1
Q
A protocol outline includes what 6 things?
A
RQ, background & significance, study design, study population, measurements and statistics
2
Q
A RQ is the __ of the study that must be __ and meet ___
A
objective of the study; focused; FINER criteria
3
Q
Background & significance section of the protocol should include (6):
A
- rationale
- what is already known
- why is the RQ/study important
- what answers the study will provide
- cites previous research (uncertainties that still remain)
- how this study will address uncertainties from prior research
4
Q
2 types of study design:
A
- observational
2. experimental
5
Q
What are the types of observational studies?
A
cross sectional; case-control; cohort
6
Q
What are the types of experimental studies?
A
randomized controlled trial (RCT)
7
Q
The study population is defined by:
A
inclusion/exclusion criteria
8
Q
Measurements section of protocol includes (4):
A
- what variables will be measured
- predictor variable(s)
- outcome variable(s)
- confounding variables
9
Q
A predictor variable typically occurs __. It could have been ___. It is the ___ variable.
A
first; causal; independent
intervention
10
Q
An outcome variable is the __ of the study, aka the ___; it is the ___ variable; ___ per research question
A
goal of the study; endpoint; dependent; one endpoint per RQ
ex= MI, death, stroke, decrease in BP
11
Q
What are confounding variables?
A
anything that can influence/effect the outcome; needs to be controlled;
12
Q
What is the difference between internal and external validity?
A
Internal validity= are the results accurate?
External validity= can the results be applied to the general population?
13
Q
What are the two types of errors in research?
A
- random error: due to chance
2. systematic error: due to bias
14
Q
Random error is due to __. It is distorted in what direction?
A
chance; equally likely to distort study in either direction
15
Q
Systematic error is due to __. It is distorted in what direction?
A
Bias; distorts study in one direction
16
Q
What are the three ways to reduce random error in research?
A
improve study design, increase sample size, increase precision
- sample size has no effect on systematic error
- *precision= consistency (hitting same spot on dart board)
17
Q
What are the two ways to reduce systematic error?
A
improve study design, and increase accuracy
**accuracy= hitting bulls eye
18
Q
What are the three characteristics of a good researcher?
A
skeptical, creative, tenacious
19
Q
To determine if a RQ is feasible, look at (4):
A
number of subjects; technical expertise; cost in time and money; scope
20
Q
Good clinical research __, __ and __ previous findings or provides __
A
confirms, refutes, extends; new findings
21
Q
To confirm that a RQ is novel (3):
A
literature review, consulting experts, search funded research (ex= NIH)
22
Q
An ethical RQ is the key to __. Don’t want study to (2):
A
IRB approval; pose unacceptable risk or invade privacy
23
Q
A good RQ is relevant to (3):
A
scientific knowledge; clinical and healthy policy guidelines; future research
24
Q
Advantages (2) and disadvantages (2) to multiple RQs:
A
adv: efficiency, back up plan
disadv: increases complexity, drawing statistical inferences
25
Three types of RQs:
1. descriptive
2. relational
3. comparative
26
How many variables/groups does a descriptive RQ have?
one variable, one group
27
How many variables/groups does a relational RQ have?
2 or more variables, one group
28
How many variables/groups does a comparative RQ have?
2 or more variables; two or more groups
29
A relational RQ finds __
associations
30
A comparative RQ can determine __ and makes __
cause; predictions
31
Key words that indicate a one sided RQ:
more common, greater/less than
32
Key words that indicate a two sided RQ:
Association or difference
33
A hypothesis is a specific version of the RQ that: __; A form of RQ that establishes: __
summarizes key elements of a study; basis of statistical significance testing
34
The null hypothesis says __ and is the basis for ___. It estimates ___
Says there is no association or difference between groups; Basis for stats significance testing. Estimates probability that association or difference observed is due to chance
35
The null hypothesis estimates the probability that:
association or difference observed is due to chance
36
The alternative hypothesis says __
There is no association/difference between groups; not tested directly; "accepted" by default when Ho rejected
37
Type I error
Rejecting Ho when it is true; false-positive; Alpha= level of significance
38
Type II error
Failing to reject Ho when it is false; False-negative; Beta= level of significance
39
The effect size is an estimate of ___. Greek letter= ___. Typicall based on __
The size of an association or difference that you expect to see; delta; prior studies
40
Power is the probability of ___; or the ability to ___; power formula= __. Inadequate power=
Probability of correctly rejecting Ho if effect in population is greater than or equal to effect size; the ability to find a difference/association when it truly exists; power= 1- beta; inadequate power= type II error
41
P value is the probability of __ if __ is true. It is determined by ___. It is the calculated chance that ___ occurred. Reject Ho if:
statistically signifiant result if Ho is true; determined by statistical tests; calculated chance that type I error occured; p less than or equal to 0.05
42
With multiple hypotheses, the likelihood that at least one will achieve significance on basis of __ increases. Need to adjust __ which is done by:
- chance
| - need to adjust alpha: Bonferroni= divided alpha by number of hypotheses; smaller alpha means increased sample size
43
Smaller alpha (with multiple hypotheses) means increased ___
sample size
44
An observational study design allows the study of __
associations
45
An experimental study design allows the manipulation of ___
predictor variables
46
5 criteria for endpoints:
defined, appropriate, objective, repeatable, measurable
47
A surrogate endpoint is a marker for ___. It must be biologically ___ and associated with ___. Examples (2)=
marker for the risk of an outcome; plausible; associated with outcome of interest
ex= lipid levels for CAD; bone density for risk of fracture
48
A composite endpoint combines ____. Examples:
combines the data of more than one outcome into a single analysis
ex: heart failure trial- hospitalization, worsening symptoms and death
easier to find significance
49
How are confounding variables controlled for in an experimental study?
Randomization
50
How are confounding variables controlled for in an observational study?
statistical methods
51
What are examples of confounding variables?
income level, intake of X, smoking status, diabetes
52
What are the (5) key elements of a RQ?
1. predictor variable(s)
2. outcome variable
3. study population
4. data source (if relevant)
5. timeframe (if applicable)
53
Categorical data includes what phrase in the RQ?
"is there an association"
54
Continuous data includes what phrase in the RQ?
"is there a difference"
55
What type of study design provides the strongest evidence ?
Randomized Controlled Trial
56
The weakest observational study design is ___
Cross-sectional study
57
The strongest observational study design is ___
cohort study
58
___ (type of study design) studies a population at one point in time
cross-sectional study
59
A cross-sectional study is useful for measuring ___ or studying ___
1. measuring prevalence of disease or event
| 2. studying associations
60
Serial surveys are a series of ___ studies. It looks at ___ that are observed ___ to show ___
cross-sectional studies; single population; observed at several points in time; shows changing patterns over time
61
A descriptive report documents ___. It can be ___ or ___. It is ___ generating. Often ___ in research. No ___.
documents experience; prospective or retrospective; hypothesis generating; often 1st step in research; no control
62
Examples of a descriptive report include:
1. case report/ case study - one patient
| 2. case series- more than one patient
63
The benefits to cross-sectional studies and descriptive reports include (4):
1. quick, easy and inexpensive
2. identify ideas for future research
3. determine need for further study
4. no loss to follow-up
64
The drawbacks to cross-sectional studies and descriptive reports include (3):
1. can't infer cause and effect or sequence
2. lots of opportunity for bias
3. no controls
65
A case control study identifies ___ based on __/___. It is always ___. Outcomes expressed as ___
identifies subjects based on an outcome or disease; always retrospective; outcomes expressed as odds ratio
66
Case control study schematic has __ on left side and ___ on right side with arrows pointing ___
Left= exposed/unexposed
Right= outcome/no outcome
Arrows point BACK from outcome to exposure
67
Benefits of case control studies: good for ___; useful in diseases with ___; smaller __ compared to cohort study; can investigate multiple ____; inexpensive and ___; __ generating; useful when data for ___ is unavailable
1. good for rare outcomes/diseases
2. useful in diseases with long latencies
3. smaller sample size compared to cohort study
4. can investigate multiple predictors
5. inexpensive and quick
6. hypothesis generating
7. useful when data for entire cohort unavailable
68
Drawbacks of case-control: Matching __ to __ is challenging; ___ and bias; not __ and ___; not good for ___; can't compute __ or __; temporal relationship difficult to determine between __ and ___; can only study ___
1. matching controls to cases is challenging
2. counfounding and bias
3. not cause and effect
4. not good for rare exposure
5. can't compute incidence or prevalence
6. temporal relationship difficult to determine b/w disease and exposure
7. can only study one outcome
69
Controlling for sampling bias in case-control studies: select __ in same way; matching; use ___ sample of cases; use ___ groups
1. select cases and controls in the same way
2. matching (age, race, gender, etc)
3. use population-based sample of cases
4. use 2+ control groups (from 2 or more settings/locations)
70
Controlling for measurement bias in case-control studies (2)
1. use data/predictors recorded prior to outcome
| 2. blinding
71
Case-crossover studies are useful for :
short term effects of intermittent exposures (ex= radiation from x rays)
72
A case-cross over study is only ___ and each case serves as ___
only retrospective; serves as its own control
73
A cohort study identifies subjects based on __ or ___. The frequency of ___ is compared.
Identifies subjects based on exposure or characteristics; frequency of developing disease or outcome is compared
74
In a prospective cohort study, begin with __ and then look to __
Predictor (exposed/unexposed) and then look to outcome (outcome/no outcome)
75
Benefits of a cohort study: useful for studying __. Multiple __ and __. Good for __ exposures. Less __ than case-control. Able to compute __. Able to identify __.
1. useful for studying frequent outcomes of diseases
2. multiple outcomes and predictors
3. good for uncommon/rare exposures
4. less bias than case-control
5. Able to compute incidence
6. Able to identify temporal trends
76
How is incidence calculated?
who have disease/ # at risk for disease
77
Drawbacks of cohort study: more ___; need larger ___; loss of subjects to ___ (prospective); can't control __/__. Potential for __ from outcome info. __ effect (Prospective)
1. more time and expensive
2. need larger sample size
3. loss of subjects to follow up (prospective)
4. can't control intervention/exposure
5. potential for bias from outcome info
6. Hawthorne effect: changing behavior since you know you're being followed
78
A nested case control study takes a sample of __
cohort study
79
Nested case-control studies are used with __ variables. They are __ to measure and are assessed __
used with predictor variables; expensive to measure; assessed at the end of the study
80
The advantage of a nested case-control study is ___ where you're able to estimate __ and __. Allows for a cohort to be compared to ___
controls represent cohort in general; estimate incidence and prevalence; cohort can be compared for more than one type of outcome
81
Benefits of nested studies: useful for ___; avoids __
Useful for costly measurements (can use data from a previously performed cohort study); avoids bias of conventional case-control
82
Drawbacks of Nested Studies: RQs not always ___
amendable to storing material for later analysis
83
Multiple cohort studies includes 2 or more ___ which are grouped by ___
2 or more separate samples or subjects; exposed group, unexposed group, low-level of exposure group
84
Multiple cohort studies measures ___ and assesses ___
measures predictor variables; assess outcomes
85
Multiple cohort studies are often the only feasible approach for (3):
1. rare exposures
2. potential occupational exposures
3. potential environmental hazard exposure
86
The drawback of multiple cohort studies is __
confounding (cohorts assembled from different populations). Populations may differ beyond just predictor variables which can influence the outcome
87
Odds ratio is defined as ___. It is used with ___ and used in ___ and __ studies.
(# events occurred)/ (# events that didn't occur)
used with logistic regression
used in case control and retrospective studies
88
An odds ratio is interpreted as ___. The significance is interpretedy by :
- the odds of the event is X times higher
| - significance interpreted by 95% confidence intervals ("significant" doesn't include 1)
89
Relative risk is the risk of __ relative to ___. It is used in __ and __ studies. It can't be calculated for __ study
- risk of an event relative to an exposure
- used in RCTs and cohort studies
- can't be calculated for case-control
90
The formula for relative risk is:
RR= (probability of outcome if risk factor present)/ (probability of outcome if risk factor absent)
91
Explanations for associations in observational designs (5):
1. chance (random error)
2. bias (systematic error)
3. effect-cause
4. confounding
5. cause-effect
92
Minimize chance (random error) by:
- increasing sample size
- increasing precison
- calculate p values and CIs
93
Minimize bias (systematic error) by:
- avoid sampling/measurement bias
- collect/obtain additional data
- check consistency with other studies
94
A population is defined in research in three different ways:
- clinically
- temporally
- demographically
95
The target population is ___. It is defined in what two ways?
A large set of people throughout the world to which results will be generalized; defined clinically and demographically
96
An accessible population is a ___ and is defined in what two ways?
- subset of target population available for study
| - defined geographically and temporally
97
A study sample is:
A subset of accessible population who participates in study
98
How is the study population defined?
By inclusion/exclusion criteria
99
The inclusion criteria defines ___ of the target population pertaining to the RQ. In what four ways can they be specified?
# Define main characteristics of target population pertaining to RQ
- clinically
- demographically
- temporally
- geographically
100
The inclusion criteria are the basis for:
Inferences to target population (generalizability; external validity)
101
Common inclusion criteria=
age, disease state, time frame, geographics
102
Exclusion criteria should include subjects with characteristics that could interfere with (3):
- quality of data
- acceptability of treatment
- success of follow up efforts
103
Increase generalizability for your study by (3):
- data collection by mail/telephone
- collaboration
- database research
104
What are the four types of probability samples?
1. simple random sample
2. systematic sample
3. stratified random sample
4. cluster sample
105
What are the two types of samples? Which one is the gold standard?
Convenience samples and probability samples; probability sample= gold standard for ensuring generalizability back to study population
106
A simple random sample is taken by:
assigning numbers to individuals in population and selecting subset of them at random (can use a table of random numbers)
107
When is a simple random sample used?
when selecting a representative subset from a population that is larger than what you need
108
A systematic sample is taken by:
numbering the population first and then selecting the sample in a predetermined periodic process (ex= selecting every 2nd person from list)
109
Systematic samples are susceptible to:
errors: allows investigator to predict and perhaps manipulate who's selected for sample
110
What is the advantage of a systematic sample over a simple random sample?
none
111
A stratified random sample is made by:
dividing population into subgroups by characteristics and selecting randomly from each subset or strata (ex= race or sex)
112
With a stratified random sample, how can you be selective?
Can weight different subsets to draw disproportionately from less common subgroups in population (or subgroups that are of more interest to investigator)
113
What is a cluster sample? When is it useful?
A random sample of natural groupings of those in population; useful for widely dispersed populations
114
What are the 6 criteria for endpoints?
defined, appropriate, objective, repeatable, sensitive, specific
115
What are the two measurement classifications? Which one is stronger?
categorical and continuous; continuous
116
Two classes of categorical data=
nominal and ordinal
117
Precision is defined as the degree of:
reproducibility, reliability and consistency
118
Precision is threatened by ___ such as (3):
random error
| -observer, instrument and subject variability
119
How is precision assessed?
By repeated measurements
120
Precision increases __ to ___
Increases power to detect effects
121
What are the five steps to enhancing precision?
1. standardize measurements
2. training observers
3. refining instruments
4. automating measurements
5. repetition
122
Accuracy is defined as the degree to which:
a variable actually represents what it is intended to represent (increasing accuracy also increases precision)
123
How is accuracy assessed? What does accuracy increase?
Assessed by comparison to reference standard; increases validity of conclusions of study
124
What is accuracy threatened by?
Systematic error
| -observer, subject and instrument bias
125
Validity is defined as:
How well measurement represents phenomena of interest
126
Validity is assessed by (3):
content, construct and criterion-related validity
127
What 7 things enhance accuracy?
1. standardize measurement methods
2. training observers
3. refining instruments
4. automating measurements
5. making unobtrusive measurements
6. calibrating instruments
7. blinding
Advanced Study Design
flashcards
Decks in class (1)
# Cards
