NBDHE Review · Epidemiology (Community Health and Research Principles)
Research Study Designs in Epidemiology
On this page 6 sections
In 30 seconds
Research methodology questions on the NBDHE require you to identify study designs from descriptions, recognize each design's strengths and limitations, and match research questions to the appropriate study type. The four major observational designs — cross-sectional, case-control, cohort, and randomized controlled trial (RCT) — form a hierarchy of evidence quality driven by the temporal relationship between exposure and outcome. You must understand why the RCT is the gold standard for establishing causation, what each design can and cannot demonstrate, and the specific biases that threaten validity in each design.
The college version
Core Review
The Hierarchy of Evidence
Epidemiological study designs form a hierarchy based on their ability to establish causal relationships. From strongest to weakest for establishing causality:
- Randomized Controlled Trial (RCT) — gold standard for causation
- Cohort study — strong for causation, temporal sequence clear
- Case-control study — moderate; efficient for rare outcomes
- Cross-sectional study — weakest; cannot establish temporality
Randomized Controlled Trial (RCT)
Design: Subjects are randomly assigned to either an intervention (experimental) group or a control group. Both groups are followed forward in time (prospective), and outcomes are compared.
Key features:
- Randomization: Each subject has an equal chance of being assigned to any group. This is the defining feature that distinguishes RCTs from all other designs.
- Randomization controls for BOTH known and unknown confounders by balancing them between groups (on average).
- Prospective: The study moves forward in time from exposure assignment to outcome.
- Blinding: Ideally double-blind (neither subject nor investigator knows group assignment), but in dental studies, this can be challenging (e.g., you cannot blind a subject to whether they received a sealant).
- Allocation concealment: The person enrolling subjects does not know the next assignment. This is distinct from blinding and prevents selection bias at enrollment.
Strengths:
- Strongest evidence for causality — randomization addresses confounding
- Temporal sequence is clear (intervention → outcome)
- Can calculate incidence and relative risk directly
Limitations:
- Expensive and time-consuming
- May have limited generalizability (external validity) if the study population is highly selected
- Ethical constraints: Cannot randomize subjects to harmful exposures
- Attrition (loss to follow-up) may introduce bias
Dental example: Randomly assigning 200 schoolchildren to receive either fluoride varnish or placebo varnish, then following both groups for 2 years to compare caries incidence.
Cohort Study
Design: A group of individuals (the cohort) is identified based on exposure status and followed forward in time to see who develops the outcome. The key: subjects are selected by EXPOSURE status, then followed for OUTCOME.
Key features:
- Prospective (most common): Exposures measured at baseline, subjects followed into the future
- Retrospective cohort (historical cohort): Both exposure and outcome have already occurred, but exposure data was collected before outcome (e.g., using occupational records). The temporal direction of the study design is still exposure → outcome.
- Can calculate incidence, relative risk, and attributable risk
Strengths:
- Clear temporal sequence (exposure → outcome)
- Can study multiple outcomes from a single exposure
- Can calculate incidence rates (new cases in a defined population over time)
- Less susceptible to recall bias than case-control studies (exposure measured before outcome)
Limitations:
- Inefficient for rare outcomes (must follow many subjects for a long time to accumulate enough outcome events)
- Expensive and time-consuming for diseases with long latency
- Loss to follow-up threatens validity
- Confounding still possible (no randomization)
Dental example: Following a cohort of 500 dental hygienists for 10 years, measuring their occupational exposure to aerosol-generating procedures and comparing respiratory disease incidence between high-exposure and low-exposure groups.
Case-Control Study
Design: Subjects are selected based on OUTCOME status. Cases (those with the disease/outcome) are compared to controls (those without the disease/outcome), and the investigator looks BACKWARD in time to compare prior exposure between the two groups.
Key features:
- Retrospective by design: Start with outcome, look back for exposure
- Controls must be selected from the same source population that gave rise to the cases
- The measure of association is the odds ratio (OR), which approximates the relative risk when the outcome is rare
- Efficient for rare outcomes (you deliberately recruit subjects with the outcome)
Strengths:
- Efficient for studying rare diseases or outcomes (don't need to wait for them to develop)
- Relatively quick and inexpensive
- Can study multiple exposures for a single outcome
Limitations:
- Cannot calculate incidence or absolute risk (the investigator controls the ratio of cases to controls)
- Highly susceptible to recall bias (cases may remember/report exposures differently than controls)
- Selection bias: Controls may not represent the exposure distribution of the source population
- Temporal sequence can be ambiguous (did the exposure precede the outcome?)
Dental example: Identifying 100 patients with oral cancer (cases) and 100 age-matched patients without oral cancer (controls), then reviewing their tobacco and alcohol use histories.
Cross-Sectional Study
Design: Exposure and outcome are assessed simultaneously in a defined population at a single point in time. It is a "snapshot" or "prevalence survey."
Key features:
- No follow-up: Everything is measured once
- Measures prevalence, not incidence
- Cannot establish temporal sequence — the chicken-and-egg problem
Strengths:
- Quick and inexpensive
- Useful for hypothesis generation and needs assessment
- Good for describing disease burden in a population
Limitations:
- Cannot establish temporality (did exposure precede outcome or vice versa?)
- Cannot calculate incidence or relative risk
- Susceptible to prevalence-incidence bias (Neyman bias): If the disease has high mortality or short duration, prevalent cases may not represent all cases
- Only identifies associations, not causal relationships
Dental example: Screening 1,000 third-graders in a school district for DMFT and surveying their families about dietary habits at the same visit. Finding an association between high sugar consumption and high DMFT does not prove sugar caused the caries.
Comparison Table
| Feature | RCT | Cohort | Case-Control | Cross-Sectional |
|---|---|---|---|---|
| Selection by | Neither | Exposure | Outcome | Neither |
| Temporal direction | Prospective | Prospective (usually) | Retrospective | Simultaneous |
| Causality evidence | Strongest | Strong | Moderate | Weakest |
| Incidence calculation | Yes | Yes | No | No |
| Measure | Relative Risk | Relative Risk | Odds Ratio | Prevalence Ratio |
| Rare outcome | Inefficient | Inefficient | Efficient | Inefficient |
| Cost/Time | High | High | Low-Moderate | Low |
| Confounding | Controlled by randomization | Potential | Potential | Potential |
| Recall bias | Low | Low | High | Moderate |
| Loss to follow-up | Concern | Concern | N/A | N/A |
Clinical/Board Application
Board-style question: "A researcher wants to determine whether occupational nitrous oxide exposure is associated with spontaneous abortion in dental hygienists. Because spontaneous abortion is a relatively rare outcome, which study design is most appropriate?"
Answer: Case-control study. Rare outcomes are most efficiently studied by identifying cases (hygienists who experienced spontaneous abortion) and controls and comparing exposure histories. A cohort study would require following thousands of hygienists for years to accumulate enough cases.
Common Traps
- Trap: Confusing a retrospective cohort with a case-control study. The defining difference: in a cohort study, subjects are selected by exposure status regardless of outcome; in a case-control study, subjects are selected by outcome status. A study using historical records to identify workers exposed to a chemical (exposure-based selection) and then checking cancer registries for outcomes is a retrospective COHORT study, not a case-control study.
- Trap: Thinking case-control studies are unethical. They are observational — no intervention is assigned. They can be perfectly ethical.
- Trap: Assuming a cross-sectional study that finds an association has proven causation. Without temporality, it has not.

Eli explains
The same idea, in plain words
Explain it like I’m 10
Imagine four different ways to study whether eating candy causes cavities:
RCT (best): Flip a coin to assign kids to "candy" or "no candy" groups and watch what happens for 2 years. Because you randomly assigned who gets candy, differences in cavities at the end are probably caused by the candy.
Cohort: Find kids who already eat lots of candy and kids who don't. Follow both groups for years and see who gets more cavities. You know candy came first.
Case-control: Find kids with lots of cavities and kids without. Ask both groups how much candy they ate in the past. This is backward-looking — kids with cavities might remember (or report) their candy eating differently.
Cross-sectional: Check kids' teeth and ask about candy at the same moment. Bad teeth and candy eating go together, but you can't tell which caused which.
Key takeaways
- RCT: Randomization is the defining feature — controls for known AND unknown confounders
- Cohort: Exposure → Outcome; follow forward; can calculate incidence
- Case-control: Outcome → Exposure; look backward; efficient for rare diseases; odds ratio
- Cross-sectional: Snapshot; prevalence only; cannot prove causation
- Only RCTs can establish causation with high confidence
- Recall bias hits case-control studies hardest
- Loss to follow-up threatens cohort studies and RCTs
- Q1: What is the defining feature that distinguishes an RCT from all other study designs?
- A. Random assignment of subjects to intervention and control groups ✓ — Randomization is unique to RCTs and is what allows them to control for both known and unknown confounders, making them the gold standard for causal inference.
- B. Prospective data collection
- C. Use of a control group
- D. Blinding of investigators
- Q2: A researcher identifies 50 patients with erosive tooth wear and 50 age-matched controls without erosive wear. Both groups complete a detailed dietary questionnaire about past consumption of acidic beverages. This study design is:
- A. Prospective cohort
- B. Case-control ✓ — Subjects are selected based on outcome (erosive wear = cases, no erosive wear = controls), and the investigator looks backward at past exposure (dietary acidic beverages). This is the classic case-control structure.
- C. Cross-sectional
- D. Randomized controlled trial
- Q3: Which study design is most susceptible to recall bias?
- A. RCT
- B. Prospective cohort
- C. Case-control ✓ — Case-control studies are highly susceptible to recall bias because subjects with the outcome (cases) may remember and report past exposures differently than subjects without the outcome (controls), especially when recall is relied upon (e.g., dietary habits, environmental exposures).
- D. Cross-sectional
Quick check
3 questions here. Answers stay hidden until you check.
A researcher identifies 50 patients with erosive tooth wear and 50 age-matched controls without erosive wear. Both groups complete a detailed dietary questionnaire about past consumption of acidic beverages. This study design is:
Which study design is most susceptible to recall bias?
Study toolsYou’ll learn to
You’ll learn to
- Identify the four major epidemiological study designs from description or scenario
- Explain the temporal relationship between exposure and outcome for each design
- Compare the strengths, limitations, and susceptibility to bias for each study type
- Match research questions to the most appropriate study design
- Differentiate prospective from retrospective data collection
- Explain why correlation does not equal causation in observational studies
Educational content only. It is not medical, legal or professional advice. Found an error? Tell us.
