Research final Flashcards
experimental research
based on logical structure or design
inferences about cause and effect
designed to control for confounding variables
extraneous variables
an factor not related to purpose of the study which may affect dependent variables
when uncontrolled, become confounding variables
true experimental designs (essential 3)
independent variable manipulated by experimenter
include control or comparison group
random assignment
random assignment
each participant has equal chance of being assigned to any group
helps control for extraneous variables and prognostic indictors
simple random assignment
equal chance of either group
may have unequal groups
block random assignment
divided equally into blocks
stratified random assignment
when certain attributes may be confounding
cluster random assignment
each cluster randomly assigned a treatment
all members of a cluster get same treatment
concealed allocation
when a participant is determined to be eligible, researchers do not know group assignment
statistical conclusion validity
is there a relationship between IV and DV?
internal validity (3 components)
is evidence of a casual relationship between IV and DP
temporal precedence
covariation of cause and effect
no plausible alternative explanation
construct validity
to what constructs can results be generalized
external validity
can results be generalized to other persons, settings, or times
order of experimental design validity
statistical conclusion validity
internal validity
construct validity
external validity
internal threats
history
maturation
attrition
testing
instrumentation
regression to mean
selection
social interaction
internal validity: social threats
diffusion/imitation - imitate experimental group
compensatory equalization - treat control differently
compensatory rivalry - control works extra hard
demoralization - control gives up
ruling out threats to internal validity
random assignment
blinding
threats to construct validity
operational definitions
comprehensive measurements
time frame
multiple treatment interactions
threats to external validity
influence of selection
influence of settings
influence of history
noncompliance
refuse assigned treatment after allocation
cross over to another group
withdraw from study
missing data
need to account
concerning if >20%
especially if related to treatment
per protocol analysis
analyze only who complete
intention to treat analysis
analyze with the group they were assigned to
quasi-experimental
may lack randomization
may lack comparison group
may lack both
between subjects design
assigned to independent groups
within subjects design
participants act as their own controls
factorial
how we describe designs that have 2+ IV
independent groups
different people in each level of the IV
between subjects
one way designs
only one IV
single factor
pretest-posttest control group design
both groups measured before and after treatment
cause and effect
strong internal validity
posttest only
no pretest
when pretest is impractical, contraindicated or reactive
strong internal validity
assume groups are randomly assigned
two way factorial design
two or more independent variables
randomized block design
1 attribute IV is not randomized
same number of subjects in each level
repeated measures
same people in each level of IV
within subject
one way repeated measures design
subjects exposed to all levels
no control group
effects of repeated measures
practice effects - get better
carryover effects - still present
order effects - latin square
crossover designs
randomized to treatment square
control for order effects
only when condition is stable
considerations for wash out period
mixed design
one IV is RM
one IV is IG
PEDro
11 questions, 10 points
question 1 is not scored
Quasi-experimental designs
lack random assignment, comparison group, or both
time series designs
single group - time is IV
one group pretest-posttest design
all receive same treatment
no comparison group - limits validity
IV is time - two levels - pre and posttest
when are one group pretest-posttest designs defendable?
behavior of control group over time has been documented
ethical implications of withholding treatment
time interval from pretest to posttest is very short
- limits confounding and other threats to internal validity
repeated measures
NOT true experiment, but can be viewed in similar light
no comparison group, limits in and ex validity
interrupted time series design
multiple measures of DV
interrupted by treatment
no comparison, limits validity
nonequivalent group designs
not formed by randomization - existing or self-selected groups
nonequivalent pre/posttest control group design
no RA
intact groups
control over threats to internal validity
historical controls
receive a different treatment during an earlier time period
nonequivalent posttest only control group design
no RA
no pretest
exploratory
downsides of group experimentation
require control groups and large numbers
time intensive
too few measurements
treatments are standardized
usually not feasible for clinicians
averaging results and losing individual
single subject design studies
within subject
allows for C&E inferences due to control phase and planning
consistent with EBP on individual patients
characteristics of SSDS
not a case report/study
controlled manipulation of an IV
extended baseline phase before intervention
intervention introduced when condition is stable
continuous assessment
baseline phase
evaluating for stability and trend
length of phases
extend until stability id reached
minimum of 3-4 data point for each phase
target behavior
patient focused
needs to be quantitative
Single subject AB design
baseline followed by treatment
can observe changes
unsure if confounding, maturation, or learned effect
limitations of AB
no control
cannot conclude causality
SS BC design
control intervention (B) followed by experimental intervention (C)
ABA design
can determine if intervention truly caused change
behavior must be reversible
multiple baseline across subjects
most common
same intervention, varying baseline phases
multiple baseline across settings
one individual monitored in multiple settings with same intervention
multiple baseline across behviors
one treatment, multiple clinical behaviors
limitations of SS designs
ethical issues with withhold treatment in baseline
little control over threats to internal validity
limited generalizability
statistic are in infancy
exploratory research
relationship among 2 variables
observational
are exploratory
data collected as naturally exist
no manipulation of variables
longitudinal studies
prospective - exposed/unexposed beforehand
retrospective - outcome is known, look in records
cross-sectional studies
snapshot
less time and money
epidemiology: measuring risk
studying determinants of disease, injury, dysfunction
dichotomous variables
case control and cohort studies
study risk factors
association between disease and exposure
cohort studies
relative risk
grouped by exposure
case-control studies
odds risk
grouped by outcome
select all from same population
interpretation of RR and OR
= 1 null
> 1 postive association, harmful
< 1 negative association, protective
factors to consider for causality (5)
time sequence
strength of association
biological credibility
consistency
dose-response relationship
challenges for case control studies
no randomization
collecting from med records
observation bias
recall bias
challenges for cohort studies
time
bias and attrition
misclassification of exposure
outcome may not occur in sufficient numbers
cross sectional studies
snapshot in time
all variables measured at same time
used in diagnostic
used for efficiency
do not know time sequence
developmental research
natural history of phenomenon
normative research
describe typical or standard values
descriptive surveys
questionnaires or responses to interview
overall picture of characteristics, attitudes, behaviors characteristics or risk factors for disease/dysfunction
case report/case series
detailed description of condition or treatment response
purpose
- understanding unusual pt conditions
- innovative therapies
- future research directives
retrospective
does not meet IRB definition of research
diagnostic tests - 3 purposes
focus examination
identify problems that require physician referral
assist in classification
studies for diagnostic accuracy
index test - test being studied
gold standard - accurate indication of true status
diagnosis - EBP perspective
probabilities and limiting uncertainty
pretest probability
baseline probability of certain condition before any testing
posttest probability
revised likelihood of the diagnosis on outcome of test
test threshold
probability below
test will not be ordered
possibility of diagnosis is so remote
treatment threshold
probability above
test will not be ordered
possibility of diagnosis so great
basic structure of diagnostic study
series of pts
index test
gold standard
compare results of IT and GS
key methodological aspects of diagnostic studies
was there a comparison to gold standard?
were the individuals interpreting each test’s results unaware of the other test’s results?
were subjects with all levels of disorder included?
did all subjects undergo both tests?
100% sensitivity
all true positives
rule out when test negative
good screening test are highly sensitive
100% specificity
all true negatives
rule in when test positive
more important for diagnostic special tests
positive predictive values
identify pts with the disease from all positive test results
negative
identify pts without the disease from all negative test results
likelihood ratios
sensitivity information combined with specificity information
if diagnostic test positive use +LR
if diagnostic test negative use -LR
clinical prediction rules
diagnosis - rule in
screening - rule out
factors that predict response to treatment
CPR for diagnosis
improve accuracy of diagnostic assessments
CPR for prognosis
can we expect certain outcomes based on a cluster of findings
validating CPRs
deriving the model
validation of the rule
impact analysis
issues with CPRs
most dont move past derivation
qualitative research (%)
4% of what was published in PTJ in 2018
qualitative
understanding
discovering frameworks
interview/observation
texual (words)
theory generating
quality of informant > sample size
subjective
model of analysis: fidelity to text or words of interviewees
qualitative research
all interactions are inherently social phenomena
inductive process - narrow to broad
occurs in natural setting
purposes of qualitative
generating theories
developing theories to explain observed phenomena
investigating complex phenomena
- sociocultural influences
- organizational processes
- exploring special populations
why doesnt the research question apply to qualitative research?
PICO do not apply
no specific hypothesis based on question
ethnography
describes cultural characteristics and behaviors in specific groups
- investigators immerse themselves
- participants called inormants
ground theory
individual responses contribute to understanding theoretical relationships that can explain behavior
- constant comparison
phenomenology
seeks to explain how events and circumstances influence perspectives and behaviors
qualitative only methods
observation
- field observation
- participant observation
interviews
- unstructured
- structured
- semi
- focus groups
written documents
mixed methods
convergent - simultaneous, combine results
sequential - one before the other
embedded - simultaneous, do not combine results
multiphase - over time
sampling in qualitative
non-probability - convenience, purposive, snowball
sample size may be large or small
credibility
are the results believable
like validity
transferability
can results be applied to people in similar circumstances
like generalizability
dependability
how stable are data over time
like reliability
confirmability
are finding due to beliefs and experiences or bias
triangulation
more than one source
credibility, dependability and confirmability
member checking
confirm interpretation with others
credibility
negative case analysis
explain conflicts that emerge from preliminary data
credibility
thick description
improves ability to make comparisons
transferability
purposive sampling
choosing participants that will make good informants
transferability
audit trail
documenting decisions so another researcher can confirm
dependability and confirmability
reflexivity
examining how researcher’s beliefs may influence interpretation of data
confirmability
scoping review
selective review of literature that is less systematic
exploratory assessments of available literature on broad topic
answer background questions
summarize in general
do not critically appraise
scoping review methods
PICO too specific
methods
- systematic search
- more expansive selection criteria for designs
- no critical appraisal
results
- report search and number of articles
- describe narratively
discussion
- review results and limitations
- may discuss implications for clinical practice of relevant to objective of review
systematic review
utilizes exacting search strategies to make certain that maximum extent of relevant research has been considered
original articles are methodologically appraised and synthesized
what is a systematic review
searching
appraising
summarizing
objective and transparent process
process of systematic reveiw
search and selection for articles well defined
results of included studies are qualitatively
bias in SR
publication bias
access clinical trials registries
access the grey literature
- conferences, abstracts, websites, dissertations
cochran risk of bias
selection bias - random assignment, allocation concealment
performance bias - blinding of pts and investigators
detection bias - blinding of assessors
attrition bias- incomplete outcome data
reporting bias- selective reporting
other bias
data synthesis
tables
- study design
- participants
- interventions
- outcomes
- study quality
address clinical homo and heterogeneity
reporting on SR
follow typical article format
provide enough detail to reproduce
PRISMA flowchart
include articles in table or appendix
appraisal of SR
AMSTAR and AMSTAR-2
meta-analysis
quantitatively combines results of studies that are not result of systematic literature review
capable of performing a statistical analysis of the pooled results of relevant studies
when can/should you do a meta-analysis?
when more than one study has estimated an effect
when there are no differences in the study characteristics
when the outcome has been measured in similar ways
meta analysis characteristics
effect size from individual studies
pooled effect size
displayed as forest plot
each square is treatment effect for that study
diamond is pooled effect across all studies
point estimate
single value that represents the best estimate of population value
confidence interval
a range of values that we are confident contains the population value
- width concerns precision of the estimate
if heterogeneous
analyze separately