CRQ Flashcards
Hierarchy of evidence
Meta analysis Rct Non randomised CT Cohort study Case control study Cross sectional surgery Case series Case report
Sensitivity
Chance of it being correctly positive
Ie does have the disease and is positive on the test
A/ a+c
Positive predictive value
If positive on a test what is the chance they actually are positive for the disease?
A/a+b
Specificity
If they don’t have the disease what is the chance the test correctly identifies them as negative?
D/b+d
Negative predictive value
If the patient is negative on the test what is the chance they actually are a true negative without the disease
D/c+d
What does CONSORT diagram stand for?
Consolidated standards of reporting trials
To improve quality of rct
Patient flow
Includes those lost to follow up
Evidence based minimum set of recommendations for reporting rcts
Standardised way to present report findings
What does PICO stand for?
Formulating a question for a study
Patient or problem (population)
Intervention
Comparison
Outcome
Case report
Experience of one person
Anecdotal
Prone to chance and bias
Case series
Group of people studied
Useful for rare diseases
5 children who presented to EF with abdo epilepsy etc
Cohort study
Consequences of exposure to a risk factor
Same group of people followed
One with risk exposure and one without
No one has the disease at start of the trial
Can do retrospectively
Takes long time (usually years)
E.g 500 people who smoke cannabis monitored for 15 years to see if at greater risk of developing schizophrenia
Case control study
Compares those with disease and those without
and looks at a risk factor
50 women with hepatitis and 50 women without queries about ear piercing to see if a risk factor
Cross sectional study
Surveyed group of people at one point in time
Prevalence of exposure and outcome in a population
E.g population questionnaire examining prevalence of stroke risk factors
RCT
Minimises bias
Measures efficacy
Compare two treatments
Kids with fever given either paracetamol or ibuprofen to determine which is better at reducing fever
Endpoints
Study should say how endpoint selected e.g morbidity, plaque developed, major CV risk factor
Validity
How well does it measure what it was supposed to
Measures variable
Criteria compared to what is know to be valid
Reliability
How consistent a test is on repeated measurement
How true is it to true value
Consistency of measurement
Measurement error either:
systematic (dif readings on same machine)
Random error- dif in weight readings
Incidence
Rate of occurrence of new cases over a period of time
Measures risk of disease
Number of new cases over time period divided by population size
Prevalence
Proportion with disease at a given time
Number with disease divided by population size
Variance
Sum of all the differences between the values plus the mean squared
Divided by
Total number of observations - 1
( work out mean Minus mean from each value Square each result Sum of squared results Divide by number of results -1
Standard deviation
Degree of spread from the mean
Square root of variance
2 standard deviations above and below =95%
Z score
Value of an observation into number of standard deviations frommean where it lies
Observation -mean divided by standard deviation
95% confidence interval
95% of means in the sample like in 1.96 standard error
So 95% of the time the population mean would lie in this range
Equation for risk
Number of times an event likely to occur
Divided by
Total number of events possible
Ie one in 6 fall ill
Risk = 1/6 = 0.167 or 16.7%
Equation for odds
Ratio of number of times likely to occur
Divided by number of times not likely to occur
1/5 = 0.2 odds of falling ill
Control event rate equation in 2x2
C/c+d
Experimental event rate
Absolute risk of outcome in experimental group
A/a+b
Absolute risk reduction
CER - EER
If 0.4 shows 40% reduction in risk from control group
Relative risk
EER/CER
If <1 reduced risk
Relative risk reduction
CER-EER/ CER
Number needed to treat
1/ARR
Odds ratio
A x d
Divided by
B x c
> 1 more likelihood of developing that condition with risk factor
P value
Usually less than 0.05
Probability of obtaining the result by chance
Type 1 error
False positive result
Wrongful rejection of null hypothesis
Usually bias, confounders
Type 2 error
False negative
Wrongful acceptance of null hypothesis
Small sample size
Power
Probability that a type 2 error will not be made
Ability to detect smallest difference between the two groups
Bigger sample size = more power
0.8 power generally accepted so only 20% chance of type 2 error
Sample size
Alpha value that is as
Variability if decreased this decreases power
Chi squared test
Compare significance of categorical data
The 2x 2 boxes
Unpaired data
Tests hiv acquisition between iud or dmpa
Ie tests null hypothesis that the two groups are the same
Test for normal distribute data
T test
Anova
Tests for non normal data
Wilcoxons rank
Mann Whitney u
Anova
Kruskal Wallis
Four stages listed in consort diagram for rct
Enrolment
Intervention allocation
Follow up
Analysis of data
Forest plot
Used to present results of meta analysis
Funnel plot
Used to identify publication bias
Fishers exact test
Significance of independent categorical data
Alternative to chi squared test when small sample size
Test from funnel plot and in metaanalysis to test for asymmetry
Egger test
Therapeutic papers
Eg rct looking at high flow oxygen on mortality in copd
Objective Design Setting Population Method Intervention Control Outcome Results Conclusions
Diagnostic papers
Eg diagnosis of intersussepyipn by uss in ed
Objective Design Setting Population Methods Test under ix Gold standard Results Conclusion
Meta analysis
Paper identification Secelection Funnel plot Forest plot Stat analysis (heterogeneity)
