Unnecessary Investigations Junior Doctors Order

Before ordering, ask what you would do differently; if nothing changes, the test is not indicated.

The only question worth asking before you order a test

Ask one thing before any request: what will I do differently depending on the result. If the honest answer is nothing, the test is not indicated. A normal value adds little, an abnormal value you would not act on confuses the picture rather than clarifying it.

In practice, that question is harder at 2am with a sick patient in front of you. It steadies if you name your working differentials and keep narrowing what it could be. For each contender, decide which result would move it, and which would not.

Diagnosica bakes this into practice cases. The library holds 74 cases, every case reviewed and signed off by a doctor, mapping 1,737 investigations as essential true or essential false. Of these, 92 are tagged red herrings, each with a reason stating what the result does not change.

Why juniors over-order

On a busy take, you're not yet sure which single result will change your working diagnosis, so you tick the whole panel and let the numbers narrow it for you.

Defensiveness is rational too. An omitted test is a visible fault at handover, a redundant one is invisible, so you over order to avoid the conversation tomorrow. You call them unnecessary investigations in hindsight, but in the moment they felt safer.

Culture matters. Nobody gets pulled aside for sending too much, and you've likely seen praise for thoroughness but only challenge for a miss, so you learn safety as action and volume.

Across Diagnosica's 74 signed-off cases, there are exactly 92 investigations tagged as red herrings, each with a written reason from the doctor who signed the case off. Read a few and you'll recognise the pattern above.

None of that makes the extra result mean anything on its own. Panels don't think for you, reasoning under uncertainty is how you build the pause where you ask what a single result would actually change.

A test that is sensitive but useless here

Martin Rowe, 55, is an AI patient in our library with two days of sudden breathlessness and chest pain. He is usually fit and thinks he is having a heart attack.

Three weeks earlier he had left knee keyhole surgery and has been less mobile since, and five days ago he flew home from Australia on a very long-haul flight. He has coughed up small amounts of blood on two occasions, only on a specific probe, and his observations are heart rate 112, respiratory rate 24, blood pressure 120 over 78, oxygen saturation 91 per cent, temperature 37.1.

The reflex is to send a D-dimer when you fear PE but are not certain, yet in this picture even a negative would not change where you go next.

The case author puts it plainly: "Sensitive but not specific: useful to rule OUT PE only when clinical probability is low/intermediate. With his high probability (Wells) it should not be used to decide. Proceed straight to imaging; a raised value here is expected and non-diagnostic."

That matches the NICE guidance on suspected VTE. It uses the 2-level Wells score, above 4 points is PE likely and a CT pulmonary angiogram immediately if possible, 4 or less is PE unlikely and that is where a D-dimer belongs, with the result available within 4 hours if possible.

In Martin's case the D-dimer comes back markedly raised, expected and non-diagnostic. A normal result would not have moved you either, because pre-test probability dominates what this assay can tell you.

On Diagnosica you can run a case and order the tests; the debrief shows which results added nothing in that scenario.

Where a D-dimer helps at low or intermediate pretest probability, and where it cannot decide.

A test that is tempting and changes nothing

Same patient, Martin Rowe, 55, knee surgery three weeks ago and a long-haul flight five days ago. The tempting add-on is a thrombophilia screen.

Here the result reads "Not sent acutely." It is flagged essential false, tagged a red herring.

"Tempting in a young-ish patient with a clot, but it does not change acute management, is unreliable during an acute clot/on anticoagulation, and is not indicated here where the PE is clearly PROVOKED (recent surgery + long-haul flight)."

NG158 defines a provoked clot as one with a recent transient major risk factor within 3 months, such as surgery, trauma or significant immobility. Martin meets that.

Recommendation 1.9.2 is blunt: "Do not offer thrombophilia testing to people who have had provoked DVT or PE." You will find it in the NICE thrombophilia recommendation.

At 3am thoroughness is a strong pull. Here the result would add nothing acutely, which is why the case marks it as a red herring.

The reflexive panel

A reflexive panel is ordering a bundle of tests without tying each one to a clear clinical question. The problem is rarely the test itself. It is that none is anchored to a decision point, so results pile up without changing the working diagnosis or the next diagnostic step.

Here is the real cluster from our atrial fibrillation case, with the signed-off reasons.

D-dimer What the signed-off reason says it does not answer: non-specific and frequently raised in AF; without PE features it mostly adds confusion rather than clarity
Lipid profile What the signed-off reason says it does not answer: Part of longer-term cardiovascular risk assessment but does not change the acute management of AF.
Ethanol level What the signed-off reason says it does not answer: Not needed to manage the AF.
Digoxin level What the signed-off reason says it does not answer: Only useful to check for toxicity in a patient already taking digoxin; irrelevant at first presentation.
24-hour ambulatory ECG (Holter) What the signed-off reason says it does not answer: unnecessary here because AF is already documented in front of you
Cardiac MRI What the signed-off reason says it does not answer: the echocardiogram already answers the acute structural questions
CT pulmonary angiogram What the signed-off reason says it does not answer: here there are no DVT features and a clear alternative explanation, so a CTPA mostly adds radiation and contrast risk
24-hour ambulatory blood pressure What the signed-off reason says it does not answer: A tool for diagnosing/monitoring hypertension in the outpatient setting, not for acute AF.

The pattern to notice is that the rhythm was already on the ECG in front of them. Nothing in that column could move the diagnosis.

Six tests ordered in one atrial fibrillation case, each with the reason it answers nothing.

When more testing actively harms

Incidental findings matter because they hijack the plan. A CT head sent to reassure in a primary headache can start a second workup you did not need. NICE guidance on headaches states: "Do not refer people diagnosed with tension-type headache, migraine, cluster headache or medication overuse headache for neuroimaging solely for reassurance." The signed-off reason in our case file adds, "it carries radiation and often causes more anxiety through incidental findings."

Delay harms too. Colonoscopy "is NOT indicated in acute infectious diarrhoea. It is invasive, risks perforating an acutely inflamed gut and needs bowel prep." While you organise that, the real answer is not being chased and the story stays vague. Slowing the search is how diagnoses end up in the ones that get missed.

False reassurance is quieter but bites. In the breathlessness case, with high clinical probability a raised D-dimer was, in the signed-off reason, "expected and non-diagnostic". Ordering a test you would not act on invites you to point at a normal number and relax, even when your pre-test judgement is strong.

What "essential" actually means on a case

Every investigation in the library wears two labels, an essential flag and a relevance grade, and they're separate. They're independent, and that's the useful part.

Across 74 cases, 1,737 investigations. 1,096 essential true and 641 essential false. Relevance grades are relevant, reasonable, reasonable but normal, normal, and red herring.

In the breathlessness case the chest X-ray is graded reasonable but normal, not a red herring. Its signed-off reason reads: "Usually normal or non-specific in PE, but a sensible early film to exclude pneumonia/pneumothorax as mimics. It does not diagnose PE and is normal here."

The CT pulmonary angiogram is graded relevant and flagged as the KEY test, and its reason opens "The definitive diagnostic test".

A test can be reasonable and not be the one that decides anything. A red herring answers a question this patient never raised. See it in a marked chest pain station.

Make it a habit you can use tonight: state the question first, then pick the test that answers it. If you can't state the question, you've found a test you don't need yet.

Practise ordering with a reason

Ordering with a reason is a habit you need reps for. If you're starting FY1, see FY1 on the wards.

In a Diagnosica case you take a history from an AI patient, then order investigations as step two. An AI senior debrief follows. Diagnosica does not do physical examination practice.

The product is live and early, rough edges expected. Free tier is one case a week. The library has 74 cases, every one reviewed and signed off by a doctor. "It doesn't give advice about real patients, and it's not a diagnostic system. Treat every output as educational."

If you want structured reps without stakes, start a case free, run one weekly, and see how your reasoning holds up in debrief.

Last reviewed: 13 August 2026. This post cites NICE NG158 and NICE CG150, both of which are updated from time to time. Check the guideline itself before relying on a recommendation.

Clinical review pending.

Educational use only, not medical advice. AI-generated; verify clinically against primary sources.