Biostatistics & Public Health

Study design, test characteristics, bias and ethics — fully learnable points that many candidates leave on the table.

Exam weight 4–6% · 2 QBank items · 3 flashcards

Topic completion0 / 4

Diagnostic test characteristics

Sensitivity rules OUT when negative (SNOUT); specificity rules IN when positive (SPIN).
PPV rises with disease prevalence; sensitivity and specificity do not change with prevalence.
Lowering the cutoff increases sensitivity and decreases specificity; area under the ROC curve summarises overall accuracy.
Likelihood ratio positive = sensitivity / (1 − specificity); LR+ above 10 substantially raises post-test probability.

Case-control gives an odds ratio; cohort gives relative risk; only randomised trials support causal claims.
Absolute risk reduction = control event rate − treatment event rate; NNT = 1/ARR.
Lead-time bias makes screening look beneficial by detecting disease earlier without changing outcome.
Confounding is addressed by randomisation, restriction, matching, stratification and multivariable adjustment.

Type I error (α) is a false positive; type II error (β) is a false negative; power = 1 − β and rises with sample size.
A 95% confidence interval that crosses 1 for a ratio measure (or 0 for a difference) is not significant.
t-test compares 2 means, ANOVA 3 or more, chi-square compares proportions.
Positive skew: mean > median > mode; in a normal distribution all three are equal.

Four principles: autonomy, beneficence, nonmaleficence, justice; autonomy usually wins in a competent adult.
Minors can consent independently for emergency care, sexually transmitted infection, contraception, pregnancy and substance-use care in most jurisdictions.
Decision-making capacity requires understanding, appreciation, reasoning and a consistent stable choice — it is decision-specific.
For an angry or grieving patient, explore the emotion first; never respond defensively or transfer care prematurely.