Revision notes · Lecture 1 · Free
Introduction to health research
In one line: health research is any systematic study that produces knowledge to improve health. Before it starts, it faces three reviews: scientific, ethics and regulatory.
What this lecture's questions test
Out of the 30 questions on this lecture:
| Idea | Questions |
|---|---|
| The three reviews, and which approval is mandatory | 7 |
| Types and scope of health research | 6 |
| Planning a sound study | 6 |
| Confounders and effect modifiers | 4 |
| Pilot studies, and what is not a study design | 4 |
| Random error and bias | 3 |
1. The three reviews
| Review | Asks | Looks at |
|---|---|---|
| Scientific | Is the science sound? | Question, design, sample size, methods |
| Ethics | Are participants protected? | Consent, risks and benefits, privacy, the researcher's competence, care during and after |
| Regulatory | Are the laws followed? | Funds and how they are used, samples or data sent abroad, intellectual property |
- Only ethics committee approval is mandatory for every study on people. The others depend on the study.
- Peer review of a manuscript happens after the study, so it is not needed before starting one.
- Setting up an institute's scientific committee is administration, not a step in your study.
Remember it: the science → scientific. The people → ethics. Money, borders, patents → regulatory.
2. Types of research
- Bench (basic): in the laboratory, on animals or cells.
- Clinical: on patients.
- Translational: the bridge from bench to bedside.
- Implementation: does a programme already known to work actually work on the ground? Acceptability, feasibility, delivery.
- Health research covers disease burden, prevention and programme evaluation. It confirms what is known, finds what is new, and draws on biomedical science, biostatistics and social science.
3. Random error or bias?
| Random error | Systematic error (bias) | |
|---|---|---|
| Cause | Chance | A flaw in design, conduct or analysis |
| Direction | Either way | One direction |
| Reduced by | A bigger sample (never eliminated) | Better design; sample size does not help |
Trap: "better design eliminates random error" is false. Design fixes bias; sample size shrinks random error.
4. Confounder or effect modifier?
| Confounder | Effect modifier | |
|---|---|---|
| What it does | Distorts the association: linked to the exposure and, on its own, to the outcome | Changes the strength of a real association from one group to another |
| Deal with it by | Design (randomise, restrict, match) or analysis (stratify, adjust) | Stratify and report each group separately; never adjust it away |
| Example | Smoking, in a study of alcohol and lung cancer | Age, when a vaccine protects children far better than older adults |
- Known effect modifiers come from a thorough literature review, not guesswork.
5. Pilot studies and study designs
- A pilot study is a small rehearsal of the main study: it tests the tools, the sampling frame and feasibility, before main data collection. Ethics approval does not wait for it.
- "Which is not a study design?" → pilot study or translational study. Those describe the size or purpose of a study, not its design.
6. Planning a sound study
- The sample must be representative (external validity) and adequate (power). An adequate sample size is a prerequisite.
- Inclusion and exclusion criteria must be specific and measurable.
- Give every team member clear roles and responsibilities.
- Plan on justification, a focused question and standard case definitions. Never financial gain.
- Does a new vaccine programme reduce cases? → a field (randomised) trial.
Watch out: one question in this lecture combines a confounder statement with an effect-modifier statement that ends "…may be excluded from the study". The key and standard teaching disagree; the app explains it on that question.
Last look before the exam
- Only ethics approval is always mandatory.
- Intellectual property, samples or data abroad, funds → regulatory.
- Bigger sample → less random error. Better design → less bias.
- Confounder: adjust for it. Effect modifier: stratify and report.
- Pilot and translational studies are not study designs.