Lecture 1 Flashcards
The process of evidence-based practice?
1) Ask an answerable question
2) Access the evidence
3) Appraise the evidence
4) Apply the appropriate evidence
5) Assess the process
How is a bias caused?
- extraneous variables may effect study variables
- may MASK or INCREASE effect of independent variable
- bias matters as we may not get the truth.
What does bias threaten?
Validity. Also affects confidence in study findings.
What are types of bias in quantitative research studies?
- systematic error (something that happens b/c of some things that are done systematically)
- can occur from: choice of study design, methods, analysis, reporting, interpretation
- conscious or subconscious
What is the goal in quantitative research?
- minimize extent of bias that may affect the outcome – what criteria of merit does this address?
- look at how different designs minimize bias
8 threats to internal validity (Campbell and Stanley)
*review threats to external validity
- history
- maturation
- testing
- instrumentation
- statistical regression
- selection biases
- experimental mortality
- selection - maturation interaction
True Experimental Designs
Randomized Control Trial
- elements: randomization, control group, manipulation (when the investigator puts the intervention in place) *** quiz
Quasi Experimental Design
Variations on RCT, but w/o randomization
- no random assignment of participants to groups –> participants from ‘intact’ groups available to the researcher (ex: exercise class)
or ‘matched’ groups (age, gender, height, weight)
Pre-experimental Designs
Non-randomized, used to collect preliminary information.
- characteristics: no random assignment, may not have control
- elements: intervention and outcome
- can’t say intervention caused the outcome
Notation symbols
R
R = random sample selection
*any r is random allocation
Notation symbols
r
r = random assignment
*any r is random allocation
Notation symbols
O
O = outcome - observation or measurement
Notation symbols
X
X = intervention or exposure to an experimental variable for which the effects are to be measured (*manipulation)
When are quasi-experimental designs most valuable?
When you can’t randomly allocate, better to do a group all together, ethics issues
When are pre-experimental designs most valuable?
- answering a level 1 question - can describe intervention and outcome, can’t say intervention caused O
- pilot testing an intervention protocol or measurement approach (exploratory)
Elements of Observational Studies (epidemological studies)
- researcher does not provide the intervention (no manipulation)
- comparison group
- naturally - occurring (not randomized groups)
What level question do you answer with descriptive studies?
Level 1
What level question do you answer with analytic studies?
Level 3 - cause and effect
What are observational studies called?
- cohort study - fwds in time
- case - control study - backwards in time
- cross- sectional (analytic) - interaction
Why are they different?
the direction of the outcome they provide
What is a cohort study?
- prospective - measure fwd in time - outcome measured after intervention
What is a case-control study?
- retrospective - start with outcome, then look back in time if they were exposed to the treatment
What is a case study or case series?
prospective
- target population - treatment, treatment effect
- measure these two things w/ no control
What is the world view for quantitative research?
- only one reality - looking for the truth
What are the 4 criteria of merit in quantitative research?
- internal validity - within the study, have you reached a valid answer. Controlled all you can control.
- external validity - can you transfer the findings ?
- reliability - research can be repeated, get the same outcome
- objectivity - minimizing bias, distance between researcher and population
Hierarchy of evidence for Therapy
- background info
- unfiltered information: case controlled studies, cohort studies, RCT
- filtered information: critically appraised individual articles, critically-appraised topics (evidence syntheses), systematic reviews
- ebm pyramid
Question on the quiz - bias decreases as you go up the pyramid - quality of evidence increases
What is a problem with counterbalance designs?
- adherence; acute conditions can change w/ different interventions
- need people w/ a chronic condition that doesn’t change over time so they go back to baseline
Interrupted Time series
- consistent patterns before and after interventions show no change - intervention not successful
Factorial designs
- randomized or observational
- previous designs involved testing one independent variable
- evaluate effect of two or more independent variables and the effect of the interventions on different factors/levels – allocation to groups may be randomized or may be occurring naturally
Non - experimental designs
- survey research
- correlation research
- secondary analysis of data
Survey research
provides quantitative (numeric) description of trends, attitudes, or opinions
- mailed, emailed, phone
- level __
Correlational research (passive observation designs)
- examines relationships among variables
- regression studies - explain or predict changes in dependent variable on baiss of changes on other factors
- level 2 questions
Secondary analysis of data
- see ex post facto designs
- level __
Sources of bias in quantitative studies about ‘therapy’
- allocation to groups
- sample selection for non-randomized group
- methods
- analyses and interpretation
reporting
What is random allocation?
- determined by chance
What should be indicated in the article to ensure that appropriate randomization has occurred?
- method of allocating interventions to participants - ie, sequence generation, flip coin
- concealment of allocation sequence to ensure adherence
- evidence that randomization worked
Why is random allocation done?
To avoid systematic differences between groups.
- both groups should have similarities in characteristics
Double blind study
- researcher and subject don’t know what group they’re in
single blind study
- subject doesn’t know what group they’re in - researcher knows
blind assessment
- person doing the assessment doesn’t know what group the participants are in
- why?
How could bias caused by lack of blinding affect outcomes?
- increase or decrease effect you might see
How can blinding reduce biases?
- may help equalize attention/group
- may help avoid differential attrition (control group may not be motivated to stay in the study)
- may help avoid expectation of better/worse outcomes
What is the magnitude and direction of the bias, ie how could it affect the results?
Think about this…
Which criteria of merit will these affect?
Think about this…
What is reported that indicates efforts to minimize bias?
Think about this…
Elements of quantitative studies
- study design - determined by question and chosen to minimize bias - analytic, descriptive - best design to reduce bias
- methods - care taken to minimize bias - allocation to groups or sample selection, quantity of intervention/group, outcome assessment - outcome measures, testing, equipment, interviewers, assessors
- adherence to the protocol - minimize bias
