RESS Flashcards
What must a sample be when trying to make assumptions about a population?
Representative of that population
Name the 2 types of categorical data, and describe each.
Nominal: no natural ordering e.g. sex, eye colour
Ordinal: have categories that are ordered, e.g. stage in disease: absent, mild, severe
Name the two types of numerical data, and describe each.
Discrete: only whole number values, e.g. number of hospital visits.
Continuous: values with no limitations, e.g. weight
Which average and spread should be used for normally distributed data?
Mean and standard deviation
Which average and spread should be used for skewed data?
Median and inter-quartile range
What is incidence?
Number of new cases of a disease, measured over specified period.
Number of new cases/number at risk.
How do you use incidence to calculate the number at risk?
Number at risk halfway through the specified period
What leads to inaccuracies in using incidence to calculate risk, and how would you overcome this
If the number at risk varies over time. Calculate person-time risk instead, which is to add up the length of time that all people are at risk.
What is prevalence?
Number of people with disease at a specific time.
Number of people with disease / total population
What is the case fatality rate?
Number of people who die from a disease / number of people with the disease. Measured over specified time period.
Mortality rate
Number of people who die from disease / number of people in the population, over specified period
Risk
Number of new cases / number at risk. E.g. number of hospital-acquired infections in cancer inpatients / total number of cancer inpatients
Risk ratio
Comparison of risk between two groups - exposure and control.
Risk in exposed group / risk in unexposed group. Also called relative risk
When are odds used?
Where risk cannot be calculated, e.g. in case control studies
Odds calculation
Probability of an event / probability event does not occur
Odds ratio
Comparison of odds between two groups. Odds in exposed group / odds in unexposed group
What does a RR or OR of 1 mean
No difference in effect between the two groups
What does a RR or OR of less than 1 mean
The risk/odds in the exposed group is less than in the control group i.e. the exposure is protective
What does a RR or OR of more than 1 mean
Risk/odds in the exposed group is more than that in the control group i.e. the exposure is harmful
In a normal distribution, 95% data lie within ________ standard deviations
1.96
Standard error calculation
Standard deviation / square root of sample size
95% confidence interval calculation
mean ± (1.96 x standard error)
Null hypothesis
There will be no relationship between exposure and outcome
What is “no effect”
Differs for different comparisons. Difference: no effect = 0
Ratio: no effect = 1
How do you use confidence intervals to reject the null hypothesis?
If the confidence interval does not cross “no effect”, then the effect is statistically significant, and the null hypothesis is rejected.
What is meant by p<0.05
The effect is statistically significant.
Correlation
Linear association between two continuous variables - one exposure, one outcome.
Graph to show correlation
Scatterplot
Linear regression
Extends correlation to consider other confounding variables. Allows you to give and equation for line of best fit to make predictions.
Chi-squared is used to
Determine the association between categorical variables
Primary prevention
Remove the cause
Secondary prevention
Screen for the disease
Tertiary prevention
Prevent the disease by treating clinical cases
Sensitivity
True positive.
How well does the test detect the condition?
= number who correctly test positive / total number with disease
Specificity
True negative
How good is the test at correctly excluding people without the disease.
= number who correctly test negative / total number of people without the disease
Positive predictive value
If a person tests positive, what is the probability they have the condition?
Number who correctly test positive / total number who test positive
Negative predictive value
If a person tests negative, what is the probability that they do not have the condition?
Number who correctly test negative / total number who test negative
Test accuracy
Proportion of all tests that have given the correct result.
(Number who correctly test positive + number who correctly test negative)/total number of tests