top of page
Search

Likelihood Ratios: The Secret to Diagnosis (MRCP Part 1)

TL;DR: 

Likelihood ratios translate test results into meaningful changes in disease probability and are repeatedly examined in MRCP Part 1. They outperform sensitivity and specificity for bedside decision-making and allow rapid rule-in or rule-out without complex maths. Mastering a few thresholds and traps can secure easy marks across multiple systems.


Why likelihood ratios matter in MRCP Part 1

MRCP examiners favour tools that assess clinical reasoning. Likelihood ratios (LRs) do exactly that: they ask whether a test result meaningfully changes the probability of disease in this patient. Candidates who rely only on sensitivity/specificity often miss these questions because those metrics describe test performance, not diagnostic impact.

LRs are especially useful because they:

  • Are independent of prevalence

  • Apply directly to individual patients

  • Work for both positive and negative results


What are likelihood ratios?

A likelihood ratio compares how likely a test result is in patients with disease versus without disease.

  • Positive likelihood ratio (LR+): how much a positive result increases disease probabilityLR+ = Sensitivity ÷ (1 − Specificity)

  • Negative likelihood ratio (LR−): how much a negative result decreases disease probabilityLR− = (1 − Sensitivity) ÷ Specificity

You are not expected to calculate these in the exam; interpretation is the key skill.


The high-yield thresholds (memorise this)

Likelihood ratio

Effect on probability

Exam meaning

>10

Large increase

Rules disease in

5–10

Moderate increase

Strongly supports diagnosis

2–5

Small increase

Weak evidence

1

No change

Useless test

0.5–0.2

Small decrease

Weak exclusion

0.1–0.2

Moderate decrease

Helps rule out

<0.1

Large decrease

Rules disease out

Exam pearl: LR+ >10 and LR− <0.1 should trigger an instant decision.


Medical exam revision desk setup for MRCP Part 1 study and diagnostic test interpretation

Five most tested subtopics

1) Likelihood ratios vs sensitivity and specificity

  • Sensitivity/specificity → intrinsic test properties

  • Likelihood ratios → clinical usefulnessIf asked which statistic best guides diagnosis at the bedside, the answer is likelihood ratio.

2) Pre-test and post-test probability

Examiners expect direction, not calculation:

  • Moderate/high pre-test probability + LR+ >10 → diagnosis very likely

  • Low/moderate pre-test probability + LR− <0.1 → disease effectively excluded

Graphical aids (e.g. Fagan nomogram) may be referenced, but calculations are rarely required.

3) Screening vs confirmatory tests

  • Screening tests: high sensitivity → low LR− (good to rule out)

  • Confirmatory tests: high specificity → high LR+ (good to rule in)

Typical phrasing: “Which test best excludes disease?” → choose the option with the lowest LR−.

4) Combining tests

When tests are independent, their LRs can be multiplied. Two moderate tests may outperform one weak test. This appears in sequencing questions (e.g. initial screen followed by confirmation).

5) Common clinical contexts

Frequently examined examples include:

  • D-dimer for suspected pulmonary embolism

  • High-sensitivity troponin in acute coronary syndrome

  • ANA and anti-dsDNA in SLE

  • PSA in prostate cancer screening

Focus on whether the test changes probability enough to alter management.


One-minute MCQ case

A 55-year-old man presents with pleuritic chest pain. Clinical probability of pulmonary embolism is moderate. A D-dimer has an LR− of 0.08. What does this result imply?

Answer: Pulmonary embolism is effectively ruled out.

Explanation: An LR− <0.1 produces a large reduction in post-test probability. In MRCP questions, this is sufficient to exclude disease in low-to-moderate risk patients without further imaging.


Five examiner traps to avoid

  • Confusing likelihood ratios with predictive values

  • Choosing sensitivity when the question asks about clinical usefulness

  • Ignoring pre-test probability

  • Assuming a positive test always rules disease in

  • Forgetting that LR = 1 means no diagnostic value


Practical study checklist

  • Memorise LR+ >10 and LR− <0.1 thresholds

  • Practise interpreting LRs without calculations

  • Link screening tests to LR− and confirmatory tests to LR+

  • Answer questions by the direction of probability change

  • Reinforce with timed MCQs from the Crack Medicine QBank: https://crackmedicine.com/qbank/

For structured revision, see the MRCP Part 1 overview (https://crackmedicine.com/mrcp-part-1/) and consolidate performance with mock tests (https://crackmedicine.com/mock-tests/). Targeted lectures are available at https://crackmedicine.com/lectures/.


FAQs

What is the difference between likelihood ratio and sensitivity?

Sensitivity measures how often a test is positive in disease. Likelihood ratios show how much a result changes disease probability in an individual patient.

Do I need to calculate likelihood ratios in MRCP Part 1?

No. Interpretation and threshold recognition are sufficient.

Why are likelihood ratios preferred to predictive values?

They are independent of disease prevalence and apply across populations.

What LR values rule disease in or out?

LR+ >10 rules in disease; LR− <0.1 rules it out.


Ready to start?

If likelihood ratios feel abstract, they need practice, not rereading. Apply these principles now using high-yield MCQs from the Crack Medicine MRCP QBank and validate progress with a full mock test.


Sources

 
 
 

Comments

bottom of page