Measures of Disease Frequency and Association/Evidence in Therapeutics

Question 1

How is prevalence calculated?

Answer 1

Range: 0-1 (or 0-100%)
Prevalence of disease among N subjects=(#With disease)/N

Refers to the amount of disease that exists within a population at a point in time

Question 2

How are odds calculated?

Answer 2

prevalence/(1-prevalence)

Measure of likelihood of a particular outcome; how likely that condition does versus does not exist

DOES NOT TELL US HOW LIKELY A CONDITION EXISTS IN A POPULATION!! THIS IS PREVALENCE

Question 3

What are the types of incidence measures?

Answer 3

1. Incidence rate: instantaneous risk of developing the disease at a point in time
2. Incidence proportion: proportion of population that develops the disease during a fixed follow-up period, does NOT account for timing of disease occurence

Question 4

How is incidence rate calculated?

Answer 4

(#new cases of disease observed)/(#at risk for total amount of time, in 100 person-yr)

Eg. In Drug A, 6 people experienced the disease. 2 people experienced it in 1 year, 4 people experienced it in 2 years. What is the incidence rate?

(2people x 1yr) +(4people x 2yr)=2+8=10 person-yr
6 incidents/10 person-yrs = 0.6 person-yrs x 100 –> 60 per 100 person-yrs

Question 5

How is incidence proportion calculated?

Answer 5

(#new cases of disease during a follow-up period)/(#at risk during the period)
assumes no loss to follow-up or competing risks

Question 6

How is risk ratio (RR) calculated?

Answer 6

Incidence proportion (drug)/incidence proportion (placebo)
RR>1 = drug is harmful
RR<1 = drug is effective

Question 7

How is risk difference (RD) calculated?

Answer 7

Incidence proportion (drug) - incidence proportion (placebo)
RD = (+), drug is harmful
RD = (-), drug is beneficial

Question 8

How is disease odds ratio (OR) calculated?

Answer 8

[Incidence proportion drug/(1-incidence proportion drug)]/[incidence proportion placebo/(1-incidence proportion placebo)]
OR<1, drug is effective
OR>1, drug is harmful

Question 9

How is the Number-Needed-to-Treat (NNT) calculated?

Answer 9

patients who need to be treated in order to prevent one event

NNT=1/|RD|
(RD=risk difference)

Question 10

Fixed range (posterior distribution) with a 95% chance of finding an unobserved random parameter, based on Bayesian analysis.
-Fixed range, estimated parameter=random variable

Answer 10

95% Credible interval

Question 11

Range where 95% of measurements in the interval will capture the fixed but unknown true value of the parameter
-Range=random variables, fixed parameter

Answer 11

95% Confidence Interval

Question 12

Comparison of how well two hypotheses (effect sizes) represent the data
Eg. How likely outcome A is expected compared to outcome B

Answer 12

Likelihood Ratio

Question 13

The probability of obtaining a result equal to or more extreme than what was observed, assuming the null hypothesis is true

Answer 13

P-value
-usually written as inequalities (eg. P<0.05)
-small p-value = repeat the experiment