A few thoughts on mediocre
information handling at a premier hospital.
As I describe in an earlier blog entry,
I am at home, in the process of recovering from cardiac surgery at
Massachusetts General Hospital (MGH), one of the premier hospitals
in the country. I believe I have gotten excellent care, but
thought it might also be useful to record some of my less favorable
impressions of the way computers are and are not used in the care
process.
As I have known from previous encounters with MGH, it deploys a
large variety of poorly-integrated electronic medical record systems
to support its operations, although much of record-keeping is still
manual and on paper. The electronic systems include
eMAR, an electronic medication administration system, used in
almost all Partners Healthcare inpatient settings,
a GE cardiac monitoring system
a lab system that permits retrieval of lab results,
a PACS system that allows access to imaging data that
originates at MGH or Partners,
LMR, a longitudinal medical record, used for outpatient
clinics,
OnCall, another clinical record system used mainly in
outpatient clinics, but also said to support cardiology and
cardiac surgery
A pharmacy system to support in-patient drug orders
and many others.
During my 18 days of hospitalization, the only ones of these I saw
used extensively were the eMAR and the cardiac monitor. The nurses
use eMAR regularly to check on what drugs I should be getting and to
record that they have given them. At least in the units where I was,
each bedside has its own computer terminal that normally runs eMAR.
It includes a bluetooth bar-code scanner that allows the nurses to
identify the unit dose packaged drug that is about to be given, as
well as the fact that it’s being given to the right patient by
scanning the patient’s wrist-band. Even this system is
“fiddly”, so my impression was that it took several minutes for a
nurse to enter all the information and electronic signatures needed
to enter data, and occasionally I saw nurses frankly frustrated by
the system. For example, my first night in the observation
unit, my nurse was clearly unfamiliar enough with the system that it
took her 15 minutes to give me a single pill. She explained that she
normally worked in the ED proper, where they do not use it, and
although she had taken a training course, she hadn’t become
comfortable with the system.
I later spoke with one my my very friendly nurses, to ask about
problems she perceived with the system. One she mentioned as a
frequent cause of complaint is that the eMAR and the pharmacy system
are not well integrated. Thus, if a doctors places or changes an
order, the eMAR believes that the new drugs should be given right
away, but normally it takes the pharmacy several hours to deliver
the order, by which time eMAR has timed it out. This then requires
exception reporting, perhaps re-confirmation, and a much longer
interaction.
On another occasion I noted that contingent orders are not handled
well in this system. Before surgery, my heart rate was low, which
made it difficult to give me beta blockers. These, despite their
beneficial effect of reducing demand on the heart, also further slow
the heart beat. So the initial order for a short-acting beta
blocker (metoprolol) stood on my meds list for a week, but with a
comment that it should only be given if my heart rate exceeded some
threshold, which I think was 55. During this time, I never got the
drug because that comment always excluded it, and each time the
nurse had to enter explanatory evidence. After a while, my
docs thought that perhaps they could get a little of the metoprolol
into me by giving me very small doses (6.25mg) at 8am and 4pm, but
not near night-time when I was most bradycardic. However, as
far as I could tell, they meant this order to be a fall-back to the
original order. Looking over the nurse’s shoulder, it seemed
to me that there were now two unrelated metoprolol orders on my
list, requiring her to make the logical connection between them. An
algorithmic specification, of nested if-then-else statements, was
clearly not among the system’s representational capabilities.
The only other uses I saw of the terminal at my bedside was when a
nurse looked up a lab result I had specifically asked about, and
when one of my doctors looked up a chest x-ray to tell me it was
normal.
The cardiac monitoring system is completely separate from the
others, with its own central monitoring stations, ECG printers, and
even wireless networking hubs to permit monitoring of ambulatory
patients. I was reminded that 20 years ago, one of my VI-A
(industrial co-op) students worked with HP on an infrared system to
monitor the location of each wireless transmitter after a patient
had been seen dying on the central monitor but could not be located
on the unit to save his life. I’m sad to report that two
decades later, the current GE system is still vulnerable to just
such an event. If the patient wearing it leaves the ward, its signal
drops out completely. On the ward, the signal is monitored, but the
monitor has no idea where the patient actually is.
The other high-tech equipment that showed up in my room from time to
time was a bank of continuous infusion pumps for IV fluids and
drugs. Again, we have known how to connect such pumps by
serial lines and networks for decades, but they are not so
connected. The eMAR system is used to track IV’s being
administered, but there is no feedback from the pump to any other
system. If the pump senses a problem (obstruction, bag empty,
etc.), it simply beeps until the patient calls the nurse.
All other tracking by nurses, including vital signs, temperature,
pulse oxymetry, and fluid input/output are done completely manually,
entered on paper flow-sheets in the patient’s chart. Even
when, say, an SaO2 reading or blood pressure is made by an
instrument connected to the GE monitor, it only becomes part of the
patient’s record by being manually transcribed by the nurse.
One morning my chart accidentally wandered away from my room, and
for the next several hours various scraps of paper accumulated near
my bed until the chart was found and could be updated.
Before my surgery, I wandered around the ward and was curious to
understand what was being done at all the various computers and
workstations in addition to the ones by the patients’
bedsides. One nurse, sitting in at one of these in the
hallway, explained to me that she was typing up her notes in
Microsoft Word so she could report to the incoming shift on her
patients. These machines are, apparently, networked to some
ward-level file server, where her word files are stored, but those
do not become in any sense part of the patient record. They do
not show up in LMR, OnCall, or any other institutional system. It is
only by being printed and placed (as a paper sheet) in the chart
that they acquire any significance.
I also noticed the attending, residents, physicians’ assistants and
others looking up data on central terminals, but my impression is
that they then wrote notes on paper to record their thoughts. Oddly
as a patient, these events were rather distant from what I normally
encountered, so I got little evidence. Even when I saw my team
round, often they would stand outside in the hall discussing my
case, and only one of the junior people would come in to deliver the
word. A happy exception was my surgeon, who was quite
courteous and stopped by when she was rounding, and the cardiac
Fellow, who was quite helpful throughout my stay. I suspect that
being Professor X rather than Joe X may have made a difference.
I was also taken aback a couple of times when I would ask one of the
nurses to let me know the value of a lab result or the
interpretation of some test and she would blow me off, as if it were
none of my business. Generally, the nurses were quite friendly
and helpful, and sometimes would even use circumlocutions to answer
such questions, but clearly they had been taught that the patient
has no right to know his own data. I also asked my attending at one
point whether MGH had yet implemented a patient portal such as the
one at BIDMC that provides patients access to such data, but he
thought not, except on certain wards or for patients whose primary
care providers practiced at MGH (which is not my case). So, I
was effectively denied access to my own data, though someone did
helpfully point out the cumbersome and expensive procedures whereby
I could make an official request for a copy of my chart after my
discharge.
Finally, I should comment on the difficulty I perceived in bringing
data into the system. While waiting in the ED when I first arrived,
it had occurred to me that MGH might find it useful to see the
results of the ECG and x-rays done at MIT that afternoon. I
found that the ECG, at least in the form of a Xerox copy of the
paper tracing, had been brought over by the ambulance crew and had
become part of my record. I had phoned the doctor who saw me
at MIT and asked if he could make my x-rays available to the MGH
crew, but he could not think how he might accomplish this. So
I asked him to simply email the digital images to me, because I had
my laptop with me and MGH kindly provides gratis WiFi to their
patients and visitors. Sure enough, I got the x-rays by email, and
downloaded the Osirix software that makes them conveniently
viewable. However, it turned out to be impossible to get anyone to
look at these! In fact, in the middle of the night, when the
more sensitive troponin test revealed my stressed cardiac cells, I
was admitted, moved to the Step-Down Cardiac unit, and re-x-rayed,
even after my nurse had tried to take my digital x-rays (that I had
put on a USB memory stick) to someone who would interpret
them. Two days later, one of Zak’s colleagues, who was
informally helping to educate me, did look at them and agreed that
they had been perfectly adequate x-rays. I don’t know what an x-ray
costs, but I suspect it must be a couple of hundred dollars, to pay
for the technician who took it, the portable x-ray unit that had to
be wheeled to my room, and the radiologist who reads them.
Plus I got an extra dose of radiation that I had no need for.
I spoke about this with several people, including one of the
visiting MIT docs who stopped by to see me daily and one of my
colleagues at MGH who was not involved in my care but checked in to
see me often. He, being the most cynical, pointed out that if
they simply looked at my x-ray, they could not charge me for it,
whereas by taking their own, it’s a billable event. He thought
that there might be a suitable billing code for re-interpretation of
an existing x-ray, which might have worked, but this is clearly not
part of the workflow. Others speculated that MGH would be leery to
rely on an x-ray with poorly established provenance—after all, I
might have Photoshopped it! This seems pretty silly,
especially because DICOM contains adequate cryptographic mechanisms
to prove provenance, if used properly.
My MIT visitor also pointed out the digital x-rays taken at MIT are
shipped (electronically) to the Mt. Auburn for interpetation, which
of course is part of CareGroup, not Partners, and thus a competing
institution. Furthermore, even those images are not stored at the
Mt. Auburn because MIT has calculated that they can run their own
server to store them much more cheaply at MIT than to pay the
something like $5/image fee at Mt. Auburn. However, the MIT
system is only accessible within MIT. The two images total 5.4MB,
Given the cost of raw storage at under $100/TB, the marginal cost of
storing these data is zero, as Google and Microsoft well
understand. Yet some artificial pricing structure helps make
my data unavailable to my doctors.
I was pleased a couple of days later when one of the residents
preparing for my surgery was asking me about whether I had any
problems with other internal organs, and I was able to show her a
spiral CT scan of my kidneys downward that had been taken on the
occasion of an unexplained episode of gross hematuria about a month
or two ago. Of course she did not need to see this, as it was
a pretty simple screening question, but I felt like I had gotten my
money’s worth from carrying the 300MB data set around on my laptop. Back to Blog
