All posts for the month March, 2013

Medicine is about telling stories. The story is how the patient was yesterday, how they are now, and how you think they will be tomorrow. EMR’s are tables of data. Tables of data do not tell stories, hence the problems.


When you think about how we practiced medicine just 5 years ago, you would go examine a patient you knew (since there was a good chance you took care of them yesterday and the day before), you already had a feel of their trajectory so you pulled the data you needed to confirm or throw out your impression of their progress. You then grabbed a sheet of paper and wrote that story. If you were a good doctor, you scanned the other data that may relate to their progress, just to make sure something wasn’t sneaking up on the patient.


How do we do it today? The patient is admitted and 30 demographic fields are checked and filled in, then ~50 clinical fields from the nurse, then 20 fields from the MD,  then, depending on the clinical status of the patient, another 50 or so lab, study, and consult fields are filled in. This data is placed in tables in a database so it can be displayed and retrieved.


When the doctor comes in, you review 20-30 of these 120 fields and the fields you look at are greatly affected by how they are displayed (the things right in front of you are seen first, the ones under the bottom of the screen are not. There is no synthesis of data, just upper and lower limits to the field data that will turn the font red is the datapoint is on either side of the cutoff.


It is amazing today to see people valiantly take this process, and try to tell a story. Nurses try to do it, but they are hampered by the number of fields they need to fill in during their shift.  I see this every day, sitting in front of the computer checking boxes instead of doing assessments. This would be OK if the box checking actually improved your situational awareness of the patient’s condition, but it almost never does.


Medical students have essentially given up trying to learn when they need to present. There is so much data they need to deliver, especially in the ICU, that they have almost no inkling of how to formulate a hypothesis and tell a story. I am constantly asking students “what do you think is going on?” and I constantly get blank stares.


As residents move through training, they adapt to this process change, but it takes more time for them to become facile with assessment. They too often have to sift through reams of data, and through that search they forget what they were looking for in the first place, or miss the kernel of truth in the mountain of data.


I just saw a YouTube video for a new product that allows you to render your patients monitor on your iPhone. Great idea, but more un-synthesized and uncorrelated data for you to deal with. Are you supposed to look into your iPhone and switch between patients monitors while walking down the hall? What are the nurses for? How do you reconcile the data on your screen with the true state of the patient? How many false positive and false negative will there be for detecting serious deviations from the predicted course.


What do we need out of EMRs and patient data systems? SITUATIONAL AWARENESS. The “state of the patient” at this point in time. This will truly be different for every patient, but it is not impossible to codify. What I find funny, is that when I tell physicians we need to put these things down on paper, they say “that’s impossible, it comes from experience and it’s a gestalt” but many studies have shown that is not the case. People think in algorithms (if A, then B, if not B, then C). We all do it. What you need is the data to tell you that the state of the patient is “A” (that is a story) then you decide that you want the state in 8 hours to be “B”. You picture in your mind how you are going to get to B and you tell that story to the nurse, who implements the plan. It is very possible to quantify these states, but there needs to be room for judgment since patients will reflect their state differently. That doesn’t mean that we cant do a “Mad lib” for a certain state, where you would have to fill in the blanks to tell the story. If the story makes sense, you are at that state, if not, then not.


For instance, you have a post-operative heart patient. The story is “this 78 year old woman had an MVR-CABG yesterday. She is awake and talking, her chest xray shows some lower lobe consolidation, she is afebrile, her WBC is normal, and she is making good urine. She is not doing well on spirometry (250cc) but she denies pain (her pain scale is 4)”. That is this patient’s story. An experienced practitioner reads into that story that she is doing well, but she is at risk for pulmonary complications. You fill in the story for today:

“We need PT to see her and get her out of bed and even walk her if possible, she needs to do spirometry 10 times an hour and hit a goal of 500-750cc. We may need to get a pain consult, or give her a PCA since she may be stoic but is really in discomfort. If she has a fever she needs to be cultured. Lets get a white count and a CXR if there are any problems, and definitely get both in the morning.”

This story focuses on her pulmonary toilet, but leaves room for negative deviation, where she may be manifesting signs of pneumonia, and what we will do if that happens.


How do the EMR’s get us to this situational awareness? I believe they hinder it by presenting data in an unfocused way, and by forcing practitioners to spend time obtaining data that brings little or no value. It is also the fallacy of our current regulatory environment in medicine. More discrete data allows us to do more measurement. The measurement allows you to be graded, and grading allows you to be rewarded or penalized. In the end, it makes it easier to measure what we’re doing, but harder to get better. It also makes it easier to give you a grade.


We need to work to adapt EMR’s to what we need, and less to adapt clinicians to the EMR’s. There is still little data that EMR’s improve safety beyond certain discrete areas (medication safety and accurate electronic ordering). There is no evidence that information transfer is enhanced or that the EMR’s augment our ability to quantify the state of the patient.


Thus, we must look into the way we practiced before the EMR to identify what allowed us to get situational awareness. We then need to talk to each other to align some general principles, and work with the programmers to get the EMR to enhance this process, not detract from it. Additionally, with the large number of handoffs in patient care, it is VITAL that we be able to relate accurate stories to each other, and not just regurgitate data. If we don’t do that, I think there will be serious consequences for the safety of our patients.