Ophthalmic Epidemiology
Ophthalmic epidemiology tests the arithmetic of screening and the logic of study design. You are likely to be given a 2 by 2 table, a screening scenario or a trial, and asked what the numbers mean.
Core rules
- Prevalence is the proportion with disease at a point in time; incidence is new cases per population per unit time.
- Sensitivity = true positives/(true positives + false negatives); specificity = true negatives/(true negatives + false positives).
- Positive predictive value = true positives/(true positives + false positives). It rises with prevalence; sensitivity and specificity do not.
- A test with 95% sensitivity and 95% specificity used in a low-prevalence population will have a low PPV and many false positives.
- Study designs: cross-sectional for prevalence, cohort for incidence and risk, case-control for rare outcomes, randomised controlled trial for causation.
- Bias: selection, information, lead-time and length-time bias in screening.
- Number needed to treat = 1/absolute risk reduction.
Worked example
A diabetic retinopathy screening programme uses non-mydriatic photography with 90% sensitivity and 95% specificity in a population with 10% prevalence. In 10,000 people, 1,000 have disease: 900 are detected and 100 missed. Of the 9,000 without disease, 450 are falsely positive. PPV = 900/1,350 = 67%. If the same test were used where prevalence is 1%, PPV would fall to about 15%, so the programme would generate many unnecessary referrals. This is why screening is targeted at high-risk groups.
Traps
Confusing sensitivity with PPV, assuming a good test is good in every population, and forgetting that prevalence changes PPV but not sensitivity or specificity.