Science Of Practice Flashcards
Incidence
The number of new cases arising in a given period of time in a specified population.
Prevalence
The number of cases within a defined population at a specific time
Risk
Number of cases divided by total population being studied
Relative risk
Risk in intervention group divided by risk in placebo group. Used to measure effect in a cohort study.
Confounding factor
Background variable which is different between the groups being compared and affects the outcome being studied.
Experimental studies
Individuals randomised to control and intervention groups by investigator. E.g. Double blind, single blind, and unblinded.
Crossover study
‘Within patient’ study. Each patient receives treatment then placebo in random order. For chronic incurable disorders.
Observational study
Predefined groups, results merely observed over time.
Case-control study
Retrospective comparison between groups (looking at the past)
Cross-sectional study
Comparison at the present time
Cohort study
Comparison of future differences (looking at the future)
Odds ratio
Number of cases divided by the number of non-cases. Appropriate for case-control studies.
Mode
Value which occurs most often
Median
Middle value in a ranking (50th centile)
Mean
Arithmetic average
Left skew
Tail points to the left, most data is at the ‘higher’ end. Mean < median < mode.
Right skew
Tail points to the right. Most data is lower. Mode < median < mean
Standard deviation
Measure of variance. Approx 68% of values within one SD, 95% within 2 SD.
Standard error
Standard deviation divided by the square root of the sample size. Used as a measure of how precisely the sample mean approximates to the population mean. Small for large samples. Used to construct confidence intervals.
P value
The probability of observing a difference of that magnitude of the null hypothesis is true.
Null hypothesis
The hypothesis that there is not difference between the two groups.
Type 1 error
Disbelieving the null hypothesis when it is actually true (because the p-value is low).
Type 2 error
Accepting the null hypothesis because the p-value is high while the null hypothesis is actually false. Small samples often lead to type 2 errors because there is insufficient power to detect differences of clinical importance.
Power of a study
The probability (as a percentage) of correctly rejecting the null hypothesis when it is false.
Parametric tests
Assume that the data is normally distributed (e.g. t-tests, Pearson’s coefficient of linear correlation).
Unpaired t-test
A two sample t-test used to compare the average values of two independent groups (e.g treatment vs placebo).
Non-parametric tests
For skewed data (e.g. Wilcoxon, Sign, Spearman’s Rank Correlation, Mann-Whitney U, Chi-squared)
Correlation coefficient (r)
Indicates how closely points lie to a line. -1-> +1. The closer it is to 0, the less the linear association between the 2 variables.
Regression equation
y = a + bx may be used to predict one variable from another. a is the intercept. b is the slope - the regression coefficient
Sensitivity
The proportion of true positives correctly identified by the test.
Specificity
The proportion of true negatives correctly identified by a test.
Positive predictive value
Proportion of those who test positive who actually have the disease.
Negative predictive value
The proportion who test negative who do not have the disease.
Likelihood ratio for a positive test
Sensitivity / (1 - specificity)
Can be multiples by pre-test odds to give post-test odds.
Likelihood ratio for a negative test
(1 - sensitivity) / specificity
Can be multiples by pre-test odds to give post-test odds.
