Statistics

Screening and diagnostic test interpretation

Wilson screening criteria, sensitivity and control group risk when reading diagnostic test results.

23 questions 4 source pages 1 fact-check flags

Images appear with the first question taken from each source page — tap a question to open it.

Q1What are Wilson's criteria for a screening program?▸

Condition: an important health problem; natural history understood; recognisable latent or early symptomatic stage
Test: easy to perform and interpret, acceptable, accurate, reliable, sensitive and specific
Treatment: acceptable treatment recognised, more effective if started early, policy on who should be treated
Diagnosis and treatment should be cost effective; case-findings should be a continuous process

Q2What features should the screening test and the screened population have?▸

Test: acceptable and tolerated by patients; high sensitivity to detect disease before the critical point; high specificity to reduce false positives; cost effective
Population: disease should have high prevalence to allow screening; accepted and effective treatment is available; treatment and further evaluation acceptable to the patient
Disease: significant impact on community; recognisable latent or early symptomatic phase; natural history understood

Q3Give two examples of screening programmes discussed.▸

DDH screening
Scoliosis screening in Hong Kong (P5, F1 and F3)

Q4What is the incidence of DDH and how do the Barlow/Ortolani test and ultrasound compare?▸

DDH incidence 1 in 1000
B and O test: sensitivity 67%, specificity 96%
Ultrasound: sensitivity 89%, specificity 97%, PPV 62%

Q5What did the 2009 International Hip Dysplasia Institute study show?▸

Compared no screening, selective screening and universal screening
Selective screening most cost effective

Q6Does DDH screening fulfil Wilson's criteria?▸

Fulfils most criteria
BUT there is no universal agreement on who should receive treatment
Cost of the screening program versus the full cost of delayed detection has not been fully established

Q7Describe the Hong Kong scoliosis screening programme.▸

3-tier system at P5, F1 and F3
FBT sensitivity 84%; ATR sensitivity 83% (>=15 degrees refers to specialist hospital; 5-14 degrees goes to Moire topography)
Moire topography >=2 lines leads to XR; Cobb >20 degrees refers to specialist hospital
Overall sensitivity 88%

Q8Define sensitivity and specificity.▸

Sensitivity: probability of a positive test in patients with the disease (true positive / disease positive)
Specificity: probability of a negative test in patients without the disease (true negative / disease negative)

Q9Define false positive rate, false negative rate and accuracy.▸

False positive rate = 1 - specificity
False negative rate = 1 - sensitivity
Accuracy: probability of correct identification of a disease = (true positive + true negative) / total number

Q10What are the positive and negative likelihood ratios?▸

Positive LR = sensitivity / (1 - specificity)
Negative LR = (1 - sensitivity) / specificity
LR is the likelihood of correctly predicting disease versus the probability of incorrectly predicting it

Q11What are PPV and NPV and how are they affected by prevalence?▸

PPV: probability of disease when the test is positive (true positive / test positive)
NPV: probability of no disease when the test is negative (true negative / test negative)
Sensitivity and specificity are intrinsic to the test itself, but PPV/NPV change with prevalence
prevalence affects the pretest probability
PPV is higher if prevalence is higher

Q12What is an ROC curve?▸

Plot of the true positive rate against the false positive rate (sensitivity / 1 - specificity) at different cut-off points
Shows the tradeoff between sensitivity and specificity with different cut off values: any increase in sensitivity is accompanied by a decrease in specificity
Area 0.5 = useless test (complete overlap); 1.0 = perfect test (no overlap)
AUC 0.5-0.7 marginally useful tests, 0.7-0.9 good, >0.9 excellent

Q13How do you choose a cut-off for a screening versus a confirmation test?▸

Screening test: choose high sensitivity (right side of the curve)
Confirmation test: choose high specificity (left side of the curve)

Q14What cohort study example is used to interpret results?▸

A study on the risk of DVT after given enoxaparin in TKR patients
The question asked is: how to interpret the result?

Q15What are the exposure, outcome and population in this study?▸

Exposure: enoxaparin
Outcome: DVT
Population: TKR patients

Q16How is a cohort study result interpreted?▸

List the risk ratio, risk difference, odds ratio and NNT

Q17Given a control risk of 3/45 and experimental risk of 1/50, calculate the risk difference and NNT.▸

2x2 table, outcome DVT: exposure enoxaparin +ve 1 DVT+ / 49 DVT- / 50; exposure -ve 3 DVT+ / 42 DVT- / 45; column totals 4 DVT+ / 91 DVT-
Control group risk 3/45 = 0.067; experimental group risk 1/50 = 0.02
Risk difference (absolute risk reduction) = 0.067 - 0.02 = 0.047
NNT = 1/ARR = 1/0.047 = 21.3

Q18Calculate the relative risk from these data and state when it can be used.▸

RR = 0.02/0.067 = 0.3
Risk of developing the disease with the exposure compared with the risk without the exposure
Relative risk can be calculated in a cohort study

Q19Calculate the odds ratio and state when odds can be calculated.▸

Odds in exposed group 1/49 = 0.02; odds in unexposed group 3/42 = 0.07
Odds ratio = 0.29
Odds can only be calculated in a case control study

Q20Can causality be established from these results?▸

No - these measures cannot establish causality
Causality is established through the Bradford Hill criteria

Q21List the Bradford Hill criteria.▸

Strength (effect size), consistency (reproducibility), specificity
Temporality, biological gradient (dose-response), plausibility, coherence
Experiment, analogy

Q22How is a chi-squared test calculated?▸

Chi2 = sum of (observed - expected)^2 / expected
Expected value = (row sum x column sum) / total

Q23Calculate the expected values for the male/female 2x2 example.▸

Disease male = 10x14/21 = 6.67; disease female = 10x7/21 = 3.33
Control male = 14x11/21 = 7.33; control female = 7x11/21 = 3.67
Expected value = (row sum x column sum) / total number

Fact check

Selective screening is the most cost-effective DDH screening strategy (International Hip Dysplasia Institute 2009 JBJS) — imprecise — The 2009 JBJS study was a decision analysis of hip outcomes (not costs) that favoured physical examination for all with selective ultrasonography; a later systematic review found screening cost evidence inconclusive/heterogeneous — (medium confidence) — source

This site is private

Screening and diagnostic test interpretation

Fact check