Biostatistics
Biostats & Evidence-Based Medicine
Biostatistics

Biostats & Evidence-Based Medicine

Sensitivity, specificity, PPV, NPV, study types, biases.

Select any text to highlight it or make a flashcard.

◆Diagnostic test statistics

  • •Sensitivity = TP/(TP+FN) — rules OUT disease when negative (SnNOUT)
  • •Specificity = TN/(TN+FP) — rules IN disease when positive (SpPIN)
  • •PPV = TP/(TP+FP) — depends on prevalence
  • •NPV = TN/(TN+FN) — depends on prevalence
  • •LR+ = sens/(1-spec); LR- = (1-sens)/spec
  • •Pre-test probability × LR = post-test odds

◆Study designs (strongest to weakest)

  • •Meta-analysis > systematic review > RCT > cohort > case-control > cross-sectional > case series/report
  • •RCT: gold standard for causation; randomization eliminates confounding
  • •Cohort: prospective; can calculate INCIDENCE + RELATIVE RISK
  • •Case-control: retrospective; calculates ODDS RATIO; good for rare diseases
  • •Cross-sectional: prevalence; snapshot in time

◆Biases

  • •Selection bias: non-random selection (Berkson, healthy worker)
  • •Recall bias: case-control studies (cases remember exposure better)
  • •Lead-time bias: screening makes disease appear longer just because diagnosed earlier
  • •Length bias: slowly progressing disease over-represented in screening
  • •Confounding: third variable associated with both exposure and outcome
  • •Effect modification: relationship varies across subgroups (NOT a bias)
  • •Hawthorne effect: subjects change behavior because being observed
  • •Pygmalion effect: researcher's expectations affect outcome

◆Hypothesis testing

  • •Type I error (α): false positive — reject true null
  • •Type II error (β): false negative — fail to reject false null; Power = 1-β
  • •Increasing sample size ↑ power
  • •p-value <0.05 = statistically significant by convention
  • •Confidence interval: if it includes 1 (for ratios) or 0 (for differences), result is not significant

High-yield pearls

  • ◆Use sensitive test for SCREENING (rule out); specific test for CONFIRMATION (rule in)
  • ◆Number needed to treat (NNT) = 1/absolute risk reduction (ARR)
  • ◆Number needed to harm (NNH) = 1/absolute risk increase
  • ◆Odds ratio approximates relative risk when disease is rare
Done reading?
Track your progress by marking this complete.
Next in Biostatistics