Slides PDF · BG / EN · Reading
Read the preparation text before the practical. Work in groups of three or four. Each of five tasks takes 15 minutes: 75 minutes after 15 minutes of preparation. All data are hypothetical. Open the answer only after agreeing on your analysis.
Task 1. Assemble the Plovdiv cohort — 15 min
In an invented five-year cohort, first periodontitis occurs in 40 of 200 adults with diabetes and 48 of 600 without diabetes. All are initially free of the endpoint and all outcomes are known. Build the table, calculate both risks, RR and RD, and interpret them.
Diabetes: 40 cases and 160 non-cases; no diabetes: 48 cases and 552 non-cases. Risks are 20% and 8% over five years. RR = 2.5; RD = 12 percentage points. The exposed group has 2.5 times the observed risk and 12 additional cases per 100. This is a crude association, requiring assessment of confounding and bias before a causal interpretation.
Task 2. Distinguish attributable denominators — 15 min
Use the same cohort. Calculate AF among exposed people, overall risk, population risk difference and PAF. Explain why the two fractions differ and what a preventable-burden claim requires.
AF = (0.20−0.08)/0.20 = 60%. Overall risk = 88/800 = 11%. Population risk difference = 11%−8% = 3 percentage points. PAF = (11−8)/11 = 27.3%. At the reference risk, 800 people would have 64 cases rather than 88, an excess of 24. The denominators are 40 exposed cases versus 88 population cases. A causal interpretation needs valid causal risks and a defined intervention; the arithmetic alone does not prove preventability.
Task 3. Interpret an additional preventive programme — 15 min
In an invented individually randomised three-year trial, 15/300 children with an additional programme and 30/300 with standard prevention develop at least one new cavitated lesion. Calculate RR, relative reduction, ARR and NNT. If the same RR applied to a group with standard risk 2%, what would ARR and NNT be?
Risks 5% and 10%; RR = 0.5; relative reduction 50%; ARR = 5 percentage points; NNT = 20 over three years versus standard prevention. At baseline risk 2%, a transferable RR 0.5 implies programme risk 1%, ARR 1 percentage point and NNT 100. This is a conditional calculation, not proof of transferability.
Task 4. Audit time and missing outcomes — 15 min
A separate invented cohort has 18 events in 450 exposed person-years and 12 in 600 unexposed person-years. Compare rates. Another cohort starts with 100 people; only 80 outcomes are known, including 8 cases. Explain the possible full-cohort risk.
Rates are 4 and 2 per 100 person-years; rate ratio 2; rate difference 2 cases per 100 person-years. They are not probabilities. In the second cohort, known-outcome risk is 10%; full-cohort bounds are 8–28%, depending on the 20 unknown outcomes. Neither complete-case analysis nor person-time automatically removes informative-loss bias.
Task 5. Appraise confounding and uncertainty — 15 min
For new tooth loss over five years, younger exposed people have 10/100 cases and younger unexposed 40/400; older groups have 160/400 and 40/100. Calculate stratum and crude RRs. Then explain what a separate RR 4 with approximate 95% CI 2.64–6.06 does and does not establish.
Both stratum RRs are 1. Crude RR = (170/500)/(80/500) = 2.125. Equal age weights give 25% in both groups. In this stated model age causes exposure and outcome; the crude association reflects age composition. The separate interval describes sampling uncertainty under its assumptions. It does not establish causation, remove bias or give a range of individual probabilities.
A complete answer includes the design, denominator, calculation, units, period and a justified interpretation. Assess sampling, measurement, confounding and uncertainty separately.