Preparatory Mindset
Medical research is the foundation of evidence-based medicine. The exam tests: types of research (qualitative vs quantitative, descriptive vs analytical vs experimental), key variables (independent, dependent, confounding), sampling, reliability & validity in research, basic statistics (mean, SD, p-value, CI), research ethics (Helsinki, IRB, informed consent), and the study designs (cohort, case-control, RCT, cross-sectional).
Exam mindset: know the difference between qualitative and quantitative data, descriptive vs analytical vs experimental methods, the experimental method (RCT, hypothesis testing, control), types of variables (independent, dependent, confounders), and the research cycle.
Core Concepts
1. Objectives of science (recap)
- Describe, explain, predict, control/influence — apply to research
2. Steps of scientific research (MUST KNOW)
- Observe + formulate problem
- Review literature (current state)
- Develop hypothesis (testable prediction)
- Design study (methodology, sample, variables)
- Collect data
- Analyze data (statistics)
- Interpret results
- Report & publish
3. Types of data (MUST KNOW)
- Qualitative: subjective, non-numeric (interviews, observations, narratives, themes) - Quantitative: numeric, measurable (surveys, lab measurements)
- Subjective: self-reported (pain, mood, satisfaction) - Objective: measured externally (BP, lab values)
- Continuous: infinite values (age, BP, weight) - Categorical: discrete groups (gender, blood type, disease ±)
- Qualitative vs. Quantitative:
- Subjective vs. Objective:
- Continuous vs. Categorical:
4. Key concepts (MUST KNOW)
- Definition of "operation": the way a variable is measured or defined in a study (operationalization)
- Reliability: consistency of measurement (test-retest, internal consistency — Cronbach's α)
- Validity: accuracy of measurement (content, construct, criterion)
- Bias: systematic error (selection, recall, confounding, observer bias)
- Confounding: extraneous variable that affects both exposure & outcome
- Randomization: random allocation to groups (reduces bias/confounding)
- Blinding: single (participant), double (participant + investigator), triple (+ outcomes assessor)
5. Research methods (MUST KNOW)
Descriptive methods
- Case study: detailed report of an individual (rare conditions, hypotheses)
- Survey: questionnaire/interview of a sample (cross-sectional)
- Naturalistic observation: watching behavior in natural setting
- Archival/Epidemiological: secondary data analysis
Correlational methods
- Correlation does NOT imply causation (key exam trap)
- Correlational study: examine relationship between 2 variables (Pearson r for continuous; chi-square for categorical)
- Predictive: −1 to +1; r² = proportion of variance explained
Experimental methods
- Independent variable (IV): the variable manipulated - Dependent variable (DV): the outcome measured - Control group vs. Experimental group
- True experiment: randomly assign to independent variable (treatment vs. control)
- Randomized controlled trial (RCT): gold standard for clinical research
- Quasi-experiment: no randomization (e.g., pre-post comparison)
- Crossover: each participant receives both treatments (washout between)
- Factorial: multiple IVs simultaneously
6. Sampling (MUST KNOW)
- Random sample: every member has equal chance (gold standard)
- Convenience sample: easy to access (selection bias)
- Stratified sample: ensure subgroups represented
- Sample size: powered to detect a meaningful difference (type II error if too small)
- Power (1 − β): probability of detecting a true effect
- Type I error (α): false positive (rejecting true null); usually α = 0.05
- Type II error (β): false negative (failing to reject false null)
- Confidence interval (CI): range of plausible values (95% CI = mean ± 1.96 × SE)
7. Basic statistics
- t-test: compare 2 group means - ANOVA: compare ≥3 group means - Chi-square: categorical associations - Regression: continuous predictors → continuous outcome - Correlation: linear relationship
- p < 0.05 = statistically significant (but not necessarily clinically meaningful)
- Descriptive: mean (μ, x̄), median, mode, SD (σ, s), range, IQR
- Inferential:
- p-value: probability of seeing data this extreme if null hypothesis is true
- Effect size: Cohen's d (small 0.2, medium 0.5, large 0.8)
- CIs and effect sizes are more informative than p-values alone
8. Research ethics (MUST KNOW)
- Belmont Report (1979): respect for persons, beneficence, justice
- Declaration of Helsinki: ethical principles for medical research
- Informed consent: voluntary, informed, comprehension
- IRB / Ethics Committee review before research
- Confidentiality & data security
- Vulnerable populations: special protections (children, prisoners, pregnant women)
- Conflicts of interest disclosure
- Clinical trial registration (e.g., ClinicalTrials.gov)
- Right to withdraw at any time
9. Levels of evidence (Evidence-Based Medicine)
- Level I: systematic reviews of RCTs (meta-analyses)
- Level II: individual RCTs
- Level III: cohort studies
- Level IV: case-control studies
- Level V: case series, expert opinion
- Always start from the top for clinical decisions
10. Translating research to practice
- GRADE: grading of evidence (high, moderate, low, very low)
- Clinical practice guidelines use the best evidence
- Evidence-based medicine: integration of best evidence + clinical expertise + patient values
High-Yield Points
- Qualitative vs. Quantitative: subjective vs. numeric
- Subjective vs. Objective: self-report vs. external measurement
- Reliability = consistency; Validity = accuracy
- Correlation ≠ causation
- IV = manipulated; DV = measured; Confounder = third variable
- RCT is the gold standard; conceal allocation; blind to treatment
- Type I error (α 0.05) = false positive; Type II error (β) = false negative
- Power = 1 − β; sample size determines power
- 95% CI = mean ± 1.96 × SE
- p < 0.05 = significant; but effect size matters
- Belmont Report: respect, beneficence, justice
- Informed consent: voluntary, informed, comprehension, right to withdraw
- Levels of evidence: I (meta-analysis) → V (expert opinion)
Topic Summary
Research methods are the foundation of evidence-based medicine. Scientific research follows a step-wise process: observation → literature review → hypothesis → design → data collection → analysis → interpretation → reporting. Data types are qualitative (subjective, narrative) vs. quantitative (numeric); subjective vs. objective; continuous vs. categorical. Key concepts: reliability (consistency), validity (accuracy), bias (systematic error), confounding (extraneous variable affecting both exposure and outcome). Research methods include descriptive (case study, survey, observation), correlational (Pearson r, chi-square — correlation does not imply causation), and experimental (RCT — gold standard: manipulation, randomization, control group). Sampling methods include random (gold standard), convenience, stratified; sample size determines statistical power (1 − β). Inferential statistics include t-test, ANOVA, chi-square, regression. The p-value is the probability of the data given the null hypothesis; p < 0.05 is the conventional significance threshold. Ethical research follows the Belmont Report (respect, beneficence, justice), the Declaration of Helsinki, IRB review, informed consent, confidentiality, and transparency. Hierarchy of evidence runs from meta-analyses (Level I) to expert opinion (Level V).
LMCHK OSCE Practice
- Reading a paper: identify the study design (RCT, cohort, case-control, cross-sectional), sample size, primary outcome, results (effect size, CI, p-value), and limitations.
- Critical appraisal: assess randomization, blinding, allocation concealment, follow-up completeness, and conflict of interest.
- Compute a 95% CI: for a mean of 100 and SE of 5, the CI is 100 ± 1.96 × 5 = 90.2 to 109.8.
- Interpret a p-value: "If the null hypothesis is true, there is a X% chance of observing data this extreme."
- Differentiate confounding from intermediate variable: a confounder is associated with both exposure and outcome; an intermediate variable is on the causal pathway.
- Formulate a clinical question: PICO (Population, Intervention, Comparison, Outcome).
- Research ethics case: patient with advanced cancer — informed consent for experimental therapy; discuss risks, unknown benefit, alternatives; voluntary.
- Bias identification: recall bias (cases remember more than controls), selection bias (non-random sampling), observer bias (subjective assessment).
- Sample size justification: for a small effect size, need a larger sample; for a binary outcome, the rarer outcome needs more cases.
- Evidence hierarchy: explain why a meta-analysis of RCTs is stronger than a single case report.
- EBM in clinical practice: for a 60-year-old with hypertension, look for the latest systematic review (Level I) before prescribing.
- Memorize: "Correlation ≠ causation" — the classic exam trap.
- Randomized vs. non-randomized: confounders are minimized by randomization.