MHS Revision Flashcards
Associations may be the result of…
Chance
Bias
Confounding
REAL ASSOCIATIONS
What type of study is a cohort study?
An observational study.
Alpha level is usually ____
0.05 (1 in 20)
P value = ______
The estimated probability of seeing a difference of a result equal to or more extreme than that observed in a study occurring by chance.
Confounding means _____
The effects of a proposed risk factor is mixed up with other risk factors.
Doll and hill study 1952 Looked at what?
Lung cancer, with controls of no lung Cx in a hospital.
Why is recall bias an issue?
People who have had the disease or illness tend to be able to recall more clearly.
What methoda can you employ to deal with confounding?
- ) In design, restrict the study to one stratum of the confounder, i.e. only study smokers or only non- smokers.
- ) In study design, match for potential confounders.
Residual confounding.
Occurs when all known confounders have been accounted for but there are other, unknown confounders that cannot be accounted for.
Incidence
No. of NEW events, occuring during a period of time. I.e. 1000/year.
Prevalance
Point prevalence - how many with a condition with a SNAPSHOT
Period prevalence - NO. of people with diabetes during a period of time - NOT just new cases.
Incidence Density
Looking at incidence in individuals over varied lengths of time. This is because different individuals will be “at risk” for varied lengths of time.
the standardised mortality ration is an example of ______ standardisation.
Indirect
SMR = _____/_____
Observed No of deaths/Expected number of deaths.
Define Sensitivity
How sensitive the test is to detect the presence of a disease, basically saying how likely it is the test will give positive result if a patient HAS the condition.
Define Specificity:
Is a result POSITIVE to the target condition? In other words, how likely will someone WITHOUT condition give a NEGATIVE reponse.
Low specificity will lead to Type 2 error. (saying that there is no difference between condition and not when there is)
Define POSITIVE PREDICTIVE VALUE.
How VALUABLE is a POSITIVE test result predicting a condition. AKA: how many of the people that tested positive actually HAVE the disease?
Calculated by (true positive/sum positive results)
Define NEGATIVE PREDICTED VALUE.
What proportion of patients with negative values are TRULY FREE of the disease?
i.e. (true negative/sum negative)
Type ____ error is usually considered the “worst” in research. Type ____ error is worse in medicine, as it will mean patients requiring treatment will not receive it.
- ) Type 1
2. ) Type 2
How do you calculate the efficiency of a test?
E = ((TP+TN)/(TP+FP+FN+TN))
Predictive values are influenced by ______
The prevalence of the disorder in the population.
Sensitivity and specificity are influenced by _____
The cut off point chosen for continuously distributed test parameters.
NOT influenced by prevalence.
How would you calculate prevalence?
Total with/Sum total of pop
If prevalence < the PPV _____ and the NPV ____.
Increases PPV
Decreases NPV
how would you calculate absolute risk?
(No/(populationxyears)) x 1000 (if per 1000 in pop)
How would you equate attributable risk?
Absolute risk (with drug/etc) - Absolute risk (without drug/etc)
How do you calculate relative risk?
Absolute risk (with drug) / Absolute risk (without drug)
