Monday, March 12, 2018

Can data in electronic medical records be trusted?

If the subject is respiratory rates, the answer is “No” according to a recent study. The authors reviewed the records of 28,500 patients over 220,000 hospital days and found recorded maximum respiratory rates “were not normally distributed but were right skewed with values clustered at 18 to 20 breaths per minute.”

The figure below shows maximum respiratory rates compared to heart rates which were normally distributed.
Reminder: Most sources say the normal respiratory rate of an adult ranges from 12 to 20 breaths per minute.

Other unusual findings were as follows. During 75% of hospital days, the maximum respiratory rate was 18 or 20, minimum and maximum respiratory rates were equal on 26% of days, and variations in rates decreased over the course of hospitalizations.

Respiratory rates did not vary much “even among those with cardiopulmonary compromise or immediately prior to ICU transfer.”

In their discussion, the authors pointed out that these inaccuracies may result in patients who are becoming tachypneic—say with respiratory rates rising from 12 to 14 to 20—being discovered too late to avoid intensive care.

The study is limited because it was impossible to know if individual respiratory rates were measured correctly. However, the abnormal distribution with rates of mainly 18 and 20 and similar variations regardless of clinical condition seem convincing.

The findings raise a question. Does the electronic medical record contain other inaccurate data?

I believe so. What do you think?

Thanks to Twitter friend @MaryDixonWoods for letting me know about the study


Glen said...

PERRLA, and arterial pulse are more often clicked through in EMR than performed carefully. In clinic, weight loss counseling is required by for certain BMI ratings-but I doubt if ophthalmologist spend much time on that.

Skeptical Scalpel said...

Yes, you can add most of the review of systems, sexual history, domestic abuse, and smoking history too.

Anonymous said...

Heart rate is accurate because it is accurately captured: by way of the pulse-oximetry machine. Respiratory rate is usually just made up by the person who input data.

Don't throw out the baby with the bathwater.

Skeptical Scalpel said...

Anon, I agree with you. As I said in the post, the respiratory rate issue prompted questions about other EMR data. It's so easy to click boxes just to get through the fields. The first two comments mentioned some of the other areas of doubt. Someone is collecting all this data the doctors and nurses input. It's too bad no useful studies will result.

Jeremy Reppy said...

The value and accuracy of EMR is heavily dependent on the person providing the information. Many medical providers “copy and paste” large sections of material from one treatment note to the next, so it can be difficult to tell what information is new.

Anonymous said...

Fair points about "EMR fatigue" and the potential for careless data input, but I think the results for the respiratory rate distribution make sense.

If the breath rates were measured by counting breaths during 30 seconds, you would get peaks on even numbers at the most commonly measured values (in this case 16, 18, and 20) rather than a normal distribution.

Even if the breath rates were calculated exactly, the rounding would produce peaks at whole numbers in the same region of the graph rather than a nice normal distribution. In contrast, the heart rates yield a smoother, more normal-looking distribution because there are more possible values on the x-axis.

In the end, I think it is impossible to tell from this data whether breath rates were measured correctly (this was probably done correctly but perfunctorily) and whether this impacted patient care. Per the graph, there were few tachypneic patients in the sample, so it is quite possible that patient care was not affected at all.

Skeptical Scalpel said...

Jeremy, I agree. I have blogged about the problem. Type "copy and paste" into the search field in the upper right area of my blog to read those posts.

Anon, you are right about the even numbers, but that's not the point. If the normal rate of breathing is 12 to 20 breaths per minute, why are the values clustered around 18 and 20? From the figure I posted, you can see that 22 and 24 have peaks as well. One would think they would be clustered around 16 or so.

william reichert said...

I believe the medical record , not just the ELECTRONIC medical record , is not to be trusted. For example, I examined a man with normal pedal pulses who I found had undergone an AK amputation in the past.
I examined a man complaining of abdominal pain with a recorded normal rectal exam who had a large , hard rectal impaction. And , most scary, I was asked to consult on a lady who had been putting out 6 liters of urine a day for a week.
The medical administration record did not note any diuretics being given.After ruling out other causes I went back to the bedside and discovered that a bag of dopamine was running
for the last 6 days.No note of it in the chart.
TheEMR is nice but over reliance on it as a statement of truth about a patient can be misleading. Be careful. Discovering errors like this may not make you any friends.

Anonymous said...

I'd also note surgical blood loss as a made-up number.

Skeptical Scalpel said...

I always said you should double the surgeon’s estimate and divide the anesthesiologist’s estimate by 2. The amount of actual blood loss is somewhere between the two numbers.

Post a Comment