Study Design in Epidemiology

Loading...

Study design in Epidemiology Kristen Reyher Centre for Veterinary Epidemiological Research Atlantic Veterinary College University of Prince Edward Island

The theory behind study design • We want to design studies to – Identify causal factors for disease, so that we can • Focus on target points to work towards prevention • Minimise harmful effects of treatments or management changes

1

Study types Explanatory (Analytical)

Descriptive

Case Report

Case Series

Survey

Crosssectional

Cohort

Casecontrol

Observational

Laboratory

Experimental

Controlled Trials

2

Descriptive vs. Analytical • Descriptive – Describe characteristics – Do not make comparisons • case report • case series • survey

• Analytical (explanatory) – Seek to make comparisons • inference about exposures (risk factors, treatments) and outcomes (disease, death, production) – Experimental vs. observational

3

Descriptive studies 1 • Case study – Report on one or a few cases – Usually a rare condition – Limited to ‘real world’ conditions? – Any conclusions about cause or outcome are author’s conjecture

4

Descriptive studies 2 • Case series – Describe (often unusual) clinical course of condition of interest – Might provide information about prognosis if cases are representative of all cases – Again, no direct data but features might help build hypotheses

5

Descriptive study 3 • Survey – Estimate the frequency and distribution of outcomes – Provides some data (say about disease in a population) – Need to take care re: sampling (Signe) and design of questions – Surveys including exposures and outcomes = cross-sectional analytic studies!

6

Analytical studies • Experimental – Investigator can allocate study subjects – Advantages • stronger evidence of causation • control of confounders through randomisation

– Disadvantages • limited range of hypotheses • may not be “do-able”

7

Analytical studies • Observational – No allocation of study subjects • Do not confuse random sampling with random allocation! • Observation in a real-world setting

– Advantage • Complex web of causation might not be otherwise reproducible – practically – ethically – economically

8

Observational studies • Prospective vs. Retrospective – Has outcome occurred before study starts? • yes = retrospective • no = prospective

– Advantage of prospective • data quality • better able to study incidence

9

Observational studies • Classified by subject selection – Cross-sectional studies – Cohort studies – Case-control studies

10

Observational study 1 • Cross-sectional – Most frequent study design in vet epi = straightforward – Random sample of subjects from a population • Try to represent population in sample

– Non-directional in time = ‘snapshot’ – Simultaneously classify according to • Disease status (or outcome) • Study factor or risk factor – determinant – exposure 11

Cross-sectional

F-/D+ F+/D+ Study Population F+/DF-/D-

Time

Past

Present

Future

12

Cross-sectional studies • Limitations – Only suitable for chronic conditions occurring at a moderate level in the population – Only quantifies prevalence of exposure and outcome • May over-represent factors affecting incidence and duration • Can confuse protective risk factors

– Reverse-causation • Best for time-invariant exposures (sex, breed, housing) • Can confuse procedures implemented in response to disease

13

Cross-sectional studies • Example: – 100 dairy herds selected randomly from the Dairy Herd Improvement (DHI) register to answer a questionnaire assessing treatment of dry cows – Each farm is classified according to exposure (which dry cow therapy used) and according to outcome (>30% fresh cow mastitis or <30% fresh cow mastitis)

14

Cross-sectional Used CefaDri / >30% mastitis

Used DryClox / >30% mastitis

F+/D+ 100

F-/D+

farms

F+/DUsed DryClox / <30% mastitis

F-/D-

Used DryClox / <30% mastitis

Time

Past

Present

Future

15

Cross-sectional studies • Pros – Representative of population – Potentially efficient – Low cost – Rapid

• Cons – Must verify that risk factor came before the disease

16

Observational study 2 • Cohort – Identify subjects • with exposure • without exposure

– Follow the groups through time to determine if disease develops • usually prospective

17

Cohort D+ exposed

D-

D+ nonexposed

DTime

Past

Present

Future

18

Cohort studies – special case • Single cohort = longitudinal study – Follows an entire population through time – Record all exposures of interest • Investigate multiple exposures at once

– Record all outcomes of interest • Outcomes must follow exposure!

– Useful in measuring incidence of disease

19

Cohort Studies • Example: – 500 healthy cows (D-) from farms using dry cow therapy (E+) were randomly selected from a list of cows in DHI databases across Canada – 500 healthy cows (D-) from farms not using dry cow therapy (E-) were also randomly selected from the same list of cows – Followed for two years – Assessed for clinical mastitis during that time

20

Cohort Farms using dry cow therapy

Mastitis

No mastitis

Mastitis Farms not using dry cow therapy

No mastitis

Time

Past

Present

Future

21

Cohort studies • Pros: – Less susceptible to bias compared to case-control – More control over quality of data – No confusion on time order of exposure and disease

• Cons: – – – –

Expensive Time-consuming Potential losses to follow-up Only works for diseases common in a population

22

Observational study 3 • Case-Control – Identify subjects • with disease • without disease

– Compare histories of risk factor (exposure) • Usually retrospective

23

Case-control exposed

D+ nonexposed

exposed

Dnonexposed

Time

Past

Present

Future

24

Case-control studies • Used for rare diseases • Relatively quick and inexpensive (if quality data is accessible)

25

Case-control studies • Limitations – Finding source of cases – Defining a case – Appropriate controls are often difficult to identify. These should be animals that would have been cases if they got the disease (but not always as straightforward as that sounds!)

26

Case-control • Example: – A rare mastitis is being studied. – 50 farms in Quebec has confirmed cases of this type of mastitis (D+) – 50 comparable farms in Quebec with no confirmed cases (D-) are also identified for the study – All 100 farmers are asked about management practices (type of dry cow therapies) used on their farms (exposure) 27

Case-control Used CefaDri Used DryClox

50 farms with mastitis

Used CefaDri Used DryClox

50 without mastitis

Time

Past

Present

Future

28

Case-control studies • Pros: – Rare diseases – Potentially efficient – Low cost – Potential for rapid completion

• Cons: – Highly susceptible to bias related to selection of controls

29

Summary of observational studies Study type

Cross-sectional

Cohort

Minimal cost





Short time (little to no follow-up)



 

Control selection difficult Representative of population

Case-control

 

Good for rare disease Good for rare exposure



Time sequence known



30

Experimental studies • Laboratory-based • Randomised controlled trials

31

Experimental study 1 • Laboratory-based – Carried out under strictly controlled conditions – Investigator has almost complete control over experimental conditions – Evidence of association of exposure and outcome is the best evidence of causation – Relavance to ‘real-world’ conditions often doubtful

32

Experimental study 2 • Randomised controlled trials – Covered by Signe next!

33

Characteristics of various study types

Type of study

Level of difficulty

Level of investigator control

Strength of “proof” of causal association

Relevance to “real-world” situations

Descriptive Case report

very easy

very low

n/a

low to high

Case series

easy

very low

n/a

low to high

Survey

moderate

moderate

n/a

high

Explanatory - experimental Laboratory

moderate

very high

very high

low

Controlled trial

moderate

high

very high

high

Explanatory - observational Cross-sectional

moderate

low

low

moderate

Cohort

difficult

high

high

high

Case-control

moderate

moderate

moderate

high

n/a = not applicable

34

Loading...

Study Design in Epidemiology

Study design in Epidemiology Kristen Reyher Centre for Veterinary Epidemiological Research Atlantic Veterinary College University of Prince Edward Isl...

340KB Sizes 14 Downloads 11 Views

Recommend Documents

Study Designs in Epidemiology - uOttawa uOttawa
When it's not acceptable to do an experiment: "How much does not wearing a condom increase the likelihood of HIV infecti

EPIDEMIOLOGY THROUGH CELLULAR AUTOMATA Case of Study
Abstract. This paper performs the utilization of cellular automata computational analysis as the dynamic model of spatia

Epidemiology
Jul 4, 2012 - STATISTICS TEXT BOOKS PDF "Nonparametric Statistics is a short and sweet introduction to the five most fam

epidemiology - UNAM
each quiz you must answer at least 20 of the questions correctly; this indicates that you have a ..... Page 6. Principle

Compartmental Models in Epidemiology - Springer
In epidemiological terms this can be stated as follows: If the av- erage number of secondary infections caused ... are t

Study of Environmental Radioactivity Level and Epidemiology i..|INIS
Original TitleStudi Tingkat Radioaktivitas Lingkungan dan Epidemiologi Lingkungan pada Area Pertambangan Timah Pulau Ban

Basic epidemiology
Confidence intervals. Hypothesis tests, p-values, statistical power. The p-value. Statistical power. Basic methods t-tes

Descriptive epidemiology
epidemiology is divided into :. 1. Descriptive epidemiology .. 2. Analytic epidemiology . ... the typical study design f

Epidemiology - ACIAR
Semua hewan terserang adalah sapi Bali yang berasal dari Propinsi Sulawesi Selatan dan. Nusa Tenggara Timur. Kasus terse

Principles of Epidemiology: Lesson 6 Exercise Answers|Self-Study
May 18, 2012 - Exercise Answers. Exercise 6.1. Nine cases of cancer in a community represents a cluster — a group of c