Epidemiology Flashcards
Positive predictive value
vs negative predictable value
Testing true positive over all positive results (chance of having a TRUE positive. chance of patient having the disease if the test is positive)
= true pos / (true pos+false pos)
-> measured horizontally across the table (1st row).
Vs testing true negative over all negative results (chance of patient not having the disease if the
= true negative / (true neg + false neg)
-> measured horizontally across the table (2nd row).
Sensitivity
Sen = T+/(T+ + F-)
How good is the test for the condition of interest (what is the true positive rate?)
A good screening test - useful for RULING OUT a condition
Measured vertically down a table (left column)
Specificity
T- /(T- + F+)
True negative rate
Positive result in a test with high specificity is useful for ruling IN a condition.
Measured vertically down a table (right column)
Likelihood ratios - what are they used for?
Positive
vs negative
Used to determined the value of performing a diagnostic test
Positive LR: probability of a true positive (w disease) to false positives (no disease)
= sensitivity / (100%-specificity)
Neg LR: likelihood of a negative test is probability of a false negative (w disease) to a true negative (no disease)
= (100 - sensitivity) / specificity
Incidence of a disease
Cases per number of people
o Number of new events that have occurred in a specific time interval
o Likelihood of developing a new event in that time interval – time interval is the denominator
Prevalence of a disease
o Number of individuals with a given disease at a given point in time
o Influenced by disease duration
o Prevalence > incidence for chronic disease
- Prevalence= incidence rate x average disease duration
- Prevalence = 15 per 100 000 x 5 years
What is the most important property of a screening test?
Sensitivity
-> you screen to RULE OUT the disease
Relative risk
Ratio of outcomes between two exposures of treatment groups
- probability of an event occurring in exposed group vs non exposed group
Odds vs probability
Odds = event occurring : event not occurring
- odds of having a girl is 1:1
Probability = event occurring / total events (50%)
- prob of having a girl is 50%
Absolute risk reduction
Real difference in absolute terms between the two exposure or treatment groups
= (events in controls / total controls) - (events in treatment/total)treatment
NNT formula
number needed to treat
= 1 / (probability exposed - probability in non-exposed)
= 100 / Absolute risk reduction (%)
Number needed to harm formula
1/(probability in exposed - probability in non-exposed)
= 100 / absolute risk increase %
Type 1 error
Rejecting null hypothesis when it is actually TRUE
OR false positive
this is the p value = 0.05
Type 2 error
False negative
Failing to reject a null hypothesis when it should have been rejected
Dependent on the POWER of a study
Power of a study
-what is this determined by?
(determined by population size, size of effect, variance within population that you are interested in)
what phase of clinical trials do paediatric patients get studied?
phase 4 (post marketing surveillance phase)
What is phase 3 clinical trials designed to test?
efficacy of an experimental therapy against standard tx (gold standard or placebo)
Cohort studies define
Divide population on whether they are exposed to a treatment or not.
Case-control studies defined
divide populations based on whether or not they have a disease
Validity of a study
what are the various aspects of this
Internal - well designed and thus minimises risk of bias
- randomisation (computer/independent statistician)
- blinding (don’t know what groups ppl are in)
- allocation concealment
- appropriate comparator
- intention to treat
- published protocol
External
- applicable to different units/environments
- -> way to do this is via Multi-center studies
- -> wide inclusion criteria (inclusive of children w comorbidities and limiting exclusion criteria)
WHO principles of screening
Condition
o The condition should be an important health problem
o There should be a recognisable latent or early symptomatic sage
o The natural history of the condition, including development from latent to declared disease, should be adequately understood
Test
o There should be a suitable test or examination
o The test should be acceptable to the population
Treatment
o There should be an accepted treatment for patients with recognized disease
Screening program
o There should be an agreed policy on whom to treat as patients
o Facilities for diagnosis and treatment should be available
o The cost of case-finding (including diagnosis and treatment of patients) should be economically balanced in relation to possible expenditure on medical care as a whole
o Case-finding should be a continuing process and not a ‘once and for all’ project
Conditions for which there is evidence for screening
- Congenital hypothyroidism
- CF
- Neonatal hearing
- PKU
- Galactosaemia
Sensitivity
- Ability to detect the disease
- Proportion of true positives (ie. proportion of individuals with disease that the test identifies)
- A sensitive test has a low number of false negatives (Sensitive – Negatives – Rules Out disease)
Specificity
- Ability to identify those WITHOUT the disease
- Proportion of true negatives (ie. proportion of individuals without the disease that the test identifies)
- A specific test has a low rate of false positives (Specific – Positives – Rules In disease)
Positive predictive value
What happens to the PPV in low prevalence situations populations (compared with higher prevalence)
- The likelihood of having a disease if a test is positive
- Proportion of tests the at are truly positive
- True positive/ total positive results
*** PPV always drops if low prevalence
Negative predictive value, define
- The likelihood that a patient with a negative test is free from disease
- Proportion of tests that are truly negative
- True negative/ total negative
- *** Usually negative predictive value is the largest percentage
Cross-sectional/prevalence study
- Assess frequency of a disease and RF at a certain time point (cross-section in time) - measures disease prevalence and risk factor associations.
Case-control:
Case control study
Retrospective study comparing two groups 1. Case (disease) vs 2. Controls (no disease) and their prior exposures/risk factors Risk of recall and selection bias etc
Cohort studies
Can be observational, prospective or retrospective in deisgn
Compare group with given exposure/RF to group without and then look at outcomes in risk of disease between both
Good for rare exposure and common outcomes
expensive and time consuming
RCTs
Prospective
Random allocation to treatment (intervention) vs control groups -> compare outcomes
Gold standard study design
What is a hazard ratio
- A hazard ratio is similar to odds ratio , but it is also time dependent
- Usually represented as a Kalplan Meier curve – the chance of an event occurring with reference to time – eg time to death after starting smoking, vs. time to death in a control group that doesn’t smoke
- How to interpret the hazard ratio:
o If the HR = 1, the event rate is the same in both treatment arms
o If the HR = 2, the event rate has occurred twice as often in the treatment group compared to control group
o If the HR = 0.5, the event rate has ocured half as often in the treatment group compared to the control group
If the same treatment and endpoint are used for a study comparing a treatment vs placebo in DIFFERENT STUDY POPULATIONS/CENTRES, which measure of treatment effect is most likely to remain approximately the same?
IE what measure tells us what the benefit for a treatment is across different populations?
RELATIVE risk reduction
What is an ecological study
Correlation/observational study comparing different groups or at different time periods
eg: assess the effects of in utero exposure to smoking, alcohol and cocaine on childhood growth.
How to calculate incidence
Number of new events in a year
15 new cases/(population size - number of pre-existing cases)
Interpretation of 95% CI (x-y)
Range of results within which we can be 95% certain that true answer lies within
The narrower the range the more precise the study’s estimates and the more confidence that study results represent true findings and not due to chance
If the confidence interval includes 1 = findings are not significant
If the confidence interval is narrow = there are ↓ variance in results or ↑ observations
If the confidence interval is large = there is ↑ variance or ↓ observations
Infant mortality rate - what are the units?
deaths per 1000 babies
calculate prevalence from incidence - what is the formula?
survival x annual incidence
How do you interpret an odds ratio?
OR > 1 - positive association
OR = 1 - no association
OR < 1 - negative / inverse association
