PH Flashcards
(98 cards)
Causes of associations of outcome in a study
1) Bias
2) Chance
3) Confounding
4) Reverse causality
5) True association
What is a bias
Systematic error resulting in deviation from true effect
Types of bias
Selection bias (non-response by certain groups, loss to follow up etc)
Allocation bias (Groups with differing traits allocated to diff groups)
Information bias
- Measurement bias (diff equipment gives diff reading)
- Observation bias (observers expectations influence
- Recall bias (memory)
- Reporting bias (don’t report the truth)
Publication bias (negative results less likely to get published
Bradford Hill criteria (9 things to prove a relationship is causal)
Strength of association (high relative risk)
Consistently shown across studies
Temporality
Dose response
Reversibility
Biological plausibility
etc
Randomised control trial
2 Pros and 2 Cons
Low risk of bias and confounding + can show causality
Time consuming
Expensive
Case control:
What is it?
2 Pros + 2 Cons
Observational study comparing those with disease (case) to those without (control). Looks retrospectively at exposures.
Pro: Quick, good for rare diseases
Cons: difficult finding appropriate controls, Selection and information bias prone
Cross sectional study:
Basics
Pros + Cons
Collects data from population at a point in time (snapshot)
Pro: large sample size, Provides prevalence data
Cons: Risk of reverse causality (which came first)
What is reverse causality?
Outcome mau have been caused by exposure
E.g. in depressed people who are obese which caused which
Cohort study
Basic
Pros + Cons
Longitudinal study of similar groups getting different Tx/RFs
Follows them over time
Can follow up rare exposures
Allow Rfs to be identified
Takes a long time
High drop out rate
50 cases in 1000 people over ten years. What is the incidence per year?
10yrs
= 0.5% per year
How to calculate the relative risk (ratio)
(% with disease in exposed) / (% with disease in control)
How to calculate attributable risk of smoking
(% with disease in exposed) - (% with disease in control)
How to measure number needed to treat (the number which would save 1)
1/Attributable risk
%diseaseexposed - %diseaseunexposed
Wilson + Jungner screening criteria
INASEP
Important disease
Natural Hx under
Acceptable intervention
Simple + Safe
Effective Tx with early detection
Policy of who should get tx
Disadvantage of screening
Overdetection of subclinical disease
False +ve: worry and exposure to harmful further testing
False -ve: more dangerous as gives false sense of health
Positive predictive value
% of positives who are positive
Negative predictive value
% of negative who are negative
Sensitivity
% of those with the disease who are detected
Specificity
% of those without the disease who are negative
Lead time Vs Length time bias
Lead time = false sense of increased survival time due to early detection
Length time = a disease with a slower progression/low aggression more likely to be picked up by screening giving false idea screening is reason for good prognosis
PROGRESS mnemonic for health inequality
Place of residence
Race/Ethnicity
Occupation
Gender
Religion
Education
Socioeconomic status
Social capital resources
Definition of health:
A state of complete physical, mental and social wellbeing. No merely the absence of disease or infirmity
Causes of Errors
System Error
- staffing
- Equiptment unavailability
Human Error
- memory
- skill
- timing
Types of errors & model
Latent (system), Active (human)
Swiss cheese model