MPH Epidemiology & Biostatistics

Solved Previous Year Papers | Dec 2019 + Class Test 2 (Institute of Public Health, Kalyani)

📄 PAPER 1 (Dec 2019) – Institute of Public Health, Kalyani

Q1(a) Design a case-control study: CHD and smoking (with diagram & 2×2 table)

Case-Control Study: Observational study starting with cases (disease present) and controls (disease absent), then looking back for past exposure.

Steps: 1) Select CHD cases from hospitals. 2) Select controls (no CHD) from same community. 3) Assess smoking history. 4) Compare exposure frequency.

Diagram: CHD present (Cases) ← Smoking history ; CHD absent (Controls) ← Smoking history.

	CHD Cases	Controls	Total
Smoker	a	b	a+b
Non-smoker	c	d	c+d
Total	a+c	b+d	N

Odds Ratio (OR) = \( \frac{a \times d}{b \times c} \) → OR>1 indicates smoking increases CHD risk.

Q1(b) Define Attributable Risk (AR) and Population Attributable Risk (PAR)

Attributable Risk (AR) = \( I_e - I_o \) : excess disease among exposed due to exposure.
Population Attributable Risk (PAR) = \( I_t - I_o \) : excess disease in total population due to exposure.

Q1(c) Cohort: smokers incidence 9/1000, non-smokers 1/1000, smokers=45%

(i) AR among smokers = \( \frac{9}{1000} - \frac{1}{1000} = \frac{8}{1000} \) → 8 per 1000.
(ii) \( I_t = 0.45\times\frac{9}{1000} + 0.55\times\frac{1}{1000} = \frac{4.6}{1000} \) → PAR = \( \frac{4.6}{1000} - \frac{1}{1000} = \frac{3.6}{1000} \) → 3.6 per 1000.

Q2(a) Define Confounding with example

Confounder: third factor associated with both exposure and outcome, not on causal pathway.
Example: Age confounds smoking-lung cancer association (older people smoke more & have higher cancer risk).

Q2(b) How confounding is controlled?

Design stage: Randomization, Restriction, Matching.
Analysis stage: Stratification, Multivariate regression.

Q2(c) Stratified analysis & comment

Crude OR = \( \frac{60\times164}{36\times140} = 1.95 \).
Age <40: OR = 1.0 ; Age ≥40: OR = 1.0 → crude OR differs → Age is a confounder.

Q3(a) Difference: Prospective vs Retrospective Cohort

Feature	Prospective	Retrospective
Direction	Forward	Backward (past records)
Cost	High	Low
Bias	Less	More recall/record bias

Q3(b) Smokers 3000 (CHD=90), non-smokers 5000 (CHD=100)

Incidence smokers = 90/3000 = 0.03 ; non-smokers = 100/5000 = 0.02.
RR = 0.03/0.02 = 1.5 → smokers have 1.5× higher CHD risk.

Q3(c) Type I and Type II errors

Type I error (α): Reject true H₀ (false positive).
Type II error (β): Fail to reject false H₀ (false negative).

Q4(a) Flow diagram for causation decision

1) Is association real? (chance, bias, confounding?) → 2) If real, is it causal? (temporality, strength, consistency) → 3) If causal, implement prevention.

Q4(b) Interaction (additive & multiplicative models)

Baseline R₀₀=3, R₁₀=7, R₀₁=8.
Additive expected = \( 3 + (7-3)+(8-3) = 12 \).
Multiplicative expected = \( 3 \times (7/3) \times (8/3) \approx 18.67 \).
If observed = 12 → no additive interaction; if observed = 19 → no multiplicative interaction.

Q4(c) Hill’s criteria for causation

Strength, Consistency, Specificity, Temporality (most important), Biological gradient, Plausibility, Coherence, Experiment, Analogy.

📄 PAPER 2 (Class Test 2) – Screening, Kappa, Survival

Q1 Sensitivity, Specificity & PPV (new prevalence 2%)

From 2×2: TP=20, FN=5, TN=380, FP=95.
Sensitivity = 20/25 = 80% ; Specificity = 380/475 = 80%.
New population 10,000, prevalence 2% → Disease=200, Non-disease=9800.
TP=160, FN=40 ; TN=7840, FP=1960 → PPV = \( \frac{160}{160+1960} = 7.55\% \).

Q2 Net Specificity – Test A then Test B

Test A: Se=80%, Sp=60% ; Test B: Se=90%, Sp=90% ; Population 4000, prev 5% (D=200, ND=3800).
After A: TP=160, FN=40 ; TN=2280, FP=1520.
Test B on positives: among 160 TP → new TP=144, among 1520 FP → new TN=1368, FP=152.
Final TN = 2280+1368 = 3648 ; Final FP=152 → Net Specificity = \( \frac{3648}{3648+152}=96\% \).

Q3 Willis Rogers phenomenon + Five‑year survival

Willis Rogers (stage migration bias): Improved diagnostics shift patients to more severe stage → survival appears improved in each stage without true benefit.
5‑year survival: Only 2010 cohort has complete 5-year follow-up (alive in 2015 = 32/160) → 20%.

Q4 Reliability & Kappa Statistics

	A: Grade II	A: Grade III	Total
B: Grade II	82	6	88
B: Grade III	8	54	62
Total	90	60	150

\( P_o = \frac{82+54}{150} = 0.9067 \).
\( P_e = \left(\frac{90}{150}\times\frac{88}{150}\right) + \left(\frac{60}{150}\times\frac{62}{150}\right) = 0.352 + 0.1653 = 0.5173 \).
\( \kappa = \frac{0.9067-0.5173}{1-0.5173} = 0.806 \approx 0.81 \) → almost perfect agreement.

📌 EXAM CRITICAL FORMULAS – MUST REMEMBER:
\( OR = \frac{ad}{bc} \)    \( RR = \frac{I_e}{I_o} \)    \( AR = I_e - I_o \)    \( PAR = I_t - I_o \)
Sensitivity = \( \frac{TP}{TP+FN} \)    Specificity = \( \frac{TN}{TN+FP} \)
PPV = \( \frac{TP}{TP+FP} \)    NPV = \( \frac{TN}{TN+FN} \)    \( \kappa = \frac{P_o-P_e}{1-P_e} \)