Research Flashcards
Studies in order of best evidence
Systematic reviews>RTCs> cohort study, case control, cross sectional, case series, case reports, ideas/opinions
Meta analysis
Type of systematic review - has an estimate of effect size by comparing multiple RTCs to determine the effectiveness of a treatment; can minimize the issue of a small sample size
Cohort studies
Observational longitudinal study - looks at a specific group of people with a risk factor and follows to see if they get the disease or not; can be done prospectively or retrospectively; difficult bc a lot of lifestyle factors can influence outcomes
Case control study
Retrospective observational study; individuals with a disease are matched with a comparison group of those without the disease - looking for differences in exposture and occurrence of disease = the odd’s ratio
Cross sectional study
Observational study - data collection done only at one point in time and all participants are tested at relateively the same time. Describes relationships between a disease and factors of interest that exist in that population at a given time; can look at prevalence - but can’t look at newly occurring conditions, and doesn’t give a causal relationship
Case report or series
In depth description of an individual’s condition or response to treatment; cannot test hypotheses or estabilish cause & effect
Descriptive research
Analyzing with the goal of classifying and understanding a clinical phenomenon EX: developmental, normative, qualitative, case report, series
Experimental research
Comparing 2 or more conditions for the purpose of determining cause and effect EX: RTCs, quasi experimental studies, single subject designs
Exploratory research
Examines the dimensions of a phenomenon of interest and its relationships to other factors; EX: cohort studies, case control, historical research, methodological
Qualitative research
Data from observation/interviews focused on meaning and interpretations to gain an understanding in thoughts and opinions or develop hypotheses; global, probing, small sample size, non statistical, exploratory or investigative - findings cannot be used to make generalizations
Quantitative
Data or measurements that are analyzed via statistics with the goal of quantifying data to generalize results onto a population; non-probing, specific, large sample size, objective observer, used to recommend a final course of action
Respect for persons
Refers to individuals rights to make autonomous decisions about their health care
Beneficent
The obligation of the researcher to provide for the well being of their subjects by maximizing benefits and minimizing the possible harm
minimal risk
The magnitude of harm or discomfort anticipated in the research is not greater than that ordinarily encourntered in daily living
Vulnerable populations
Typically minors, those with diminished capacity to consent, pregnant women, human fetuses, neonate, non-English speaking, prisoners, students
Continuous data
Can assume any value along a continuous scale that covers a range of values without gaps or interruptions
Discrete data
Measured in whole units: HR, number of clinic visits, etc.
Dischotomous data
Type of discrete data; limited to only two values: gender, smoking vs non smoking, etc.
Qualitative data
Categorical; non numeric
Quantitative data
Measurements or numerical value
Nominal scale
Classification scale; Ea object and person can only be assigned to one category; ex: blood type, breath sound
Ordinal scale
Ranking scale, based on the property of the variable; muscle grading tests, level of assistance, joint laxity scale
Interval scale
Measurement scale where intervals are equal and there is no true zero point; Ex: temperature on the F or C scale
Ratio
measurement scale where the intervals are equal and there is a true zero: ex; ROM, distance walked, time to complete an activity
Alternate forms reliability
Parallel forms reliability; 2 things are assessing the same thing consistently and accurately; ex: different NPTE tests can be administered throughout the year as long as their reliability is the same
Internal consistency
The extent that items or elements that contribute to a measurement reflect one dimension; ex: a functional assessment scale should only include things that look at pt’s function
Intrarater reliability
Consistency of one person repeating the same measure
Inter-rather reliability
Consistency of a test being measured by different people
Test-retest reliability
Consistency or equivalence of repeated measurements performed on the same person
Face validity
Degree of which a measurement tests what it is supposed to
Content validity
Degree in which a measurement reflects meaningful elements; ex: location, and type of pain, not just number of pain
Construct validity
Degree of which a theoretical construct is measure by a test or measure; ex: MMT to innervation status of a muscle.
Criterion related validity
Validity of a measure is determined by comparing it to a gold standard
Concurrent validity
Criterion related validity; measurement is compared to a gold standard at the same time as it is being tested by something else
Predictive validity
Criterion related validity; measurement is considered valid because it predicts future behavior or events; ex: GRE to predict grad school success
Prescriptive validity
Criterion related validity; the measurement suggests the form of treatment the pt should receive, then measured by the successful outcome of the treatment
Sampling error
Chance difference between statistical calculated from a sample and the true value of the parameter in the population; inherent in the use of sampling
Sampling with replacement vs without
Ea unit sampled is put back in the population before the next is drawn, ea unit truly has an equal chance of being selected; not used with humans; without: no placed back, reduces the size of the population
Probability sampling
Sampling that uses random selection
Systematic sampling
Every ‘n’th number in a population is selected; simplicity
Stratified random sampling
Population is dived into homogenous subgroups then simple random sampling from each group; this assumes that the sample will be representative of key subgroups as well as the overall population
Cluster sampling
Population divided into clusters, usually based on geography; then random samples of the clusters are taken; less costly and more efficient than simple random sampling
Downfall to simple random sampling
May not be representative of population; not statistically efficient
Non probability sampling
Any method of sampling that does not include random selection
Convenience sampling
Using participants that are convenient to the researcher
Purposive sampling
Selected based on key criteria
Quota sampling
Subjects from ea subgroup are not randomly chosen, convenience sampling is used to choose from ea group
Snowball sampling
Subjects identify other subjects - used when the characteristic being studied is rare
Completely randomized design
Parallel group design; subjects are randomly assigned to different groups and ea group receives a unique intervention and then results are compared at the end
Crossover design
Research design where subjects receive both treatments in random order separated by a period of no treatment; ea subject serves as their own control
Factorial design
Research design in two or more independent variables are investigated with different subjects assigned to different combinations of the varibales
Pre-test/post-test control group design
Compares the outcomes of two or more groups formed by random assignment by testing all groups before & after treatment; basic format of a RCT
Posttest only control group
Compares the outcomes of two or more groups formed by random selection, by testing all groups only after the treatment
Repeated measures design
Subjects are tested under all conditions, ea person acts as their own control: within subjects design
Sequential clinical trial
Research design where the data is analyzed as it becomes available, so the trial can be stopped as soon as the evidence is sufficent to show a significant difference
Single subject design
Drawing conclusions about the effects of a treatment based on the responses of a single patient
Quasi-experimental design
Research without a control group, random assignment, or both
One group pre-test/post-test design
Research where measurements are made on group before and after treatment
One way repeated measures design
Extension of the one group protest-posttest; measurements are made on one group and multiple specific time intervals
Time series design
Multiple measurements are made before and after treatment to observe patterns or trends during pre/post treatment periods
What are the 3 types of quasi-experimental designs?
One group pre-test/post-test
One way repeated measures over time
Time series design
Triple blinding
The subject, certain members of the team, and data analyzer are unaware of the research hypothesis
Active control group
A known effective treatment is provided
Matching/pairing
Identify pairs of subjects that have identical characteristics; i.e. Twin studies
Intention to treat analysis
Analysis of all subjects randomly assigned to one of the treatments, regardless of whether they received or completed that treatment; preserves the balance of subject groups achieved through randomization
External validity
Degree to which results are generalizable to populations beyond those in the study; threats: treatment interaction within setting and time
Internal validity
Degree in which an intervention is the cause of the outcome measured in the study, and not due to extraneous factors. Threats: maturation, attrition, testing, instrumentation, regression toward the mean
Hawthorne effect
An untreated subject experiences a change simply from participating in a study
Alternate hypothesis
The experiemental hypothesis; Ha or H1
Null hypothesis
Statistical hypothesis
Independent variable
Caused the effect
Dependent variable
Response or outcome caused by the independent variable
P-value
Probability that a particular result could have happened by chance
Alpha level
The significance level; usually 0.05; the probability of rejecting the null hypothesis when it is true or of committing a type I error
If the p-value is less than vs. greater than the alpha level
Less than: reject the null hypothesis; greater than: do not reject
Type I error
Concluding there is a significant difference when there is not; incorrectly rejecting the null
False positive
Type II error
Stating there is not a difference when there is; incorrectly not rejecting the null
False negative
Effect size
Measure of the magnitude of the difference between two treatments or values; the larger the ES, the more likely it will be statistically significant
Effect size index
Difference between 2 groups divided by the SD
<0.1 trivial effect; >0.5 large effect
Minimal detectable difference or change (MDC)
Minimum chance in a patients condition beyond a measurement of error
Box & whisker plot - 5 values
Min score, Q1, median (Q2), upper quartile (Q3), max score
Forest plot
Used in meta-analysis to show results of individual studies in addition to cumulative summary through a diamond shape
Histogram
Displays distribution of data through plotting frequency
Kurtosis
Describes the peak of a distribution: high kurtosis has a sharp peak, while a low has a flatter peak
Bell curve
68% of data within 1SD above and below the mean
95% within 2
99% within 3
Measuring skewness - positive, negative
Positive: mean/median are to the R of the mode (near the tail)
Negative: to the L of the mode (near the tail- tail to L)
Coefficient of variation
The ration of the standard deviation of a distribution to the mean
Variance
SD squared, usually not reported
Analysis of variance (ANOVA)
Used to test the equality between two or more populations; one way: separated into groups based on only one characteristic vs. two way: separated based on 2 characteristics or factor
Repeated measures ANOVA
All individuals are measured under a different number of conditions
Regression analysis
Predicts how a change in one or more independent variables affects the dependent variable
Interclass correlation (ICC)
Assess both the degree of correspondence and agreement among scores
Pearson’s r
From -1.0 to +1.0; indicates the direction and magnitude of the relationship
Dependent/paired t-test vs. independent t test
Compares the means of two groups that are correlated vs. compares two independent groups
One sample t test
Compares to an expected or reference mean
One tailed test
Statically symmetrical distribution, used only when a test can have only one effect: ex: helpful vs. not helpful
Z-test
Used when the population is normally distributed
Chi square test
No parametric statistics, compares between categorical varibales
Kruskal Wallis test
Used to determine if 3 or more independent samples come from the same population; nonparametic version of one way anaylsis
Mann Whitney test
Compares two independent samples with ordinal level data; independent t test comparison
Spearman rank correlation coefficient
Correlational test for ordinal values; comparable to pearson’s r
Wilcoxon signed rank test
Compares two dependent samples with ordinal data; comparable to paired t
Sensitivity vs. specificity
Sen: with high sensitivity negative result = rule out (SnNout)
Sp: high specificty, positive rules in = SpPin
Relative risk
Measure of risk for those exposed to the same risk factor; <1.0 increased risk
Odds ratio
Measure of the odds of something happening; <1.0 exposure may reduce risk, >1.0 exposure may increase risk
Number needed to treat
Number of patients that need to be treated to prevent one bad outcome; the higher the NNT, the less effective the treatment
