Public Health Flashcards
Best type of study for therapy?
Randomised controlled trial RCT > cohort
Best type of study for diagnosis?
Prospective, blind comparison to a gold standard
Best type of study for etiology/harm?
RCT > cohort
Best type of study for prognosis?
Cohort > case control
Best type of study for prevention?
RCT > cohort study
Best type of study for clinical exam?
Prospective, blind comparison to gold standard
Name the 4 phases of diagnostic research
• Phase 1: compare test results of target disorder and normal people (ideal circumstances)
• phase 2: sensitivity and specificity
• phase 3: if suspect target disorder, does test distinguish Those with and without (usual circumstances)
• phase 4: does diagnostic test better prognosis
How calculate positive predictive value
Positive disease / positive test
How calculate negative predictive value
Negative disease / negative test
How calculate study prevalence
Patients with target disease / total patients
How calculate sensitivity
Positive test / positive disease
SnOUT
How calculate specificity
Negative test / negative disease
Rule in spIN
Define pre test probability
The background prevalence in group of people you’re testing
How calculate likelihood ratio
(Positive test/positive disease) / (positive test / negative disease)
= sensitivity/ (1-specificity),
Useful because post test odds of having disease = pretest odds of disease x Lr
What is a fagan monogram and what is needed
Graphical tool for estimating how much the result on a diagnostic test changes the probability that a patient has a disease. For this, you need the pre-test probability and the likelihood ratio
How is study validity determined (8)
Primary guides
• Assignment of patients to treatments * randomized * and randomization concealed
• all patients properly *accounted for and attributed *
• * follow up * complete
• *Intention to treat analysis *: patients analyzed in the groups to which they were randomised
Secondary guides
• were all involved blind to treatment
• were groups * similar * at start of trial
• were groups * treated equally *
• is * sample size * determined prior to avoid type 2 error, and numbers reached?
Validity = truthful, free of bias (systematic error)
Which 2 forms of error in research must be excluded to prove validity
• Random error (chance, p values)
• systematic error (bias)
How calculate relative risk
risk of negative outcome on target drug / risk on placebo
(Risk ratio in % )
How calculate absolute risk
Risk of negative outcome on placebo - risk on target drug
Eg 10% -5% = absolute risk is reduced by 5% (5 less patients out of 100 will get disease)
How calculate number needed to treat NNT
NNT to get 1 less negative outcome/disease = 100 / absolute risk
Eg if absolute risk is reduced by 5%, need to treat 20 patients so that 1 less patient suffers the negative outcome eg stroke
Give better idea about clinical significance
Define statistically significant (2)
P < 0,05
95% Confidence interval cannot contain 1
How interpret rr 0,5 (95% CI 0,2 - 0,7)
I am 95% confident that if I use this target drug in the population of people with this disease then the rr will be between 0,2 - 0,7 so the relative risk reduction will be between 30 and 80%
0,5 = sample results (study result)
1 is not in the 95% CI therefore statistically significant
Define clinically significant
RRR (relative risk reduction) 25% or more
But use mostly absolute risk reduction and NNT to decide
What tells us about precision in a study?
Confidence intervals
How express uncertainty in meta-analysis
Or = 1
0 (treatment better ) or ( 1 ) - 2 (treatment worse)
