Epidemiology and Medical Statistics Training Course

Participants can select and justify a study design for a defined health question.

• Cross-sectional studies estimate prevalence and examine exposure-outcome associations
• Cohort designs measure incidence and establish temporal exposure sequences
• Case-control studies efficiently investigate rare outcomes and long latency
• Randomised controlled trials estimate intervention effects under controlled allocation
• Ecological studies use group-level data while recognising ecological fallacy
• Design selection considers feasibility, ethics, validity, time and cost

Participants can assess whether health data sources and variables are fit for purpose.

• Routine health information systems assessed for completeness, timeliness and consistency
• Surveys, registries and electronic records compared for population coverage
• Exposure, outcome, confounder and modifier variables defined before analysis
• Nominal, ordinal, interval and ratio scales determine suitable analyses
• Measurement error classified as random, systematic, differential or nondifferential
• Data dictionaries document coding rules, units, missing values and provenance

Participants can plan sampling approaches and interpret survey estimates appropriately.

• Target populations, sampling frames and eligibility criteria specified precisely
• Simple random, systematic and stratified sampling compared for representativeness
• Cluster sampling accounts for geographic grouping and design effects
• Sample size calculations use expected prevalence, precision and confidence level
• Non-response bias assessed through response patterns and participant characteristics
• Survey weights adjust estimates for selection probabilities and population structure

Participants can prepare a documented and analysable health dataset.

• Data cleaning plans specify range checks, duplicates and logical consistency
• Missing data patterns distinguished from valid zero and unknown values
• Outliers investigated using clinical plausibility, distributions and source verification
• Unique identifiers support linkage while protecting personal health information
• Reproducible analysis files separate raw data, derived data and outputs
• Quality assurance logs record corrections, reasons, dates and responsible reviewers

Participants can summarise and present health data accurately for decision-makers.

• Frequency tables report counts, percentages, rates and valid denominators
• Mean, median, variance and interquartile range summarise continuous variables
• Histograms and boxplots reveal distributions, skewness and unusual observations
• Epidemic curves show outbreak timing, exposure windows and transmission patterns
• Age-standardised rates support fair comparisons between populations with different structures
• Tables and charts label populations, units, periods and uncertainty clearly

Participants can interpret probability, sampling variation and inferential statistics in health studies.

• Probability distributions describe expected variation in health event data
• Normal, binomial and Poisson distributions matched to common health outcomes
• Standard errors quantify sampling variability around estimated population parameters
• Confidence intervals express plausible ranges for effects and disease measures
• P-values interpreted alongside effect size, precision and study quality
• Type I and Type II errors inform statistical decision thresholds

Participants can choose and interpret common statistical tests for medical data.

• Null and alternative hypotheses stated for clinically meaningful comparisons
• Chi-square tests compare categorical outcomes across independent study groups
• Fisher's exact test handles sparse cells in contingency tables
• Independent and paired t-tests compare means under stated assumptions
• Analysis of variance compares means across three or more groups
• Non-parametric tests address ordinal data and non-normal distributions

Participants can calculate association measures and compare populations using standardisation.

• Risk difference quantifies absolute change attributable to an exposure
• Risk ratio compares cumulative incidence between exposed and unexposed groups
• Odds ratios estimate associations in case-control and logistic models
• Rate ratios compare person-time incidence across exposure categories
• Attributable fraction estimates preventable burden among exposed populations
• Direct and indirect standardisation adjust comparisons for age structure

Participants can identify major threats to validity and plan appropriate controls.

• Selection bias arises when study participation relates to exposure and outcome
• Information bias results from misclassification, recall error or observer effects
• Confounding occurs when a third variable distorts an association
• Directed acyclic graphs clarify causal assumptions and adjustment decisions
• Stratified analysis detects confounding and examines effect modification
• Restriction, matching and randomisation reduce bias during study design

Participants can build and interpret regression models for common health outcomes.

• Linear regression models continuous outcomes using estimated mean differences
• Logistic regression estimates adjusted odds ratios for binary outcomes
• Poisson regression models counts, rates and person-time denominators
• Model specification includes predictors, interactions, offsets and reference categories
• Residuals, influential observations and multicollinearity inform model diagnostics
• Adjusted estimates interpreted with confidence intervals and clinical relevance

Participants can interpret survival data and compare time-to-event outcomes.

• Time origins, event definitions and censoring rules specified explicitly
• Kaplan-Meier curves estimate event-free probability over follow-up time
• Log-rank tests compare survival distributions between defined study groups
• Cox proportional hazards models estimate adjusted hazard ratios
• Proportional hazards assumptions checked using plots and time interactions
• Competing risks considered when alternative events prevent the outcome

Participants can evaluate screening and diagnostic tests using appropriate performance measures.

• Sensitivity measures detection among people with the target condition
• Specificity measures correct exclusion among people without the condition
• Positive and negative predictive values depend on disease prevalence
• Likelihood ratios update diagnostic probability following positive or negative tests
• Receiver operating characteristic curves compare thresholds for continuous test results
• Screening programmes assessed for benefits, harms, feasibility and equity

Participants can structure an outbreak investigation and interpret surveillance signals.

• Outbreak verification confirms excess cases beyond expected baseline occurrence
• Line lists capture person, place, time, exposure and outcome details
• Epidemic curves generate hypotheses about source and transmission mode
• Attack rates compare illness occurrence among defined exposure groups
• Contact tracing identifies transmission chains and priority follow-up actions
• Surveillance thresholds trigger investigation, verification and risk communication

Participants can assess whether health interventions achieved intended outcomes.

• Logic models link inputs, activities, outputs, outcomes and intended impact
• Process indicators assess reach, fidelity, dose and implementation quality
• Before-after designs separate measured change from plausible programme effects
• Difference-in-differences estimates intervention effects using comparison populations
• Interrupted time series analyses assess changes after policy implementation
• Equity analyses compare outcomes by sex, age, location and income

Participants can judge the validity, relevance and applicability of health research.

• Research questions assessed against population, exposure, comparator and outcome
• Methods sections examined for selection, measurement and confounding controls
• Results tables checked for denominators, missing data and adjusted estimates
• Internal validity distinguished from external validity and transferability
• STROBE reporting items guide review of observational study reporting
• CONSORT reporting items guide review of randomised trial reporting

Participants can translate multiple sources of evidence into qualified recommendations.

• Systematic review questions specify eligibility criteria and outcome priorities
• Forest plots display pooled effects and individual study uncertainty
• Heterogeneity assessed using clinical differences, I-squared and study methods
• Publication bias considered through protocol registration and funnel plots
• GRADE criteria rate certainty across risk, consistency and precision
• Recommendations distinguish evidence findings, assumptions, values and resource implications

Participants can apply ethical and governance principles to epidemiological data work.

• Ethical review assesses risk, consent, benefit and participant protections
• Data minimisation limits collection to variables necessary for stated purposes
• De-identification reduces disclosure risk in health datasets and reports
• Access controls assign permissions according to data sensitivity and role
• Data-sharing agreements specify purpose, retention, safeguards and responsibilities
• Results communication avoids stigma, re-identification and unsupported causal claims

Participants can present epidemiological findings clearly to technical and non-technical audiences.

• Key messages state the finding, uncertainty, limitation and decision implication
• Absolute risks presented alongside relative measures to prevent misinterpretation
• Executive briefs separate evidence, interpretation, recommendation and implementation considerations
• Technical appendices document methods, assumptions, definitions and sensitivity analyses
• Risk communication addresses concern without exaggerating certainty or harm
• Briefing slides use annotated charts and decision-relevant comparison points

Participants can build, review and present a complete epidemiological analysis pack for a public health decision.

• Decision question defined with population, outcome, exposure and comparator
• Study design justified against feasibility, bias risks and available data
• Analysis plan specifies measures, models, subgroup analyses and assumptions
• Results tables and visualisations present findings with appropriate uncertainty
• Validity review documents bias, confounding, missing data and limitations
• Decision briefing presents recommendations, evidence boundaries and implementation considerations