Now that EHR adoption is well under way and meaningful use is in full-swing, the effects of EHR systems on every day clinical practice are becoming more obvious. According to studies that look at how providers spend their time, clinicians are increasingly spending more time with computers and less time with patients. Is this merely an unintended consequence of computerization, a symptom of poor design, or a combination of both?
The first study, reported in the paper, In the Wake of the 2003 and 2011 Duty Hours Regulations, How Do Internal Medicine Interns Spend Their Time?, was conducted at Johns Hopkins and the University of Maryland medical centers, and looked at how interns spend their time. Here are the results as reported by the authors:
Interns were observed for a total of 873 hours. Interns spent 12 % of their time in direct patient care, 64 % in indirect patient care, 15 % in educational activities, and 9 % in miscellaneous activities. Computer use occupied 40 % of interns’ time. There was no significant difference in time spent in these activities between the two sites.
Interns spend only 12% their time in direct patient care compared with 40% spent using computers?! How should these numbers be interpreted?
An earlier paper, Internal Medicine Residents’ Time Study: Paperwork Versus Patient Care, reports similar findings. The authors offer the following results:
Residents reported spending most of their time at workstations (43%) and less time in patient rooms (20%). By task, residents spent 39% of time on indirect patient care that must be completed by a physician, 31% on structured education, 17% on direct patient care, 9% on indirect patient care that may be delegated to other health care workers, and 4% on personal activities. From these data we estimated that residents spend 34 minutes per patient per day completing indirect patient care tasks compared with 15 minutes per patient per day in direct patient care.
Both papers seem to corroborate the results of a review conducted by Poissant, Pereira, Tamblyn, and Kawasumi published in 2005. The paper, The Impact of Electronic Health Records on Time Efficiency of Physicians and Nurses: A Systematic Review, noted large increases in CPOE-related documentation times.
The final paper looks specifically at emergency department EHR use. Like the studies that focused on interns, it reports that ED physicians spend about 44% of their time on data entry. While these papers focus on inpatient systems, a recent RAND survey commissioned by the AMA, finds that outpatient providers have similar concerns and complaints.
Obviously, there are problems, but it is difficult to determine their exact nature. There are many possible reasons for poor usability and workflow disruptions. For example, do these studies indicate that computerized patient notes are simply more extensive (thus taking longer to read and write), harder to navigate, difficult to assimilate because multiple screens are required to view a complete note, or all of the above, plus additional confounding factors?
From the standpoint of software design and development, correcting the problems requires knowing what is wrong. However, such information is very difficult to obtain. Studies of the negative effects of clinical systems report what users dislike and record productivity losses, not how specific architectural/ design choices caused those negative effects.
A major impediment to building better systems is the lack of a blueprint for what constitutes a good EHR system. Take something as obvious as workflow. There are no standards for clinical workflow representation in EHR systems (or for paper-based clinical workflows for that matter). Current EHR systems were designed to be better versions of paper records, but alas, things are not working out quite as expected.
Those of you who follow this blog know that, over the last two years, I have come to the conclusion that clinical software will not improve until we have formal methods and theories to guide their design and evaluation. In fact, this conviction was the gist of the post, EHR Science: The State of the Blog. Of course, with such a grand goal, there is no one, obvious place to begin. However, looking over my notes, I think graphs, relations, and functions provide the constructs required to formulate a mathematical theory of clinical systems. And yes, sometimes I think this quest is more than a little fantastic in a Jules Verne sort of way, but hey, one has to keep busy, right? With this in mind, I was encouraged to discover Thomas Beale’s most recent post, The Real Reason Most Software Fails. As is the usual case in his posts, he cuts to the heart of the matter and offers a very cogent argument for the need for theories and formal methods. Knowing that someone else has come to the same conclusion makes the idea of a formal theory for clinical systems seem less far-fetched.
From my perspective, Petri nets and workflow patterns provide a solid foundation for recasting clinical systems in formal terms, which is why I have written so much about them. Using workflow as a starting point, I began my education in discrete mathematics with graph theory, which was not the best point to jump in because I quickly had to go back to the basics and learn about logic and sets. I’ll save you from that jarring experience and start at the beginning. My take on the mathematics of clinical concepts will appear as a series of posts (10 or so) interspersed with other topics, over the next six months. Ambitious? Yes. Fantastic? You make the call…