Flashcards in S1 Hadpop Deck (35):

1

##
When looking at relationships between two variables, what do you need to consider?

### Chance, bias, confounding factors

2

## Define bias

### Preference to a particular perspective

3

## Define confounding factor

### Factor that affects both exposure and outcome but is not in the causal pathway

4

## What is deterministic causality

### Validation of a hypothesis by systematic observations to predict with certainty future events

5

## Define stochastic causality

### Assessment of a hypothesis via systematic observations to give the likelihood of future events

6

## What is the census?

### Simultaneous recording of demographic data by the government at a particular time to all people who live in a particular territory

7

## Features of Census

###
Run by government

Covers defined area

Personal enumeration

Simultaneous throughout defined area

Universal coverage

Regular intervals

8

## Three population characteristics the census looks at

###
Population size

Population characteristics

Population structure

SCS

9

## Crude birth rate

### Number of live births per 1,000 population

10

## General fertility rate

### Number of live births per 1,000 females aged 15-44

11

## Total period fertility rate

### Average number of children that would be born to a hypothetical woman in her life

12

## Fecundity

###
Physical ability to reproduce

E.g. More sterilisation and hysterectomies reduces fecundity

13

## Fertility

### Realisation of the potential of births

14

## Crude death rate

### Number of deaths per 1,000 of the population

15

## Age specific death rate

### Number of deaths per 1,000 in an age group

16

## Standardised mortality rate (SMR)

### Compares observed number of deaths with number of expected if the age sex distributions of the populations were identical so it adjusts for age sex confounding

17

## Define incidence

###
Number of cases that have occurred over a period of time

New events/ person x time (years)

18

## Define prevelance

###
Number of people affected by the disease

Number of cases/ number of people

P = I x L

19

## P<=0.05 means results are

###
Statistically significant

Enough evidence to reject the hypothesis

This is when the null hypothesis value is outside the 95% confidence interval

20

## Cohort study features

###
Recruit outcome free individuals

Exposed and unexposed groups

Looks forward in time

Calculate IRR (incidence rate ratios) to give relative risk (exposed/unexposed)

21

## Cohort study internal v external comparison

###
Internal comparison

- divide cohort into degrees of exposure

- calculate IRR and e.f.

- more random variation as it is smaller group

External comparison

- compare to external reference pop

- calculate SMR as often no incidence data

22

## Healthy worker effect

### If people are in work they are more likely to be healthy compared to the rest of the population so likely to get SMR under 100

23

## Case control study

###
- Get a group of cases

- Classify into whether they are have exposed or not

- Work out odds ratio (AD/BC) - precision affected by number of controls

24

## Types of bias can include

###
Selection

Information

Recall

25

## Cohort study v Case control study

###
Cohort more time consuming and expensive

Cohort better for rare exposures

Case control better for rare outcomes

Case control prone to selection and information bias (population of sampling not defined so cannot work out IR etc)

26

## Cause and effect relationship - Henle-Koch's postulates

###
- Agent must be present in every case of the disease

- Agent must not be found in cases of any other disease

- Agent must be capable of reproducing disease in experimental animals and must be recovered from experimental disease produced

27

## Bradford Hill Criteria for Causality

###
SCS DRT CBA

Strength of association

Consistency of association

Specificity of association

Dose response

Reversibility

Temporal sequence

Coherence of theory

Biological plausibility

Analogy

28

## Steps involved in randomised controlled trials (RCTs)

###
- Define disease, treatments, possible bias and confounders etc

- Identify source of eligible patients, recruit and consent them

- Allocate participants to treatments fairly, random allocation minimises allocation bias and confounding

- Follow up patients in identical ways

- Compare outcomes fairly

- Minimise losses to follow up and non compliance with treatment

29

## What is the placebo effect?

###
Placebo is inert substance that appears identical to the active formulation. Placebo effect is the psychological benefit derived from being on the drug or the intervention.

Placebo should only be used if no standard treatment is available

30

## Explanatory (as treated) v Pragmatic (intention to treat) analysis

###
Explanatory (as treated)

- those who completed follow up and complied with treatments analysed

- loses randomisation as non compilers systematically different to compliers

Pragmatic trial (intention to treat)

- Analyses according to original allocation to treatment groups

- Looks at likely effects of using treatments in routine clinical practice and preserves effects of randomisation (minimises selection bias and confounding)

31

## Issues to consider in a clinical trial (RCT):

###
SCEVV

Scientifically robust

Clinical equipoise

Ethical recruitment

Valid consent

Voluntariness

32

## What is a systematic review?

###
Overview of primary studies that use explicit and reproducible methods

May do meta analysis (quantitative synthesis of results of two or more primary studies that addressed hypothesis in same way)

33

## What is a meta analaysis

###
Quantitative synthesis of the results of two or more primary studies that addressed same hypothesis in the same way.

In meta analysis the odds ratio and 95% CIs are calculated for all the studies and then combined to give pooled estimate odds ratio. Can create forest plot.

34

## Fixed effect model v random effects model for pooled estimate odds ratio in forest plot

###
Fixed effect model assumes studies estimating same effect size

Random effects model assumes studies are estimating similar effect size, not the same. Wider 95% confidence interval, more equal weighting of studies

35