Cases & Commentaries

EMR Entry Error: Not So Benign

Commentary By Ross Koppel, PhD

The Case

A 47-year-old man with advanced AIDS was admitted
to an academic medical center with a chief complaint of shortness
of breath. He was diagnosed with Pneumocystis jiroveci
pneumonia (PCP) and started on appropriate antibiotic therapy. On
physical examination, in addition to abnormal pulmonary findings,
the patient had multiple flat purple skin lesions on his left thigh
and several perianal lesions. Given his advanced AIDS, the medical
team was concerned about Kaposi's sarcoma and human papillomavirus
(HPV) infection, respectively. The dermatology service was
consulted, and they performed biopsies of both lesions.

The patient continued to receive treatment for
PCP and was
slowly improving. Three days later, the intern on the team was
reviewing the patient's clinical information in the hospital's
electronic medical record (EMR). She looked up the biopsy results
and discovered that the left thigh lesion was Kaposi's sarcoma and
the perianal biopsy showed squamous cell carcinoma in
situ
. Interestingly, there was a third biopsy result in the
electronic record, labeled "right neck" and reported as "basal cell
carcinoma." The intern didn't recall any neck lesions (or
discussion of a third biopsy), but questioned her memory as it had
been a busy call night. She noted the results and went to see other
patients.

The patient's primary care
doctor (who was not directly caring for the patient in the
hospital) visited the patient and looked at the medical record
before seeing him. He noted the PCP diagnosis, a low CD4 count, and
biopsy evidence of three separate cancers. Given the patient's
end-stage AIDS and these new diagnoses, the primary care doctor met
with the patient and recommended hospice care. He told the patient
that, with "cancer in three places," his overall prognosis was
poor.

That afternoon, the inpatient medical team
recognized the error—the neck biopsy had been performed on
another patient and accidentally entered into this patient's
medical record. The team and the primary care doctor all met with
the patient to disclose the mistake, but clearly the error had
caused the patient tremendous pain and mental anguish.

On further investigation, it became clear that
the dermatopathology department was unaware of the error. Their
department used a standalone software program to track and report
biopsy results, a system whose results were electronically "dumped"
into the hospital's EMR. But the department physicians and staff
didn't have access to the hospital's EMR. In fact, when called and
asked if they had seen the error in X (the name of the EMR), the
pathologist responded, "What is X?" Eventually, it was determined
that the third, incorrect biopsy result had been entered into the
pathology software under the wrong patient identifier and then
uploaded into the hospital's EMR.

The Commentary

This case is an opportunity to examine patient
identification mix-ups within electronic medical records (EMRs) and
their impact on patient safety. A naïve view is that this case
demonstrates the dangers of EMRs. I argue that the EMR did not fail
here. Rather, the errors were in (i) weak linkages among
computer systems, (ii) insufficient safeguards against patient
misidentification, and (iii) poor hospital work-processes and data
fragmentation.

The Case for
EMRs

Health care providers and systems have been
increasingly urged to adopt EMRs. The federal government and many
payers seek to subsidize and/or reward physicians' EMR use and to
penalize "failure" to use EMRs. These incentives markedly increased
recently with the passage of the Obama administration's stimulus
package, which set aside $19 billion to promote health care
information technology.(1)

Some aspects of this growing
emphasis are understandable and justified. EMRs, also called
electronic health records (EHRs), offer many benefits. They
facilitate:

  • Easier and more accurate record keeping
    and scheduling;
  • Automation of lab orders and
    integration of lab reports;
  • Links to pharmacies (including
    electronic prescribing) and computerized physician order entry
    (CPOE);
  • Chronic disease management
    tools;
  • Integration of decision support
    systems (DSS)—alerts and reminders for providers to improve
    the quality of care, reduce medication costs, eliminate redundant
    tests, and prevent errors;
  • Epidemiological analysis of data
    from targeted or broad populations of patients; and
  • Potentially safer and less
    expensive care.

Despite their many current
and potential benefits, EMRs are found in few hospitals or
physicians' offices. Experts estimate that their prevalence ranges
from less than 2% of hospitals (for fully operational EMRs) to as
high as 16% for EMRs in physicians' offices.(2-4) This sluggish implementation means that, although
EMRs carry a tidal wave of expectations, to date they have left
only small puddles in doctors' offices and hospitals. Supporters of
electronic records blame the low adoption rate on physician
stinginess, technophobia, and timidity, but the reality is more
complex. Clinicians are also reacting to well-documented EMR
implementation and usability difficulties, which I will review
below.(2,3)

Potential Challenges with
EMRs

Despite their possible benefits and the push for
implementation, EHRs are associated with a number of potential
problems. This case in particular raises some systemic issues about
EMRs in situ—independent of any faults with the EMR
itself.

First, sending data across
computer systems is often a perilous journey, with possible
distortions and uncertain arrivals. As in this case, these sorts of
errors are hard to track—there is rarely an alert for
information not received, or received but inaccurate. Also, what
about the other patient whose dermatopathology findings never
arrived?

Second, EMR installation (like
most health care information technology [HIT] installations)
usually takes many months to a few years. The EMRs must work with
the hospital's several other computer systems, which are often
implemented and updated on overlapping schedules. As
implementations and updates proceed in parallel, software and work
processes change in ways that create barriers to smooth and
fail-safe communication. Systems that harmonized on Monday might be
incommunicado on Wednesday. And, in a complex environment with
multiple systems, re-programming or software changes to fix one
issue can easily create more problems.

Third, unintended consequences
are the rule in EMR implementation. There is no map of the myriad
hospital and office processes that affect, and are affected by, EMR
use, data input, and data output. Despite intensive planning,
unforeseen problems inevitably arise, e.g., a requirement to
provide a diagnostic code to order a lab test may have tragic
consequences when the "made-up" (but erroneous) diagnosis becomes
embedded as part of the patient's medical history.

All HIT implementations require
never-ending vigilance and on-the-floor observation to "get it
right." Clinicians are also inventive creators of workarounds when
faced with system barriers—their need to help patients
supersedes HIT protocols.(5)
Hence clinicians, for example, will reach into one patient's
medication supply to grab a prepared IV bag for another patient in
urgent need. While the clinicians undoubtedly feel that they are
supporting their patient's needs, this type of "swap" can play
havoc with the EMRs, which are often integrated with the hospital's
CPOE, pharmacy dispensing system, and medication barcoding
system.(4)

Next, most EMRs undergo massive
customization during implementation (3,4,6), involving how information is displayed, order
sets, warnings on drug–drug interactions and dosages,
permission requirements, and linkages to other hospital IT systems.
With millions of lines of computer code, error possibilities are
staggering. Linkages with other new systems escalate error odds
further.

Another problem with EMR
implementations involves the absence of standard definitions and
processes. For example, there may be more than 20 ways of writing
the same patient's name in any one hospital (Table). In addition, long names may be truncated by
EMRs, patient record numbers or room numbers may be attached to
names as suffixes, and innumerable misspellings and
transliterations are commonplace in EMRs and can set up medical
mistakes. The variety of ID numbers for each patient is also
disconcerting. Name and ID matching errors are common. In my own
studies of a CPOE system (7),
almost all residents reported having accidently entered orders in
the wrong patient's electronic chart at one time. While these house
officers uniformly reported catching their errors before closing
the electronic chart, it goes without saying that they probably
didn't catch the errors they did not notice. Barcoding offers some
protection yet generates scores of other error-risks.(5)

Also, data displayed in
electronic records always appear neat and tidy (as I'm sure the
display of the neck biopsy results appeared in this case). These
entries offer no clues that might reveal a patient identification
error. For example, data displayed in the chart are usually without
additional identifying information such as room number, referring
physician, or other illnesses—items that might trigger a
viewer's questions. (Note: remember that paper-based records came
with many problems, too.)

Finally, many health care
organizations keep their electronic systems in silos. In this case,
the inability of the dermatopathologists to check the patient's
record is remarkable. If the hospital had installed the full EMR in
that lab, dermatopathologists could review cases and see, for
example, that this patient had no neck lesions. One could imagine
that the cost of making the core EMR system available to different
departments and labs would be far outweighed by the benefits to
patients and clinicians.

Some might argue that improved
interoperability—creating seamless connections among computer
systems—would help prevent problems like the one in this
case. But interoperability could also create a river of information
that would flood clinicians with massive amounts of data, not just
what is needed to improve patient care. More data do not
automatically equal better care. In this case, the (wrong) datum
seamlessly moved from the dermatopathology lab's computer system to
the EMR. It was then correctly placed into the appropriate spot in
the EMR where it was easily read by the patient's physicians. It
has been said that computers can be extraordinarily efficient error
propagators—incorrect data flow as easily as valid
data.

Moreover, having more computer
systems feeding information to EMRs also increases the possibility
of misplaced information. Thus, the efforts of industry-led
certification groups to enhance information exchange across HIT
platforms may unintentionally produce EMRs populated with
information that is neither well integrated nor well displayed. To
date, most vendors have focused on integrating data within their
own suites of programs, not across platforms. Patient information
that is displayed to physicians in unfamiliar formats may be of
limited value. Inadequate integration of patient information from
several systems may produce a patchwork of confusion rather than a
gain in clinical insights.

This case demonstrates the
phenomenon of an error that is associated with an EMR, but one in
which the EMR software itself did not make an error. Such
EMR-related errors are common and must be addressed if we are to
fully benefit from this valuable technology. Common EMR-related
errors include: (i) the often labyrinthine EMR user interfaces,
where essential information might be on screens seven clicks away,
or where finding it may require endless scrolling and searching;
(ii) graphic displays (user interfaces) that obscure or separate
essential data; (iii) the challenges of accurately entering
clinical data while directly caring for (and talking with)
patients; (iv) disruptive decision-support warnings or unwanted
suggestions (of which about 85% are ignored), leading to "alert
fatigue" (8);
and (v) internal software mistakes, such as incorrectly using body
weight in pounds to compute suggested dosages rather than the
kilograms the software expects.

Although the present case was
not a true EMR error, we have found many examples of errors like
those listed above in our studies of EMRs.

Solutions

Although some question the wisdom of the proposed
vendor/government-supported forced march toward widespread EMR
implementation, experts in the field are unequivocal in supporting
EMRs as patient safety and efficiency tools. Our recommendations
are:

More EMRs and better integrated EMRs.
Integration of the dermatopathology lab into the hospital's EMR
system would have reduced the probability of this error because the
physician entering the lab results would have been able to see the
patient's record.

Better EMRs. The good news is that many
of the EMR graphic display sins and other difficulties are largely
addressable. If vendors were more aggressive in repairing clunky
interfaces and functions, they would not have to pressure
clinicians to use EMRs via governmental rules and subsidies.
Instead of blaming clinicians' character flaws for the tepid
adoption of EMRs, vendors should look to their own failures in
quality, design, and responsiveness.

Smarter EMR implementations.
Implementations are more wrenching than they need to be. Each
implementation is treated de novo; there is little
learning or sharing among medical facilities. This isolation must
stop. Hospitals and clinicians should own the process of
implementation and should benefit from each other's experiences.
Agencies such as AHRQ can help by facilitating research and
information exchange.

Attention to EMR-generated errors. Most
research on EMRs has focused on documenting the advantages over
traditional paper systems. Let's declare victory in that battle,
and now examine the neglected reality of EMR-generated errors. As
these errors are revealed, we must seek solutions for them.

Less unnecessary customization.
Customization is often a marketing ploy and is always a
double-edged sword. EMRs that are well designed from the get-go
would eliminate many of these struggles. Some customization may be
necessary, as in the need for weight-based algorithms for
medications in children's hospitals. But each children's hospital
need not build these complex algorithms in isolation. Even systems
that require customization will ultimately benefit from standard
user interfaces and core processes.

Unique patient identification. No
technical fix can eliminate the need for careful name and patient
identification. Because the United States prohibits a unique
medical ID, as EMRs and interoperability grow, identity errors will
increase, causing more avoidable errors and death. There is no
facile solution for this dilemma. Privacy advocates offer good
arguments about the dangers of a unique medical ID. On the other
hand, ID errors kill, injure, and cost billions of dollars. We must
search for a reasonable solution.

That the medical team found the error is perhaps
the most encouraging part of this case. It illustrates the caring
and professionalism of dedicated clinicians—despite, or in
addition to, the many benefits and promises of EMRs.

Take-Home
Points

  • Medical informaticists have focused on
    demonstrating EMRs' many advantages over paper patient charts, but
    have generally ignored EMR-related errors.
  • Interoperability will offer much
    information from many sources. We must ensure, however, that EMR
    information is integrated and displayed in ways that work for
    clinicians.
  • With the efforts toward
    interoperability, EMR information mix-ups may increase unless we
    find a workable way to achieve unambiguous patient
    identification.
  • The apparent neatness of information in
    EMRs may obscure data of uncertain accuracy. Similarly, poorly
    displayed data or data unnecessarily spread across many screens can
    generate errors.
  • EMRs are undoubtedly better than paper
    records. The funding to encourage EMR use, however, should not
    force us to prematurely implement inadequate EMRs.

Ross Koppel,
PhD
Principal Investigator
Study of Hospital Workplace Culture and Medication Errors

Center for Clinical
Epidemiology and Biostatistics
School of Medicine

University of
Pennsylvania
Professor

Sociology Department
University of Pennsylvania

References

1. Werner E. Questions surround health IT money.
Washington Post. March 23, 2009. [Available at]

2. Jha AK, DesRoches CM, Campbell EG, et al. Use
of electronic health records in U.S. hospitals. N Engl J Med.
2009;360. [go to PubMed]

3. Hoffman S, Podgurski A. Finding a cure: the
case for regulation and oversight of electronic health record
systems. Harv J Technol. 2008;22:104-164. [Available at]

4. Harrison MI, Koppel R, Bar-Lev
S. Unintended consequences of information technologies in health
care—an interactive sociotechnical analysis. J Am Med Inform
Assoc. 2007;14:542-549. [go
to PubMed]

5. Koppel R, Wetterneck T, Telles
JL, Karsh B-T. Workarounds to barcode medication administration
systems: their occurrences, causes, and threats to patient safety.
J Am Med Inform Assoc. 2008;15:408-423. [go to
PubMed]

6. Dixon BE, Zafar A. Inpatient
Computerized Provider Order Entry: Findings from the AHRQ Health IT
Portfolio. Rockville, MD: Agency for Healthcare Research and
Quality; 2009. AHRQ Publication No. 09-0031-EF.

7. Koppel R, Metlay JP, Cohen A,
et al. Role of computerized physician order entry systems in
facilitating medication errors. JAMA. 2005;293:1197-1203. [go to
PubMed]

8. Sinsky CA. E-nirvana:
are we there yet? Fam Pract Manag. 2008;15:6-8. [Available at]


Table


Table. Example of how a
hospital might write one patient's name 22 different ways.

  • J N Tobias
  • J Noel Tobias
  • J Tobias
  • JNTobias
  • Jonah N Tobias
  • Jonah Noel Tobias
  • Jonah Tobias
  • JonahNTobias
  • JTobias
  • Tobias J
  • Tobias J N
  • Tobias, J
  • Tobias, J N
  • Tobias, J Noel
  • Tobias, Jonah
  • Tobias, Jonah N
  • Tobias, Jonah Noel
  • TobiasJ
  • TobiasJN
  • Tobias J N
  • TobiasJonahN
  • TobiasJonahNoel