Test and treatment thresholds
- Concept
Doctors never reach certainty, so they need a rule for when doubt is small enough to stop, to test or to treat. The threshold model is that rule.
What it is
The threshold model puts the probability of disease on a line from 0 to 100% with two cut points. Below the test threshold neither testing nor treatment is worthwhile; above the test-treatment threshold treatment starts without testing; in between, a test is done and its result decides.
Key facts
- The model was described by Pauker and Kassirer in 1980 and is still the backbone of testing pathways.
- B = gain from treating a patient who has the disease; H = loss from treating one who does not.
- Order of the lines: test threshold < treatment threshold < test-treatment threshold.
- A more accurate test lowers the test threshold and raises the test-treatment threshold: the zone widens.
- A test with its own risk or cost (radiation, contrast, an invasive biopsy) needs more doubt to be worth it: the zone narrows.
- Analysis of suspected tuberculosis: a benefit/harm ratio near 36 gives a treatment threshold of about 2.7%.
Worked example
- Made-up disease: treating a sick patient gains B = 9 points; treating a healthy one costs H = 1 point.
- Treatment threshold = H / (H + B) = 1 / 10 = 10%: above 10%, treating beats not treating.
- Made-up harmless test: LR+ 9, LR− 0.1. At the 10% line the odds are 0.1 / 0.9 ≈ 0.111.
- Test threshold: where a positive just reaches 10%: odds 0.111 / 9 ≈ 0.012, about 1.2%.
- Test-treatment threshold: where a negative just falls to 10%: odds 0.111 / 0.1 ≈ 1.11, about 53%.
- Zones: below 1.2% do nothing; 1.2–53% test and follow the result; above 53% treat without testing.
- Weaker test, LR+ 3 and LR− 0.3: 0.111 / 3 = 0.037 (3.6%) and 0.111 / 0.3 = 0.37 (27%). The zone shrinks to 3.6–27%.
- Harsher treatment, H = 3: line = 3 / 12 = 25%, odds 0.333. With the first test the zone moves to 3.6–77%.
- Reading: benefit and harm place the treatment line; the test's strength sets how far either side of it testing pays.
What moves the lines
| Change | Treatment threshold | Testing zone | Example |
|---|---|---|---|
| Treatment more harmful | Rises | Shifts up | Anticoagulation in a patient with a high bleeding risk: all thresholds rise |
| Untreated disease more dangerous, treatment more effective | Falls | Shifts down | Suspected tuberculosis: treatment worthwhile from about 3% |
| Test more accurate | No change | Widens | A test with higher LR+ and lower LR− can settle more patients |
| Test riskier or more costly | No change | Narrows (the test threshold rises) | CT pulmonary angiography: radiation and contrast; low-probability, PERC-negative patients are not tested |
| Test fast, cheap and specific | No change | Starts lower | A rapid streptococcal antigen test lowers the test threshold for sore throat |
| Patient fears the treatment's harm more | Rises | Shifts up | Patient values enter as the weight given to harm against benefit |
In practice
- Pulmonary embolism: the test threshold has been put near 2%; a low-probability patient who meets all PERC criteria is below it and is not tested.
- Pulmonary embolism: low or intermediate probability goes to D-dimer, high probability straight to CT pulmonary angiography (the testing zone).
- Pharyngitis: in a vignette study, GPs set the test threshold near 16% and the treatment threshold near 88%; a quick specific antigen test keeps the test threshold low.
- Appendicitis: in the same study the thresholds were about 52% (test) and 69% (treat); the Alvarado score (appendicitis) and the McIsaac score (pharyngitis) set the probability.
- Most errors are in the probability, not the line: GPs underestimated appendicitis and overestimated streptococcal pharyngitis; in a bacteriuria vignette most clinicians overestimated infection and would treat.
Common pitfalls
- Using 50% as the treatment line by habit instead of weighing benefit against harm.
- Testing patients below the test threshold, where false positives and test harms outweigh the few cases found.
- Testing patients above the test-treatment threshold, where even a negative would not stop treatment.
- Reusing a published threshold for a patient whose bleeding risk or values differ from the study's.
Common traps
- The treatment threshold is 50%: it is H / (H + B); a safe treatment for a dangerous disease is right at a few percent (about 2.7% for suspected tuberculosis in one analysis)
- Treating without certainty is a mistake: above the treatment threshold treating is the better bet even though some treated patients turn out not to have the disease
- A better test lowers the treatment threshold: test accuracy only widens the testing zone; the treatment threshold depends on benefit, harm and the patient's values
- A low probability calls for a test to be sure: below the test threshold a test finds few cases and its false positives and risks do more harm than good
- Overtreatment means the thresholds are wrong: often the line is sensible and the probability estimate is too high, as in the bacteriuria vignette where most clinicians would treat
Test yourself
Think of the answer, then tap the question to see it.
Treating a sick patient gains 4 points and treating a healthy one costs 1 point. What is the treatment threshold?
H / (H + B) = 1 / 5 = 20%: above 20%, treating is the better bet.
Why can the treatment threshold for suspected tuberculosis sit near 3%?
The benefit of treating a true case is large compared with the harm of treating a non-case (ratio about 36 in one analysis), so 1 / (1 + 36) ≈ 2.7%.
A new test is safer and more accurate than the old one. Which threshold does not move, and what happens to the others?
The treatment threshold stays: it depends on benefit, harm and values. The testing zone widens: the test threshold falls and the test-treatment threshold rises.
A low-probability patient with suspected pulmonary embolism meets all PERC criteria. Why not order a D-dimer anyway?
The probability is already below the test threshold (about 2%): false positives leading to CT, with radiation and contrast, would outweigh the few emboli found.
A patient's bleeding risk is high. What happens to the probability at which anticoagulation becomes worthwhile?
It rises: more harm (H) raises H / (H + B), so more certainty is needed before treating, and the testing zone shifts up with it.
Elsewhere in DurgiPedia
- Suspected pulmonary embolism: probability first, then the test
- Approach to sore throat: Centor and McIsaac
- Imaging in suspected appendicitis
Connections
Educational content on epidemiological and statistical methods; the examples use made-up numbers to teach the calculation. Not a clinical or legal guide: current Colombian rules are in the official sources.
Sources: The threshold model revisited (Journal of Evaluation in Clinical Practice, 2019); Decision theoretical foundations of clinical practice guidelines: an extension of the ASH thrombophilia guidelines (Blood Advances, 2024); Rational decision making in medicine: Implications for overuse and underuse (Journal of Evaluation in Clinical Practice, 2018); Correlation of clinical decision-making with probability of disease: A web-based study among general practitioners (PLoS ONE, 2020); Pulmonary embolism: the diagnosis, risk-stratification, treatment and disposition of emergency department patients (Clinical and Experimental Emergency Medicine, 2016); Clinician Testing and Treatment Thresholds for Management of Urinary Tract Infection (Open Forum Infectious Diseases, 2023).
Spotted a mistake?We are always improving for the community. Tell us what is wrong and our team will review it.Report a mistake
Representative model, for educational purposes only.
