Flashcards in Use of evidence Deck (33):

1

## Checking association is not due to chance

###
p values, less than 0.05 is statistically significant

95% CI is range true value lies within, 95% sure

2

## Checking association is not due to confounders

### Can use multivariable analysis to adjust for confounding factors present

3

## Types of bias, have to be addressed during study design stage

###
Measurement bias

Observer bias

Loss to follow up

4

## Bradford Hill viewpoints for causation

###
Temporal relationship (rf must precede disease)

Biological plausibility

Consistency

Association strength

Biological gradient (dose-response relationship)

Specificity e.g. rf inc association of one disease, not many

Coherence - association seen across different measures

Experimental - association replicated in lab/trials

Reversibility - removal of rf results in reduced disease risk

5

## Hierarchy of study designs for causality

###
Systematic reviews/meta-analyses

RCT w/ definitive result

RCT w/out definitive result

Cohort studies

Case-control studies

X-sectional studies

Ecological studies

Case reports

6

## Critical appraisal of epidemiological findings

###
Bias (selection, measurement (observer/responder))

Confounding - dealt with by stratification by confounder status (e.g. age) or statistical modelling

Chance - p value and CI

Causality - Bradford Hill criteria

Generalisability - look at reverse causality being possible and whether these results could apply to your patient or more to another group

7

## Types of outcomes of studies

###
Continuous

Categorical

Binary (type of categorical)

Time to event (type of binary)

8

## RCT analysis of treatment effect

###
Risk ratio= risk in intervention/risk in control

Risk reduction= 1- risk ratio

Risk difference (ARR)= risk in control-risk in intervention

NNT= 1/ARR

Odds ratio= Odds in intervention/odds in control

Rate ratio= rate in intervention/rate in control (for time to event results)

Difference in means= mean in intervention-mean in control (for continuous results)

9

## Observational study appraisal

###
Measurement/adjusting for confounders

Is interpretation of result reasonable given confounders

10

## Systematic review process

###
Find relevant evidence in unbiased manner

Appraise each paper for methodological quality

Synthesise results in met-analysis

11

## Cochrane collaboration in EBM

### Cochrane database has register of all published RCTs and publishes SRs

12

## PRISMA

### Preferred reporting items for systematic review and meta-analysis, has checklist to help authors construct SRs

13

## How to gather evidence for EBM

###
Posing a question using PICO (population, intervention, comparator, outcome)

Search for evidence (published and unpublished)

Appraise evidence (Forest plot and pooled estimate with 95% CI)

Applying evidence, see if more RCTs needed

14

## SR limitations

###
Varying methodological quality assessed by risk of bias graph

Heterogeneity test (p>0.05 means no difference) for differences in results

Publication bias - statistically significant more likely to be published than insignificant

15

## Assessing a measuring instrument

###
Validity - check sensitivity and specificity, measures continuum, correlation coefficient between a gold standard test and this

Reliability - Same results on repeated trials by same or different observers, kappa statistic measures how much more agreement than by chance alone

Responsiveness - Shows when true changes have occurred

16

## Measures for malnutrition

###
Stunting - low height for age

Underweight - extremely low weight for age

Wasting - extremely low weight for height

17

## Rapid weight gain during childhood can increase risk of

### CHD/T2DM in later life

18

## Growth stunting at population level indicates

### Long-term nutritional deficiencies, individual stunting may be genetic

19

## In europe growth patterns reflected by

### incidence of wars

20

## Ways to calculate gentiles for growth

###
Rank data- all data in order and no assumptions about distribution

Normal distribution - mean and SD used (mean ±1SD is 16th/84th centile, mean ±1.96SD is 2.5/97.5th centile)

21

## Z score for height/weight

### z=(value-mean)/SD, shows how many SDs away from average of a certain population an individual is

22

## Interpreting z scores for height

###
0 is average

<0 is below average

>0 is above average

Adjusted for parental height

23

## Use of z scores

###
Osteoporosis/penia - used to assess bone mineral density for age/gender/ethnicity

Educational attainment

Paediatric cardiology measurements

24

## Risk definition

###
Proportion of people with specific event in specific time period

New cases/number in study population= risk (cumulative incidence)

25

## Incidence rate definition

###
Number of events per person-time at risk

New cases/person-time at risk= incidence rate

26

## Low prevalence can be due to

###
low incidence

rapid fatal disease progression

rapid cure

27

## High prevalence can be due to

###
high incidence

prolonged survival without cure

28

## point prevalence

### no. of people with condition/no. in study at one point in time

29

## odds

### no.of people with condition/no. without

30

## Standardised mortality ratio

###
observed no. of deaths/expected no. of deaths

Compares death rate in standard population vs pop of interest

31

## X sectional study

###
Defined population surveyed

Can't measure incidence and reverse causality a problem

32

## Case control study

###
Cases and control groups selected and look at previous exposure to rf

Can't assess risk/incidence, selection bias, recall bias but can use odds ratio to estimate risk ratio

33