A Symptom Is Not a Sign. A Sign Is Not a Clue. And an Inference Is Not a Fact.

How clinical expertise can make us better investigators — and occasionally make us too certain of what we think we know.

There is a peculiar problem that comes with expertise.

The better we become at recognizing patterns, the less conscious we become of the steps we took to recognize them.

For a clinician, this is usually a very good thing. A nurse with twenty years of critical care experience does not approach a deteriorating patient the way a first-semester nursing student does. She does not need to consciously work through every possibility before recognizing that something is wrong. She sees the blood pressure. The heart rate. The skin. The urine output. Perhaps something subtler that she cannot immediately articulate.

She recognizes the pattern.

That ability is one of the great advantages clinicians bring to legal nurse consulting and medical record investigation.

It can also be one of our biases.

Because investigation demands something that clinical practice does not always demand of us: that we slow the reasoning process down, take it apart, and distinguish very carefully between what we observed, what someone reported, what we inferred, and what we can actually establish as fact.

Those are not the same things.

Consider a patient who reports crushing chest pain.

The pain is a symptom. It is the patient's reported experience. If the nurse documents the complaint, we can establish that the patient reported crushing chest pain at a particular time. That is not quite the same proposition as independently establishing the physiological nature or cause of the pain.

Now suppose the same patient is diaphoretic, has a blood pressure of 82/48, and has ST-segment elevation on an ECG.

Those are signs. They are observations or measurements rather than subjective experiences.

But even here, the investigator should resist becoming too comfortable.

Who observed the diaphoresis? When?

Was 82/48 a single measurement or part of a trend? Automatic cuff or manual? Was it repeated? Was there a later correction?

When was the ECG performed? When was it interpreted? By whom?

The numbers may be sitting plainly on the page. Their significance is not necessarily sitting there with them.

And then there is the clue.

I think we sometimes use that word too casually.

A clue is not simply another name for a sign.

A symptom can be a clue. A sign can be a clue. A timestamp can be a clue. So can a medication discontinuation, an amended note, a copied-forward phrase, a contradictory flowsheet entry, a gap in documentation, or an order that appears at a rather inconvenient time.

Sometimes the most interesting clue is something that is not there at all.

What makes something a clue is not its clinical classification. It is its relationship to a question we are trying to answer.

And that is where the clinical mind can get itself into trouble.

Imagine the record shows tachycardia, hypotension, cool skin, and declining urine output.

An experienced clinician may recognize the pattern almost instantaneously.

Shock.

That may be an entirely reasonable clinical conclusion. It may ultimately prove to be the best explanation of the evidence.

But shock was not one of the observations.

It was an inference drawn from them.

That distinction sounds painfully obvious when written out. In practice, it is remarkably easy to lose.

Clinical expertise teaches us to recognize patterns.

Investigation requires us to take those patterns apart again.

That is harder than it sounds.

The experienced clinician has spent years learning to compress information. We see constellations rather than isolated stars. Eventually the intermediate reasoning becomes so familiar that we barely notice ourselves doing it.

A novice may reason:

Observation A.

Observation B.

Observation C.

A, B, and C are consistent with D.

Therefore, D should be considered.

The expert sees A, B, and C and thinks:

D.

Again, that is often precisely what expertise is supposed to accomplish.

At the bedside, speed matters.

In an investigation, however, the compressed reasoning itself needs to be examined.

What exactly supports D?

What contradicts it?

What else could explain A, B, and C?

What evidence would I expect to find if D were true?

Is that evidence present?

More importantly, what evidence would cause me to abandon D?

That last question is one we probably do not ask ourselves often enough.

Once we have constructed an explanation, the human mind becomes remarkably good at finding additional evidence that fits it. Clinical training does not immunize us against anchoring, confirmation bias, premature closure, or hindsight bias. In some circumstances, expertise may make those problems more difficult to recognize because the inference feels less like an inference.

It feels like recognition.

And recognition feels very much like knowing.

This is one reason I believe medical record investigation requires a slightly different mental posture from medical record review.

The question cannot immediately be, What does this mean?

First:

What do I actually know?

Take a seemingly simple statement from a record:

“The patient was confused.”

What exactly have we established?

Perhaps the nurse documented that the patient was confused. That is certainly evidence.

But suppose elsewhere the record tells us that the patient repeatedly attempted to climb out of bed, removed a telemetry lead, and asked where his wife was despite having been told she had gone home.

Now we have something different.

We have documented behaviors.

From those behaviors, someone may characterize the patient as confused.

From confusion, a clinician may infer delirium.

From delirium, someone may eventually infer medication-induced delirium.

Look at how far we have traveled.

Documented behavior.

Clinical characterization.

Clinical inference.

Causal inference.

Each step may be reasonable.

Each step may even be correct.

But they are still steps.

The danger comes when we collapse them into one another and present the final conclusion as though it had been sitting there in the record all along.

This is why one of the simplest exercises in investigative thinking is also one of the most useful.

Take a section of a medical record and separate what you find into three categories:

Observation. Inference. Support.

Then impose one rather unforgiving rule.

Nothing goes into the Observation column if it requires you to interpret what happened.

You will probably discover that this is considerably harder than it sounds.

Experienced clinicians are trained to interpret. We are supposed to interpret. Patients would be poorly served if every nurse stood at the bedside contemplating epistemology while the blood pressure continued to fall.

But the legal nurse consultant is operating in a different environment.

We have the luxury — and I would argue the obligation — to slow the process down.

The medical record is not merely a clinical narrative waiting to be summarized. It is evidence created by different people, at different times, through different systems, for different purposes, under imperfect conditions.

Some of it is observation.

Some is recollection.

Some is measurement.

Some is interpretation.

Some is inference.

Some is copied from somewhere else.

Some may be wrong.

And some of what matters most may be missing.

Our clinical training helps us understand all of it.

Our investigative discipline should prevent us from pretending those things are equivalent.

Sherlock Holmes famously made much of the distinction between seeing and observing. I think there is another distinction worth adding for those of us who investigate clinical evidence.

Observation is not inference.

Inference is not fact.

And expertise, valuable as it is, does not exempt us from showing our work.

Perhaps it makes showing our work even more important.

The clinician asks:

What does this pattern mean?

The investigator first asks:

What do I actually know?

The best Legal Nurse Consultants need to be able to do both.