overall.” However, relatively few of these focused on EHRs, even fewer on ambulatory EHRs, and fewer yet on provider experience and the EHR role during the clinical encounter, as opposed to administrative and data interchange functions. The authors acknowledged the potential for publication bias, and they also acknowledged that studies by adopters would be more likely to focus on assessing benefits than adverse effects and to lead to an underestimate of negative outcomes. Nonetheless, they stated, “the benefits found in the published articles are real.”

We would not accept such an argument about a new medical treatment whose initial trials made no effort to measure adverse effects and it is unclear why we should accept it for publications about HIT. Given the massive resource and career investments in adopting EHRs and other HIT systems, it will not be possible to get a realistic estimate of how publications in this area “accentuate the positive” and “eliminate the negative” unless multiple, externally-funded evaluations are performed. For EHRs, these need to focus on actual usability in clinical care, not just on “meaningful use.”

One contributor to the paucity of publications about commercial systems may be “gag clauses” inserted in the contracts of many customers. Such clauses do not suggest an industry confident it is providing good value for money. Point-of-care reminders to perform needed services ought to be one of the “no-brainer” successes for EHRs, yet a review found evidence for very modest benefits of such systems, with a median of only 5.6 percent improvement and the one high-performing outlier being a home-grown system. The clearest example of a health care system in the U.S. demonstrating the role of an EHR in improvements in quality and lower costs comes from the Veterans Administration’s development of VistA, a home-grown, noncommercial, originally provider-generated system.

How can it be that we are being pushed so hard, excoriated as laggards for moving so slowly to adopt these systems, yet so many are disliked and dysfunctional? One obvious explanation is one of the health information technology fallacies identified by Karsh and coauthors: those who make most of the purchasing decisions and reap most of the benefits from EHRs are managers, so there is little incentive to focus on improving the provider experience. The systems are focused on capturing “clickable data” to justify billing, preferably at higher levels than before.

However, use of checkboxes and radio buttons ends up producing pages of pseudo-English – phrases like, “Auscultation could be described as rales” – with remarkably low information density that could be described as “More filling! Less taste!” Providers end up either swallowing hard, using clicks that document things that may not be exactly true and not documenting things that don’t fit in the boxes, or typing free text that the systems cannot capture as data and hence are no more useful for capturing and justifying charges than a dictated note. CMS meaningful use criteria so far have virtually nothing to do with facilitating use by providers. The new proposed stage 2 criteria remain mute on provider usability – only functions and data are specified.

Fortunately, the free market will rescue us by letting the cream rise to the top, right? Alas, the EHR market has virtually none of the attributes of a free market. As noted above, buyers have limited information about the products they are considering and typically are not the end-users, limiting impetus to improve functionality for providers. A free market also requires low barriers to entry and exit – but high adoption costs lead to “EHR lock,” with no simple solution if buyer’s remorse sets in. EHR vendors have a strong disincentive to lowering barriers to exit, so it is no wonder that trying to exchange even limited amounts of data between systems is so difficult. In fact, most vendors do not even provide their customers with data dictionaries for the underlying databases.

Further market distortions come from the strong federal incentives pushing the adoption of EHRs. EHRs seem like they should be closer to utilities or public services than commodities – a natural monopoly, or at least oligopoly. Personal computer software has reached this state. Imagine the inefficiency if every employee changing jobs had to learn an entirely new operating system and suite of office applications – and if documents from a different system could be opened only with great difficulty, losing formatting and much content. Also, many EHR systems do nothing out of the box and extensive customization is needed to make them usable at all. Small (and sometimes large) organizations struggle simply to get basic functionality working and can rarely support programmers to do extensive customization. Larger organizations may make significant investments in customizations, but because the Balkanized EHR market makes most innovations difficult or impossible to transport, EHR customization by providers is often a “What happens in Vegas, stays in Vegas” phenomenon.

How Should We Move Forward?

I have seen a few systems that made me think, “This system wouldn’t slow me down much – in fact, I might even be able to be a little more efficient.”  These have almost all been systems designed for small practices, typically by physicians, and tend to have a clean, simple interface and focus on implementing a limited set of tasks efficiently. Despite the potential volunteer bias, the Family Practice Management user survey results have pretty good face validity. The top-rated EHRs are almost all those targeted to small practices where the buyers are likely to be the users.

There are two notable exceptions to this – VistA and EpicCare Ambulatory, the expensive system used by many of the nation’s largest integrated delivery systems such as Kaiser, which colleagues tell me has a steep learning curve but a lot of useful functionality once mastered. Medical students returning from rotations at Veterans’ Administration Hospitals often rave about how good VistA is – something I have never heard with any other EHR. While I have not used it in clinical care, I have examined the demonstration client available on the web and been impressed by the simple, clean interface – quite unlike most other EHRs I have used or seen.

There may still be time to change course from “pick an EHR, any EHR” to one more likely to yield better health and lower costs for Americans. It is probably too late simply to say “never mind” and remove the federal incentives to adopt EHRs. If one could guarantee achieving some of the potential benefits of EHRs, this would also be unwise. I believe that the greatest savings would likely accrue from standardizing on a single EHR, with VistA being obvious choice. It probably has the best evidence for both improving quality and lowering costs, it works in settings from small offices to large hospitals, and the software is open-source and free.

I do not believe that a mandate for the adoption of a national EHR, VistA or otherwise, will ever be politically feasible. However, there is no reason the government could not actively choose to facilitate the adoption of VistA. Each regional HIT Extension Center currently chooses which EHRs it will support. If, instead, they were all to support the adoption of VistA, organizations would still be free to adopt any EHR they wanted, but might be much more likely to give VistA serious consideration despite its suspect status as free software.

Having a large population of users of one EHR nationwide would not only simplify health information exchange (the proposed stage 2 meaningful use criteria related to exchange of information specifically exclude transfers between entities using the same EHR from the measurements – the barriers there are too low) and lower personnel training costs, it would create a tool for dramatically more powerful and inexpensive drug and device postmarketing surveillance, comparative effectiveness research, and identification of potential participants for clinical trials. With rising health care costs a major contributor to the nation’s fiscal crisis, choosing a policy that, by subsidizing the purchase of any “certified” EHR, locks in substantially higher costs for the foreseeable future in addition to the other adverse effects discussed above seems remarkably short-sighted and wrongheaded.

It is late, but perhaps not too late, to alter the course we are following as a nation to one more likely to achieve the full potential of EHRs to improve care, save money and lives, and reduce stress on overburdened providers. We need to take a hard look at whether the emperor’s new clothes and our current EHR policies may have more in common than we really want.