Types of Epidemiology: Descriptive and Analytical Epidemiology
Epidemiology is commonly divided into descriptive and analytical approaches. Descriptive epidemiology shows how disease is distributed by person, place and time; analytical epidemiology compares groups to test why the pattern occurs.
Learning objectives
- Differentiate descriptive from analytical epidemiology.
- Describe disease patterns using person, place and time.
- Formulate an epidemiological hypothesis from descriptive findings.
- Compare major observational and experimental study designs.
- Interpret basic measures of association while recognising bias and confounding.
Descriptive epidemiology
Descriptive epidemiology organises health data to answer who, where and when. It measures disease frequency, identifies high-risk groups, detects trends and generates hypotheses for further investigation. It does not by itself prove why an association exists.
Person
Age, sex, occupation, education, income, behaviour, immunity, genetics, pregnancy, comorbidity and access to care.
Place
Country, district, village, household, workplace, school, health facility or environmental exposure zone.
Time
Long-term trends, seasonal variation, cyclic patterns, outbreaks and time since exposure.
Common descriptive designs
| Design | Description | Strength | Important limitation |
|---|---|---|---|
| Case report | Detailed description of one unusual patient or event. | Early signal of a new disease, adverse effect or presentation. | No comparison group; cannot estimate risk. |
| Case series | Description of several patients with a similar condition. | Defines clinical spectrum and generates hypotheses. | Selection bias and no denominator. |
| Cross-sectional survey | Exposure and outcome measured at one time or during a short period. | Estimates prevalence and health-service needs. | Temporality may be unclear. |
| Ecological study | Compares exposure and disease measurements for groups or areas. | Uses population data and explores geographical patterns. | Ecological fallacy: group relationships may not apply to individuals. |
| Routine surveillance analysis | Uses continuously or regularly collected reports. | Detects trends and outbreaks. | Depends on reporting completeness and consistent definitions. |
Presenting descriptive data
- Person: frequency tables, age-specific rates and stratified charts.
- Place: spot maps, area-rate maps and facility catchment comparisons.
- Time: epidemic curves, seasonal graphs and long-term trend lines.
Why denominators matter
Counts can mislead. A district with 200 cases may have lower risk than a district with 100 cases if its population is much larger. Compare populations using proportions or rates with clearly defined numerators, denominators and time periods.
Analytical epidemiology
Analytical epidemiology tests specific hypotheses by comparing groups. It asks whether an exposure is associated with an outcome, how strong that association is and whether alternative explanations such as chance, bias or confounding could account for the finding.
Formulating a hypothesis
A useful hypothesis identifies the population, exposure, comparison group and outcome. Example: “Among households in Village A, people drinking from Well X have a higher risk of acute watery diarrhoea than people using protected boreholes.”
Major analytical study designs
| Design | Starting point | Main measure | Best use |
|---|---|---|---|
| Case-control | Select people with the outcome (cases) and without it (controls), then compare previous exposure. | Odds ratio | Rare diseases, long-latency outcomes and outbreak investigations. |
| Cohort | Group participants by exposure and follow or reconstruct outcomes. | Risk ratio, rate ratio, risk difference | Rare exposures, multiple outcomes and demonstrating temporality. |
| Analytical cross-sectional | Measure exposure and outcome together, then compare prevalence. | Prevalence ratio or prevalence odds ratio | Common conditions and rapid population assessments. |
| Randomised controlled trial | Randomly allocate eligible participants to intervention or control. | Risk ratio, risk difference and treatment effect | Evaluating efficacy of preventive or therapeutic interventions. |
| Community trial | Allocate interventions to groups or communities. | Population-level effect measures | Health education, sanitation, vector control and service programmes. |
Case-control studies
Investigators select cases using a standard case definition and controls from the same source population. They then determine previous exposure in both groups. Controls should represent the exposure distribution that would have occurred among cases if they had not developed the disease.
Odds ratio
For a 2 × 2 table with exposed cases a, unexposed cases c, exposed controls b and unexposed controls d:
Odds ratio = ad ÷ bc
OR = 1 suggests no association; OR > 1 suggests increased odds with exposure; OR < 1 suggests a protective association. Interpretation also requires a confidence interval and assessment of bias and confounding.
Cohort studies
A cohort study compares disease occurrence among exposed and unexposed people. A prospective cohort follows participants forward; a retrospective cohort uses reliable past records to reconstruct exposure and outcomes.
- Risk ratio: risk in exposed ÷ risk in unexposed.
- Risk difference: risk in exposed − risk in unexposed.
- Attributable fraction: proportion of disease among exposed participants associated with the exposure.
Experimental epidemiology
In an experimental study the investigator assigns an intervention. Randomisation improves comparability between groups, allocation concealment prevents foreknowledge of assignments, and blinding reduces differences in measurement or care. Ethical approval, informed consent, safety monitoring and clinical equipoise are essential.
Descriptive versus analytical epidemiology
| Feature | Descriptive | Analytical |
|---|---|---|
| Main questions | Who? Where? When? What is the burden? | Why? How? Is exposure associated with outcome? |
| Comparison group | Often absent | Essential |
| Primary role | Describe patterns and generate hypotheses | Test hypotheses and estimate associations |
| Typical output | Counts, proportions, rates, maps and curves | Risk ratios, odds ratios, rate ratios and effect estimates |
| Causal inference | Limited | Stronger, but still requires careful assessment |
Threats to valid interpretation
Chance
Random variation can create an apparent association. Confidence intervals and statistical tests help describe uncertainty.
Bias
Systematic error in selection, information collection or analysis that moves results away from the truth.
Confounding
A third factor associated with both exposure and outcome distorts the exposure–outcome relationship.
Effect modification
The true association differs across levels of another variable, such as age or sex, and should be reported rather than controlled away.
Common examples of bias
- Selection bias: participants included in the study differ systematically from those not included.
- Recall bias: cases remember past exposures differently from controls.
- Interviewer bias: knowledge of disease or exposure status influences questioning.
- Misclassification: exposure or outcome is placed in the wrong category.
- Loss to follow-up: outcome information is missing differently across cohort groups.
From description to action: worked example
A health centre line-list shows many typhoid cases among adolescents from one school during the same week. Descriptive analysis by person, place and time generates the hypothesis that a school food or water source is involved. Investigators then conduct a retrospective cohort study among students, calculate attack rates for each meal and compare exposed with unexposed groups. A high risk ratio for one food item supports targeted environmental investigation and control.
Choosing a study design
- Use a cross-sectional survey to estimate prevalence.
- Use a case-control study for a rare disease or when rapid outbreak investigation is required.
- Use a cohort study when exposure is clearly defined and incidence can be measured.
- Use a randomised trial when assigning the intervention is ethical and feasible.
- Use an ecological design for group-level hypotheses, but avoid individual-level conclusions.
Exam-focused summary
- Descriptive epidemiology characterises person, place and time and generates hypotheses.
- Analytical epidemiology requires comparison groups and tests hypotheses.
- Case-control studies start with disease status and commonly estimate an odds ratio.
- Cohort studies start with exposure status and can measure incidence and risk ratios.
- Association does not prove causation; assess temporality, strength, consistency, plausibility, chance, bias and confounding.
Review questions
- Differentiate descriptive and analytical epidemiology.
- Describe person, place and time variables in a malaria investigation.
- Compare case-control and cohort studies.
- Calculate and interpret an odds ratio from a 2 × 2 table.
- Explain selection bias, information bias and confounding.
References and further reading
Educational note: Use current Uganda Ministry of Health surveillance definitions and ethical requirements in actual public-health work.
