NNT (number needed to treat)

The number of people you would have to treat to prevent one adverse outcome.

NNT=1/ARR

Relative Risk (RR)

How many times more likely one group is to have an outcome compared to another group.

RR=EER/CER

Incidence

Incidence: the ** rate** of new cases of a disease

Prevalence

The **proportion** of the population that could be affected by a disease that is affected by a disease.

Sensitivity

How good a test is at calling true positives positive (the true positive rate)

Sn= true positives/all those who have the disease

Specificity

How good a test is at calling negative those who are actually negative (**the true negative rate)**

Sp= true negatives/ all those who don't have the disease

Positive predictive value (PPV)

If someone has a positive test result, the percent chance that they are actually positive

PPV= true positives/all positive results

***N.B: depends on prevelance. As the disease prevelance goes down, the PPV goes down***

Negative predictive value (NPV)

If someone has a negative test result, the percent chance they are truly negative

NPV= true negatives/all negative test results

****NB depends on prevalence. As prevalence of a disease goes down, the NPV goes up***

Absolute risk reduction (ARR)

EER= exposed event rate= # events/# ppl in exposure group

CER= control event rate= #events/# ppl in control group

What do confidence intervals tell us?

Tell us if something is statistically significant (the confidence interval does not include zero if you are looking at a difference, and does not include 1 if you are looking at an OR or RR) and the precision of the study (good precision=narrow interval)

Accuracy

The number of people correctly classified using a test (true positives + true negatives/all test results)

What does a P-value tell you?

It is the probability that the event occured by chance.

Likelihood ratio for a positive test

How much more likely a positive test result is in someone with as opposed to without the condition

LR+= sens/(1-spec)

(probability of an individual with the disease testing positive)/(probability of an individual without the disease testing positive)

Likelyhood ratio for a negative test

How much more likely a negative test result is to be found in a person with vs. without the disease

LR-= (1-sens)/spec

(probability of an individual with the disease testing negative)/(probability of an individual without the disease testing negative)

Relative risk reduction

EER= exposed event rate= #events/#ppl in exposed group

CER= control event rate= # events/#ppl in control group

RRR: (CER-EER)/CER

A study where a variable is manipulated is called _____ and a study where variables are not manipulated are called_____

Experimental (e.g. RCTs)

Obersvational (e.g. case reports, case-control, cross-sectional, cohort)

What are a few defining differences between quantitative research and qualitative research?

Qualitative:

- generates hypotheses (inductive)
- asks open-ended questions
- data=text, images...
- answers "why?"

Quantitiative

- answers hypotheses (dedcutive)
- asks closed ended questions
- data= numbers etc..
- answers "why?, how much?"

How do you use likelihood ratios to predict whether or not a person has a disease?

You need to use a nomogram. Pre-test probability is estimated or you can use the prevalence and combined with the LR+ or LR- (depending on whether they have a positive or a negative test result) to give a post-test probability.

Odds Ratio

Used in case-control studies.

OR= [(ill+exposed)/(ill+unexposed)]/[(not ill+exposed)/(not ill +not exposed)]

OR=(a/c)/(b/d)

What is type 1 error?

Rejecting the null hypothesis when the null hypothesis is true (alpha)

What is type 2 error?

Failing to reject the null hypothesis when the null hypothesis is false.

Draw the hypothesis testing table