EAB Flashcards
How do you calculate incidence rate?
total no of new cases in given period/total population at risk
How do you calculate prevalence rate?
all new and exisiting cases/total population
How do you calculate attributable risk?
(incidence exposed - incidence unexposed to risk) / incidence
What are the main concerns of experimental study designs?
unethical and feasability
What are the main concerns of observational study designs?
confounding bias, selection bias and measurement errors
What are examples of selection bias?
self-selection, attrition, non-response
What are examples of information bias?
false positives/negatives,, reporting bias, error or omission of details
What is intention to treat?
in RCTs, always do analysis with patients in their original groups
- even if switched treatments or drop out
What are the features of case control studies?
- observational
- retrospective
What are the pros and cons of case control studies?
Pros: quick, inexpensive
Cons: usually biased or data is incomplete
What are the features of cohort studies?
- observational
- moniters healthy indviduals over time for disease risk/rates
- prospective
What are the pros and cons of cohort studies?
Pros: less bias than case control
Cons: long follow up, need large groups
What are the features of cross-sectional studies?
- observational
- data is collected at 1 point in time
- useful for frequencies and attitudes
What are the 2 types of quantitative data?
continuous or discrete
What is dichotomous or binary data?
categorical data with only 2 categories
e.g. gender
What is categorical data with more than 2 categories?
Ordered/nominal - e.g. stages of cancer
Unordered - e.g. marital status
How is variance determined?
SD^2
How is standard deviation measured?
(value-mean)^2
What type of skew is normally seen in medicine?
positive skew (tail on right hand side)
What does normal distribution mean?
- bell shaped curve
- 95% of data lies within 2SD of mean
- 68% of data lies within 1SD of mean
what does the confidence interval indicate?
range within the true mean is likely to lie
What is sampling error?
different samples will give different estimates of the mean
What does standard error indicate?
the extent of sampling error
How do you calculate sampling error?
SE = SD/√N
What does a 95% confidence interval mean?
that the true mean is expected to like within 1.96 SE of the estimated mean in 95% of calculation
What are the assumptions when calculating confidence intervals?
- normal data
- large population
- randomly chosen samples
- observations are independent of each other
What is the difference between statistical and clinical significance?
Statistical means that P<0.05 - i.e likely to be true difference
Clinically means that the difference is large enough to have a clinical impact
What does a t test assess?
if there is a difference in means of 2 difference samples
What is the difference between a paired and unpaired t-test?
paired = data is matched e.g. BP before and after treatment unpaired = different samples e.g. mean height in boys and girls
How is the t test calculated?
mean difference/SE
What are the assumptions of a t test?
- quantitative data
- normally distributed
- equal variances
- randomly selected samples
- observations are independent
What does the Chi-squared test assess?
association between categorical variables
How is degrees of freedom calculated in chi-square tests?
(no. of rows - 1 ) x (no. of columns - 1)
What does the Chi-squared test calculate?
compares expected frequencies for no association against observed frequencies
- large difference indicates and an association
What are the assumptions of the Chi-squared test?
large sample
use of frequencies (not percentage)
80% of expected frequencies are >5
How do population pyramids differ for high and low income countries?
Low-income have high birth and death rates - lower percentage of older age groups
High income have more rectangular shapes with more elderly
What is demographic transition:
Concept for population growth:
- low income countries have high birth and death rates
- in development, get initial reduction in death rates -> rapid growth
- Birth rate will later decrease, giving stable population with lower birth:death rate equilibrium
How often is the census conducted?
every 10 years
How are areas divided up in census?
enumeration districts
What groups are considered hard to reach?
disabled, elderly, students, migrants/non-english speakers, travellers
What measures estimate populations inbetween censuses?
population estimates annually - cohort component method
population projections every 2 years
What are the indicies of multiple deprivation (IMD)?
- income
- employment
- health and disability
- education, skills and training
- barriers to housing and services
- living environment
- crime
What are super-output areas?
areas divided to calculate the level of deprivation using IMD
What is avoidable mortality?
deaths from causes that are considered avoidable with effective healthcare/public health
What is the bradford-hill criteria?
the minimum conditions a study needs to provide adequate evidence of a causal relationship
What are the components of the bradford-hill criteria?
- strength of association
- consistency
- specificity
- temporality
- dose-response relationship
- biological plausability
- coherence
- experimental evidence
- analogy
What types of assoctions are case control studies good for?
rare outcomes and multiple risk factors
What is non-response bias?
people who are more likely to respond are more likely e.g. to go for a test/check up
What is recall bias?
cases remember exposure differently than controls
what is an example diagnostic bias?
contraceptive pill use makes detection of uterine cancer more likely
What is interviewer bias?
cases and controls are asked about exposure differently
What is confounding?
there is an alternative factor that is contributing to/causing the disease e.g. age
What are the criteria for a confounding factor?
- is associated with the exposure and disease independently
- is a causal risk factor
- must not be an intermediate cause
What is interaction?
when association differs according to the level of a third factor
What are the assumptions of cohort studies?
representative
well defined absence of exposure
outcome comparability
How do you calculate relative risk?
number of new cases/number at risk at start of period
What are the values for relative risk?
1 = no association >1 = positive association (risk factor) <1 = negative association (protective factor)
How do you calculate relative rate?
number of new cases/total person time at risk
What is the rate ratio?
relative rate in expose/relative rate in unexposed
What is equipose?
when there is no existing evidence that the intervention being tested is superior to existing treatment or effective
What is the power of an RCT?
probability of a type II error (false negative)
- usually use power of 80-90%
What is the significance of an RCT?
probability of a type I error (false positive)
- use significance of 5%
What is the criteria for choosing a primary outcome?
- clinically relevant
- not subjective
- easy and accurate to measure
- independent of treatment
How do you calculate absolute risk reduction (ARR)?
risk in control - risk in intervention
How do you calculate relative risk reduction (RRR)?
ARR /risk in control
How do you calculate number needed to treat?
1/ARR
What is NNT (number needed to treat)?
the number of patients that would need to receive treatment to prevent an adverse event in 1 patient
