Occam's Razor vs Hickam's Dictum: One Diagnosis or Two?

On a Tuesday ward round, one senior taps the obs chart and says, "One diagnosis, find it." Another shrugs, "He can have two things." They are quoting different saints of clinical reasoning.

Occam's razor says the simplest diagnosis is the most likely to be correct. Hickam's dictum replies, 'A man can have as many diseases as he damn well pleases'. Use them in sequence. Start simple, stay open.

We'll unpack what each means, how to use them in order, then run three cases to test you, one unify, one split, one with real comorbidity.

What is Occam's razor in medicine?

Occam's razor in medicine says you start by seeking a single diagnosis that explains all the patient's complaints and findings. The simplest unifying cause is usually the likeliest. It is a bias toward parsimony, not a rule, and you abandon it if the facts do not fit.

The Latin often quoted is 'Entia non sunt multiplicanda, praeter necessitatem', which translates as 'things should not be multiplied unless necessary'.

The idea traces to William of Ockham's principle of parsimony in philosophy, prefer the explanation that commits you to the fewest assumptions.

Miller in 1998 wrote that Osler applied it to medicine, fit all of a patient's symptoms and findings into one diagnosis, though the attribution is disputed and could not be confirmed in his textbook.

Why it works on the wards, in younger, previously well patients a single new problem usually explains a cluster of new symptoms, so you start there and save the splits for when the data forces your hand.

What is Hickam's dictum?

Hickam's dictum is the counterweight to Occam’s razor: when you’re working up a case, multiple separate diseases are more likely than a single unifying one. It cautions against forcing one label onto every finding, and it bites hardest in older or comorbid patients.

John Hickam, 1914 to 1970, has the line doctors swap between wards and coffee: "A man can have as many diseases as he damn well pleases" from Miller 1998. A later print softened it to "darn well pleases" in Hilliard et al. 2004. Both wordings sit in LITFL's Hickam's dictum entry.

Mani et al., BMJ 2011, frame it as a guard against over-reliance on Occam, especially helpful in older patients and those with multiple comorbidities. The bluntness is the point, it stops you tidying away a second diagnosis because it feels untidy.

Occam's razor vs Hickam's dictum: which should you use?

Use them in sequence. Start with Occam's razor to seek one diagnosis that explains most findings. Then, before you commit, apply Hickam's dictum and ask what else might also be true. The balance shifts with context, unifying early helps, but older multimorbid patients often need you to split.

Presuming one disease process often gives a cognitive advantage, Blaser et al. found. It speeds pattern recognition and reduces noisy branches.

The failure mode is Crabtree's bludgeon, the urge to force messy data to fit the explanation you like, described by Blaser et al.

Hickam earns its keep in atypical presentations, multiple disease processes and unique populations, Blaser et al. conclude. Older patients and those with multiple comorbidities benefit in particular, reported by Mani et al. via LITFL.

A metacognitive checklist and deliberate analytic thinking help you switch gears, and both aphorisms are imperfect, according to Blaser et al. Keep the list wide before you prune it, see keeping the differential honest.

• Best fit. Occam: classic pattern with coherent data. Hickam: atypical story or mixed signals.
• Risk if overused. Occam: Crabtree’s bludgeon, missed coexisting disease. Hickam: diagnostic scatter, lost signal.
• Who it suits. Occam: younger, single-issue presentations. Hickam: older patients and multimorbidity.

The four ways a second diagnosis shows up

Here is a practical four-bucket checklist for second diagnoses, lifted from a 2024 JGIM analysis.

1. Incidentaloma, n=21, 25.3%. Found on the way past, unrelated to why the patient came in.
2. Pre-existing disease, n=26, 31.3%. Already known in the history, helpful context but not today’s problem.
3. Causally linked to the primary diagnosis, n=33, 39.8%. Part of the same chain, either a trigger or a complication.
4. Causally unrelated second disease contributing to the acute presentation, n=3, 3.6%. Two active problems at once, sharing time but not cause.

In every one of the 83 reports, there was still a primary diagnosis that explained the presenting illness.

All 220 NEJM cases reported a unifying final diagnosis.

Among 265 survey respondents, 79% (95% CI 74 to 84%) picked the coincident, causally unrelated second disease as the most unlikely or surprising combination.

The authors’ conclusion was blunt and helpful: multiple diagnoses commonly occur, but most were causally linked, incidentalomas or known pre-existing disease, and time, probability and causation reconcile the alleged conflict. Hickam is real but usually polite, the second disease tends to be old, related, or incidental.

Case 1: back pain, tiredness and thirst

Alan Brightwell is 68, a retired teacher and a non-smoker, and he opens with, "I've had back pain and I've been exhausted for about four months now." He has been told twice it is arthritis and old age, and the physio and painkillers changed nothing.

At first pass it reads like four separate problems: a bad back, tiredness, thirst and infections. He volunteers that he is very tired, breathless on doing anything, and looks pale, but you have to ask to learn the back pain is constant and at night, he has lost a stone, there have been three chest infections in six months, he is very thirsty and passing more urine, constipated and muddled at times, bruises more easily, and his urine is frothy with slightly swollen ankles. No leg weakness, numbness or bladder or bowel trouble.

One unifier earns its keep. Myeloma declares itself through CRAB, and this story ticks through it: hypercalcaemia explains thirst, polyuria, constipation and confusion, renal impairment fits the frothy urine and ankle swelling, anaemia accounts for fatigue, breathlessness and pallor, and bone involvement meets constant night pain, with recurrent infections on top. Occam wins. The trap this case punishes is treating the label anaemia on its own.

Old age is a poor dustbin diagnosis. Serious disease in disguise likes to hide there, and myeloma is a repeat offender.

Even so, you still have to check your unifier actually holds every finding. In a published case, a 66-year-old woman with vomiting, weight loss, shoulder pain, renal failure, anaemia and a lytic clavicle had myeloma as the working diagnosis, but the shoulder biopsy showed follicular thyroid carcinoma and the renal failure was probably a chronic, separate process, from a 2021 Diagnosis case report.

Case 2: shoulder pain and a weak hand

Raymond Doyle is 62, a school caretaker who spent twenty years as a shipyard welder, and he has smoked twenty a day since seventeen, about forty-five years, still smoking. His opener is blunt: “I’ve had this pain in my right shoulder for months now and nothing shifts it, it’s worst at night.”

He has collected labels. First frozen shoulder from GP and physio, ten weeks of physio with no benefit, then wear and tear in the neck, a trapped nerve.

One apical lung lesion fits the lot here, a Pancoast tumour that can hurt the shoulder, batter C8 and T1, and hit the sympathetic chain. Moving the shoulder does not bring the pain on and he can reach the top shelf, one of those negatives that earn their place.

The pain runs down the inside of the arm into the little and ring fingers, and his right hand is weak, clumsy, thinner. That is lower brachial plexus territory, while C5 or C6 wear and tear would be thumb and index, not the ulnar side.

Family spotted a droopy right eyelid and a smaller pupil, and he has stopped sweating on that side of his face. Add a stone and a half lost in four months and a father who died of lung cancer at 66, the pattern tightens.

A chest X-ray three months ago was called fine. It does not close the question, the apex is the blind spot of a plain film.

Splitting is attractive when each piece has a friendly local name, frozen shoulder for pain, cubital tunnel for the hand, a tired eye for the droop. Neat, and wrong.

One or two Take a history from an AI patient who only answers when asked, then commit to one diagnosis or two. Start a case

Case 3: a muddled 79-year-old on a blood thinner

“My dad’s gone slow and muddled over the last six weeks, and he sleeps all day now,” Mark says. Raymond Ellis is 79, lives alone, drinks more than he admits, and takes a blood thinner for an irregular heartbeat.

Before Easter he shopped and did the crossword. Now his right hand is clumsy, his right foot scuffs, he has veered with two near falls, and he has wet himself twice. Headache is worse on waking and when he bends or coughs. No fever, cough or urinary burning. Downhill, steady, not swinging hour to hour.

The surgery called it “his dementia” and “his age”, with no scan and no blood tests, which is textbook how slow diagnoses get missed. Tempo is the discriminator, and the son owns the tempo.

This is Hickam in the polite JGIM sense. More than one thing is true. One new diagnosis, a chronic subdural haematoma, fits the six weeks of drift and focality. The background problems explain why it happened and why the story sounds messy.

Ask “has he hurt his head?” and Mark says no. Ask about trips, slips and knocks, and a fall two months earlier appears, head catching the door frame. A substantial minority recall no head injury, so you keep asking.

The blood thinner turns a trivial knock into a large collection. Alcohol muddies the picture and shows on liver tests, but it does not explain a six week focal decline. A clean urine and normal CRP leave the “delirium from infection” label with nothing to stand on. Often completely reversible.

• Pre-existing disease. The alcohol, liver tests back the history but it does not explain a six week focal decline, plus his high blood pressure and acid reflux sitting quietly in the background.
• Causally linked. The anticoagulation for an irregular heartbeat, which turned a trivial knock into a large collection.
• Incidentaloma. Nothing in this story.
• Causally unrelated new disease. Nothing earns this slot; the “his dementia” label and the “delirium from infection” idea both fail against the tempo and the clean urine and normal CRP.

A three-question check before you settle on one diagnosis

1. Does my one diagnosis explain every finding, including the boring ones? In the myeloma case, hypercalcaemia, renal impairment, anaemia and bone pain lived under one roof, the CRAB unifier. That is the Occam's razor vs Hickam's dictum starting point, but Blaser et al.'s metacognitive checklist and engaging analytic thinking keep it honest.
2. Am I splitting because each piece has a familiar local label? In the Pancoast case, shoulder pain, little and ring finger symptoms, and Horner's sat on the same apical lesion. Calling them cervical radiculopathy here and an ophthalmic thing there was the trap.
3. For each extra finding, is it old, caused by the main problem, incidental, or a genuinely separate new disease, and if it's the last one, have I earned that? In the subdural case, being anticoagulated was causally related background, and the six-week tempo made the dementia label wrong. A truly unrelated second disease contributing to the acute presentation was 3 of 83 reports, 3.6%, in JGIM 2024, and 79% of surveyed clinicians found that combination the most surprising.

A red herring is different from a second diagnosis. See telling clues from red herrings.

Practise the call on a live patient

Reading three cases is one thing. Making the call when answers only surface if you ask is another. In Diagnosica you speak or type to an AI patient on a video avatar, choosing what to pursue next.

Order investigations and see results and imaging. Commit to a diagnosis and a management plan, then get a scorecard with competency scores and teaching points.

Those three sit in a library of more than 130 across 18 specialties, each case signed off by a doctor. Browse on the web app or the iPhone and Android apps.

For a daily habit, the two-minute quick case gives one line then five clues, you guess after each, then the answer and a one-line teaching point. It is a tidy way to test one diagnosis or two, and to play diagnosis games with real stakes.

No physical examination practice. It doesn't give advice about real patients, and it's not a diagnostic system. Treat every output as educational.

Run a case Run a case free and try the three-question check to see how you balance simplicity, openness, and links between problems. Start a case free

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

Clinical review pending.