L2&3 - Observational Health Research 1 & 2 Flashcards
What are the aims of health research?
Aim to answer/contribute:
- better screening and diagnosis
- better therapies and treatments
- better prevention through the identification of risk factors and subsequent work in education
What is experimental research?
The researcher manipulates 1+ variables and observes the outcomes
What is observational research?
No manipulation of the situation, just observation of the outcome measures within pre-existing groups
How is the efficacy of intervention typically tested?
Experimental research design - RCT
How are risk factors for a disease typically investigated?
Observational research design.
Not the best but experimental design here would be unethical.
What is an association with variables?
A change in one variable causing a change in another
What are categorical variables?
Variables that take 1 value from a fixed selection. e.g. hair colour, favourite film, smoker/non smoker
What are non categorical variables?
Represented numerically e.g. height, weight, age, test score
How are associations between categorical variables shown?
Bar chart
How are associations between non-categorical variables shown?
Scattergrams
What does a positive correlation show?
When the value of one variable increases, the value of the other variable increases
What does a negative correlation show?
When the value of one variable increases, the value of the other variable decreases
What does no correlation look like?
The scatter is random and no line of best fit can be drawn
What is a non-linear association?
A random scatterplot with no line of best fit, showing no correlation between the variables
How is the strength of a correlation determined?
- ‘Correlation coefficient r’ ranges from -1 to 1.
- 0 means no correlation
- -1 means perfect negative
- 1 means perfect positive
How do we test for a correlation?
Pearson’s Correlation Coefficient tests the null hypothesis that there is no linear relationship
H0 >> r=0
What are the alternative terms for an exposure variable?
- Risk factor
- Predictor variable
- Independent variable
- Explanatory variable
What are the alternative terms for an outcome variable?
- Predicted variable
- Dependent variable
- Response variable
What is a confounder?
- A confounding variable is associated with both the exposure and outcome variables.
- If not controlled for, the confounder can show an association between the outcome and exposure variables that is not causal.
What are some key confounders in health studies?
- Age
- Gender
- Socio-economic Status
- Anything evidenced in your lit. review
- Anything that could plausibly be a confounder, even if the lit doesn’t mention it
How can confounders be controlled?
- Use inclusion/exclusion criteria
- Through statistical analysis
What is the principle of fair subject selection?
Emanuel et al, 2000 – “groups of individuals should not be excluded from the opportunity to participate in research without a good scientific reason or susceptibility to risk that justifies their exclusion”
What are some traditionally underrepresented groups in health studies?
- Women
- The elderly
- People from ethnic minority groups
- Children
- ‘Real-world patients’ - those with co-morbidities
What is a cross-sectional study?
- Takes all data from one point in time
- Samples participants blindly -
- Simultaneously collects data about outcome and exposure variables
- No follow up
- Allows prevalence to be calculated
- Observational Design
What is a chi-squared test?
Chi-squared tests the null hypothesis that there is no association between 2 variables
What are the requirements/assumptions for chi-squared?
- Both variables must be categorical (nominal or ordinal)
- Data points for each variable must be independent (i.e. only one measurement per variable per person)
- Each cell must have at least 5 expected cases
What is a cohort-study design?
- Follow groups of people overtime, and compare how many develop the outcome of interest
What does a cohort-study design look like?
How do you calculate the risk of disease?
Risk of disease in the specified group = number of cases/total at risk x 100
How do you calculate comparative risk?
Risk of disease in exposed group/risk of disease in the non-exposed group = x amount higher risk
What is a case-control study design?
Identify the cases and the controls (with no disease), work out the exposure in both groups, then compare the exposed to non-exposed.
It is often used to investigate rare outcomes.
What does a case-control study look like?
What is a prospective cohort?
Both exposure and outcome variables are determined in the future, after the study has begun
What is a retrospective cohort?
Both the exposure and outcome data were collected before the study began
What is a combined retrospective and prospective cohort?
Exposure data is collected in the past, but outcome data is collected after the study has begun
What are the advantages of prospective cohorts?
- Establish temporal relationships
- The research has better control over the data collection methods (as they do it theirself), meaning they have better control over confounders (when compared to a retrospective cohort).
What are the disadvantages of a prospective cohort?
- Outcome may take a long time to develop making the project very lengthy
- If the outcome is rare, you may need to recruit a high number of participants
- Very expensive
- Hard to gain funding for
What are the advantages of retrospective cohorts?
- Can establish temporal relationships
- Studies are shorter because they use existing data
- Cheaper
- Easier to win funding for
What are the disadvantages of retrospective cohorts?
- No control over the data collection methods
- This means data for exposure and outcome variables can be incomplete
- Data about confounders may be lacking
What are 3 reasons for false results?
- Chance through sampling
- Confounding
- Bias
How can you control for confounding?
- Widen exclusion criteria
- Use statistical analysis (only if the confounder is known in advance)
- Stratified analysis compares like with like
What is bias?
Systematic error in the design, conduct or analysis of a study
What are the two types of bias?
- Selection Bias
- Information Bias
What is selection bias?
When cases and control patients are selected in a way that the association is observable in the study, but doesn’t exist in the real population
What is information bias?
When the methods used for data collection are inadequate, leading to incorrect results about exposure and/or disease outcomes
What is recall bias?
When the participant misremembers an event
What does association but not causation show?
A pattern in the observed data
What does chi-squared compare?
Patterns in the observed data with patterns in the expected data (the population)
What does a P value/alpha value of <0.05 mean?
Significant because it shows the probability of finding those results due to chance is less than 5%.
The p value shows if the relationship is statistically relevant - if it exists within the population, not just the sample.
