Research Design Flashcards
Research Design
- analogous to PT plan of care
- details re:
- methods by which investigator will interact with subjects
- timing of interactions
- total length of time of study
- type of outcomes to be measured
- extent to which unwanted influences can be minimized
General Features of Research Design
- approach-paradigme: qualitative or quantitative
- study design
- grouping of subjects
- control components: minimize bias
- time elements
True Experimental, Quasi Experimental, and Non Experimental Approaches
- true: randomized, 2+ groups, manipulation
- quasi: non-randomized, 1+ groups, manipulation
- non: no manipulation, may only have 1 group, non-random assignment (predetermined based on naturally occurring subject characteristics), observational studies
Random Selection vs. Random Assignment
-????
Chart
- review table 5.1 in Jewel
- know this chart
Quantitative Research Paradigm
- assumptions: investigator and object of study are independent
- causes and effects can be determined and differentiated
- objectivity (value-rree)
Qualitative Research Paradigm
- assumptions: investigator and subject are interdependent
- impossible to distinguish between causes and effects
- subjectivity (value-bound)
Design Elements: Purpose-Descriptive
- to describe a variable of interest without manipulation of the subjects
- ex: what is the functional status of patients at intervals following TKR?
Design Elements: Purpose-Relationship
- analysis of relationships
- ex: what is the relationship between knee ROM and functional status at intervals following TKR?
Design Elements: Purpose-Differences
- analysis of differences
- ex: what is difference in functional status at hospital discharge between patients who receive PT twice a day and patients who receive PT once a day following TKR?
Variables
- represent a concept
- can have more than one value or level
- 3 main: dependent, independent, extraneous (bias in study, factors you can’t control)
Independent Variable
- aka factor or predictor
- intervention studies: the presumed cause of (or influence upon) a measured effect; variable that is manipulated
- predictive validity and prognosis studies: variable that is presumed to predict an outcome of interest
- level: number of forms or values the independent variable takes in a study. ex treatment group: heat/exercise, manual therapy, HEP so this is 1 independent variable with 3 levels
Dependent Variable
- outcome of interest
- variable that is hypothesized to be caused or predicted by the independent variable
Extraneous Variable
- confounds relationship between independent and dependent variable
- need to be anticipated and controlled for if possible
- also those that you can’t control
Design Elements-Groups
- between subjects design: comparison of independent groups of subjects, measures are compared between 2 or more groups (subject appears in only one level of independent variable)
- within subjects design: repeated measures within same subjects, individual baselines measures are compared to subsequent measures
Design Elements-Control
- bias: results that systematically deviate from truth
- control options: subjects, equipment, environmental conditions, testing and/or training administration, data collection and recording, communication
Design Type-Degree of Control
- experimental is maximal
- quasi is moderate
- non is minimal
Research Designs-Questions About Interventions
- can be experimental or quasi-experimental
- prospective
- may be cross-sectional or longitudinal
- efficacy: intervention produces outcome under ideal conditions
- effectiveness: intervention produces outcome under usual clinical conditions
- treatment must precede outcome of interest
- randomized clinical trial: at least 2 comparison groups, random assignment, blind study, possibly double blind, large subject pool preferred
- factorial design: number of independent variables is included-single factor (one-way, effect of 1 IV), two factor (effect of 2 IVs as well as interaction with one another), multi-factor > 2 IVs
- single system design: not a case report, controlled manipulation of an IV, standardized, repeated measures methodology, at least two levels of IV, n=1, patient is his/her own control, conducted prospectively
Design Options
- A-B: A=baseline phase, B=treatment phase
- does not control for concurrent influenced during treatment phase
- A(1)-B-A(2): A(2)=withdrawal phase, can be used to assess duration of effects as well as to tease out extraneous influences, assumes effects are reversible, ethical considerations
- can only compare adjacent time phases
Group vs. Single-System Design Limitations
- group: power-sample size, subjects comparable on DV at start, large differences between groups at finish-measurement, external validity, practicality
- single system: ethical considerations-treatment delays, treatment withdrawal, costs of care-internal validity, external validity, statistical analysis
Quasi-Experimental Designs-Questions About Interventions
- time series design: single group, repeated measures over time, before and after experimental intervention
- non equivalent control group design: naturally occurring groups, baseline characteristics unlikely to be equal between groups
Research Designs for Questions About Diagnostic Tests and Clinical Measures
- designs are usually non-experimental and cross-sectional
- determine measurement validity and reliability for variable of interest and compare to the “gold standard” or “reference standard”
- random assignment is not a priority since subject must have the condition under study
- clear delineation and application of subject selection criteria
- protocols to ensure consistency in all procedures
- verification of examiner’s competence in performing test of interest
- identification of a “gold standard” as a basis for comparison
- masking examiners from subject grouping or test result
Methodological Studies of Clinical Measures
- development of new or modification of existing measurement instruments
- non experimental
- often cross-sectional but can be longitudinal if objective is to detect change over time
- repeated administrations determine reliability
- comparison to a superior instrument determine validity
- control achieved by: subject selection and recruitment, protocols for development and administration of instruments, statistical adjustment to counter extraneous factors
Research Designs for Questions About Prognostic Factors
- designs assess relationship between prognostic factor and outcome
- nonexperimental
- control achieved through subject selection and statistical adjustment
- causal links between prognostic factor and outcome factor CANNOT be established directly
Cohort Designs
- used for questions about prognostic factors
- same group of subjects followed over time, additional groups can be used for comparison
- group shares common characteristic
- observational descriptive designs provide results to statistical analysis of relationship between prognostic factor and outcomes
- time to ensure outcome occurs is essential
- studies may be prospective or retrospective
- prospective studies have more control
- retrospective studies assure outcome has occurred
Case-Control Designs
- retrospective approach
- subjects with outcome are compared to a control group known to be free of outcome
- goal of study is relative frequency of exposure to the prognostic factor-risk factor identification
Research Designs for Clinical Prediction Rules
- CPR studies are non-experimental
- intended use determine specific features
- diagnostic CPR will have features of diagnostic test studies, as will intervention CPR have features of intervention studies
- as CPR develops, methods to identify clinical indicators will also change to those allowing more control
Research Designs for Outcome Research
- focuses on impact of clinical practice in real world, i.e. end of results care
- nonexperimental designs, primarily observational format with less control features
- results reflect real-world conditions with confounding conditions
- used as basis for assessing quality of care across settings and disciplines
- frequently retrospective in approach, may be cross-sectional or longitudinal
- multiple groups with differing characteristics can be used
- self-reported outcome measures used for studies about disability, health status, satisfaction, or quality of life: operational definitions and development procedures for the measure must be considered
Secondary Analyses
- reports that study other studies
- narrative reviews: analyze individual studies-a low level of evidence
- systematic reviews: true research studies with defined designed features-the highest level of evidence (Cochrane Collaboration is best known source, PEDro also has PT related reviews)
- design elements and controls:…
- specific research question
- inclusion and exclusion criteria for studies
- elaborate and thorough search strategies
- standardized review and abstracting protocols and trained reviewers
- pre-established quality criteria to rate value of individual studies (hierarchies)
Meta-Analysis
- form of systematic reviews
- emphasis on pooling data for additional statistical analyses
- pooling increases sample size and statistical power of analysis
Qualitative Studies
- focus on subject’s thoughts, perceptions, opinions, beliefs, and/or attitudes
- data reported in narrative rather than statistical format
- data collection methods include interviews, surveys, diaries, observations, etc.
- results focus on themes or patterns of data